LASER &PHOTONICS REPRINT
Laser Photonics Rev., L1 L5 (2014) / DOI 10.1002/lpor.201300157 LASER & PHOTONICS Abstract An 8-channel hybrid (de)multiplexer to simultaneously achieve mode- and polarization-division-(de)multiplexing is proposed and demonstrated experimentally on a siliconon-insulator platform to improve the link capacity of an onchip optical interconnect. The present hybrid (de)multiplexer has four channels for each polarization. A polarization beam splitter based on a three-waveguide coupler is used to combine/separate the fundamental modes of TE- and TMpolarizations (TE 0 and TM 0 ). Six asymmetric directional couplers are cascaded for (de)multiplexing the high-order modes (TE 1,TE 2,TE 3,TM 1,TM 2, and TM 3 ). The experimental results show all eight channels have low loss and low crosstalk (< 10 db) over a 30 nm wavelength range. LETTER ARTICLE On-chip silicon 8-channel hybrid (de)multiplexer enabling simultaneous mode- and polarization-division-multiplexing Jian Wang, Sailing He, and Daoxin Dai As it is well known, wavelength-division-multiplexing (WDM) provides a promising way to expand the link capacity with N wavelengths and has been used very widely [1]. In order to improve the link capacity further to satisfy the increasing bandwidth demand for optical communication, great efforts have been made to enhance the capacity for single wavelength carrier by utilizing various technologies, e.g., multi-core space-division multiplexing (SDM) [2], multimode SDM [3], polarization-division multiplexing (PDM) [4]. Among them, the multimode SDM has become a focus not only for optical fiber communications [2], but also for on-chip optical interconnects [5, 6] and one of the most important components is a mode (de)multiplexer, which is used to separate/combine the channels carried by different eigenmodes. Some structures, like multi-mode interferometers [7], asymmetric Y-junctions [8], and adiabatic couplers [9] with good performances, have been proposed to realize two-channel mode multiplexers. However, these structures are not easy to extend for working with more mode channels. Asymmetrical directional couplers (ADCs) provide another good option to realize mode (de)multiplexer [10 12] because of the simple structure, the scalability and the flexibility to excite any desired highorder mode from a launched fundamental mode. We have proposed and demonstrated a compact 4-channel silicon mode multiplexer operating for TM polarization with low loss and low crosstalk previously [11]. Recently a gratingassist ADC was designed to realize 4-channel narrow-band mode multiplexer [13]. In Ref. [14], a three-channel multiplexer combining ADCs and microrings for simultaneous mode- and wavelength-division-multiplexing has also been recently realized to expand the link capacity. In this paper, we propose and demonstrate an 8- channel hybrid (de)multiplexer, which simultaneously enables mode- and polarization-division-(de)multiplexing. As silicon photonics is very appealing regarding the CMOS compatibility, we choose silicon-on-insulator (SOI) nanowires with a 220 nm-thick top silicon layer to realize the present 8-channel hybrid (de)multiplexer, which has four channels for TE- and TM-polarization modes respectively, as shown in Fig. 1(a). A polarization beam splitter (PBS) is used to combine/separate the fundamental mode of TE- and TM-polarization (TE 0 and TM 0 ) while the highorder modes (TE 1,TE 2,TE 3,TM 1,TM 2, and TM 3 )are (de)multiplexed by using six cascaded ADCs. The PBS is achieved by using a three-waveguide coupler consisting of two narrow waveguides (w 1 ) and a wide waveguide (w 2 ) [15] (see Fig. 1(b)). According to the phase matching condition [15], here we choose w 1 = 0.4 μm, and w 2 = 1.035 μm as an example. The simulated light propagation in the designed PBS is given in Fig. 2(a) and 2(e). It can be seen that the TE 0 mode launched from port I 4 arrives at the bus waveguide without any change almost while the TM 0 mode launched from port I 5 is cross-coupled to the TM 0 mode of the bus waveguide with the assistance of the TM 1 mode in the wide middle waveguide. The ADCs for six higher-order modes are cascaded and the bus waveguide consists of seven straight bus sections of differing widths (w bi ), with an adiabatic taper at a1.8 angle between the two adjacent straight bus sections, as shown in Fig. 1(a). The width w bi for the i- th section of the bus waveguide is determined according to the phase-matching condition n eff_m (w bi ) = n eff_0 (w ai ), where n eff_m (w bi ) and n eff_0 (w ai ) are the effective indices of Centre for Optical and Electromagnetic Research, State Key Laboratory for Modern Optical Instrumentation, Zhejiang Provincial Key Laboratory for Sensing Technologies, Zhejiang University, Zijingang Campus, Hangzhou 310058, China Corresponding author: e-mail: dxdai@zju.edu.cn
