Scalable Multimode Waveguide Bend Based on Tapered Couplers

To satisfy the ever-increasing bandwidth demands, silicon-based optical interconnects can offer a promising solution, thanks to the advantages of broad bandwidth, high speed, and excellent CMOS compatibility. Various advanced multiplexing technologies have been employed to further improve the transmission capacity, such as space division multiplexing, polarization division multiplexing, and wavelength division multiplexing. Among them, mode division multiplexing utilizing each mode to serve as an independent data channel has been drawing much attention. To construct the densely integrated mode division multiplexing system, multimode waveguide bends are one of the important functional components. Up to now, a number of structures, such as photonic crystal, multimode interference, and subwavelength grating, have been used to realize different multimode waveguide bends. Although these multimode waveguide bends can have good performance, they only support 2 to 4 modes and are not easy to expand. Thus, scalable multimode waveguide bends, which can support more modes, are highly desired. In this paper, a scalable multimode waveguide bend based on tapered couplers is presented. As an example, the multimode waveguide bend supporting TE0, TE1, TE2, TE3 and TE4 modes are designed and analyzed in detail. The proposed multimode waveguide bend is a symmetrical structure. It is composed of single-mode bend waveguides and a pair of five-mode multiplexer and demultiplexer. By utilizing the matching of the effective modal indexes, the mutual conversion between the fundamental mode and the high-order modes can be realized in each tapered coupling region, and thus multimode bending transmission is achieved. To improve the performance, the particle swarm optimization algorithm and finite difference time domain method are used to optimize the structural parameters of the tapered coupling region. Simulation results show that, for the designed multimode waveguide bend, within a bandwidth from 1 500 nm to 1 600 nm, when the TE0 mode is input, the insertion loss is smaller than 0.087 dB, and the crosstalk is less than -20.72 dB. As the TE1 mode is launched into the input port, the corresponding insertion loss and crosstalk are respectively lower than 0.27 dB and - 19.81 dB. The corresponding insertion loss and crosstalk are less than 0.40 dB and - 19.60 dB when the TE2 mode is injected into the input port. As the TE3 mode is input, the corresponding insertion loss and crosstalk are lower than 0.78 dB and - 20.34 dB. When the TE4 mode is launched into the input port, the corresponding insertion loss and crosstalk are smaller than 0.75 dB and - 19.03 dB. In order to verify the feasibility of our design, the optimized multimode waveguide bend is experimentally demonstrated on a silicon-on-insulator platform by using a commercial complementary metal-oxide-semiconductor compatible process. The experimental results reveal that, for the fabricated device, within a bandwidth from 1 520 nm to 1 600 nm, the insertion loss is lower than 1.71 dB, and the crosstalk is less than - 10.60 dB when the TE0 mode is injected into the input port. When the TE1 mode is launched into the input port, the insertion loss and crosstalk are respectively lower than 3.04 dB and - 11.35 dB. When the TE2 mode is input, the insertion loss is smaller than 2.9 dB, and the crosstalk is less than - 10.92 dB. The corresponding insertion loss and crosstalk are smaller than 3.16 dB and - 10.35 dB when the TE3 mode is injected into the input port. When the TE4 mode is input, the insertion loss is lower than 4.00 dB and the crosstalk is less than - 11.45 dB. The main reasons for the inconsistency between the experimental results and the simulation ones are as follows:firstly, the scattering loss can be induced by the sidewall roughness of the fabricated waveguide. Secondly, due to fabrication errors, the actual width of the fabricated waveguide could deviate from the optimal value, so that the insertion loss and crosstalk of the fabricated device would become worse. Thirdly, owing to the bandwidth limitation of the adopted light source, the measured bandwidth is not as wide as the simulated one. In future work, high-quality fabrication processes are required to further improve the crosstalk and insertion loss. To sum up, with the above characteristics, our presented scalable multimode waveguide bend based on tapered couplers can offer an attractive option for constructing the densely integrated mode division multiplexing system to improve the transmission capacity.