Open Access
Add to BibSonomy
Abstract
We experimentally demonstrate a double-layer platform of silicon nitride and silicon for ultralow-crosstalk multiport optical switches. By using a silicon nitride overpass with a large gap of 1.5 µm, we achieve a crosstalk of less than −50 dB and −45 dB almost entirely in the C-band for 4 × 4 and 16 × 16 switches, respectively. To demonstrate the scalability of the platform, we also measured a 32 × 32 passive test device and show that a worst-case crosstalk of less than −50 dB is feasible with appropriate gate switches.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Optical path switches are expected to play a key role in next-generation telecom and datacom optical networks [1,2]. Among various types of optical switches, silicon photonics switches are considered promising candidates because they have a fast switching time, small footprint, long-term reliability and can be mass produced [3–10]. Recently, we have developed a compact, low-loss 32 × 32 silicon photonics switch which has an average loss of 10.8 dB [5] and an SOA-integrated silicon photonics switch to compensate for the loss [6]. Another limiting factor in large port-count silicon photonics switches is crosstalk between the paths. For reducing crosstalk, the switch-and-select topology [7] is a good option because gate switches can be used to suppress the crosstalk caused by the element switches. However, since the largest number of waveguide crossings on a path is (N − 1)2 in the switch-and-select topology, where N is the port count, the number of crossings becomes very large when N is large (e.g. a maximum of 961 crossings exist if N is 32) [7]. Therefore, the development of both ultralow crosstalk and ultralow loss waveguide crossings is very important.
The realization of ultralow-crosstalk and ultralow-loss waveguide crossings is difficult with a single-layer optical circuit because of the large index change at the crossing point. Using mode expanders, the crosstalk can be suppressed by approximately −40 dB with an insertion loss of approximately 0.03 dB to 0.2 dB [8,9]. Applying these crossings to large port-count switch-and-select-type switches, such as 32 × 32, results in a worst-case crosstalk of −25 dB, due only to the crossings, which is a similar value to switches with path-independent insertion loss (PILOSS) topology. Hence there is no benefit in using the switch-and-select topology. Moreover, the total loss in the worst-case path (961 crossings) would be ∼27 dB, which is too large. To overcome these issues, we have proposed an overpass on an isolated optical layer, as schematically shown in Fig. 1 [11], where silicon nitride (SiN) is used as a second layer over a silicon (Si) waveguide layer. This method does not affect the mass productivity of the silicon photonics platform because it is fully compatible with CMOS processes. Recently, polymer-based overpasses have been demonstrated, which offer the flexibility to connect any two points by directly written polymer waveguides [12]. However, these overpasses have challenges in mass production because they require considerable post-processing. We consider that thin (∼100 nm) SiN waveguides are not suited for loss and crosstalk suppression because a large spacing is required between the waveguides and layers due to the large mode field diameter [13]. Although several studies have utilized thick (∼400 nm) SiN waveguides, their interlayer spacing has been smaller than the wavelength [14,15], which is not sufficient to suppress both the loss and the crosstalk [16]. Recently, a tri-layer platform has been developed to realize both low interlayer coupling loss and a large gap between the first and third layers [17]. This method is potentially promising, although the formation of three layers increases the complexity of the process. Other techniques of forming multi-layer optical circuits include amorphous-silicon-based multi-layer stacking techniques [18,19]. In this paper, we propose a SiN/Si double-layer platform with thick (∼400 nm) SiN waveguides and a very large interlayer separation (∼1.5 µm), and demonstrate a 4 × 4 strictly nonblocking switch exhibiting −50 dB crosstalk for the worst case. We also study experimentally the scalability of our scheme to 16 × 16 and 32 × 32 switches.
Fig. 1. Schematic illustration of a double-layer platform showing Si underpasses and SiN overpasses.
