Open Access
Add to BibSonomyAbstract
A high-performance silicon-based polarizing beam splitter (PBS) is proposed and demonstrated experimentally by using an improved structure with cascaded bent directional couplers. The measured extinction ratio (ER) is >35dB and the excess losses (EL) is <0.35dB around the central wavelength for both polarizations. The present PBS has a compact footprint of ~6.9 × 20μm2. The measured bandwidths for an ER of >20dB, >25dB and >30dB are ~135nm, ~95nm and ~70nm, respectively, while the measured EL is <1dB and <0.5dB in a bandwidth of ~140nm and ~85nm, respectively. The fabrication tolerance of the core-width variation is as large as ± 40nm, which makes the fabrication very easy.
© 2017 Optical Society of America
1. Introduction
On-chip polarization-handling is very important for many applications, including polarization transparent devices [1], polarization-division multiplexing [2], and quantum photonics integrated circuits [3]. A polarizing beam splitter (PBS) is one of the most important polarization-handling devices for separating/combining TE- and TM-polarizations, and has attracted intensive attention. It is becoming very critical to develop compact high- performance PBSs with a low excess loss (EL) and a high extinction ratio (ER) in a broad bandwidth.
Various structures have been demonstrated successfully for realizing on-chip PBSs [4], including directional couplers (DCs) [5–15], multimode interferometers [16,17], gratings [18], and Mach-Zehnder interferometers [19]. Among these structures, asymmetric DCs have been regarded as one of the most attractive options [4], because of the performance excellence and design simplicity. In particularity, an ultra-small PBS based on a bent DC was proposed for the first time in 2011 [10] and was realized experimentally in 2013 [11]. For this type of PBS, the ER of TE polarization is not very high yet because of some undesired residual cross-coupling in the DC [12]. A solution to improve the ER is cascading an additional TM-pass polarizer [12], which helps filter out the undesired cross-coupled power. More recently, another improved PBS was realized with a novel design consisting of two cascaded bent DCs [15]. Even though it has a high ER (>25dB) over a ~30nm bandwidth, which is one of the best results reported for an ultra-compact PBS on silicon, it is still desired to further improve the bandwidth.
In this paper, we propose and demonstrate an ultra-broadband high-performance PBS based on SOI nanowires with a SiO2 upper-cladding by introducing an improved structure with cascaded bent DCs. There are two additional bent DCs cascaded at the through- and cross-ports of the first bent DC, respectively, so that the residual power of the undesired polarizations can be filtered out efficiently. The measured bandwidths for ERs of >20dB, >25dB and >30dB are as large as ~135nm, ~95nm and ~70nm, respectively. The measured EL is <1dB and <0.5dB in a broad band of ~140nm and ~85nm, respectively. The footprint is as small as ~6.9 × 20μm2 and the fabrication tolerance of the core-width variation is as large as ± 40nm. To the best of our knowledge, this is one of the best PBS on silicon reported.
2. Structure and design
Figure 1(a)-1(b) show the schematic configuration of the proposed PBS, which consists of three cascaded DCs, i.e., DCs #1, #2 and #3. The PBS is configured to make DCs #2 and #3 work simultaneously as a part of the decoupling region of DC #1, so that the PBS footprint is almost as compact as the previous PBS designed with a single bent DC [10,11]. For the design of DC #1 and #2, the core widths (w1, w2) and the bending radii (R1, R2) are determined according to the phase-matching condition [10] for TM polarization, so that an efficient cross-coupling is achieved when choosing the angle-lengths θ1 and θ2 of the coupling region appropriately. For TE polarization, there is a large phase mismatch introduced automatically due to the huge waveguide birefringence, and thus the cross-coupling becomes weak. The undesired weak cross-coupled power of TE-polarized light in DC #1 is then filtered out efficiently by DC #2, and thus the ER for TE polarization is improved greatly. For DC #3, we choose the same widths for the two cores in order to make the design simplified. In this case, TM-polarization mode has much shorter coupling length than TE polarization because of the weaker mode confinement in ~500 × 220 nm2 SOI nanowires. Therefore, when choosing the gap width and the coupling length appropriately, the undesired power of TM polarization can be cross-coupled and filtered out from the through port of the PBS. Meanwhile the cross-coupling of TE polarization is weak enough to avoid any notable excess loss.
