Open Access
Add to BibSonomy
Abstract
We demonstrate a 2 × 2 16-ch silicon photonics wavelength-selective switch consisting of contra-directional couplers and thermo-optic Mach-Zehnder switches. The contra-directional couplers are based on sidewall corrugated Bragg gratings with an unlimited free spectral range and thus enable the device to operate over a very wide wavelength range of the C- and L-band. We obtain a fiber-to-fiber insertion loss of 9.2 dB, an on-chip loss of 5.4 dB, a 3-dB bandwidth of 4.2 nm, a bar-port extinction of 15.0 dB, and a cross-port extinction of 23.0 dB, all in 16-ch average. An average trimming power for the bar states and an average switching power for the cross states are 3.8 mW (σ = 2.3 mW) and 15.6 mW (σ = 2.1 mW), respectively. The wavelength dependence of the spectral responses and the resonant effect from the apodization are discussed in detail to show how to further improve the spectral performance.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Recent datacenter networks require an ultrafast data transmission rate of 400 Gbps or even faster [1]. Optical transceivers for such data rate require multiplexing technologies including wavelength division multiplexing (WDM). Wavelength selective switches (WSS) will thus be a key device for the future large-scale and energy-efficient datacenter networks to flexibly aggregate and disaggregate each wavelength channel [2]. Silicon photonics offers highly integrated ultra-compact devices with low power consumption as well as fast switching speed, which are desirable features in datacenters and computing platforms [3–8]. A number of silicon photonics WSSs have been reported using ring resonators [9–11], arrayed waveguide gratings (AWGs) [12,13], and echelle gratings [14]. We have recently proposed and demonstrated a new type of silicon photonics WSS with an unlimited free spectral range (FSR) using Mach-Zehnder (MZ) switches and sidewall corrugated waveguide gratings or contra-directional couplers (C-DCs) [15]. The silicon photonics C-DCs consists of two coupled waveguides with sidewall corrugation that induces mode coupling between the waveguides in the counter direction at a specific wavelength satisfying the Bragg condition [16]. Due to the first order operation, the C-DC has an extremely wide FSR and is therefore suitable for broadband add-drop filters. Thanks to the large index contrast, the sidewall corrugation induces a large coupling coefficient leading to a wide and flattop spectral shape, which offers high tolerance for temperature change and is suited for Coarse WDM application. In [15], we demonstrated the element switch (2 × 2 single channel WSS) as a building block and 2 × 2 2-ch WSS, and discussed a factor limiting the inband extinction in the demonstrated devices. In this paper, we extend the number of wavelength channel to 16 over a very wide wavelength range of the C- and L-band to show the suitability of the proposed WSS for an ultra-broadband and multichannel WSS or wavelength cross connect (WXC) switch. The wavelength dependence of the spectral responses and the resonant effect from the apodization are discussed in detail to show that a better spectral performance is achievable.
2. Device and circuit design
Figure 1 shows the device structure of a 2 × 2 single channel WSS. The C-DC consists of two adjacent waveguides with widths of W1 and W2, a gap of G, and a length of L, which have sidewall corrugation or grating with a period of Λ, and modulation depths of ΔW1 and ΔW2. In this structure, contra-directional cross coupling is induced between the propagating TE modes in each waveguide with effective indexes of n1 and n2 at a specific wavelength of λB satisfying a Bragg condition of λB = (n1 + n2)Λ. The (i) wavelength, (ii) bandwidth, and (iii) strength (or extinction) of cross coupling can be designed by adjusting, respectively, (i) W1, W2, Λ, (ii) ΔW1, ΔW2, G, and (iii) L [16]. The Bragg reflection in each waveguide governed by λB = 2n1,2Λ is eliminated by giving π phase shift between gratings on each sidewall of waveguide [17]. The modulation depths of ΔW1 and ΔW2 are gaussian-apodized to suppress the sidelobes in the spectral response. The C-DCs work as the filtering device to drop and add signals and the MZ switch selects the output port of the signal dropped by the C-DC. For example, the wavelength channel with λB that entered at the input port I1 couples to the other waveguide at the C-DC and is guided to the 2 × 2 thermo-optic MZ switch. If the MZ switch is in the cross state, λB channel is guided to the lower C-DC and couples to the waveguide connected to the output port O2. If the MZ switch is in the bar state, λB channel is guided to the other side of the upper C-DC and couples back to the waveguide connected to the output port O1. We can introduce signals into the input port I1 and I2 at the same time. In this case, if the MZ switch is in the cross state, only the λB channels are cross-switched, and the other channels just pass through the C-DCs. In other words, this device works also as a WXC switch.