In Section 2, we detail the design of the interlayer spacing and interlayer couplers for the overpass. In Section 3, we demonstrate a 4 × 4 silicon photonics switch with a worst-case crosstalk of less than −50 dB with a bandwidth of more than the C-band. In Section 4, we show the scalability of our system by demonstrating a 16 × 16 switch with a worst-case crosstalk of less than −45 dB over a bandwidth of the C-band. We consider that these crosstalk values are not limited by the overpasses but by the gate switches. By measuring a passive test device for a 32 × 32 switch that contains 961 crossings, we conclude that our scheme can scale up to 32 × 32 with a worst-case crosstalk of less than −50 dB if the gate switches are appropriately designed. Conclusions are presented in Section 5.
2. Design
We first analyzed the insertion loss (IL) of the double-layer crossing, as schematically shown in Fig. 2(a), using a 3D finite difference time domain (FDTD) method. The transmitted power through the SiN waveguide is calculated for various values of the gap between the Si and SiN layers. The Si and SiN waveguides have widths of 0.43 µm and 1 µm, respectively, and heights of 0.22 µm and 0.4 µm. These waveguides cross orthogonally and are separated by a vertical gap, where the gap is defined as the distance between the upper face of the Si waveguide and the lower face of the SiN waveguide. A transverse-magnetic (TM)-like mode is used throughout this paper, because the TM-like mode is generally robust against fabrication errors, which helps the fabrication of interlayer couplers. The calculated results are shown in Fig. 2(b), showing a maximum insertion loss of approximately 0.0005 dB for a gap of 1.5 µm and above. A loss of 0.0005 dB is considered negligible for a 32 × 32 switch, because the accumulated loss from the intersections on the worst path will be only 961 × 0.0005 dB ∼ 0.5 dB. We note that the insertion loss for light travelling along the Si waveguide will be less than that of the SiN waveguide, because the modal volume of the Si waveguide is smaller than that of the SiN waveguide. Therefore, we considered the insertion loss of the light travelling in the SiN waveguide, which is not addressed in [16]. In conclusion of these discussions, we determined a gap of 1.5 µm to be appropriate.
Fig. 2. (a) Schematic illustration of the structure for IL calculations. (b) Calculated IL for a double-layer crossing where light with a wavelength of 1.55 µm passes through the SiN waveguide. A 3D FDTD method was used. (c) Calculated leakage values for the SiN waveguide to the Si waveguide (circles) and that without the Si waveguide (or analytical artifacts, cross marks). This result shows that the leakage will be less than the analytical artifacts (< −45 dB) for a gap of 1.5 µm. The waveguide dimensions were 0.43 × 0.22 µm2 and 1 × 0.4 µm2 for the Si and SiN waveguides, respectively.
We next calculated the leakage from the double-layer crossing. As schematically shown in Fig. 2(a), we analyzed the rate of light leakage from the SiN waveguide to the Si waveguide. Figure 2(c) shows the calculation results. The circles and cross marks show the leakage values with and without the Si waveguide, respectively. It can be seen that the two results overlap each other, indicating that the leakage calculated here is limited only by the analytical artifacts, and therefore the leakage in an actual device is expected to be lower than −45 dB for a gap of 1.5 µm.
For the optical coupling between the Si and SiN waveguides, we designed a directional coupler (DC)-based interlayer coupler, as schematically shown in Fig. 3(a). DCs are more advantageous than adiabatic couplers in our platform because they have high coupling efficiency even if the gap between the two layers is large. For the coupling between vertically separated waveguides, the mode field diameters are expanded using inverse tapers with lengths of 30 µm and 45 µm for the Si and SiN waveguides, respectively. The widths of the Si and SiN waveguides at the DC region were designed to be 0.17 µm and 0.45 µm, respectively, so that their effective indexes match. We calculated the tolerance of the coupling efficiency to the width of the SiN waveguide and the gap between layers, as shown in Figs. 3(b) and 3(c), respectively. These results show that a coupling loss of less than 1 dB is possible with a fabrication tolerance of ± 20 nm for the width of the SiN waveguide, which is achievable with our fabrication process. The structure also has tolerance for the gap distance between the two layers, as shown in Fig. 3(c). We note that the tolerance to the width of the SiN waveguide can further be alleviated if we apply slightly tapered structure at the DC region, although the total length will be extended.
