A wavelength selective switch (WSS) can route optical signals into any of output ports by wavelength, and is a key component of the reconfigurable optical add/drop multiplexer. We propose a wavefront control type WSS using silicon photonics technology. This consists of several arrayed waveguide gratings sharing a large slab waveguide, wavefront control waveguides and distributed Bragg reflectors. The structure, design method, operating principle, and scalability of the WSS are described and discussed. We designed and fabricated a 1 × 2 wavefront control type WSS using silicon waveguides. This has 16 channels with a channel spacing of 200 GHz. The chip size is 5 mm × 10 mm. The switching operation was achieved by shifting the phase of the light propagating in each wavefront control waveguide, and by controlling the propagation direction in the shared large slab waveguide. Our WSS has no crossing waveguide, so the loss and the variation in loss between channels were small compared to conventional waveguide type WSSs. The heater power required for switching was 183 mW per channel, and the average extinction ratios routed to Output#1 and Output#2 were 9.8 dB and 10.2 dB, respectively.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
A 1 × N wavelength selective optical switch (WSS) has one input port and N output ports, and can be used to route every signal with different wavelengths independently to any of the output ports. A typical WSS is composed of a wavelength demultiplexer, an array of switching elements, and two or more wavelength multiplexers. The incident signal is wavelength demultiplexed and each separated signal is switched to one of the wavelength multiplexers by switching elements. The switched signals are again multiplexed and guided to the output ports . A WSS is an indispensable component in highly functional reconfigurable optical add/drop multiplexers (ROADM) in wavelength division multiplexing (WDM) optical communication systems . Any wavelength signal can be selectively added or dropped to the transmitted signal in a ROADM without the optical to electrical or electrical to optical conversion. A color-less, direction-less, and contention-less (CDC) ROADM can route the signal from any of the add ports with any wavelength in any direction, and drop any wavelength signal from any direction to one of the drop ports without conflict. This is necessary for a flexible network system. The CDC ROADM architectures using WSSs and optical switches have been reported [3, 4].
WSSs are classified into free space based WSSs [5–10] and waveguide based WSSs [11–14]. The switching engine in a free space based WSS is a microelectromechanical systems (MEMS) mirror [5-6] or a liquid crystal on silicon (LCOS) [7–10]. A diffraction grating or stacked arrayed-waveguide grating (AWG) is used as the optical demultiplexer. Currently free space optics based WSSs are commercially available because of the large number of output ports available, the small transmission loss, and very low crosstalk. Free space optics based WSSs with more than 40 output ports have already been demonstrated [6, 7]. The disadvantage of free space optics type WSSs is that they need a lot of optical components and high precision assembly; therefore, the fabrication cost tends to be high. On the other hand, waveguide based WSSs are integrated on chip, are small, alignment free, and therefore cost effective. Silica waveguides [11, 12] or silicon wire waveguides [13–15] are used for waveguide based WSSs. Silicon WSSs are much smaller than silica WSSs because the larger refractive index difference confines the light more strongly in silicon waveguides. In particular, silicon WSSs can be mass-produced at low cost using CMOS compatible processes. The application of waveguide based WSSs is one of the solutions to reducing cost in increasingly complicated network systems. The drawbacks of a waveguide type WSSs are the high crosstalk and somewhat large insertion loss. Waveguide type WSSs consist of AWGs for the wavelength multiplexing and demultiplexing, and Mach-Zehnder interferometer (MZI) optical switches. The conventional design of WSS  has many waveguide crossings and this causes excess loss, variation in loss depending on the channel, and excess crosstalk. The waveguide crossings increase in proportion to the number of channels and output port; therefore they prevent number of channels and ports from being increased. To overcome these problems, we proposed a WSS design with fewer waveguide crossings by using fold-back type AWG  and a WSS design with no waveguide crossings by wavefront control configuration . 16 channel, 200-GHz spacing, 1 × 2 wavefront control type WSS was fabricated and the preliminary experimental results were reported .
