Open Access
Add to BibSonomy
Abstract
We present the design and experimental characterization of the first multistage ring-assisted Mach-Zehnder interferometer (RAMZI) lattice (de)multiplexer implemented with silicon nitride optimized for four channels with a spacing of 100 GHz in the L-band. The device comprises two RAMZI stages to provide a sharp box-like response characterized by a shape factor of 0.9, a flat passband over the entire channel, and a crosstalk level better than -14 dB. The maximally flat passband of the demultiplexer enables a passband width twice that of the maximum spectral excursion defined in the NG-PON2 standard.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Increasing the capacity of existing fiber networks using wavelength division multiplexing (WDM) is driving the demand for high performance integrated photonic filters. Low-loss, flat-top, and low crosstalk filters are needed for (de)multiplexing closely spaced wavelength channels. Different structures have been employed to implement such filters, including echelle diffraction gratings [1], arrayed-waveguide gratings [2,3], Mach-Zehnder Interferometers (MZIs) [4], and resonator-based filters [5]. Among the mentioned interference-based devices, MZIs and ring resonators are of interest due to their compact size, low insertion loss, and spectral flatness, especially for applications that require a small number of channels, such as access networks. A flat passband in filters and (de)multiplexers ensures signal integrity, reduces distortion, provides an efficient utilization of the available bandwidth, and enhances the overall system performance. Furthermore, the design flexibility of interference-based devices allows them to be easily integrated with other photonic components.
MZIs are a fundamental building blocks for implementing interference-based (de)multiplexers. While stages made of a single MZI can be used to implement lattice filter based (de)multiplexers [6], achieving low crosstalk and a flat-top response requires stages comprising multiple cascaded MZIs [7,8], resulting in an increased complexity and insertion loss. Micro-ring assisted MZIs can help reduce the size of (de)multiplexers while enhancing their performance by improving the phase control capability of the system. By introducing a pole in the transmission response of the filter, the finite impulse response of MZIs is effectively transformed into an infinite impulse response. The filter response differs from that of a typical asymmetric MZI since the pole introduced by the micro-ring resonator results in a nonlinear phase response within one arm of the MZI [9]. Extensive efforts were made to improve the passband as well as the stopband properties of these filters by employing various arrangements of rings and MZIs [10–12]. In a single stage filter, coupling a ring resonator to one or both arms of the MZI increases the order of the filter and provides a more box-like response [13].
For applications in high-density WDM systems, multiple stages must be cascaded to increase the number of channels. It has been shown that cascading lattice-form filter stages in a logarithmic configuration results in (de)multiplexers with inherently low crosstalk and flat passbands without limitations on the number of output ports other than the added propagation losses and larger device footprint [14]. The filters in each stage can be simple MZIs or ring-assisted MZIs (RAMZIs), depending on the requirements of the application. Therefore, choosing the proper design for each filter stage is critical to achieve a flat passband and a sharp roll off, which will result in sufficient rejection in the stopband [7].
Considerable progress has been made toward developing lattice (de)multiplexers on commonly used silicon-on-insulator (SOI) platforms [15–21]. However, fabricated devices either suffer from high crosstalk or do not show a perfectly flat passband, even if simulation models predict an ideal response. Moreover, the high refractive index contrast of the SOI technology and the resulting small feature size makes the devices particularly sensitive to fabrication variations and temperature fluctuations. Meanwhile, silicon nitride (SiN) photonics provide a moderate effective index, lower losses, can be integrated with SOI [22], and have a lower thermal sensitivity [23]. These advantages help create high performance passive devices.
In this article we present, to the best of our knowledge, the first SiN-based realization of a lattice (de)multiplexer using a logarithmic configuration. Two types of filters are studied theoretically, and then the optimal arrangement is implemented. Our proposed structure offers a more flat-top and box-like response in comparison to the ones using cascaded ring-less MZIs as building blocks for each stage [23,24]. Moreover, since RAMZIs are sensitive structures, whereby the coupling coefficient can be significantly affected by fabrication variations, the lower refractive index contrast of SiN photonics can improve their tolerance to fabrication variations.