Fig. 3. (a) Top view (upper) and cross-sectional view (lower) of the DC-based interlayer coupling structure. (b) Calculated coupling efficiencies showing the tolerance for the width of the SiN waveguide at the DC region. (c) Calculated coupling efficiencies showing the tolerance for the gap between layers.
3. Demonstration of double-layer 4 × 4 switch device
Here, we demonstrate a 4 × 4 silicon photonics switch using our double-layer platform. Figure 4(a) shows a microscope image of the fabricated switch. We fabricated this device using the Super Clean Room facility (SCR) of AIST equipped with a 300-mm ArF-immersion photolithography system. We used PECVD to grow the SiN waveguide material with the temperature of below 400 °C. We also performed post processing in which we slightly etched the SiN waveguide using HF to match the effective indices of the Si and SiN waveguides at the interlayer coupling structures. In later fabrications we obtained the designed shape without the post processing by adding a feedback to the fabrication process, as described in Section 4. The 4 × 4 switch consists of four 1 × 4 switches, 16 gate switches, double-layer intersections, and interlayer coupling structures. The 1 × 4 switches and gate switches consist of thermo-optic Mach–Zehnder element switches with two 3-dB directional couplers and two phase shifters for each arm. The gate switches helped suppress the crosstalk from the 1 × 4 switches. Figures 4(b) and 4(c) show the interlayer coupling structure where the focus is on the Si waveguide and the SiN waveguide, respectively. Figure 4(d) shows a cross-sectional STEM image of the interlayer coupling region.
Fig. 4. (a) Optical micrograph of a fabricated 4 × 4 switch. (b, c) Magnified image of the interlayer coupling structure focused on the (b) Si waveguide and (c) SiN waveguide. (d) Cross-sectional STEM image of the interlayer coupling region.
Before evaluating the 4 × 4 switch, we measured the IL of the double-layer intersections with the test device shown in Fig. 5(a). The measured transmitted power through the SiN waveguide with double-layer intersections is shown in Fig. 5(b) for two identically prepared samples. Due to the large variance of the measured points, we could not determine a reliable value of the loss of the intersection. We consider that this variance is caused by the post processing, which also etched the edge couplers of the sample. However, because the points are above the 0.005-dB/intersection line, as shown in Fig. 5(b), we conclude that the loss is smaller than 0.005 dB/intersection. This value is very small compared to that of single-layer intersections, which range from 0.023 to 0.2 dB [20].
Fig. 5. (a) Optical micrograph of a test device to measure the IL of double-layer intersections. (b) Measured transmitted power as a function of the number of intersections. The two samples were identically fabricated. The wavelengths of 1543 nm and 1550 nm were used when measuring the sample 1 and 2, respectively. We consider that the wavelength dependency is small in this wavelength range.
We then measured the IL spectrum of the interlayer coupling structure, as shown in Fig. 6(a). Figure 6(b) shows an optical micrograph of the fabricated interlayer structure. The results are shown in Fig. 6(c). The IL of the two coupling structures (Si → SiN → Si) was less than 1 dB for most of the C-band. This value is acceptable because only two coupling structures exist on the path between the input and output in the switch-and-select topology, as schematically shown in Fig. 8(a).
Fig. 6. (a) Schematic illustration of the interlayer coupling structure. (b) Optical micrograph of the test device for measuring the IL of the interlayer coupling structure. (c) Measured spectrum of the transmitted power.