In this paper, we have described the detailed design method of the wavefront control type WSS, and discussed the scalability of the WSS with performance estimation. We also have measured in detail the switching characteristics of 16-channel, 200-GHz spacing, 1 × 2 wavefront control type WSS and compared them to the simulated results.
2. Configuration and scalability of a 1 × N wavefront control type WSS
2.1 Configuration of a wavefront control type WSS
Figure 1 shows a schematic configuration of the 1 × N wavefront control type WSS. This consists of an AWG for the input, N AWGs for the outputs, and wavefront control waveguides with phase shifters and reflectors. The AWG for the input and AWGs for the outputs share a large slab waveguide. The incoming signal is demultiplexed at the input AWG, and each separate signal is focused on the far side of the large slab waveguide and are coupled to Mwf wavefront control waveguides. The wavefront of each signal is tilted by controlling the temperature of the heater integrated on the wavefront control waveguide; the direction of propagation of the each signal is switched in the large slab waveguide. Finally, the signals enter one of the N output AWGs, are multiplexed and coupled to the output port. When the phase shifters are off, the signals come out from Output#1. The AWG for Output #1 and Input AWG are arranged symmetrically on the large slab waveguide. It should be noted that the wavefront control WSS has no waveguide crossings and the light signals only intersect in the large slab waveguide.
2.2 Design method of the wavefront control type WSS
In the proposed WSS, the input and output AWGs are the same except for the positions where they couple to the shared large slab waveguide, and that the output AWGs operate as multiplexers when light is received from the slab waveguide. When the width of waveguide, the number of waveguides in the AWGs, the number of channels, Nch, and the channel spacing, Δν are given, the other parameters of the AWG can be designed as below . The difference in length between adjacent grating waveguides, Δl, is an integral multiple of the center wavelength, λ0, and is given by;
In Eq. (3), ng is the group index of the effective refractive index nc. The frequency free spectral range (FSR), νFSR, needs to be wider than the system bandwidth of Δν × Nch. The radius of curvature of the large slab waveguide, Lf, is obtained from the following Eq;
The light field at each waveguide was calculated according to the AWG analysis presented in . The propagation of light from one side of the slab waveguide to the other is calculated approximately using a one-dimensional Fourier transformation . The spatial distribution of the light field F(X) on the far side of the slab waveguide which has radius of curvature Lf, is obtained from the Fourier transform of the input field f(x) on the near side,Fig. 2. The focal points on each side of the slab waveguide is the origins x = 0 and X = 0.
The coupling efficiency where the slab waveguide is joined to the grating and wavefront control waveguides, was calculated with butt coupling  between the field of the fundamental mode of each waveguide and the field in the slab waveguide.
2.3 Scalability of the wavefront control type WSS
16-channel, 200-GHz-spacing WSSs with 2, 3, and 4 output ports were designed to consider the scalability of the WSS on a silicon chip. The simulation was performed with MATLAB. The parameters used in the simulations are shown in Table 1. The transmittances when all the signals are switched to Output #1 for 1 × 2, 1 × 3, and 1 × 4 WSSs with 8 or 16 wavefront control waveguides, Mwf, are shown in Fig. 3. The loss is larger for the outer wavelength channel because of the Gaussian profile of the light. The diffraction loss increases as the number of output ports increases because the phase difference between the light propagating in adjacent wavefront control waveguides becomes large for large ΔXij. For the same reason, the diffraction loss can be reduced by increasing Mwf, as shown in Fig. 3.
Table 2 shows the loss budget for the 1 × 2 wavefront control type WSS with Mwf equal to 8, 16, and 32 when the 8th wavelength channel signal is routed to Output#2. The total loss is 7.55 dB, 5.72 dB, and 5.27 dB, for Mwf of 8, 16, and 32, respectively. The coupling loss to the wavefront control waveguides and the diffraction loss from the wavefront control waveguides to the waveguide array can be reduced by increasing the number of wavefront control waveguides. The major loss comes from the mode mismatch between the waveguide array and the slab waveguides. We introduced low-loss junction structure using a rib waveguide; this is described in Sec. 3.2. The variation in loss among the channels can be reduced modifying the shapes of the tapered waveguides at the junctions between the waveguide arrays and the slab waveguide, allowing some excess loss.