2. Design
The (de)multiplexer is designed to separate four channels used for downstream transmission in use cases such as NG-PON2 access networks [25]. The center frequencies of these channels correspond to 187.5 THz, 187.6 THz, 187.7 THz, and 187.8 THz, which is in the L-band. To implement a 4-channels lattice (de)multiplexer, two filter stages with different free spectral ranges (FSRs) are required. The properties of the filter used in the first stage is critical since it determines the flatness of the transmission window and adjacent channel isolation of the device. It has been shown that incorporating more rings in the MZI enables a faster roll-off of the band edges in the transmission spectrum [26,27]. This can be achieved by creating a constant phase difference between the two arms of the MZI. The phase difference as a function of wavelength between the arms of ring-assisted MZIs with one (R1), two (R1 & R2) or three rings (R1, R2 & R3) located along the arms is illustrated in Fig. 1(a). Figure 1(b) shows the corresponding configuration of the rings. The phase difference can be minimized by integrating more rings in the MZI, at the expense of increased structure complexity.
Fig. 1. (a) Phase difference as a function of wavelength between the arms of a single ring, double rings, and triple ring-assisted MZIs, and (b) schematic showing how the rings are incorporated into the MZI.
Our proposed design implements a double RAMZI in the first stage of the (de)multiplexer as a trade-off between filtering performance and design complexity. According to our simulation study, a crosstalk level below 30 dB and a box-like response is expected when the coupling coefficients of the rings to the MZI arms are k1= 0.92, and k2= 0.42 for the upper (R1) and bottom (R2) rings, respectively. The optimum combination of coupling coefficients was found by numerically optimizing the transfer function of the double ring assisted MZI. The coefficients were varied in the range of 0 < k < 1 to observe their effects on the flatness and crosstalk level of transmission response. A propagation loss of 2 dB/cm was considered in the simulations. The crosstalk in the first stage determines the minimum achievable overall level with the (de)multiplexer. Thus, the second stage can have a simpler filter design; however, its response must not affect the transmission spectrum of the first stage. It should provide a flat response over the entire passband and a complete roll-off in the stopband of the first stage since the output of the device is the product of the transfer functions of both stages. In the following, we investigate the final transmission spectrum of the (de)multiplexer when two different types of filters are used in the second stage: a single stage MZI or a RAMZI. Figure 2(a) shows the individual transmission response of both the first and second stages, as well as the overall transmission response of the (de)multiplexer when a simple MZI filter is used in the second stage. It is clear from the figure that the band edges of the MZI response have a relatively moderate roll-off, which results in increased crosstalk. The flatness of the passband of the (de)multiplexer is also affected by the smoothly curved passband response of the MZI. Although the single stage MZI filter can be simpler to fabricate, it deteriorates the overall transmission response of the device by increasing the crosstalk level from 30 dB to 10 dB, as shown in Fig. 2(a). On the other hand, a single ring-assisted MZI in the second stage with a ring coupling coefficient of k3 = 0.84 provides a crosstalk level below -30 dB, as illustrated in Fig. 2(b). The results were obtained using the transfer function of the MZI and ring-assisted MZI and computed with MATLAB [9]. Increasing the number of rings in the second stage does not provide any significant improvements to the response of the device, thus a RAMZI with a single ring is used in the second stage of the (de)multiplexer.
Fig. 2. Transmission response of the first stage, the second stage and the complete (de)multiplexer for a second stage composed of (a) a MZI filter, and (b) a single ring-assisted MZI.
The performance of the device is highly sensitive to small deviations of the coupling coefficients from their optimized values, as shown in Fig. 3. For instance, the crosstalk level decreases from 30 dB to 20 dB when the value of k1 changes from 0.92 to 0.96.
Fig. 3. Transmission response of the (de)multiplexer for different coupling coefficients. The black line represents the optimal parameters (k1= 0.92, k2= 0.42, and k3= 0.84). The other lines show the impact of increasing one of the coefficients by 0.04 while keeping the other parameters constant (blue k1= 0.96, green k2= 0.46, and red k3= 0.88).
3. Device structure and simulation parameters
The layout of the proposed structure is shown in Fig. 4(a), which consists of two RAMZI stages. Trapezoidal SiN strip waveguides with a side-wall angle of 86°, a base width of 850 nm and a height of 435 nm are used to implement the device. The height corresponds to the SiN thickness used in the fabrication process, and the width is the largest possible to ensure single mode operation with optimal propagation loss and sensitivity to fabrication variations. Racetrack ring resonators are used to achieve and tailor the required coupling coefficients. The device is optimized for the TE polarization at a wavelength of 1600 nm.
Fig. 4. (a) Schematic of the (de)multiplexer. R1 = 189.972 um, R2 = 203.341 um, R3 = 97.889 um, R4 = 97.812 um, and (b) simulated transmission response.