Figure 7(a) shows the measured fiber-to-fiber transmission for all path settings of the fabricated 4 × 4 switch. The upper points show the fiber-to-fiber IL for the target output ports, with an average value of approximately 12 dB. We detail the breakdown of this loss in the following. The sum of the fiber-to-chip and chip-to-fiber coupling losses was 4.2 dB. The on-chip guiding loss for the Si waveguide was approximately 1 dB. The loss for the SiN waveguide and the interlayer coupling structure ranged from 0 dB to 10 dB depending on the path, because the length of the SiN waveguide varied, as shown in Fig. 8(a). The observed loss for the SiN waveguide was relatively large (∼27 dB/cm), due to the surface roughness caused by the post processing. However, this value can be reduced to less than 3 dB/cm by process optimization, as described in detail in Section 4. The other losses, including the loss at the 3-dB couplers, was approximately 1 dB. We note that the loss for the double-layer intersections was almost negligible (<0.05 dB for 9 intersections). The lower points in Fig. 7(a) show the leakages to the non-target ports, which are quite small. Figure 7(b) shows the measured spectrum of the target-output ports. These results show that the IL spectrum is almost flat in the C-band.
Fig. 7. (a) Measured results for all-path fiber-to-fiber transmission for the fabricated 4 × 4 switch. The upper points show the target-port outputs and the lower points show the non-target (leaked) outputs. (b) Measured spectrum of all the target-port outputs.
The path setting shown in Fig. 8(a) is one of the worst crosstalk paths for the main path of 1 → 4′. The number of intersections on the path is (4–1)2 = 9. Figure 8(b) shows the detailed structure of the input selector and the gate switches. Because we used DC-based 3-dB couplers for the 2 × 2 element switches, wavelength-dependent leakages to the bar port are large when the switch is in the cross state. Therefore, we designed the gate switches to become bar state to block the leaked light as shown in the blue arrow in Fig. 8(b). In the following, we describe the method to measure the crosstalk of the switch. First, the paths shown in Fig. 8(a) were prepared. We next input continuous-wave light with fixed power to port 2, 3, and 4 and measured each power leaked to output port 4′. The sum of the leaked power was then divided by the transmitted power from input port 4 to output port 4′. The measured spectrum of the crosstalk is shown in Fig. 8(c). The crosstalk is less than −50 dB over the C-band. Leakages to the main path come from (1) the leakage at the double-layer intersections and (2) the leakage at the selectors (or the gate switches). Considering the fact that the largest leakage came from 4 → 4′, (2) will be the dominant factor because the path 4 → 4′ does not contain SiN waveguide sections and has the smallest loss. Therefore, the crosstalk value can be further suppressed by inserting additional gate switches in the selectors. We note that the effect of interference between crosstalk paths could be neglected because the lengths of the crosstalk paths are much different from each other (∼500 µm), resulting in random phase between crosstalk signals in which the effect of interference could be averaged.
Fig. 8. (a) Schematic illustration of the 4 × 4 switch and the path setting for the crosstalk measurement, where the orange lines show the main path. (b) The detailed structure of the 1 × 4 switch part and the gate switch part. Leakage from the 1 × 4 switch part (blue arrow, for example) is blocked by the bar-state gate switch. (c) Measured spectrum of the crosstalk of one of the worst crosstalk paths.
4. Examination of 16 × 16 switch and 32 × 32 test device
In this section, we examine SiN/Si double-layer switches with an increased port count. We designed and fabricated a 16 × 16 switch by simply expanding the port count. Figure 9(a) shows a schematic illustration of the 16 × 16 switch. As shown, the topology is basically the same as that of the 4 × 4 switch. This test device has 16 input selectors (1 × 16 switches), 2 output selectors corresponding to output ports 8′ and 16′ (16 × 1 switches plus gate switches), and double-layer intersections. We omitted output selectors 1′−7′ and 9′−15′ from the test device due to packaging limitations. To test the performance of this switch, we assumed a fixed switching state of 1−16′, 2−1′, 3−2′, …, 15−14′, and 16−15′, as illustrated by the orange and green lines in Fig. 9(a). The orange path is the main path, which experiences both the largest crosstalk and the largest number of crossings (225) on the path. Output ports other than 8′ and 16′ were monitored without using the selector, as shown for 1′ in Fig. 9(a). Figure 9(b) shows an optical micrograph of the fabricated 16 × 16 switch chip. The chip size is 14 mm × 13 mm. 304 pads were wire-bonded to a ceramic package. A 127-µm-pitch, 34-fiber-array was bonded on the chip edge. In this device, the gate switches are connected in the output selectors as shown in Fig. 9(c). In order to reduce the leakages, two gate switches are added to the paths in which the rightmost element switches of the 1 × 16 selectors become bar state to connect, as shown in the blue arrow in Fig. 9(c).