LASER & PHOTONICS L2 J. Wang, S. He, and D. Dai: On-chip Si 8-channel hybrid (de)multiplexer for mode-/polarization-division-multiplexing Figure 1 (a) The schematic configuration of the present 8-channel hybrid multiplexer. (b) The three-waveguide PBS. the desired m-th higher-order mode of the i-th straight bus section (w bi ) and the fundamental mode of the access waveguide (w ai ), respectively. As a result, the fundamental mode of the access waveguide could be cross-coupled to the desired higher-order mode in the bus waveguide completely when the length of the coupling region is chosen optimally. In this way one can design a mode multiplexer by cascading ADCs conveniently as shown in Ref. [11]. For example, we consider the ADCs working for the TM modes (TM 1,TM 2, and TM 3 ) first. The same width (w ai = 0.4 μm) is chosen for the three access waveguides to make the design convenient. The corresponding widths for the straight bus sections of the ADCs working for the TM 1,TM 2 and TM 3 modes are chosen as w bi = 1.035 μm, 1.695 μm and 2.363 μm, respectively, according to the phase-matching condition. When the gap width between the access waveguide and the bus waveguide is w gap = 0.3 μm, the lengths for the coupling regions are determined as 4.7 μm, 6.7 μm and 9.0 μm, respectively, according to 3D finite-difference time-domain (FDTD) simulations. Figure 2(b) (d) show the simulated light propagation in the designed three ADCs for the TM 1,TM 2, and TM 3 modes, respectively. From these figures, it can be seen that the launched TM 0 mode from the access waveguide is efficiently coupled with low loss to the higher-order mode (TM 1,TM 2,orTM 3 ) as desired. In principle the ADCs for TE polarization modes (TE 1,TE 2 and TE 3 ) can be designed similarly. For example, one has the widths of the straight bus sections for the ADCs working for the TE 1,TE 2 and TE 3 modes as w bi = 0.835 μm, 1.272 μm and 1.712 μm, respectively, when choosing the width of the access waveguide as w ai = 0.4 μm. However, the present hybrid multiplexer has to deal with two polarizations with multiple eigenmodes, and the situation becomes more complicated than that for the singlepolarization operation. The widths of the access waveguides and the straight bus section should be chosen carefully so that only the desired higher-order mode at the multimode waveguide is excited. Meanwhile, the phase mismatching should be maximized between the fundamental mode in the access waveguide and any unwanted high-order mode in the bus waveguide to minimize the crosstalk. Fortunately the designed ADCs for the TM 1,TM 2,TM 3 and TE 1 modes have low crosstalks. Light propagations in these four ADCs simulated by a 3D-FDTD method are shown in Fig. 2(b) 2(d) and 2(f), respectively. However, we note that the ADCs working for the TE 2 and TE 3 modes need to be modified in order to improve the performance. For the ADC for the TE 2 mode designed above (with w ai = 0.4 μm and w bi = 1.272 μm), the phase mismatch between the TM 1 mode in the bus waveguide and the TM 0 mode in the access waveguide is not large. Consequently the TM 1 mode in the straight bus section is coupled partially to the TM 0 mode in the access waveguide, which causes an excess loss for the TM 1 -mode channel and some crosstalk from the TM 1 -mode channel to the TE 2 -mode channel when mode demultiplexing. In order to solve this problem, the width of the access waveguide is modified optimally to be w ai = 0.406 μm so that one can choose the length L c of the coupling region as L c = L π_te2 = 4L π_tm1, in which L π_te2 and L π_tm1 are the beat lengths for the TE 2 mode and the TM 1 mode, respectively. As a result, the TM 1 mode in the straight bus section will partially couple to the access waveguide first and couple back to the straight bus section Figure 2 Light propagation in the coupling regions of the present hybrid multiplexer when the launched TM 0 or TE 0 mode of the access waveguide couples to the TM 0 (a), TM 1 (b), TM 2 (c), TM 3 (d), TE 0 (e), TE 1 (f), TE 2 (g) and TE 3 (h) mode in the bus waveguide. Insets: the electric (magnetic) field for the desired eigen-mode in the bus waveguide.