Noticing that DC #3 locates at the decoupled region of DC #1 and DC #2 in the present structure, the PBS is designed with two steps. The first step is for the part including DCs #1 and #2 only, and the second step is for the whole structure including all three DCs. For the first step, the design procedure is similar to that given in our previous papers [10,15], while the difference is that the SOI nanowire used here has a SiO2 upper-cladding (instead of air-cladding). As a result, the parameters should be re-determined optimally. As an example, ~500 × 220 nm2 SOI nanowires are used in the present design. The separation S between the bent waveguides is chosen to be 750nm (S = ΔR = R1−R2 = 750nm), so that the cross-coupling in bent DCs is weak for TE polarization. The radius is chosen as R2 = 15μm to minimize the bending loss for both polarizations and have sufficient phase-mismatching for TE polarization. As mentioned above, the widths (w1 and w2) and the bending radii (R1 and R2) of the bent waveguides are chosen optimally to satisfy the phase-matching condition [10], i.e. neff1 R1 = neff2 R2, where neff1 and neff2 are the effective indices of the two bent waveguides. Figures 2(a)-(b) show the calculated neffi Ri (i = 1, 2) of the bent waveguides for both polarizations. According to the phase-matching condition for TM polarization, the waveguide core-widths are chosen as w1 = 430nm and w2 = 555nm [see Fig. 2(b)]. For TE polarization, there is a large phase mismatch [see Fig. 2(a)]. With this design, the gap width between the two bent waveguides is wg1 = ΔR−(w1 + w2)/2 = 257.5nm, which makes the fabrication easy regarding the lithography process, the dry-etching process as well as the deposition/filling process of the SiO2 upper-cladding.
The arc-lengths (i.e., θ1 and θ2) of the coupling region in DCs #1 and #2 are optimized to realize complete cross-coupling for TM polarization at the desired central wavelength. Figures 3(a)-3(b) show the calculated transmissions of TE- and TM-polarizations for the structure consisting of DCs #1 and #2 only as the arc-length θ1 varies. Here the radii R3 and R6 are chosen to guarantee a negligible loss for TE and TM polarizations, respectively, while R4 and θ4 are chosen to form a low loss and short S-bend waveguide. As an example, we choose the following parameters: R3 = 4μm, R4 = 8.4μm, R6 = 8μm, θ2 = θ1 + 4°, θ4 = 16.3° and θ6 = 4°. From Figs. 3(a)-3(b), it can be seen that the transmissions of TE-polarization is insensitive to θ1 while the central wavelength for TM polarization is blue-shifted as θ1 increases due to wavelength dependence of the evanescent coupling. In order to make the central wavelength be ~1550nm, we choose θ1 = 28°, and the corresponding bandwidth for achieving an ER of >25dB is ~24nm only (for the case with DCs #1 and #2). In order to improve the bandwidth, DC #3 is cascaded to filter out the residual power of TM-polarization at the through port when the operation wavelength deviates from the central wavelength.
Fig. 3 Calculated transmissions for the structure consisting of DC #1 and DC #2 only (shwon in the inset) as θ1 varies. (a) TE polarization; (b) TM polarization.