Figure 2 shows a microscope image of the 2 × 2 16-ch WSS circuit where sixteen 2 × 2 single channel WSSs with different λB are cascaded. The structure dimensions of the C-DCs are summarized in Table 1. We refined the design parameters to fit our waveguide dimensions from the design in Ref. [16]. The center wavelength of each MZ switch (or the wavelength of 3-dB coupling in directional couplers (DCs)) is designed to be λB of the corresponding C-DC so that the wavelength dependence of each MZ switch does not degrade the extinction of the WSS [15]. A DC-based polarizer is inserted at each input and output port to make the polarization alignment easier at the measurement and to eliminate the TM polarization component. The chip was fabricated using our 300-mm CMOS R&D foundry with 45-nm ArF immersion lithography [18].
Table 1. Structure dimensions of C-DCs in 2 × 2 16-ch WSS.
3. Measured results
Figure 3 describes the measurement setup for the 2 × 2 16-ch WSS. The silicon chip containing the 2 × 2 16-ch WSS was die-bonded and wire-bonded to a chip carrier and then the inverse-taper type spot-size converters on the chip edge were but-coupled with a high NA fiber array. The thermo-optic micro-heaters on all the MZ switches were controlled by a multi-channel current source. An ASE light source was connected to the port I1 or I2 of the 2 × 2 16-ch WSS through a fiber-based polarizer and polarization controller. The C + L band light was polarized and aligned to the TE polarization of the device. The output light from the port O1 or O2 was measured by an optical spectrum analyzer.
Figure 4 shows the measured fiber-to-fiber transmission spectra for all the four connections (I1 to O1, I1 to O2, I2 to O1, and I2 to O2) when one of the sixteen MZ switches is in the cross state and the other MZ switches are in the bar state. We can see that the 16 channels are distributed over a wide wavelength range of the C and L band. There is a silicon waveguide without any device but polarizers on the same chip, and its transmission spectrum was also measured and plotted in the figures. The fiber-to-fiber transmission of the silicon waveguide was −6.5 dB at 1550 nm, from which the fiber-to-chip coupling loss was estimated as 1.92 dB/facet by subtracting the loss of waveguide (1.46 dB) and polarizers (1.2 dB). These values are included in Table 2.
Fig. 4. Measured fiber-to-fiber transmission spectra for all the four connections (I1 to O1, I1 to O2, I2 to O1, and I2 to O2) when one of the sixteen MZ switches is in the cross state and the other MZ switches are in the bar state.
Table 2. Loss values for each component at 1550 nm.
Figure 5 summarizes the wavelength dependences (values of 16 channels) of fiber-to-fiber transmission, grating period Λ, 3-dB bandwidth, and extinction, read from Fig. 4. The averaged value of the fiber-to-fiber transmission is −9.2 dB. The average on-chip loss is therefore estimated to be 5.4 dB. In the plot of fiber-to-fiber transmission, we observe a trend of a higher loss at a longer wavelength except for some scatter around short wavelengths. This trend is caused by the wavelength dependence of the DC-based on-chip polarizers whose center wavelength was designed for around 1550 nm. Indeed, if we subtract the wavelength dependence of the silicon waveguide without any device but polarizers from that of the device, the values are almost flat over the measured wavelength range (see the excess loss plotted in the inset). Since the fiber-to-fiber transmission around 1550 nm is read as –8.6 dB from Fig. 5, the on-chip loss of the 2 × 2 16-ch WSS except for the polarizers is estimated to be 8.6–1.92 × 2–1.2 = 3.56 dB. In principle, the light of each channel passes one MZ switch, two C-DCs with contra-directional coupling, and fifteen C-DCs without contra-directional coupling. Since the loss of our MZ switch is about 0.13 dB [4], we can roughly estimate the loss of a C-DC as 3.43/17 = 0.20 dB, which is fairly small. These loss values at 1550 nm are summarized in Table 2.