LETTER ARTICLE Laser Photonics Rev. (2014) L3 Figure 3 When different order modes are launched at the input end of the bus waveguide, (a) the calculated responses at port O 2 (corresponding to the ADC working for the TE 2 -mode channel); Inset: light propagation in the optimized ADC with the TM 1 -mode input from the bus waveguide; (b) The calculated responses at port O 2 when there is a widthvariation w =+/ 10 nm. again, as shown in Fig. 3(a). Meanwhile the TE 2 mode in the bus waveguide will be coupled efficiently to the TE 0 mode of the access waveguide (port O 2 ). Figure 3(a) shows the calculated responses at port O 2 (which is dropped by the ADC working for the TE 2 mode channel) for the cases of choosing w ai = 0.4 μm and w ai = 0.406 μm (the optimized design), when different order modes are launched at the input end of the bus waveguide. It can be seen that the crosstalk from the TM 1 mode has been reduced significantly with the optimal design (w ai = 0.406 μm). Figure 3(b) shows the calculated responses at port O 2 when the optimized ADC with a width-variation w = +/ 10 nm. It can be seen that the designed ADC can tolerate a width-variation w =±10 nm for < 10 db crosstalk and <2 db excess loss. For the ADC designed for the TE 3 mode (w ai = 0.4 μm and w bi = 1.712 μm), the phase mismatch is not large between the TM 0 mode in the access waveguide and the TM 1 as well as TM 2 modes in the bus waveguide. The access waveguide width is then modified to be w a = 0.379 μm so that the crosstalk from the TM 1 -mode channel to the TE 3 -mode channel can be depressed to < 20 db and the crosstalk from the TE 3 -mode channel to the TM 2 -mode channel is estimated to be about 11 db when mode demultiplexing. The excess loss of the TE 3 -mode channel is about 0.4 db around λ = 1550 nm. Further numerical simulation shows that the present hybrid demultiplexer can tolerate a width-variation w =±5 10 nm for < 10 db crosstalk and <2 db excess loss. The optimal widths for all the waveguides of the designed ADCs are summarized as shown in Table 1. The fabrication includes an E-beam lithography step, an inductively-coupled plasma etching process, and a deposition process for 1 μm-thick SiO 2 upper-cladding layer via the plasma enhanced chemical vapor deposition tech- Table 1 The parameters of the designed ADCs. (Unit: μm) TM 1 TM 2 TM 3 TE 1 TE 2 TE 3 w ai 0.4 0.4 0.4 0.4 0.406 0.379 w bi 1.035 1.695 2.363 0.835 1.290 1.631 w gap 0.3 0.3 0.3 0.2 0.2 0.2 L c 4.7 6.7 9.0 15.5 21.3 17.6 nology. Figure 4(a) shows the fabricated photonic integrated circuit (PIC), which includes an 8-channel hybrid multiplexer (with input ports I 1 I 8 ), a 2.363 μmwide and 100 μm-long multimode bus waveguide, and a hybrid demultiplexer (with output ports O 1 O 8 ). Figure 4(b) shows the SEM images of the coupling regions for the PBS and the six cascaded ADCs of the hybrid (de)multiplexer. In order to characterize the present hybrid (de)multiplexer, the end-fire coupling method with lensd fibers was used for coupling light to/from the chip. A tunable laser (Agilent 81940A) and a power meter (Agilent 81635A) were used to measure the output power from ports O 1 O 8. For the measurement, the TE 0 mode of the access waveguide is launched at ports I 1 I 4 and dominantly couples to the TE 3,TE 2,TE 1,TE 0 mode channels in the bus waveguide respectively, while the TM 0 mode is launched at