For example, the switching characteristics of the 1 × 4 WSS were simulated using the design parameters in Table 1. The number of wavefront control waveguides for each channel was 14, and the radius of curvature of the large slab waveguides, Lf, was 4165.7 µm. The transmittances to Output#1, Output#2, Output#3, and Output#4 when channels #1, #5, #9, #13 were switched to Output #1, channels #2, #6, #10, #14 to Output #2, channels #3, #7, #11, #15 to Output #3, and channels #4, #8, #12, #16 to Output #4 are shown in Fig. 4.
The characteristics of the 200-GHz spacing, 16-channel WSS with various numbers of output ports is shown in Table 3. The radius of curvature of the shared slab waveguide, the chip size, transmission loss, crosstalk, and extinction ratio for the center wavelength of each channel were calculated assuming the parameters shown in Table 1, and the worst values were used. The number of wavefront control waveguides was determined such that the maximum loss at the channel wavelength was less than 10 dB and the minimum crosstalk was less than −30 dB. The beam deflecting distance ∆Xij is proportional to the number of output ports; therefore, to maintain the same performance, the number of wavefront control waveguides should be proportional to the number of output ports. Using a similar deign method, it was found that a 1 × 20 WSS requires 80 wavefront control waveguides for each channel, and the chip size is about 30 mm × 30 mm, which can be laid out in the area covered by photolithography. In the case of 100 GHz spacing and 32 channels, the radius of curvature of the slab waveguide Lf is twice as large as that for 16 channels according to Eq. (4), so a WSS which has 20 output ports can be integrated on the 30 mm × 30 mm sized chip.
Table 4 summarizes the length of a larger slab waveguide, the chip size, the maximum loss difference, the minimum loss, and the largest crosstalk for each 1 × 2 WSS, when all channel switched to Output #1. The number of wavefront control waveguides is fixed at 8. As shown in Table 4, 200-GHz, 100-GHz, 50-GHz, and 25-GHz spacing WSSs have 16-, 32-, 64-, and 128- wavelength channels, respectively, when the covering frequency range is 3200 GHz. The results shows 25-GHz spacing, 128-channel, 1 × 2 WSS can be laid out on the one-shot area of photo lithography.
When channel spacing become small and the number of channel get increased, the length of large slab waveguide and chip size gets large, as shown Table 4. According to Eq. (6), the length of large slab waveguide become large, the phase difference between the light signals reflected by adjacent waveguides δφ gets small. Therefore the minimum loss was smallest when channel spacing was 25 GHz. However the maximum loss difference and the crosstalk of the 25 GHz-spacing WSS were largest, because the loss at the outermost channel gets large with increasing the number of channels.
The simulation results were obtained using free spectral range 4800GHz. The loss difference could be improved by designing the WSS with larger FSR than 4800 GHz at the expense of the compactness.
3. Monolithic 1 × 2 wavefront control type WSS
3. 1 Device design
We designed a 200-GHz-spacing, 16-channel, 1 × 2 wavefront control type WSS according to the design method described in the previous section. A silicon on insulator (SOI) substrate with a silicon layer thickness of 220 nm was used. The layout of the mask for the 1 × 2 silicon WSS is shown in Fig. 5(a). The WSS has one input AWG, two output AWGs, and 128 wavefront control waveguides with TiN heaters for the phase shifters. There are 16 heaters in the wavefront control waveguide array. Each heater controls the phase of eight waveguides for one wavelength channel. The lengths of the heaters vary in order to apply linear phase changes to the light propagating through the wavefront control waveguides, as shown in Fig. 5(b). Every signal was routed to Output Port#1 when the corresponding heater was OFF, and switched to Output Port#2 when the heater was ON. The heater located on the AWGs, as shown in Fig. 5(c), is for turning the center wavelength. The design parameters are summarized in Table 5.