Fig. 5. Wavelength dependences (values of 16 channels) of fiber-to-fiber transmission, grating period Λ, 3-dB bandwidth, and extinction, read from Fig. 4. Blue circles, blue crosses, red crosses, red circles represent port connections of I1-O1 (Bar), I1-O2 (Cross), I2-O1 (Cross), I2-O2 (Bar), respectively. Inset: Excess loss calculated by subtracting the losses of the silicon waveguide with polarizers from those of the 2 × 2 16-ch WSS.
In the plot of grating period Λ, we observe a linear relation with the channel wavelength λB, which represents the first order approximation of the Bragg condition, λB = (n1 + n2)Λ. Using this relation, we can design C-DCs with any λB around this wavelength range. As for the 3-dB bandwidth, the averaged value is 4.2 nm. There is a dependence of being wider for a longer wavelength. Since the 3-dB bandwidth is proportional to the coupling coefficient of the C-DC [16], it is considered that the dependence is caused by increase in the coupling coefficient at a longer wavelength. Effective indexes in silicon waveguides decrease as the wavelength increases, which means that the guided mode becomes larger and penetrates more into the cladding. Therefore, the mode overlap with the sidewall gratings becomes larger, resulting in a larger coupling coefficient. By introducing channel dependent modulation depths of ΔW1 and ΔW2 and adjusting the coupling coefficient for each channel, we can obtain a uniform and desirable 3-dB bandwidth. As demonstrated here, we can design the channel spacing and bandwidth depending on the application. Specifically, the bandwidth can be much narrower (< 1 nm) or much wider (> 10 nm) by adjusting ΔW1, ΔW2, and G, as theoretically discussed in Ref. [16].
In the plot of extinction, we have opposite trends in the bar (I1-O1, I2-O2) and cross (I1-O2, I2-O1) connections. For the bar connections, a larger extinction at a longer wavelength will be again a result of the increase in coupling coefficient at a longer wavelength since a stronger contra-directional coupling arises with a larger coupling coefficient [16]. Once we have determined the coupling coefficient for desirable 3-dB bandwidth, a sufficiently longer length of the C-DCs will give a larger extinction. For the cross connections, it is considered that light with a longer wavelength outside the Bragg condition that is supposed to go through the C-DC leaked to the cross port and degraded the extinction. The larger mode overlap at a longer wavelength discussed earlier will also cause some co-directional coupling or leakage at the C-DC. Making the difference between the effective indexes of n1 and n2 of the two waveguide modes larger will further suppress the leakage of co-directional coupling in the C-DCs. The averaged values of bar-port extinction, and cross-port extinction are 15.0 dB, and 23.0 dB, respectively.
Figure 6 demonstrates an example wavelength selective switching with cross state at channels 1, 4, 7, 10, 13, and 16. Any other combination of wavelength channels is possible since all the micro-heaters on the 2 × 2 16-ch WSS chip are fully controllable. We observe a few peaks and dips at the longer wavelength side in each channel’s drop band in Fig. 6 (also in Fig. 4). This is a resonant effect caused by the apodization, which will be discussed in detail in the next section. We also observe smaller ripples in each drop band, which should be a result of interference by leakages, such as the residual transmission and co-directional coupling at C-DCs. These leakages are also a source of the extinction degradation as discussed above, and a sufficient length and a larger difference between two waveguide widths of C-DCs will improve the ripples.