ports I 5 I 8 and dominantly couples to the TM 0,TM 1,TM 2,TM 3 mode channels in the bus waveguide respectively. The polarization of the input light is aligned by rotating the polarization maintaining fiber at the input side, and 20 db extinction-ratio is achieved at the central wavelength. Figure 4(c) (j) show the measured transmission responses at all output ports (O 1 O 8 ) when light is launched at each input port (I 1 I 8 ) respectively. The transmissions of the straight waveguide (without MUXer/deMUXer) fabricated on the same chip are also shown by the dark thin curves in Fig. 4(c) and 4(g) for TM and TE polarizations, respectively, for comparison. Particularly, for TE-polarization modes, some notable ripples are observed, which are mainly caused by the Fabry-Perot effect due to the high reflection at the chip-facets. The insertion loss of a straight waveguide for TM- and TE-polarizations is about 10 db and 20 db, respectively, which is mainly from the fiber-chip-fiber coupling loss and is consistent with the theoretical estimation. The scattering losses (due to the roughness at the sidewall) for all the guided modes of TE and TM polarizations in the multimode bus waveguide should be low since their field amplitude is low at the sidewall [16]. With the transmissions of a straight waveguide as the reference, the present hybrid (de)multiplexer has an excess loss (around 1555 nm) of about 0.7 db, 0.2 db, 0.3 db, 0.7 db, 0.2 db, 2dB,1.5dBand2dBfortheTM 0,TM 1,TM 2,TM 3,TE 0, TE 1,TE 2 and TE 3 -mode channels respectively. The excess loss is mainly caused by the scattering loss and incomplete
LASER & PHOTONICS L4 J. Wang, S. He, and D. Dai: On-chip Si 8-channel hybrid (de)multiplexer for mode-/polarization-division-multiplexing Figure 4 (a) The optical image of the fabricated silicon PIC including a hybrid multiplexer, the multimode bus waveguide and a hybrid demultiplexer. (b) SEM images of the coupling regions for the PBS and the six cascaded ADCs. The transmission responses at the eight output ports (O1 O8 ) when the TM0 or TE0 mode is launched at port (c) I5, (d) I6, (e) I7, (f) I8, (g) I4, (h) I3, (i) I2 and (j) I1, respectively and dominantly couples to the TM0, TM1, TM2, TM3, and TE0, TE1, TE2, TE3 mode channels in the bus waveguide respectively. The transmission responses of the straight waveguide without MUXer/deMUXer for TM and TE polarization modes are also shown by the black curves in (c) and (g). mode-coupling in the ADCs due to the fabrication deviations. The fabrication tolerance can be improved by utilizing tapered ADCs [12]. From Fig. 4(c) (j), it can also be seen that the crosstalk is relatively low ( 10 15 db) over a 30 nm wavelength range. Here the crosstalk is estimated by the difference between the powers from the major output port (Oi ) and any other output port (Oj, j = i) for a fixed input port (Ij ). Due to the reciprocity, one can also read from Fig. 4(c) (j) the crosstalk defined by the difference between the powers at a fixed output port (Oi ) when light is launched from the major input port (Ii ) and another input port (Ij, j = i). The crosstalks of the channels for the TM0, TM1, TM2 and TM3 modes are respectively about 19 db, 20 db, 20 db and 17.5 db around 1555 nm. In contrast, the channels for the TE0, TE1, TE2 and TE3 modes have relatively large crosstalks (i.e., 16 db, 12 db, 14 db and 11 db, respectively), which