A commercial software package (Ansys Lumerical, Vancouver, Canada) was used to simulate the response of the (de)multiplexer. The wavelength-dependent coupling coefficients and waveguide characteristics are extracted from 3D finite-difference time-domain and finite difference eigenmode solver simulations. The complete device is simulated using Lumerical Interconnect.
The 100-GHz FSR required for the (de)multiplexer determines the circumference of the rings, which is calculated using the equation $FSR = \lambda /{n_g}L$, where L and ${n_g}$ are the ring circumference and the group index of the waveguide, respectively. The circumference of the rings in the second stage is half the one of those in the first stage. Furthermore, the circumference of the ring resonators is twice the length of the longer arm of the MZI in each stage. This is required to obtain the desired phase shifts and interference effects [9]. Since the coupling length varies as a function of the coupling coefficient, the rings have different radii. In the first stage, the rings have radii of R1 = 189.972 µm and R2 = 203.341 µm, respectively. To achieve the required coupling coefficients, we used a straight directional coupler section with gaps of 1 µm and 1.1 µm, and with coupling lengths of 155 µm and 113 µm for the top and bottom rings, respectively. The top and bottom rings in the second stage have radii of R3 = 97.889 µm and R4 = 97.812 µm, respectively. The rings have coupling lengths of 68.5 µm, and the gap between the waveguides is 0.8 µm. Finally, 3-dB multi-mode interferometers (MMIs) are used in the MZIs and between the two stages since they are more tolerant to wavelength and fabrication variations than directional couplers [24]. The MMI couplers are implemented with a rectangular core of 163 µm in length, 15.5 µm in width, and 0.435 µm in thickness. Furthermore, the output ports are terminated with a 50 µm long taper transitioning from an opening width of 3.8 µm to a waveguide width of 0.85 µm.
The simulated transmission response of the (de)multiplexer is shown in Fig. 4(b). Each channel has a flat passband with an extinction ratio in excess of 30 dB. A shape factor (SF) of 0.87, defined as the ratio of the 1-dB bandwidth over the 10-dB bandwidth, confirms the nearly square passbands. Such a box-like response allows to use the entire passband of the device. According to the ITU standard, the maximum spectral excursion (MSE) should stay within a range of ±20 GHz around the central frequency for NG-PON2 filters operating on a 100 GHz grid [28]. The flat passband of the current design offers a width that is twice that requirement.
4. Fabrication
The fabrication process started by forming a 3.2 µm thick SiO2 layer onto a silicon wafer using tetra-ethyl-ortho-silicate (TEOS) low-pressure chemical vapor deposition (LPCVD). Subsequently, a 435 nm thick SiN layer was deposited through LPCVD. The pattern of the SiN waveguide was then defined using UV lithography with a stepper and followed by dry etching. Finally, a 3.2 µm thick SiO2 cladding was deposited utilizing TEOS plasma enhanced chemical vapor deposition (PECVD).
Despite the lower index contrast provided by the SiN waveguides, ring resonators remain sensitive to fabrication variations. These imperfections can be compensated by placing metal heaters on the racetrack resonators and along the long arm of the MZIs to rectify the phase mismatch between the optical paths. Heaters made of aluminum copper (AlCu) were therefore built over the rings, the coupling regions, and the long arm of the MZIs. A 250 nm thick layer of AlCu was sputtered onto the top-cladding and then wet-etched to create 15 µm wide heater wires. The thickness of the top cladding is sufficient to isolate the electromagnetic field of the optical mode from the metal heaters and thus no additional losses are introduced. The fabrication was performed by our industrial partner in a commercial foundry. After fabrication, the heaters were wire-bonded out to a printed circuit board to be controlled by current sources. Figure 5 shows the fabricated device. The heaters and their bonding pads are depicted in a light-blue color.
5. Results and discussion
The response of the prototype was characterized by launching TE-polarized light into the input port through a surface grating coupler. The light from a tunable laser (EXFO T100S-HP) was coupled to a fiber array with a 30° polish angle, which needed to be precisely aligned at an angle of 8.0° with respect to the surface gratings. The input and the four output ports were connected to different fibers in the array, which allowed to measure the transmitted spectrum of each output simultaneously with a component tester (EXFO CT440). As expected, fabrication variations affected the initial performance of the (de)multiplexer, which is shown in Fig. 6(a). The poor response was mainly due to the optical length imbalance between the rings and the long arm of the MZIs. By applying voltages between 0.3 V–2 V to each heater, the fabrication imperfections could be mitigated. Tuning the rings and the long arm of the first stage flattened the passband, whereas heating the rings and long arms of the second stage compensated the mismatch between the center frequencies of both stages. These adjustments were made while monitoring the outputs until the response was optimized. The current passing through the heaters was limited to 300 mA to avoid damaging them. The power consumption of each section is summarized in Table 1, with a total power consumption of 1.576 W.