Fig. 6. Example wavelength selective switching with cross state at channels 1, 4, 7, 10, 13, and 16. The other channels are at bar state.
Figure 7 shows power consumption distribution of the 16 MZ switches of the 2 × 2 16-ch WSS. Each MZ switch is set to the bar or cross state by applying current to one of two micro-heaters on each arm. Although the MZ switches were designed as the “normally bar” switch that is in the bar state without current injection, a trimming power is necessary to compensate a phase error due to a slight fabrication error. The average trimming power for the bar states and the average switching power for the cross states were 3.8 mW (σ = 2.3 mW) and 15.6 mW (2.1 mW), respectively.
Fig. 7. Power consumption distribution for the bar and cross states of the 16 MZ switches of the 2 × 2 16-ch WSS.
4. Discussion: resonant effect caused by apodization
In Figs. 4 and 6, we observed a few peaks and dips at the longer wavelength side in each channel’s drop band. A similar spectral response was studied for apodized fiber Bragg gratings [19]. A simple apodization may induce a nonuniformity in the average effective index over the longitudinal direction of gratings. In a fiber Bragg grating, the grating is formed by periodic increase of the core glass refractive index with ultraviolet light irradiation. Therefore, as depicted in Fig. 8(a), the average effective index at the center of the apodized grating is larger than that at the both edges, which creates a shift of the stopband at the center toward the longer wavelength direction. As a result, a Fabry-Perot type resonance occurs at the shorter wavelength side by the reflections at the both edges. In our C-DCs, a gaussian apodization was applied to the sidewall corrugation so that the center of the modulation unchanged (or zero dc component). However, it is considered that the effective index reduction by a concave is larger than the increase by the same amount of convex, which creates a shift of the stopband at the center toward the shorter wavelength direction and a resonance at the longer wavelength side as shown in Fig. 8(b). We can introduce an additional gaussian apodization for the dc component of the modulation, i.e. gradual widening of the waveguide width, to eliminate the peaks and dips at the longer wavelength side in each channel’s drop band of our C-DCs, as described in Fig. 8(c)
Fig. 8. Schematic diagram of longitudinal distribution of effective index change and stopband of (a) apodized fiber Bragg grating, (b) fabricated C-DC, and (c) width- and modulation-apodized C-DC.
We performed FDTD simulations (RSoft) to confirm the above-mentioned solution for the resonant effect caused by the apodization. We first reproduced the fabricated C-DC (see Table 1) in the FDTD simulation except that, due to the limitation of our computational power, the simulations were performed in two dimension using the effective index method and the length of C-DC was shortened to 200 µm. The spatial grid size was 10 nm in the longitudinal direction and 2 nm in the transverse direction, and the boundary surrounding the calculation region was the perfectly matched layers. The calculated spectrum for the contra-directional coupling is shown in Fig. 9(a), where we can clearly see the resonant peaks at the longer wavelength side although the amount of coupling is smaller than the measured spectra due to the shorter length. Figures 9(b) and (c) are the calculated spectra for the cases with a gaussian width-increase apodization of 7 and 10 nm and that of 20 and 30 nm for W1 and W2. We observe no resonant peak for the width-increase apodization of 7 and 10 nm in Fig. 9(b), and some resonant peaks at the shorter wavelength side for that of 20 and 30 nm in Fig. 9(c), which correspond to a good compensation as in Fig. 8(c) and an over compensation as in Fig. 8(a), respectively. This width control can be implemented in our fabrication as demonstrated in various device fabrications [18].
Fig. 9. Calculated spectra of the contra-directional coupling for (a) fabricated C-DC, (b) width-increase apodization of 7 and 10 nm for W1 and W2, (c) that of 20 and 30 nm. Due to the limitation in computational power, the simulations were performed in two dimensions using the effective index method and the length of C-DC was shortened to 200 µm.
