We have experimentally demonstrated a compact polarization beam splitter (PBS) based on the silicon nitride/silicon-on-insulator platform using the recently proposed augmented-low-index-guiding (ALIG) waveguide structure. The two orthogonal polarizations are split in an asymmetric multimode interference (MMI) section, which was 1.6 μm wide and 4.8 μm long. The device works well over the entire C-band wavelength range and has a measured low insertion loss of less than 1 dB. The polarization extinction ratio at the Bar Port is approximately 17 dB and at the Cross Port is approximately 25 dB. The design of the device is robust and has a good fabrication tolerance.
© 2017 Optical Society of America
Waveguides with a high refractive index contrast between the core and cladding generally result in a small footprint and have attracted much interest in recent years . Such waveguides typically have a high birefringence and the propagation properties of the transverse electric (TE) and the transverse magnetic (TM) modes are quite different. Hence, on-chip polarization control is an important issue for photonics integrated circuits. One useful device used to control the polarization on a chip is the polarization beam splitter (PBS) which divides the input light into two orthogonal polarization components .
Many different types of waveguide-based PBSs have been proposed using a number of different schemes. PBSs based on directional couplers [3–7] tend to be sensitive to wavelength and changes in dimensions. Multiple coupling sections have also been investigated and a 40 μm long, wideband PBS based on cascaded coupling sections was reported in µm . A PBS based on grating-assisted contradirectional coupling with good dimension tolerance, and a length of 13.76 μm to 30.96 μm has also been demonstrated . While PBSs based on plasmonic waveguides suffer from high insertion losses and are more difficult to fabricate [10–12]. Photonic crystal based PBS have been demonstrated and suffer from reduced transmission due to reflection and scattering . A compact PBS with a footprint of only 2.4 × 2.4 μm2 has been designed using a nonlinear search algorithm , but the corresponding extinction ratios are less than 12 dB, and the device is discretized into hundreds of 120 nm × 120 nm silicon/air pillars which could introduce additional sources of scattering and require high precision lithography. PBSs based on the Mach-Zehnder interferometer , mode evolution , and symmetric multimode interference (MMI) structures [17–19] require lengths of hundreds of microns to millimeters. PBSs based on an asymmetrical MMI waveguides with a short device length have been designed using silicon slot waveguides  and hybrid plasmonic waveguides .
We theoretically proposed a compact PBS based on the augmented-low-index guiding (ALIG) structure in , which utilizes a short asymmetric MMI section. We have experimentally demonstrated this device and have shown that the TM mode is directed to the Bar Port of the ALIG waveguide, while the TE mode forms a mirror image and is directed to the Cross Port via the MMI section. The proposed PBS has a very small footprint, low insertion loss, high polarization extinction ratio (PER), and good fabrication tolerance.
2. Device design
The ALIG waveguide consists of a thinned silicon layer and a thin silicon nitride layer working together as the core. The TE and TM polarizations can then be controlled separately by controlling the waveguide’s dimensions in different layers. For example, when the width of the silicon layer is changed, the mode distribution of the TE mode changes accordingly; while the TM mode remains undisturbed. Based on this principle, we propose a compact PBS. Figures 1(a) and 1(b) depict the three-dimensional view and the top view of the proposed PBS, respectively. The device consists of an input ALIG waveguide, followed by an asymmetric MMI section, and two output waveguides for different polarizations. The TM mode is guided straight to the Bar Port, while the TE mode is guided to the Cross Port via the principle of multi-mode interference in the MMI section. In order to achieve a smooth transition for the TE mode from the input waveguide to the MMI section, a horizontal taper is used at the input side. At the output side, another taper is used to collect the TE mode and change the width of the cross waveguide. This segment is then followed by a curved waveguide.
To keep with the standard integrated photonic design protocols, the whole device is covered with a silica upper cladding. The input and output waveguides from both the Bar and the Cross Ports are tapered to the dimensions of the standard single mode ALIG waveguide for on-chip routing to the edge of the wafer.