3.2 Low-loss joint structure
Rib waveguides with a height of 40-nm are used where the slab waveguide is jointed the waveguide array in order to reduce reflected light and coupling loss, as shown in Fig. 6(a) and 6(b). The minimum spacing between the mesas of the rib waveguides is 250 nm due to the fabrication limit. To reduce the joint loss, a tapered rib waveguide was used where the slab joins the waveguide array, as shown in Fig. 6(a). The widths of the rib structure at the end of the slab waveguide, A, and at the ends of the taper B, C are 1.65 μm, 0.8 μm, and 0.44 μm respectively, as shown in Fig. 6(b). The design parameters, L and Wrib were optimized to obtain high coupling efficiency. Figure 7(a) shows the coupling efficiency from the slab waveguide to the waveguide array as a function of L calculated by the finite difference time domain (FDTD) method. The coupling efficiency reaches its maximum value at L = 5.0 μm. Then the coupling powers of the fundamental and second order modes around L = 5.0 µm when Wrib = 1.1, 1.2, 1.3 μm were calculated, as shown in Fig. 7(b). When Wrib = 1.2 µm and L = 4.9 µm, the power coupled to the second order mode was minimized. Consequently, L and Wrib were determined to be 4.9 µm and 1.2 µm, respectively.
3. 3 Design of the wavefront control waveguides
A schematic of one channel of the wavefront control waveguides is shown in Fig. 8. A distributed Bragg reflector (DBR) is located at the end of the waveguide. Each wavelength channel has 8 wavefront control waveguides and an integrated TiN heater. The difference in heater length between adjacent waveguides, Δl, is 35.4 μm. The heater is 5 µm wide and 0.1 µm thick. The light signal couples to the group of 8 waveguides. When the heater is OFF, the light signal is reflected by the DBRs and propagates to output AWG #1. When the heater is ON, the wavefront is tilted and the light signal is routed to output AWG#2. For example, the phase shift required for the switching the light with a wavelength of λ8 to the output AWG#2, δϕ82, is 2.22 rad. Within the same wavelength channel, the lengths of the wavefront control waveguides were the same, however, the waveguide lengths were not the same between the different wavelength channels for layout convenience. For flexible grid operation, the waveguides should all be the same length.
4. Characteristics of the WSS and technical analysis
4.1 Characteristics of 1 × 2 wavefront control type WSS
The WSS chip was fabricated on a 12-inch SOI wafer at a CMOS pilot line, featuring ArF immersion photolithography, at AIST. Figure 9 shows a photograph of the fabricated chip mounted on a heat sink. Wires were bonded to the contact pads. The chip size was 5 mm × 10 mm.
The transmittance of AWG for imput was measured using one of monitor tap waveguides connected to the outer sides of wavefront control waveguides, as shown in Fig. 5. The light was inserted from the input waveguide, propagated through AWG for input and the large slab waveguide, and come out from the monitor waveguides of AWG for input. The AWG monitor #1 is the monitor waveguide, which couples to the large slab waveguide most close to wavefront control waveguides for channel 16, as shown in Fig. 5. The transmission spectrum is shown in Fig. 10. It shows two peaks corresponding to two adjacent diffraction order with a free spectral range of 4758 GHz. The peak transmittance was −13.2 dB and the crosstalk was 9.6 dB.
The average current and power per channel required for switching were 8.9 mA and 186 mW, respectively. The transmittances from the Input Port to Output Port#2 and from the Input Port to Output Port#1 when one of the wavelength channels was selected and the light was routed to Output Port#2 (the selected heater was ON), are shown in Figs. 11(a) and 11(b), respectively. The blue curves show the transmittance when no wavelength channels were selected (all of the heaters were OFF). The dotted lines in the figure show the center wavelengths of each channel. Unfortunately, channels #9, #10, and #16 were unavailable because of a broken wire. The center wavelength of the WSS was 1.539 μm and the frequency spacing was 187.5 GHz. The differences between the design values and the measured ones were due to a difference between design value and actual value of the refractive index. The wavelength channels with odd or even numbers were selected and routed to Output Port#2, as shown in Fig. 12(a). The transmittance of the even or odd numbered channels to Output Port#1 is also shown in Fig. 12(b) for reference.