Fig. 9. (a) Schematic illustration of the 16 × 16 switch and the path setting for the crosstalk measurement, where the orange lines show the main path. (b) Optical micrograph of the fabricated 16 × 16 switch. (c) Detailed structure of the output selector with 1 × 16 switch and gate switches.
The chip was fabricated in the SCR of AIST. We applied a feedback to the width of the SiN waveguide at the coupling region from the 4 × 4 switch fabrication in the following steps. First, we calculated the effective index of the SiN waveguide from the observed cross-sectional structure of Fig. 4(d) with Finite Element Method (FEM). After that, we adjusted the width of the standard channel-type SiN waveguide to match the effective index. As a result, the width of 0.48 µm was obtained for the effective index of 1.49. We did not apply the post-process and this reduced the propagation loss of the SiN waveguide from ∼27 dB/cm to ∼2.1 dB/cm. Figure 10(a) shows the result of the cutback measurement of the interlayer coupling structure. This figure shows that the IL of the interlayer coupling (Si → SiN → Si) was 2.5 dB without the post processing. Although this value is slightly larger than the 4 × 4 switch case, additional feedback in the fabrication will further improve the value. Figure 10(b) shows the results of a cutback measurement for the IL of the double-layer intersections. Compared to the results of Fig. 5(b), the improved process without the post processing resulted in smaller fluctuations. The resulting loss was 0.0032 dB/intersection. In the worst-case path, there were 225 intersections giving a loss of 0.7 dB, which is sufficiently small. Due to the fact that the process was not optimized, the gap between the SiN waveguides and Al wires partly overlaying the SiN waveguides was thinner than the design (∼1.5 µm). We measured the IL for switch paths with various overlap lengths between the SiN waveguide and the Al wires and found that the loss due to the Al wire, mainly by absorption, was approximately 10 dB/mm. A numerical calculation using FEM showed that the gap value of around 0.6 µm reproduces the observed loss value, and the loss should be as low as 1.0 dB/cm at the designed gap value of 1.5 µm. This problem did not occur in the 4 × 4 switch since we did not need to put Al wires over the SiN waveguides thanks to the small number of Al wires, as seen in Fig. 4(a). Because this loss can be eliminated by optimizing the process, in the following measurements we numerically subtracted the metal loss from the measured transmittance by the following four steps. We note that the total length of metal within the light paths was in the range of zero to 3.2 mm.
Fig. 10. Results of cutback measurements of (a) the interlayer coupling structure and (b) the double-layer intersections. Compared to Fig. 5(b), the improved process without the post processing resulted in smaller fluctuations of the coupling efficiencies to the fibers.
- Step 1. We measured the transmittance of the paths that do not have output selectors: 2−1′, 3−2′, 4−3′, 5−4′, 6−5′, 7−6′, 8−7′, 10−9′, 11−10′, 12−11′, 13−12′, 14−13′, 15−14′, and 16−15′.
- Step 2. We subtracted the propagation loss for the SiN waveguide (which has been measured from cutback measurements, as shown in Table 1) from the measured transmittance.
- Step 3. We plotted the transmittance as a function of the total length of the sections on the path where metal exists over the SiN waveguide, and calculated the metal loss per unit length from the slope. For example, the metal loss was 10.5 dB/mm for a wavelength of 1.55 µm. We note that the metal-loss prediction using this value includes prediction error with standard deviation of 1.6 dB in the overlap length ranged from zero to 3.2 mm.