The cross section of the input waveguide is shown in Fig. 2(a). Note that in the design of ALIG waveguide, the thicknesses of silicon (t1 = 120 nm) and silicon nitride (t2 = 350 nm) layers were optimized such that most of the power for the TM mode resides in the silicon nitride layer; whereas, the TE mode is mostly confined in the silicon layer. The cross section of the asymmetric MMI section is shown in Fig. 2(b). We chose W1 = 450 nm, while WMMI = 1.6 µm, which is wide enough for separation of the two orthogonal polarizations at the output side. From Fig. 2(b) it is clear that in the MMI section the silicon nitride layer has two different thicknesses; i.e. t2 = 350 nm and 0 < t3 ≤ 20 nm. Ideally, t3 = 0 nm is preferred; however, considering the fact that the change in silicon thickness dramatically affects the effective index of the TE mode, to improve fabrication tolerance, we chose 0 < t3 ≤ 20 nm to ensure that the silicon layer is not etched. Cross section of the output waveguide at the Cross Port is shown in Fig. 2(c). The width is chosen to be W2 = 600 nm. Cross section of the output waveguide at the Bar Port is the same as the input waveguide shown in Fig. 2(a).
The mode profiles are simulated using Lumerical MODE Solutions . The intensity profiles of the TE and TM modes of the input ALIG waveguide are shown in Figs. 3(a) and (b), respectively. Only fundamental modes exist in the input waveguide. For the TM mode, only 7.5% of the power resides in the silicon layer while 51% of power resides in the silicon nitride layer; thus, the TM mode is mostly confined in the lower-index silicon nitride layer.
As shown in Fig. 4, the asymmetric MMI section supports two TM and four TE modes; where the fundamental and higher order TE modes interfere, forming images in the MMI section. Moreover, the input TM0 has a very good overlap with the TM0 in the MMI section, while the TM1 in the MMI section is not excited due to the very small mode overlap. In short, the presence of the higher order TM1 mode has little impact on the operation of the PBS.
In , we have studied fabrication tolerance of the proposed PBS, and in particular the effects of the length and width of the MMI section on the device’s performance. Here, we study the effects of the thickness of the thin silicon nitride layer, t3. According to the self-imaging principle, the excited modes in a MMI section will interfere with each other due to their different propagation constants, forming a periodic mirror image along the MMI section. This periodic image has a characteristic length, referred to as the beat length (Lπ) . The beat length is calculated according to: Lπ = π/(β0-β1), where β0 and β1 are the propagation constants of the two lowest order modes. This implies that beat length is also inversely proportional to the difference of effective mode indices of the two lowest order modes. Figure 5 shows the calculated effective mode indices (neff) of the two lowest orders TE modes and the fundamental TM mode of the MMI section with varied silicon nitride thickness (t3). As the figure indicates effective mode indices for both TE0 and TE1 modes increase slightly with increasing t3, and the difference in effective mode indices between the two lowest order TE modes () is almost constant. This indicates a thin layer of silicon nitride does not have much impact on the beat length.
The full device is analyzed using 3D finite-difference-time-domain (FDTD) method . For the MMI section with WMMI = 1.6 µm, mirror image for the TE mode forms at LMMI = 4.8 μm. A 2 µm long taper La = 2 µm is used to change the silicon layer width for the input waveguide from W1 = 450 nm to Wa = 900 nm for the TE mode. At the output side, a taper with Lb = 1 µm is used to collect the TE mode and change the width of the cross waveguide from Wb = 1 μm to W2 = 600 nm. The gap is g = 100 nm. This segment is then followed by a curved waveguide with W2 = 600 nm and radius of R = 10µm. To keep with standard integrated photonic circuit design protocols, the whole device is covered with a silica upper cladding. The input and output waveguides from both Bar and Cross Port are all tapered to a single mode waveguide with t1 = 120 nm, t2 = 350 nm, and width of 600 nm for on-chip routing to the edge of the wafer. The guided power distributions for the TE and TM modes, as a function of position along the device, are shown in Figs. 6(a) and 6(b), respectively. The TE mode couples to the Cross Port due to the multimode interference, whereas The TM mode is only slightly affected by the MMI section and is coupled to the Bar Port.