partially results from some polarization crosstalk due to the polarization variation when the laser wavelength is tuned. Better measurement results are expected when the polarization extinction-ratio of the input light is improved. Ac- C 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim cording to the realization of error-free transmission (bit error rate <10 12 ) using a microring multiplexer with a crosstalk of 13 db in Ref. [14], the present hybrid (de)multiplexer should work well. It also has the potential to work with WDM filters for realizing a multi-dimensional hybrid multiplexing technology to further improve the link capacity. In conclusion, we have proposed and demonstrated a novel silicon hybrid (de)multiplexer to combine the multimode SDM and PDM technologies, which enables high link capacity significantly with a single-wavelength light carrier. The present 8-channel hybrid multiplexer has four channels (de)multiplexed for each polarization. A PBS has been used to combine/separate the TE0 and TM0 modes while the high-order modes (TE1, TE2, TE3, TM1, TM2, and TM3 ) are (de)multiplexed by using six cascaded ADCs. Our experimental results have shown that the hybrid multiplexer has relatively low loss and low crosstalk ( 10 15 db) over a 30 nm wavelength range, which makes it potential to work with WDM filters for realizing a multi-dimensional hybrid multiplexing technology for improving the link capacity further.
LETTER ARTICLE Laser Photonics Rev. (2014) L5 Acknowledgments. This project was supported by a 863 project (2011AA010301), Nature Science Foundation of China (11374263), Zhejiang provincial grant (2011C11024), the Doctoral Fund of Ministry of Education of China (20120101110094). Supporting information for this article is available free of charge under http://dx.doi.org/10.1002/lpor.201300157 Received: 1 October 2013, Revised: 18 January 2014, Accepted: 21 January 2014 Published online: 5 February 2014 Key words: hybrid, mode multiplexing, polarization, silicon. References [1] M. J. Paniccia, Optik & hotonik 6(2), 34 38 (2011). [2] D. J. Richardson, J. M. Fini, and L. E. Nelson, Nat. Photon. 7, 354 362 (2013). [3] Y. Yadin and Meir Orenstein, J. Lightwave Technol. 24(1), 380 386 (2006). [4] C. Doerr and T. Taunay, IEEE Photon. Tech. Lett. 23, 597 599 (2011). [5] M. Greenberg, M. Orenstein, Quantum Electron. Laser Sci. Conf. (QELS 05) 2, 942 944 (2005). [6] S. Bagheri, W. M. J. Green, Proc. 6th IEEE Int. Conf. Group IV Photon. 166 168 (2009). [7] T. Uematsu, Y. Ishizaka, Y. Kawaguchi, K. Saitoh, and M. Koshiba, J. Lightwave Technol. 30(15), 2421 2426 (2012). [8] J. Driscoll, R. Grote, B. Souhan, J. Dadap, M. Lu, and R. Osgood, Opt. Lett. 38, 1854 1856 (2013). [9] J. Xing, Z. Li, X. Xiao, J. Yu, and Y. Yu, Opt. Lett. 38, 3468 3470 (2013). [10] M. Greenberg and M. Orenstein, Opt. Express 13, 9381 9387 (2005). [11] D. Dai, J. Wang, and Y. Shi, Opt. Lett. 38, 1422 1424 (2013). [12] Y. Ding, J. Xu, F. Da Ros, B. Huang, H. Ou, and C. Peucheret, Opt. Express 21, 10376 10382 (2013). [13] H. Qiu, H. Yu, T. Hu, G. Jiang, H. Shao, P. Yu, J. Yang, and X. Jiang, Opt. Express 21, 17904 17911 (2013). [14] L. Luo, N. Ophir, C. Chen, L. H. Gabrielli, C. B. Poitras, K. Bergman, and M. Lipson, arxiv:1306.2378 (2013). [15] D. Dai, J. Lightwave Technol. 30(20), 3281 3287 (2012). [16] F. Payne and J. Lacey, Opt. Quantum Electron. 26, 977 986(1994).