For the present PBS, the total transmission T at the through-port for TM polarization is synthesized by the thru-transmissions (T1, T3) of DCs #1 and #3. The bandwidth of the transmission T can be maximized when the difference between the central wavelengths of DC #1 and DC #3 is optimized by adjusting the length l1. Figures 4(a)-4(b) show the calculated transmissions of TE- and TM-polarizations for the whole structure consisting of DCs #1, #2, and #3, when choosing different coupling lengths (l1 = 2, 4, 6, 8, and 10μm). Here the gap in the coupling region is chosen as wg2 = 320nm to minimize the undesired cross-coupling for TE polarization. As DC #3 is cascaded at the through port, it is not surprise to see that the transmissions at the cross port for both polarizations are insensitive to as the length l1, as shown in Figs. 4(a)-4(b). For the transmission at the through port, one sees that DC #3 introduces some EL for TE-polarization, especially in the long wavelength range [e.g., 1650nm ~1700nm, see Fig. 4(a)]. This is because that the coupling of TE-polarization in DC #3 is enhanced as the wavelength increases. Fortunately, the EL is still very low (e.g., EL<1dB when l1 = 6μm) in a broad band of 1400~1650nm. For TM-polarization, it can be seen that the transmission at the through port is depressed, in comparison with the result for the structure without DC #3 shown in Fig. 3(b). As shown in Fig. 4(b), when choosing an optimal length l1 = 6μm for DC #3, one has the largest bandwidth of a >25dB ER for TM polarization.
Figures 5(a)-5(b) give a comparison between the structures with and without DC #3. It can be seen that the bandwidth BWER>XdB for a >X ER is improved greatly with the help of DC#3. For TM polarization, the bandwidths BWER>20dB and BWER>25dB increase from 42nm to 145nm, and from 24nm to 110nm, respectively, while the EL at the cross port does not change almost. Even when it requires an ER as high as >30dB, the bandwidth is still as large as ~70nm. Meanwhile, the EL is <0.5dB. For TE polarization, DC #3 introduces a slight EL, and one has an EL of <0.5dB and <1dB in the range of λ<1600nm and λ<1650nm, respectively. Figures 5(c)-(d) show the simulated light propagation for TE- and TM-polarizations when λ = 1550nm (i.e., the central wavelength), respectively. It can be seen that efficient cross-coupling happens in DCs #1 and #2 for TM polarization. For TE-polarization, one sees that DC #2 filters out the undesired power cross-coupled in DC #1 efficiently, while some slight cross-coupling happens in DC #3. Figure 5(e)-5(f) show the simulated results for TM polarization when operating at 1500nm and 1600nm. In these cases, there is some residual power at the through port of DC #1, which prevents to achieve an ultra-high ER. Fortunately, this residual power is filtered out efficiently by DC #3, and thus an ultra-broadband PBS with a high ER is realized.
Fig. 5 Calculated results of the designed PBSs with/without DC #3. (a) TE; (b) TM. Light propagation for TE @1550nm (c),TM @1550nm (d), TM @1500nm (e), TM @1600nm (f).
3. Fabrication and measurement
The fabrication process was started from a 220 nm-thick commercial SOI wafer. The processes of electron-beam lithography and ICP (inductively coupled plasma) dry-etching were then applied to etch through the silicon core layer. Finally, a SiO2 upper-cladding was deposited on the structure with a PECVD (plasma enhanced chemical vapor deposition) process. Figures 6(a)-6(b) show the picture of the fabricated PBS together with some test structures. Focused grating couplers for TE- and TM-polarizations were used for the chip-fiber coupling because of the convenience in our lab. On the other hand, one should notice that an inverse taper is a better option regarding the broad bandwidth, particularly when using the PBS to work with fibers directly. In order to characterize the transmissions of the fabricated PBSs when TE- and TM-polarized light is launched, we fabricated two identical PBSs connected with TE- or TM-type grating couplers. A super-continuum laser and a broad-band amplified spontaneous emission (ASE) light source were used as the source. A fiber-type polarizer and polarization controller were used to adjust the polarization state of the light to be TE-/TM-polarized before launched to the chip. An optical spectrum analyzer was applied to readout the transmissions at both through- and cross-ports. Figure 6(c) shows the measured transmissions of the fabricated PBS for TE- and TM- polarizations. It can be seen that the ER at the central wavelength (~1551nm) is ~40dB for both polarizations. For TE-polariztion, the ER at the wavelength ranges of λ<1475nm and λ>1600nm becomes reduced, which is due to the limitation of the detection sensitivity of the OSA and the bandwidth of the grating couplers. Nevertheless, Fig. 6(c) and Fig. 5 show that the bandwidth for TE polariztion is very broad. For the present PBS, the bandwidth is mainly limited by the wavelength depenedence of TM polarization. For TM-polarization, the bandwidths for an ER of >20dB, >25dB and >30dB are ~135nm, ~95nm and ~70nm, respectively, which agrees well with the simulation results shown in Figs. 5(a)-5(b). To the best of our knowledge, this is the best results for a compact PBS on silicon reported.