We have calculated the transmission across the C-band (1530-1565 nm) for both TM and TE modes to either the Bar or Cross Ports. As shown in Fig. 7(a) most of the power for the TM mode is directed to the Bar Port while most of the power for the TE mode goes to the Cross Port (Fig. 7(b)). Transmissions for both modes to their corresponding ports are high (> −0.4 dB), which implies a low insertion loss. Figure 7 also shows the effect of different silicon nitride thickness, t3 (for definition of t3 see Fig. 2). Transmissions to the undesired ports are all low for t3 = 0 to 40 nm. Transmissions of the undesired TE mode to the Bar port is equal to or lower than −18 dB and transmissions of the undesired TM mode to the Cross Port are equal or lower than −20 dB (lower than −25 dB when t3 = 0 nm). From simulations we expect the PBS response to be reasonably broadband and have a good fabrication tolerance.
3. Device fabrication
The fabrication process steps used are summarized in Fig. 8. Fabrication began with a standard SOI wafer with a 220 nm Si device layer, as shown in Fig. 8(a). In order to thin the silicon device layer, a thermal oxidation process is used. The SOI wafer was put in a furnace at 1100 °C for 5 minutes of dry oxidation, followed by 9 minutes of wet oxidation, followed by another 5 minutes of dry oxidation. As a result, approximately 100 nm of the Si device layer was consumed and a layer of silicon oxide was formed at the top, as shown in Fig. 8(b). The wafer was then immersed in the Buffered oxide etch (BOE) solution to fully etch away the top silicon oxide layer, as shown in Fig. 8(c). This process resulted in a thinned device layer with smooth top surface. Next, silicon nitride was deposted using a Low Pressure Chemical Vapor Deposition (LPCVD) process at 770 °C for 105 minutes. This resulted in a 350 nm thick layer of stiochiometric silicon nitride, as shown in Fig. 8(d). At this point the ALIG waveguide core structure is complete and we can begin to fabricate the PBS.
The proposed PBS is a multilayer structure and good alignment among layers is necessary for proper operation of the device. Using a combination of Electron beam lithography (EBL) and metal deposition followed by lift-off, tungsten markers were fabricated on the sample to define the needed coordinates for subsequent fabrication steps. In the next step, a 350 nm thick layer of the positive electron beam resist, ZEP 520A, was spin-coated on the substrate, and was patterned using EBL, as shown in Figs. 8(e) and 8(f). Reactive ion etching (RIE) was used to partially etch the top silicon nitride layer to define the partial-etched section of the asymmetric MMI segment, as shown in Fig. 8(g). Gases of SF6 and C4F8 were used in this step to produce straight wall profiles. The resist was then removed using ZDMAC solution followed by an oxygen ashing process, as shown in Fig. 8(h). Later a 350 nm thick layer of hydrogen-silsesquioxane (HSQ) was spin-coated and patterned using EBL, as shown in Fig. 8(i). RIE was used again for the full etching of the whole device, as shown in Fig. 8(j). Finally, a 1 µm thick layer of HSQ was spin-coated on the whole sample and baked to form the upper cladding as shown in (Fig. 8(k)).
A number of PBSs and reference channels were fabricated on the same chip. The reference channels are single mode ALIG waveguides with the width of 600 nm. Figure 9(a) shows the Scanning Electron Microscopy (SEM) image of the top view of the full device and Fig. 9(b) shows the zoom-in picture of the MMI section.
4. Experimental results and discussions
To characterize the devices, we have used a free space end-fire coupling scheme. The PBS device is connected to the edge of the chip using a standard, single mode ALIG waveguide with t1 = 120 nm, t2 = 350 nm, and width of 600 nm. No taper was used for free space edge coupling and the input light was directly focused on the end of the waveguide. To calibrate out the coupling loss and propagation loss for the two polarizations, several single mode ALIG straight waveguide are fabricated on the same chip. By comparing the power output from the reference waveguides and the power output from Cross or Bar Port, we can get the transmission of the PBS itself.
Figure 10 shows schematic of the setup used to characterize the fabricated ALIG-PBS. A superluminescent diode (SLD) source in conjunction with an erbium-doped fiber amplifier (EDFA) was used as the C-band source. The light was first guided in a single mode fiber and then collimated in free space using a fiber collimator. A broadband PBS cube and a half-wave plate were used to control the input polarizations. A 40X objective lens was used to couple the light into the sample and another 40X objective lens was used to collect the transmitted light. An iris and an infrared camera were used at the chip’s output to ensure that light is only collected from either Bar or Cross Port. The PBS cube after the output is movable and is used to verify the polarization state. The output light was routed to an optical spectrum analyzer (OSA) and a flip mirror was used to route the output to a photodetector to measure the power.