- Step 4. We subtracted the metal loss from all the measured transmittance values.
The blue circles around −15 dB in Fig. 11(a) show the measured fiber-to-fiber transmission spectrum of the main path of 1−16′ with the metal loss subtracted. The other points in Fig. 11(a) show the leakages to the main path from non-target input ports (2−16′, 3−16′, …). As can be seen, the spectra are almost flat in the C-band and the leakage is suppressed by more than 55 dB. The fiber-to-fiber insertion loss at a wavelength of 1550 nm was ∼15 dB for the main path. We detail the breakdown of this loss in the following. The fiber-to-chip coupling loss was 3.9 dB for both edges. The propagation losses for Si and SiN waveguides were 3.4 dB and 2.8 dB, respectively. The interlayer coupling loss was 2.5 dB for both edges (Si → SiN → Si). The sum of the losses from all intersections was a maximum of 0.7 dB. Other losses were ∼3 dB, including the transmission loss for the eight MZIs and two gate switches on the path. Figure 11(b) shows the measured crosstalk spectrum of the switch. The crosstalk was below −45 dB for almost the entirety of the C-band. The crosstalk was larger by ∼5 dB than the 4 × 4 switch case [Fig. 8(c)]. We consider that the increased port count (from four 1 × 4 selectors to sixteen 1 × 16 selectors) increased the leakage from the selectors by a factor of 4 (6 dB) over the 4 × 4 switch. This observation confirms that the leakage at the selectors still dominates the crosstalk in the 16 × 16 switch. Therefore, it is expected that the crosstalk value can be further suppressed by inserting more gate switches on a path. We note that due to the prediction error of the metal loss as described in Step 3, the y-values in Figs. 11(a) and 11(b) are expected to have inaccuracy of ± 1.6 dB at least, although it is small as compared to the range of the y-axes.
Fig. 11. (a) Measured transmission spectrum of switch paths with metal losses eliminated. (b) Calculated crosstalk spectrum corresponding to the worst crosstalk path. The crosstalk was below −45 dB for most of the C-band.
Table 1. SiN loss obtained by cutback measurements.
To investigate the feasibility of a higher port count, we designed and fabricated a 32 × 32-scale passive test device, which included 961 double-layer intersections, corresponding to the largest number of intersections on a path for a 32 × 32 switch. Figure 12(a) shows an optical micrograph of the fabricated 32 × 32-scale test chip. We first measured the transmission spectrum of the SiN waveguide path, as shown by the orange path in Fig. 12(b). We then measured the Si waveguide path, as shown in Fig. 12(c). Finally, we measured the “crosstalk path” from SiN (Si) to Si (SiN) as shown in Fig. 12(d).
Fig. 12. (a) Optical micrograph of the 32 × 32-scale passive test chip. (b) Schematic illustration of the test chip, where 961 Si waveguides are crossed by a SiN waveguide. We first guided light through the SiN waveguide and then (c) through one of the Si waveguides, and finally (d) we measured the crosstalk power from SiN (Si) to Si (SiN).
The measurement results are shown in Fig. 13(a). The blue, orange, and green circles show the transmission spectra for the cases of Figs. 12(b)–12(d), respectively. Using the measured Pa value and the propagation loss for the SiN waveguides listed in Table 1, the loss from the input fiber to the crossing point of Fig. 12(d) was calculated to be (i) 10.1 dB at a wavelength of 1.55 µm. This loss includes a fiber-to-chip coupling loss of ∼2 dB, a silicon waveguiding loss of ∼0.8 dB, a SiN/Si interlayer coupling loss of ∼1.3 dB, a SiN waveguiding loss of ∼3 dB, and a loss for the SiN/Si double-layer intersections of 0.0032 × 961 ∼ 3 dB. Half of the measured Pb value corresponds to the loss from the crossing point of Fig. 12(d) to the output fiber, which was (ii) 4.4 dB. By subtracting the above (i) and (ii) values from the measured Pc value, the crosstalk for a single intersection can be calculated. The results are shown by black circles in Fig. 13(b). We multiplied these results by 31 to obtain the orange circles in Fig. 13(b), which corresponded to the crosstalk due to the crossings in the 32 × 32 switch. From these results, we can see that even in a 32 × 32 switch, a crosstalk value of below −50 dB is expected. Therefore, we conclude that the double-layer platform is scalable to large-scale switches such as 32 × 32. These results also confirm that the measured crosstalk values for the 4 × 4 and 16 × 16 switches presented in Sections 3 and 4 were not limited by the double-layer intersections.