Fig. 6 (a) Microscope pictures, (b) SEM picture, (c) measured transmissions at the through- and cross-ports of the fabricated PBS.
In order to characterize the low EL of the present PBS, we designed and fabricated some test structures with ten identical PBSs cascaded in series, as shwon in Fig. 7(a). In this way, the measurement result is robust to minimize the random errors caused by any misalignment and unstability. Figures 7(b)-7(c) show the calculated and measured ELs for the fabricated PBS. It can be seen that they agree very well. The EL at the central wavelength is <0.35dB and <0.1dB for TE- and TM-polarizations, respectively. For TE polarization, the EL is mainly caused by the undesired cross-coupling in DCs #1 and #3, which becomes strengthened as the wavelength increases. As a result, the EL increases slightly to be higher than 1.0dB in the wavelength range of λ>1650nm. For TM polarization, the bandwidth is limited due to the wavelength dependence of DCs #1 and #2. Nevertheless, one has a broad band from 1490nm to 1630nm for a <1.0dB EL. From Fig. 7(c), one sees that the present PBS has a bandwidth of ~140nm and ~85nm for achieving an EL of <1dB and <0.5dB, respectively.
In order to verify the robustness of the present PBS, we also examine the fabrication tolerance of the core widths by assuming that there is a core-width variation Δw and one has w1′ = w1 + Δw, and w2′ = w2 + Δw. Correspondingly, the gap widths are given as w3′ = w3−Δw, and w4′ = w4−Δw. In the experiment, we fabricated the PBSs with Δw = 0, ± 20nm and ± 40nm, and the measured results are shown in Figs. 8(a)-8(b). For TE-polarization, the core-width variation Δw has some influence on the weak cross-coupling in DCs. When the core width increases (i.e., Δw>0), the cross-coupling increases slightly. As a result, the ER becomes reduced and the EL increases slightly. For TM-polarization, the core-width variation Δw influences the phase-matching condition and the optimal coupling length of the DCs. Therefore, the cross-coupling becomes incomplete in DCs #1 and #2. Thus, the ER decreases due to the residual power at the through port. As shown in Fig. 8(a)-8(b), the present PBS has a > 20dB ER over a very broad wavelength band (>100nm) for both polarizations, even when the width variation Δw is as large as ± 40nm, respectively. The numerical simulation also shows that the present PBS is tolerant to some core-height variation of e.g. Δhco = ± 5nm. Such a fabrication tolerance is large enough for the mature CMOS compatible technologies, and thus makes the fabrication very easy. Finally, a comparison is given in Table 1 for the silicon-based high-performance PBSs demonstrated previously. It can be seen that the present PBS is the best one with an ultra-compact footprint, and ultra-broad bandwidth for achieving a high ER (e.g., >30dB) and a low EL (e.g., <0.5dB), as well as a large fabrication tolerance.
Table 1. Comparison of silicon-based high-performance PBSs demonstrated.