The switching characteristics are summarized in Table 6. The transmittance values were obtained by averaging over the 10% of the channel spacing at around the center wavelength for every channel. The transmittance to Output Port#1 varied from −17.2 dB to −23.8 dB and the average was −20.1 dB. The transmittance to Output Port#2 varied from −14.8 dB to −19.2 dB and the average was −16.3 dB. The extinction ratio was defined as the ratio of the optical power when the channel was selected to that when the channel was not selected, and was averaged over the 10% of the channel spacing. The extinction ratio varies from 4.1 dB to 22.5 dB, and the average of all channels for Output Port#1 was 9.8 dB and that for Output Port#2 was 10.2 dB, respectively. The average switching power that is needed for switching each channel to Output#2 was 183.6 mW. The maximum transmittance differences among available channels at Output #1 and Output #2 were 4.15 dB and 4.39 dB, respectively.
4.2 Technical analysis and future development
As shown in Figs. 11 and 12, the switching operation of 1 × 2 fabricated wavefront control type WSS were confirmed and its average extinction ratio is 10.1 dB. However, the performance was not as good as the simulation result in Table 3. In this section, we discuss the cause.
The larger transmission loss than simulation result as shown in Table 3 was caused by coupling loss at a junction of a slab lens and a channel waveguide. At a junction of a slab lens and a channel waveguide, we introduced a rib structure that is supposed to relax an abrupt jump in effective index. A part of this is patterned by KrF lithography. For insufficient optimization, at present, the largest dimensional error reaches 100 nm. This is the primary cause of the performance degradation. By correcting the design so that it is consistent with resolution of KrF lithography, it is feasible to recover the performance that we intended.
Fabricated phase errors at silicon wire waveguides also limited performance of fabricated 1 × 2 WSS chip. Especially, it is considered that degradation of the extinction ratio and crosstalk is caused by a phase error in the wavefront control waveguides. The wavefront control waveguides for each channel were designed to be the same length, but their optical path lengths were not the same because of fabricated error at each waveguides. These phase error becomes a high-order component after propagation of the slab waveguide, resulting in deterioration of extinction ratio and crosstalk to adjacent Output. The fabricated errors at arrayed waveguides occurred the difference between the effective refractive index and the mounting value and the center wavelength of fabricated WSS was shifted by 11 nm.
Figures 13(a) and 13(b) show the simulated transmittance from Input Port to Output Port#2 and the transmittance from Input Port to Output Port#1 using parameters shown in Table 5, with consideration of phase error at waveguides. The effective refractive index fluctuation of 1.1 × 10−4  was introduced for calculation of grating waveguides and wavefront control waveguides. The effective refractive index of the grating waveguides with a width of 0.915 µm was used to match the simulated channel spacing to the measured one. The red curves shows the transmittance when odd channels are selected, and blue curves shows the transmittance when even number channel is switched to Output #2. The dips for transmission curve when each channel was routed to Output Port#2 were due to the wavefront steps between adjacent wavelength channels. As mentioned in previous section, the waveguide length was not the same for different wavelength channels. In this simulation, the minimum extinction ratio was 12.2 dB, the worst crosstalk was 11.1 dB, and the maximum loss is −9.0 dB. In comparison with simulation result in Table 3, it can be found that the phase errors of arrayed waveguides and wavefront control waveguides affect the extinction ratio and the crosstalk.
If the heaters on the wavefront control waveguide can be operated individually, it becomes possible to correct the phase error in the waveguides, and improve the performance. The shift of center wavelength can be prevented by designing the WSS with the actual effective index value of arrayed waveguides, not with calculated value, and compensating closely using heaters on the waveguides.