Table 1. Power consumption of each heater used for thermal tuning
The transmission spectrum of the (de)multiplexer after tuning is shown in Fig. 6(b). The measurement results are normalized with respect to the response of the grating couplers in order to remove the wavelength dependent coupling losses. Considering the measured fabrication propagation loss of 2 dB/cm, we removed the loss of the waveguides used to route the inputs and outputs to the grating couplers, such that the overall insertion loss of the (de)multiplexer is 7.12 dB, which includes the loss caused by the propagation inside the (de)multiplexer, the MMI couplers, and the bending and coupling losses of the ring resonators. The transmission spectrum shows a very sharp roll-off at the band edges and a flat passband over the entire channel spacing. The measured shape factor and crosstalk are 0.9 and -14 dB, respectively. In comparison to previous SiN (de)multiplexers based on AWGs [29,30], cascaded MZI [23,24,31,32], and other filter types [33,34], our fabricated device shows a flatter and more box-like response. The reason for the higher crosstalk in the prototype compared to simulations is the high sensitivity of the structure to variations in coupling coefficients. Unfortunately, it was impossible to tune the coupling between the rings and their bus waveguide to the optimal values in the current configuration. This limitation is due to the short lengths and small gaps used to implement the couplers. Using larger gaps and longer coupling lengths would provide a broader tuning range, which would allow to reach the optimum coupling coefficients, but it may come at the cost of increased losses. A comparison with state-of-the-art SiN based demultiplexers is presented in Table 2, where we show that while AWGs can support more channels, they do not provide a flat top response. Cascaded MZI, cascaded MMI, and contra-directional couplers (contra-DCs) have been used for smaller channel counts. Although the crosstalk of our (de)multiplexer is higher than some of the reported devices, it provides a constant loss in all channels, and exhibits a better box-like response. We employed the xyExtract software tool [35] to calculate the shape factor of the devices listed in Table 2 by extracting data points from the measurement results presented in the other works. We also used this software to evaluate the uniformity across channels when it was not reported explicitly. As shown in Table 2, our (de)multiplexer provides a significantly higher SF than in previously reported SiN devices. Moreover, its response is highly uniform in comparison to many of the reported devices and it does not have ripples in the passband. Nonetheless, our device shows a relatively higher loss. The insertion loss of 7.12 dB is mainly due to the 4 dB loss incurred by the MMIs (around 1 dB for each MMI). Indeed, lower loss were achieved in cascaded MZI devices that used directional couplers instead of MMIs, yet with reduced response uniformity across channels. Improving the design of the MMIs could significantly reduce the insertion loss of the (de)multiplexer.
Table 2. Comparison of SiN based demultiplexers
We also investigated the impact of thermal tuning on the transmission spectrum of the device. Figure 7 shows the change in the spectrum when the current in the heaters over the rings is reduced from the optimum values. As can be seen in Fig. 7(a), reducing the current in the heater in the upper ring (R1) induces a phase discrepancy between the top (R1) and bottom (R2) rings of the first stage, predominantly altering the filter passband characteristics. A uniform temperature reduction across both rings shifts the center frequency of the passband. Furthermore, in Fig. 7(b), adjustments to the heater of the ring in the second stage (R3) significantly influence the device stopband, as deviations from the optimal setting create a misalignment between the center frequencies of the two stages.
Fig. 7. Thermal tuning effect on the ring resonators of (a) the first stage (R1 and R2), and (b) the second stage (R3).
6. Conclusion
We demonstrated a 1 × 4 SiN (de)multiplexer operating in the L-band, implemented with two cascaded ring-assisted MZI stages. The design is based on SiN to minimize losses and reduce sensitivity to fabrication variation. The prototype has a net insertion loss of 7.12 dB. Two different filter types were investigated theoretically to optimize the performance of the (de)multiplexer. It was shown that employing a ring-assisted MZI configuration in both stages can maximize the flatness, shape factor, and reduce crosstalk. A box-like response with a crosstalk level better than 14 dB, and a shape factor of 0.9 were demonstrated experimentally. Using RAMZI filters in both stages led to a better performance in terms of crosstalk and box-like response than previously reported in SiN-based cascaded MZI and AWG (de)multiplexers. We are currently working on developing the best method to compensate imperfections in the coupling region.