There are various loss mechanisms that contribute to power loss. There are losses due to coupling in and out of the waveguides, propagation losses due to the light traveling through the ALIG waveguides, and propagation losses due the ALIG-PBS itself. In order to quantify the transmission of the fabricated ALIG-PBS, we have also fabricated several reference channels on the same chip. The reference channels are all standard straight single mode ALIG waveguides with a width of 600 nm. The input and output waveguides of the ALIG-PBS were routed and tapered to match the MMI section, as shown in Fig. 9(a). The in and out coupling and propagation losses in the reference ALIG waveguides are similar to that of the ALIG-PBS, since they are on the same chip and were fabricated during the same run. The transmission efficiency of the ALIG-PBS can then be determined by comparing the transmission of the ALIG-PBS device to that of the reference waveguides.
The measured transmissions for the TE and the TM are show in Figs. 11(a) and (b) for the Bar and Cross Ports, respectively. In order to compare our experimental results to theory, we plotted our simulated results (results in Fig. 7 with t3 = 0 nm) in the same figure. As the figures indicate experimental results match the simulated transmissions well. Over the entire C-band, the measured insertion losses for both TE and TM polarizations, routed to their desired ports are less than 1 dB; while the undesired transmission of the TE mode to the Bar Port is about −18 dB and the undesired transmission of the TM mode to the Cross Port is about −25 dB.
Polarization extinction ratios (PERs), measuring the purity of a given polarization state at a desired output port, are calculated as the difference of transmissions of the two polarizations (TE and TM) at that port and are shown in Fig. 12. At the Bar Port, the TM-to-TE PER is about 17 to 18 dB while at the Cross Port the TE-to-TM PER is approximately 25 dB. The results show that the fabricated PBS works well for the entire C-band.
In conclusion, we have experimentally demonstrated a PBS based on the ALIG structure. The core of the ALIG waveguide consists of a combination of silicon and silicon nitride layers. This design allows the TE mode to be predominately confined in the silicon layer while the TM mode is predominately confined to the lower index, silicon nitride layer. This, inturn allows for the design and fabrication of compact polarization manipulation devices. In the case of the PBS the two polarizations are split in an asymmetric MMI section. The input TE mode is couples to the Cross Port via the wide multimode silicon waveguide while the TM mode is directed to the Bar Port. The PBS is compact with an MMI section of only 1.6 µm × 4.8 µm. The fabrication of such device is compatible with state-of-art silicon photonics fabrication processes. Over the entire C-band, the PBS has a low measured insertion loss of less than 1 dB for both TM and TE polarizations. The PERs were measured to be about 25 dB and 17 dB for the Cross and Bar Ports, respectively. The experimental results agree with the FDTD simulation results. With the advantages of compact size, simple structure, good fabrication tolerance, broadband response, low insertion loss and relatively high PER, the PBS has potential applications for polarization manipulation in integrated circuits.
Natural Sciences and Engineering Research Council of Canada (NSERC).
We thank the Toronto Nanofabrication Centre (TNFC) in device fabrications.