4. Conclusions
We have proposed and experimentally demonstrated an ultra-broadband high-performance PBS based on an improved structure consisting of three cascaded DCs, i.e., DCs #1, #2 and #3. For the fabricated PBS based on SOI-nanowires with a SiO2 upper-cladding, the footprint is ~6.9 × 20 μm2 only, while the measured ER and EL at the central wavelength for both polarizations is >35dB and <0.35dB, respectively. The measured bandwidths for an ER of >20dB, >25dB and >30dB are as large as ~135nm, ~95nm and ~70nm, respectively. And the measured EL is <1dB and <0.5dB in a broad band of ~140nm and ~85nm, respectively. The present PBS still works very well even when Δw = ± 40nm, which makes the fabrication very easy. To the best of our knowledge, this is one of the best silicon-based PBSs with an ultra-compact footprint, an ultra-broad band, as well as a very large fabrication tolerance.
Funding
National Natural Science Foundation of China (NSFC) (11374263, 61422510, 61431166001).
References and links
1. T. Barwicz, M. R. Watts, M. A. Popović, P. T. Rakich, L. Socci, F. X. Kärtner, E. P. Ippen, and H. I. Smith, “Polarization-transparent microphotonic devices in the strong confinement limit,” Nat. Photonics 1(1), 57–60 (2007). [CrossRef]
2. P. Dong, “Silicon photonic integrated circuits for wavelength-division multiplexing applications,” IEEE J. Sel. Top. Quantum Electron. 22(6), 370–378 (2016). [CrossRef]
3. L. T. Feng, M. Zhang, Z. Y. Zhou, M. Li, X. Xiong, L. Yu, B. S. Shi, G. P. Guo, D. X. Dai, X. F. Ren, and G. C. Guo, “On-chip coherent conversion of photonic quantum entanglement between different degrees of freedom,” Nat. Commun. 7, 11985 (2016). [CrossRef] [PubMed]
4. 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]
5. D. Dai, Z. Wang, and J. E. Bowers, “Ultrashort broadband polarization beam splitter based on an asymmetrical directional coupler,” Opt. Lett. 36(13), 2590–2592 (2011). [CrossRef] [PubMed]
6. S. Kim and M. Qi, “Copper nanorod array assisted silicon waveguide polarization beam splitter,” Opt. Express 22(8), 9508–9516 (2014). [CrossRef] [PubMed]
7. Y. Xu and J. Xiao, “Compact and high extinction ratio polarization beam splitter using subwavelength grating couplers,” Opt. Lett. 41(4), 773–776 (2016). [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. D. W. Kim, M. H. Lee, Y. Kim, and K. H. Kim, “Planar-type polarization beam splitter based on a bridged silicon waveguide coupler,” Opt. Express 23(2), 998–1004 (2015). [CrossRef] [PubMed]
10. D. Dai and J. E. Bowers, “Novel ultra-short and ultra-broadband polarization beam splitter based on a bent directional coupler,” Opt. Express 19(19), 18614–18620 (2011). [CrossRef] [PubMed]
11. 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]
12. S. Chen, H. Wu, and D. Dai, “High extinction-ratio compact polarisation beam splitter on silicon,” Electron. Lett. 52(12), 1043–1045 (2016). [CrossRef]
13. C.-W. Hsu, T.-K. Chang, J.-Y. Chen, and Y.-C. Cheng, “8.13 μm in length and CMOS compatible polarization beam splitter based on an asymmetrical directional coupler,” Appl. Opt. 55(12), 3313–3318 (2016). [CrossRef] [PubMed]
14. J. Feng, R. Akimoto, and H. Zeng, “Asymmetric silicon slot-waveguide-assisted polarizing beam splitter,” IEEE Photonics Tech. Lett. 28(12), 1294–1297 (2016). [CrossRef]
15. H. Wu and D. Dai, “Novel high-performance polarization beam splitter on silicon,” in Asia Communications and Photonics Conference 2016, OSA Technical Digest (online) (Optical Society of America, 2016), paper AF3B.4. [CrossRef]
16. 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]
17. 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]
18. 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]
19. T. K. Liang and H. K. Tsang, “Integrated polarization beam splitter in high index contrast silicon-on-insulator waveguides,” IEEE Photonics Tech. Lett. 17(2), 393–395 (2005). [CrossRef]