Funding
Natural Sciences and Engineering Research Council of Canada (CRDPJ 530551-18); AEPONYX inc (CRDPJ 530551 -18); PRIMA Quebec (R16-46-002 PSO).
Acknowledgements
The authors would like to thank AEPONYX Inc. for the support and fabrication of the prototype.
Disclosures
LA: AEPONYX (F), FN: AEPONYX (F,P), MM: AEPONYX (F,P)
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. J. Brouckaert, W. Bogaerts, P. Dumon, et al., “Planar concave grating demultiplexer fabricated on a nanophotonic silicon-on-insulator platform,” J. Lightwave Technol. 25(5), 1269–1275 (2007). [CrossRef]
2. A. van Wijk, C. R. Doerr, Z. Ali, et al., “Compact ultrabroad-bandwidth cascaded arrayed waveguide gratings,” Opt. Express 28(10), 14618–14626 (2020). [CrossRef]
3. S. Pathak, M. Vanslembrouck, P. Dumon, et al., “Optimized silicon AWG with flattened spectral response using an MMI aperture,” J. Lightwave Technol. 31(1), 87–93 (2013). [CrossRef]
4. S.-H. Jeong and Y. Tanaka, “Silicon-wire optical demultiplexers based on multistage delayed Mach–Zehnder interferometers for higher production yield,” Appl. Opt. 57(22), 6474–6480 (2018). [CrossRef]
5. S. Papaioannou, D. Fitsios, G. Dabos, et al., “On-chip dual-stream DWDM eight-channel-capable SOI-based MUX s/DEMUX s with 40-GH z channel bandwidth,” IEEE Photonics J. 7(1), 1–10 (2014). [CrossRef]
6. K. Ramachandran, N. Kumar, and V. Sahoo, “A highly accurate all-fiber MMZI current sensor,” IEEE Sens. J. 15(2), 1270–1274 (2014). [CrossRef]
7. Q. Deng, L. Liu, R. Zhang, et al., “Athermal and flat-topped silicon Mach-Zehnder filters,” Opt. Express 24(26), 29577–29582 (2016). [CrossRef]
8. S.-H. Jeong, D. Shimura, T. Simoyama, et al., “Si-nanowire-based multistage delayed Mach–Zehnder interferometer optical MUX/DeMUX fabricated by an ArF-immersion lithography process on a 300 mm SOI wafer,” Opt. Lett. 39(13), 3702–3705 (2014). [CrossRef]
9. C. K. Madsen and J. H. Zhao, “Optical filter design and analysis: a signal processing approach,” (1999).
10. C. Madsen, “A multiport frequency band selector with inherently low loss, flat passbands, and low crosstalk,” IEEE Photonics Technol. Lett. 10(12), 1766–1768 (1998). [CrossRef]
11. C. Madsen and J. H. Zhao, “A general planar waveguide autoregressive optical filter,” J. Lightwave Technol. 14(3), 437–447 (1996). [CrossRef]
12. N. Kohli, B. L. Sang, F. Nabki, et al., “Tunable bandpass filter with serially coupled ring resonators assisted MZI,” IEEE Photonics J. 13(4), 1–8 (2021). [CrossRef]
13. C. Madsen, “Efficient architectures for exactly realizing optical filters with optimum bandpass designs,” IEEE Photonics Technol. Lett. 10(8), 1136–1138 (1998). [CrossRef]
14. G. Cincotti and A. Neri, “Logarithmic wavelength demultiplexers,” J. Lightwave Technol. 21(6), 1576–1583 (2003). [CrossRef]
15. N. Ning, X. Wang, H. Yu, et al., “Comparison of silicon lattice-filter-based O-band 1× 8 (de) multiplexers with flat and Gaussian-like passbands,” IEEE Photonics J. 14(2), 1–5 (2022). [CrossRef]
16. A. Liu, L. Liao, Y. Chetrit, et al., “Wavelength division multiplexing based photonic integrated circuits on silicon-on-insulator platform,” IEEE J. Sel. Top. Quantum Electron. 16(1), 23–32 (2009). [CrossRef]
17. F. Horst, W. M. Green, S. Assefa, et al., “Cascaded Mach-Zehnder wavelength filters in silicon photonics for low loss and flat pass-band WDM (de-) multiplexing,” Opt. Express 21(10), 11652–11658 (2013). [CrossRef]
18. Z. Zhao, Z. Li, J. Niu, et al., “Eight-channel LAN WDM (de) multiplexer based on cascaded Mach–Zehnder interferometer on SOI for 400GbE,” in Photonics, (MDPI, 2022), 252.