References and links
1. D. Dai, J. Bauters, and J. E. Bowers, “Passive technologies for future large-scale photonic integrated circuits on silicon: polarization handling, light non-reciprocity and loss reduction,” Light Sci. Appl. 1(3), e1 (2012). [CrossRef]
2. D. Pérez-Galacho, R. Halir, A. Ortega-Moñux, C. Alonso-Ramos, R. Zhang, P. Runge, K. Janiak, H.-G. Bach, A. G. Steffan, and Í. Molina-Fernández, “Integrated polarization beam splitter with relaxed fabrication tolerances,” Opt. Express 21(12), 14146–14151 (2013). [CrossRef] [PubMed]
3. H. Fukuda, K. Yamada, T. Tsuchizawa, T. Watanabe, H. Shinojima, and S. Itabashi, “Ultrasmall polarization splitter based on silicon wire waveguides,” Opt. Express 14(25), 12401–12408 (2006). [CrossRef] [PubMed]
5. J. Wang, D. Liang, Y. Tang, D. Dai, and J. E. Bowers, “Realization of an ultra-short silicon polarization beam splitter with an asymmetrical bent directional coupler,” Opt. Lett. 38(1), 4–6 (2013). [CrossRef] [PubMed]
8. Z. Lu, Y. Wang, F. Zhang, N. A. F. Jaeger, and L. Chrostowski, “Wideband silicon photonic polarization beamsplitter based on point-symmetric cascaded broadband couplers,” Opt. Express 23(23), 29413–29422 (2015). [CrossRef] [PubMed]
9. Y. Zhang, Y. He, J. Wu, X. Jiang, R. Liu, C. Qiu, X. Jiang, J. Yang, C. Tremblay, and Y. Su, “High-extinction-ratio silicon polarization beam splitter with tolerance to waveguide width and coupling length variations,” Opt. Express 24(6), 6586–6593 (2016). [CrossRef] [PubMed]
11. C.-L. Zou, F.-W. Sun, C.-H. Dong, X.-F. Ren, J.-M. Cui, X.-D. Chen, Z.-F. Han, and G.-C. Guo, “Broadband integrated polarization beam splitter with surface plasmon,” Opt. Lett. 36(18), 3630–3632 (2011). [CrossRef] [PubMed]
12. T. Hu, H. Qiu, Z. Zhang, X. Guo, C. Liu, M. S. Rouifed, C. G. Littlejohns, G. T. Reed, and H. Wang, “A compact ultrabroadband polarization beam splitter utilizing a hybrid plasmonic Y-branch,” IEEE Photonics J. 8(4), 1–9 (2016). [CrossRef]
13. X. Ao, L. Liu, L. Wosinski, and S. He, “Polarization beam splitter based on a two-dimensional photonic crystal of pillar type,” Appl. Phys. Lett. 89(17), 171115 (2006). [CrossRef]
14. B. Shen, P. Wang, R. Polson, and R. Menon, “An integrated-nanophotonics polarization beamsplitter with 2.4 × 2.4 μm2 footprint,” Nat. Photonics 9(6), 378–382 (2015). [CrossRef]
15. T. K. Liang and H. K. Tsang, “Integrated polarization beam splitter in high index contrast silicon-on-insulator waveguides,” IEEE Photonics Technol. Lett. 17(2), 393–395 (2005). [CrossRef]
17. B. M. A. Rahman, N. Somasiri, C. Themistos, and K. T. V. Grattan, “Design of optical polarization splitters in a single-section deeply etched MMI waveguide,” Appl. Phys. B 73(5), 613–618 (2001). [CrossRef]
18. M. Yin, W. Yang, Y. Li, X. Wang, and H. Li, “CMOS-compatible and fabrication-tolerant MMI-based polarization beam splitter,” Opt. Commun. 335, 48–52 (2015). [CrossRef]
19. Y. Huang, Z. Tu, H. Yi, Y. Li, X. Wang, and W. Hu, “High extinction ratio polarization beam splitter with multimode interference coupler on SOI,” Opt. Commun. 307, 46–49 (2013). [CrossRef]
20. Y. Xu, J. Xiao, and X. Sun, “Compact polarization beam splitter for silicon-based slot waveguides using an asymmetrical multimode waveguide,” J. Lightwave Technol. 32(24), 4884–4890 (2014). [CrossRef]
21. X. Guan, H. Wu, Y. Shi, and D. Dai, “Extremely small polarization beam splitter based on a multimode interference coupler with a silicon hybrid plasmonic waveguide,” Opt. Lett. 39(2), 259–262 (2014). [CrossRef] [PubMed]
22. X. Sun, M. Z. Alam, J. S. Aitchison, and M. Mojahedi, “Compact and broadband polarization beam splitter based on a silicon nitride augmented low-index guiding structure,” Opt. Lett. 41(1), 163–166 (2016). [CrossRef] [PubMed]
23. Lumerical Solutions, Inc., http://www.lumerical.com.
24. L. B. Soldano and E. C. M. Pennings, “Optical multi-mode interference devices based on self-imaging: principles and applications,” J. Lightwave Technol. 13(4), 615–627 (1995). [CrossRef]