The proposed topology remains advantageous in that it can fit 1 × 2 WSS functionality to a footprint of 5 mm × 10mm. We should be able to reach the designed performance simply by improving the components of WSS without changing the topology.
In conclusion, we proposed a wavefront control type silicon wavelength selective switch. The WSS can be fabricated using silicon photonics technology and is very compact and potentially cost effective. There are no waveguide crossings in any of the optical paths, which reduce the variation in loss among the wavelength channels. Various scale WSSs were designed and the scalability was discussed. The maximum number of port counts for 200-GHz-spacing, 16-channel WSS is 20, and chip size is about 30 mm × 30 mm, which can be laid out on a single chip. For a 100-GHz spacing, 32-channel WSS the number of port is 10.
A 1 × 2 wavefront control type 16 channel WSS with 200 GHz-spacing was fabricated using ArF immersion photolithography. We successfully demonstrated wavelength selective operation of this and confirmed the wavefront control in the WSS, although we should be able to reach better characteristics by improving the fabrication conditions for higher fidelity to design. The chip size of the 1 × 2 WSS was 5 mm × 10 mm. The average switching power per wavelength channel was 183.6 mW. The averaged extinction ratios for the Output Port#1 and for Output Port#2 were 9.8 dB and 10.2 dB, respectively.
Ministry of Education, Culture, Sports, Science and Technology.
References and links
1. D. T. Neilson, C. R. Doerr, D. M. Marom, R. Ryf, and M. P. Earnshaw, “Wavelength selective switching for optical bandwidth management,” Bell Labs Tech. J. 11(2), 105–128 (2006). [CrossRef]
2. T. A. Strasser and J. L. Wagener, “Wavelength-selective switches for ROADM applications,” IEEE J. Sel. Top. Quantum Electron. 16(5), 1150–1157 (2010). [CrossRef]
3. A. Devarajan, K. Sandesha, R. Gowrishankar, B. S. Kishore, G. Prasanna, R. Johnson, and P. Voruganti, “Colorless, directionless and contentionless multi-degree ROADM architecture for mesh optical networks,” in Proceedings of 2010 Second International Conference on Communication Systems and Networks (IEEE, 2010), pp. 1–10. [CrossRef]
4. H. Yang, B. Robertson, P. Wilkinson, and D. Chu, “Low-Cost CDC ROADM Architecture Based on Stacked Wavelength Selective Switches,” J. Opt. Commun. Netw. 9(5), 375–384 (2017). [CrossRef]
5. K. Sorimoto, H. Kawashima, M. Mori, T. Hasama, H. Ishikawa, H. Tsuda, and H. Uetsuka, “MEMS mirror with slot structures suitable for flexible-grid WSS,” IEICE Electron. Express 10(3), 20120924 (2013). [CrossRef]
6. Y. Ishii, K. Hadama, J. Yamaguchi, Y. Kawajiri, E. Hashimoto, T. Matsuura, and F. Shimokawa, “MEMS-based 1x43 Wavelength-Selective Switch with Flat Passband,” in Proceeding of European Conference on Optical Communication (ECOC, 2009), paper MP1.1.