Fig. 13. (a) Measured transmission spectra for the cases shown in Figs. 12(b)–12(d). (b) Calculated crosstalk for a single crossing (black circles). The orange circles are scaled by a factor 31, corresponding to the case of the total crosstalk due to crossings in the 32 × 32 switch.
5. Conclusion
In this paper, we proposed and examined a SiN/Si double-layer optical platform for ultra-low crosstalk optical switches. We have shown that a SiN/Si gap of 1.5 µm produces a loss that is negligible compared to that for in-plane crossings and realized double-layer intersections with an ultralow-loss of 0.0032 dB/intersection. We also designed a DC-based interlayer coupler with a loss of less than 1 dB. We then designed and fabricated 4 × 4 and 16 × 16 silicon photonics switches using the double-layer platform, based on the switch-and-select topology. We obtained crosstalk values below −50 dB and −45 dB for the 4 × 4 and 16 × 16 switches, respectively, almost entirely in the C-band. We then fabricated and measured a 32 × 32-scale passive test device and found that a crosstalk of less than −50 dB is expected if the gate switches are appropriately designed. We conclude that the double-layer platform is promising to reduce the loss and the crosstalk especially with architectures comprising of a single crossing block (i.e. single Si → SiN → Si transition), such as the switch-and-select topology.
Funding
Ministry of Education, Culture, Sports, Science and Technology (MEXT) (Project for Developing Innovation Systems).
Acknowledgment
The device fabrication was supported by TIA SCR of AIST.
References
1. K. Ishii, J. Kurumida, S. Namiki, T. Hasama, and H. Ishikawa, “Energy consumption and traffic scaling of dynami optical path networks,” Proc. SPIE 8646, 86460A (2013).
2. K. Ueda, Y. Mori, H. Hasegawa, and K. Sato, “Novel Intra- and Inter-Datacenter Converged Network Exploiting Space- and Wavelength-Dimensional Switches,” in Optical Fiber Communication Conference (Optical Society of America, 2017), paper M3 K.2.
3. K. Suzuki, K. Tanizawa, T. Matsukawa, G. Cong, S.-H. Kim, S. Suda, M. Ohno, T. Chiba, H. Tadokoro, M. Yanagihara, Y. Igarashi, M. Masahara, S. Namiki, and H. Kawashima, “Ultra-compact 8 × 8 strictly-non-blocking Si-wire PILOSS switch,” Opt. Express 22(4), 3887–3894 (2014). [CrossRef]
4. H. Matsuura, S. Suda, K. Tanizawa, K. Suzuki, K. Ikeda, H. Kawashima, and S. Namiki, “Accelerating switching speed of thermo-optic MZI silicon-photonic switches with “turbo pulse” in PWM control,” in Optical Fiber Communications Conference (Optical Society of America, 2017), paper W4E.3
5. 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]
6. R. Konoike, K. Suzuki, T. Inoue, T. Matsumoto, T. Kurahashi, A. Uetake, K. Takabayashi, S. Akiyama, S. Sekiguchi, S. Namiki, H. Kawashima, and K. Ikeda, “SOA-Integrated Silicon Photonics Switch and its Lossless Multistage Transmission of High-Capacity WDM Signals,” J. Lightwave Technol. 37(1), 123–130 (2019). [CrossRef]
7. L. Chen and Y. Chen, “Compact, low-loss and low-power 8×8 broadband silicon optical switch,” Opt. Express 20(17), 18977–18985 (2012). [CrossRef]
8. P. Sanchis, P. Villalba, F. Cuesta, A. Håkansson, A. Griol, J. V. Galán, A. Brimont, and J. Martí, “Highly efficient crossing structure for silicon-on-insulator waveguides,” Opt. Lett. 34(18), 2760–2762 (2009). [CrossRef]
9. Y. Ma, Y. Zhang, S. Yang, A. Novack, R. Ding, A. E. Lim, G. Lo, T. Baehr-jones, and M. Hochberg, “Ultralow loss single layer submicron silicon waveguide crossing for SOI optical interconnect,” Opt. Express 21(24), 29374–29382 (2013). [CrossRef]
10. 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]
11. K. Suzuki, R. Konoike, K. Tanizawa, S. Suda, H. Matsuura, K. Ikeda, S. Namiki, and H. Kawashima, “Ultralow-crosstalk and broadband multi-port optical switch using SiN/Si double-layer platform,” in Proceedings of Opto-Electronics and Communications Conference (IEEE, 2017), postdeadline paper.