19. S.-H. Jeong, S. Tanaka, T. Akiyama, et al., “Flat-topped and low loss silicon-nanowire-type optical MUX/DeMUX employing multi-stage microring resonator assisted delayed Mach-Zehnder interferometers,” Opt. Express 20(23), 26000–26011 (2012). [CrossRef]
20. S.-H. Jeong, Y. Tanaka, and K. Morito, “1× 4 channel Si-nanowire microring-assisted multiple delayline-based optical MUX/DeMUX,” J. Lightwave Technol. 33(17), 3736–3743 (2015). [CrossRef]
21. D. Munk, M. Katzman, Y. Kaganovskii, et al., “Eight-channel silicon-photonic wavelength division multiplexer with 17 GHz spacing,” IEEE J. Sel. Top. Quantum Electron. 25(5), 1–10 (2019). [CrossRef]
22. R. Baets, A. Z. Subramanian, S. Clemmen, et al., “Silicon Photonics: silicon nitride versus silicon-on-insulator,” in Optical Fiber Communication Conference, (Optical Society of America, 2016), paper Th3J. 1.
23. G. Gao, D. Chen, S. Tao, et al., “Silicon nitride O-band (de) multiplexers with low thermal sensitivity,” Opt. Express 25(11), 12260–12267 (2017). [CrossRef]
24. J. Mikkelsen, A. Bois, T. Lordello, et al., “Polarization-insensitive silicon nitride Mach-Zehnder lattice wavelength demultiplexers for CWDM in the O-band,” Opt. Express 26(23), 30076–30084 (2018). [CrossRef]
25. J. S. Wey, D. Nesset, M. Valvo, et al., “Physical layer aspects of NG-PON2 standards—Part 1: Optical link design,” J. Opt. Commun. Netw. 8(1), 33–42 (2016). [CrossRef]
26. L.-W. Luo, S. Ibrahim, A. Nitkowski, et al., “High bandwidth on-chip silicon photonic interleaver,” Opt. Express 18(22), 23079–23087 (2010). [CrossRef]
27. Z. Wang, S.-J. Chang, C.-Y. Ni, et al., “A high-performance ultracompact optical interleaver based on double-ring assisted Mach–Zehnder interferometer,” IEEE Photonics Technol. Lett. 19(14), 1072–1074 (2007). [CrossRef]
28. G. Itu-T, “989.2 40-Gigabit-capable passive optical networks 2 (NG-PON2): Physical media dependent (PMD) layer specification,” Intellectual Property rights in ITU-T (2014).
29. Q. Han, M. Ménard, and W. Shi, “Superlattice arrayed waveguide grating in silicon nitride,” IEEE Photonics Technol. Lett. 32(22), 1411–1414 (2020). [CrossRef]
30. B. I. Akca and C. R. Doerr, “Interleaved silicon nitride AWG spectrometers,” IEEE Photonics Technol. Lett. 31(1), 90–93 (2018). [CrossRef]
31. S. S. Cheung and M. R. Tan, “Silicon nitride (si 3 n 4)(de-) multiplexers for 1-µm cwdm optical interconnects,” J. Lightwave Technol. 38(13), 3404–3413 (2020). [CrossRef]
32. S. Tao, Q. Huang, L. Zhu, et al., “Athermal 4-channel (de-) multiplexer in silicon nitride fabricated at low temperature,” Photonics Res. 6(7), 686–691 (2018). [CrossRef]
33. W. Shi, H. Yun, C. Lin, et al., “Ultra-compact, flat-top demultiplexer using anti-reflection contra-directional couplers for CWDM networks on silicon,” Opt. Express 21(6), 6733–6738 (2013). [CrossRef]
34. J. Menahem and D. Malka, “1× 4 wavelength demultiplexer C-band using cascaded multimode interference on SiN buried waveguide structure,” Materials 15(14), 5067 (2022). [CrossRef]
35. A. Rohatgi, “WebPlotDigitizer” (September, 2022), retrieved https://automeris.io/WebPlotDigitizer.
36. Q. Yi, S. Zheng, Z. Yan, et al., “Silicon photonic flat-top WDM (de) multiplexer based on cascaded Mach-Zehnder interferometers for the 2 µm wavelength band,” Opt. Express 30(15), 28232–28241 (2022). [CrossRef]