5. Conclusion
We demonstrated a 2 × 2 16-ch silicon photonics WSS or WXC consisting of contra-directional couplers and thermo-optic Mach-Zehnder switches. Owing to the unlimited FSR of the contra-directional couplers, the device operated over a very wide wavelength range of the C- and L-band. 16-ch averaged values of fiber-to-fiber insertion loss, on-chip loss, 3-dB bandwidth, bar-port extinction, and cross-port extinction were measured to be 9.2 dB, 5.4 dB, 4.2 nm, 15.0 dB, and 23.0 dB, respectively. The average trimming power for the bar states and the average switching power for the cross states were 3.8 mW (σ = 2.3 mW) and 15.6 mW (σ = 2.1 mW), respectively. The discussions on the wavelength dependence of the measured spectra revealed that we still have room to tailor the 3-dB bandwidth uniformity and extinction. Specifically, (i) introducing channel dependent modulation depths will provide a uniform and desirable 3-dB bandwidth; (ii) a sufficiently longer length of the C-DCs will give a larger extinction in the bar connections; (iii) a larger difference between the effective indexes of the two waveguide modes will realize a larger extinction in the cross connections. The resonant peaks and dips in the measured drop bands were caused by a nonuniformity in the average effective index over the longitudinal direction of C-DC. We proposed the waveguide width apodization to suppress the resonant effect and confirmed the suppression using two dimensional FDTD simulations. With the above improvements, we will have a larger port- and channel-count WXC with better spectral performances in the next fabrication.
Funding
Core Research for Evolutional Science and Technology (JPMJCR15N6).
Acknowledgments
The device fabrication was supported by TIA Super Clean Room of AIST.
Disclosures
The authors declare no conflicts of interest.
References
1. X. Zhou, H. Liu, and R. Urata, “Datacenter optics: requirements, technologies, and trends,” Chin. Opt. Lett. 15(5), 120008 (2017). [CrossRef]
2. Q. Cheng, S. Rumley, M. Bahadori, and K. Bergman, “Photonic switching in high performance datacenters [Invited],” Opt. Express 26(12), 16022–16043 (2018). [CrossRef]
3. K. Tanizawa, K. Suzuki, M. Toyama, M. Ohtsuka, N. Yokoyama, K. Matsumaro, M. Seki, K. Koshino, T. Sugaya, S. Suda, G. Cong, T. Kimura, K. Ikeda, S. Namiki, and H. Kawashima, “Ultra-compact 32 × 32 strictly-non-blocking Si-wire optical switch with fan-out LGA interposer,” Opt. Express 23(13), 17599–17606 (2015). [CrossRef]
4. K. Suzuki, R. Konoike, J. Hasegawa, S. Suda, H. Matsuura, K. Ikeda, S. Namiki, and H. Kawashima, “Low-Insertion-Loss and Power-Efficient 32 × 32 Silicon Photonics Switch with Extremely High-Δ Silica PLC Connector,” J. Lightwave Technol. 37(1), 116–122 (2019). [CrossRef]
5. S. Nakamura, S. Yanagimachi, H. Takeshita, A. Tajima, T. Kato, T. Hino, and K. Fukuchi, “Compact and Low-Loss 8×8 Silicon Photonic Switch Module for Transponder Aggregators in CDC-ROADM Application,” Optical Fiber Communication Conference (OFC)2015, paper M2B.6.
6. T.-J. Seok, N. Quack, S. Han, R. S. Muller, and M. C. Wu, “Large-scale broadband digital silicon photonic switches with vertical adiabatic couplers,” Optica 3(1), 64–70 (2016). [CrossRef]
7. B. G. Lee, A. V. Rylyakov, W. M. J. Green, S. Assefa, C. W. Baks, R. Rimolo-Donadio, D. M. Kuchta, M. H. Khater, T. Barwicz, C. Reinholm, E. Kiewra, S. M. Shank, C. L. Schow, and Y. A. Vlasov, “Monolithic Silicon Integration of Scaled Photonic Switch Fabrics, CMOS Logic, and Device Driver Circuits,” J. Lightwave Technol. 32(4), 743–751 (2014). [CrossRef]
8. J. Jiang, D. J. Goodwill, P. Dumais, D. Celo, C. Zhang, H. Mehrvar, M. Rad, E. Bernier, M. Li, F. Zhao, C. Zhang, J. He, Y. Ding, Y. Wei, W. Liu, X. Tu, and D. Geng, “16 × 16 silicon photonic switch with nanosecond switch time and low-crosstalk architecture,” European Conference on Optical Communications (ECOC)2019, paper M.2.A.4.