12. A. Nesic, M. Blaicher, T. Hoose, M. Lauermann, Y. Kutuvantavida, W. Freude, and C. Koos, “Hybrid 2D / 3D Photonic Integration for Non-Planar Circuit Topologies,” in Proceedings of European Conference on Optical Communication (IEEE, 2016), 2(1), pp. 707–709.
13. K. Shang, S. Pathak, B. Guan, G. Liu, and S. J. B. Yoo, “Low-loss compact multilayer silicon nitride platform for 3D photonic integrated circuits,” Opt. Express 23(16), 21334–21342 (2015). [CrossRef]
14. A. M. Jones, C. T. DeRose, A. L. Lentine, D. C. Trotter, A. L. Starbuck, and R. A. Norwood, “Ultra-low crosstalk, CMOS compatible waveguide crossings for densely integrated photonic interconnection networks,” Opt. Express 21(10), 12002–12013 (2013). [CrossRef]
15. W. D. Sacher, Y. Huang, G. Q. Lo, and J. K. S. Poon, “Multilayer silicon nitride-on-silicon integrated photonic platforms and devices,” J. Lightwave Technol. 33(4), 901–910 (2015). [CrossRef]
16. Q. Cheng, L. Y. Dai, N. C. Abrams, Y.-H. Hung, P. E. Morrissey, M. Glick, P. O’Brien, and K. Bergman, “Ultralow-crosstalk, strictly non-blocking microring-based optical switch,” Photonics Res. 7(2), 155–161 (2019). [CrossRef]
17. W. D. Sacher, J. C. Mikkelsen, P. Dumais, J. Jiang, D. Goodwill, X. Luo, Y. Huang, Y. Yang, A. Bois, P. G.-Q. Lo, E. Bernier, and J. K. S. Poon, “Tri-layer silicon nitride-on-silicon photonic platform for ultra-low-loss crossings and interlayer transitions,” Opt. Express 25(25), 30862–30875 (2017). [CrossRef]
18. J. Kang, N. Nishiyama, Y. Atsumi, T. Amemiya, and S. Arai, “Multi-stacked silicon wire waveguides and couplers toward 3D optical interconnects,” Proc. SPIE 8630, 863008 (2013).
19. J. Chiles, S. Buckley, N. Nader, S. W. Nam, R. P. Mirin, and J. M. Shainline, “Multi-planar amorphous silicon photonics with compact interplanar couplers, cross talk mitigation, and low crossing loss,” APL Photonics 2(11), 116101 (2017). [CrossRef]
20. K. Suzuki, G. Cong, K. Tanizawa, S.-H. Kim, K. Ikeda, S. Namiki, and H. Kawashima, “Multiport optical switches integrated on Si photonics platform,” IEICE Electron. Express 11(24), 20142011 (2014). [CrossRef]