7. M. Iwama, M. Takahashi, M. Kimura, Y. Uchida, J. Hasegawa, R. Kawahara, and N. Kagi, “LCOS-based Flexible Grid 1x40 Wavelength Selective Switch Using Planar Lightwave Circuit as Spot Size Converter,” in Optical Fiber Communication Conference, OSA Technical Digest (online), (Optical Society of America, 2015), paper Tu3A.8. [CrossRef]
8. B. Robertson, H. Yang, M. M. Redmond, N. Collings, J. R. Moore, J. Liu, A. M. Jeziorska-Chapman, M. Pivnenko, S. Lee, A. Wonfor, I. H. White, W. A. Crossland, and D. P. Chu, “Demonstration of multi-casting in a 1× 9 LCOS wavelength selective switch,” J. Lightwave Technol. 32(3), 402–410 (2014). [CrossRef]
9. K. Sorimoto, K. Tanizawa, H. Uetsuka, H. Kawashima, M. Mori, T. Hasama, H. Ishikawa, and H. Tsuda, “Compact and phase-error-robust multilayered AWG-based wavelength selective switch driven by a single LCOS,” Opt. Express 21(14), 17131–17149 (2013). [CrossRef] [PubMed]
10. D. M. Marom, C. R. Doerr, M. Cappuzzo, E. Chen, A. Wong-Foy, and L. Gomez, MLA, “Hybrid free-space and planar lightwave circuit wavelength-selective 1×3 switch with integrated drop-side demultiplexer,” in Proceeding of European Conference on Optical Communication (ECOC 2005), pp. 993–994.
11. T. Yoshida, H. Asakura, H. Tsuda, T. Mizuno, and H. Takahashi, “Switching Characteristics of a 100-GHz-Spacing Integrated 40-λ 1× 4 Wavelength Selective Switch,” IEEE Photonics Technol. Lett. 26(5), 451–453 (2014). [CrossRef]
12. Y. Ikuma, T. Mizuno, H. Takahashi, T. Ikeda, and H. Tsuda, “Low-loss Integrated 1×2 Gridless Wavelength Selective Switch With a Small Number of Waveguide Crossings,” in European Conference and Exhibition on Optical Communication, OSA Technical Digest (online), (Optical Society of America, 2012), paper Tu.3.E.5. [CrossRef]
13. 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]
14. F. Nakamura, H. Asakura, K. Sugiyama, K. Suzuki, K. Tanizawa, M. Ohtsuka, N. Yokoyama, K. Matsumaro, M. Seki, K. Koshino, K. Ikeda, S. Namiki, H. Kawashima, and H. Tsuda, “1×2 Silicon Wavelength Selective Switch Using Fold Back Arrayed-Waveguide Gratings,” in Advanced Photonics 2017, OSA Technical Digest (online), (Optical Society of America, 2017), paper PTu3C.3.
15. F. Nakamura, K. Muramatsu, and H. Tsuda, “Design of a Wavefront Control Type Compact Silicon Wavelength Selective Switch,” in Frontiers in Optics 2016, OSA Technical Digest (online), (Optical Society of America, 2016), paper JW4A.144.
16. F. Nakamura, K. Suzuki, K. Tanizawa, M. Ohtsuka, N. Yokoyama, K. Matsumaro, M. Seki, K. Koshino, K. Ikeda, S. Namiki, H. Kawashima, and H. Tsuda, “1 × 2 Wavefront Control Type Wavelength-Selective Switch by Silicon Waveguides,” in 19th European Conference on Integrated Optics (2017), paper MP1.1.
17. K. Okamoto, Fundamentals of Optical Waveguides, 2nd ed. (Elsevier, 2006), Chap. 9.
18. P. Munoz, D. Pastor, and J. Capmany, “Modeling and Design of Arrayed Waveguide Gratings,” J. Lightwave Technol. 20(4), 661–674 (2002). [CrossRef]
19. D. H. Bailey and P. N. Swarztrauber, “A fast method for the numerical evaluation of continuous Fourier and Laplace transforms,” SIAM J. Sci. Comput. 15(5), 1105–1110 (1994). [CrossRef]
20. Y. Kokubun, Lightwave Engineering (CRC Press, 2012).
21. K. Muramatsu, H. Asakura, K. Suzuki, K. Tanizawa, M. Toyama, M. Ohtsuka, N. Yokoyama, K. Matsumaro, M. Seki, K. Koshino, K. Ikeda, S. Namiki, H. Kawashima, and H. Tsuda, “Evaluation of the phase error in Si-wire arrayed-waveguide gratings fabricated by ArF-immersion photolithography,” IEICE Electron. Express 12(7), 1–6 (2015). [CrossRef]