9. P. Prabhathan, Z. Jing, V. M. Murukeshan, Z. Huijuan, and C. Shiyi, “Discrete and fine wavelength tunable thermo-optic WSS for low power consumption C + L band tunability,” IEEE Photonics Technol. Lett. 24(2), 152–154 (2012). [CrossRef]
10. N. Eid, R. Boeck, H. Jayatilleka, L. Chrostowski, W. Shi, and N. A. F. Jaeger, “FSR-free silicon-on-insulator microring resonator based filter with bent contra-directional couplers,” Opt. Express 24(25), 29009–29021 (2016). [CrossRef]
11. A. S. P. Khope, M. Saeidi, R. Yu, X. Wu, A. M. Netherton, Y. Liu, Z. Zhang, Y. Xia, G. Fleeman, A. Spott, S. Pinna, C. Schow, R. Helkey, L. Theogarajan, R. C. Alferness, A. A. M. Saleh, and J. E. Bowers, “Multi-wavelength selective crossbar switch,” Opt. Express 27(4), 5203–5216 (2019). [CrossRef]
12. C. R. Doerr, L. L. Buhl, L. Chen, and N. Dupuis, “Monolithic flexible-grid 1 × 2 wavelength-selective switch in silicon photonics,” J. Lightwave Technol. 30(4), 473–478 (2012). [CrossRef]
13. F. Nakamura, K. Muramatsu, K. Suzuki, K. Tanizawa, M. Ohtsuka, N. Yokoyama, K. Matsumaro, M. Seki, K. Koshino, K. Ikeda, S. Namiki, H. Kawashima, and H. Tsuda, “Integrated silicon photonic wavelength-selective switch using wavefront control waveguides,” Opt. Express 26(10), 13573–13589 (2018). [CrossRef]
14. T. J. Seok, J. Luo, Z. Huang, K. Kwon, J. Henriksson, J. Jacobs, L. Ochikubo, R. S. Muller, and M. C. Wu, “Silicon photonic wavelength cross-connect with integrated MEMS switching,” APL Photonics 4(10), 100803 (2019). [CrossRef]
15. K. Ikeda, K. Suzuki, R. Konoike, and H. Kawashima, “Silicon Photonics Wavelength Selective Switch with Unlimited Free Spectral Range,” J. Lightwave Technol. 38(12), 3268–3272 (2020). [CrossRef]
16. K. Ikeda, M. Nezhad, and Y. Fainman, “Wavelength selective coupler with vertical gratings on silicon chip,” Appl. Phys. Lett. 92(20), 201111 (2008). [CrossRef]
17. W. Shi, H. Yun, C. Lin, M. Greenberg, X. Wang, Y. Wang, S. T. Fard, J. Flueckiger, N. A. F. Jaeger, and L. Chrostowski, “Ultra-compact, flat-top demultiplexer using anti-reflection contra-directional couplers for CWDM networks on silicon,” Opt. Express 21(6), 6733–6738 (2013). [CrossRef]
18. K. Ikeda, K. Suzuki, R. Konoike, S. Namiki, and H. Kawashima, “Large-scale silicon photonics switch based on 45-nm CMOS technology,” Opt. Commun. 466, 125677 (2020). [CrossRef]
19. T. Erdogan, “Fiber Grating Spectra,” J. Lightwave Technol. 15(8), 1277–1294 (1997). [CrossRef]
