⇦Chinese Physics Letters, 2016, Vol. 33, No. 9, Article code 094201⇨ Silicon-on-Insulator-Based Broadband 1$\times$3 Adiabatic Splitter with Simultaneous Tapering of Velocity and Coupling * Yuan-Hao Gong(巩源浩), Zhi-Yong Li(李智勇), Jin-Zhong Yu(余金中), Yu-De Yu(俞育德)** Affiliations State Key Laboratory on Integrated Optoelectronics, Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083 Received 6 May 2016 ^*Supported by the National High-Technology Research and Development Program of China under Grant Nos 2015AA016904 and 2013AA014402, the National Basic Research Program of China under Grant No 2011CB301701, and the National Natural Science Foundation of China under Grant No 61275065.
^**Corresponding author. Email: yudeyu@semi.ac.cn Citation Text: Gong Y H, Li Z Y, Yu J Z and Yu Y D 2016 Chin. Phys. Lett. 33 094201 Abstract We propose and experimentally demonstrate a broadband $1\times3$ adiabatic splitter based on the silicon-on-insulator technology, with simultaneous tapering of velocity and coupling. The designed structure becomes simulated transmission uniformity of three outputs better than 0.5 dB in a broadband of 250 nm, and a large simulated fabrication tolerance is obtained. A manufactured splitter whose parameters greatly diverge from the design acquires a measured result of the worst splitting ratio better than 1.5 dB as well as an excess loss lower than 0.8 dB in a large wavelength range of 80 nm. A post-simulation based on the tested splitter obtains a result that meets the actual transmission well. DOI:10.1088/0256-307X/33/9/094201 PACS:42.82.Cr, 42.82.Ds, 42.82.Bq © 2016 Chinese Physics Society Article Text Silicon photonics is acquiring increasing attention and is considered as a promising solution for future high-speed optical interconnects on account of its high index contrast and compatibility with commercial CMOS processes.[1] Optical beam splitters are one kind of fundamental components used in many optical devices and integrated optics based on the silicon-on-insulator (SOI) technology. A high performance beam splitter should be uniform, broadband and low loss at the same time. There are only a few methods that can be used to realize beam splitting, such as self-imaging phenomena,[2] photonic crystals,[3-6] multimode interference.[7,8] Several $1\times3$ beam splitters have been demonstrated.[9-11] Beam splitters using the traditional waveguide directional coupler can hardly overcome their sensitivity to wavelength and fabrication errors. Multimode interference coupler has relatively large bandwidth and relaxed fabrication tolerances, which makes a good selection for beam splitters. However, mode interference is still wavelength dependent and will cause additional increase of the optical insertion loss when using waveguides with high index contrast. An SOI-based $2\times2$ adiabatic splitter proposed before has proved the advantages of adiabatic structure with simultaneous tapering of velocity and coupling by demonstrating its large bandwidth, low loss and good uniformity.[12] In this Letter, to discover further applications of the adiabatic structure, we propose and experimentally demonstrate a broadband $1\times3$ adiabatic beam splitter with simultaneous tapering of velocity and coupling. Figure 1 shows the schematic drawing of the proposed $1\times3$ adiabatic splitter, in which the green parts represent the waveguides while the remaining part is the etching region. The splitter consists of three SOI-based waveguides, waveguide 1 (W1), waveguide 2 (W2) and waveguide 3 (W3), placed close to each other in the same plane, and W2 is the input waveguide. The gap between W1 and W2 is gap 1 (G1), and between W2 and W3 is gap 2 (G2), $L$ shows the length of adiabatic region, and $G$ is the minimum width of G1 and G2. The left side of W2 is the input port I, and O1, O2 and O3 are three output ports. According to the coupled-mode theory,[13,14] W1, W2 and G1 can be approximatively regarded as one coupled-waveguide system, with W2, W3 and G2 as the other. For each system, the coupling strength between two waveguides is determined by the coupling coefficient and propagation constant, which depends on the gap width and waveguide parameters. The working principle of the $1\times3$ splitter is based on adiabatic evolution of the odd mode from being mainly located in W2 to being divided into three waveguides. To be adiabatic, the waveguide effective index of the odd mode should be varied very slowly to avoid exciting other modes, which means that the gap widths and the waveguide widths should be changed slowly enough. In the following simulations utilizing a 3D beam propagation method (BPM), the grid sizes for $x$, $y$ and $z$ are 20 nm, 10 nm and 20 nm, respectively. The transparent boundary condition is used. The splitter, designed for the wavelength of 1.55 μm and simulated under TE-polarized light, is comprised of three SOI-based single-mode rib waveguides. The rib height is 340 nm, and the etching depth is 200 nm. The width of the waveguide is linearly tapered from 600 nm to 480 nm for W1 and W3, and 400 nm to 500 nm for W2. The width of G2 is linearly tapered from 1100 nm to 100 nm ($G=100$ nm), the same as G1. The length of adiabatic region ($L$) is 300 μm, and the simulated transmission has been normalized.

Fig. 1. The schematic drawing of the proposed $1\times3$ adiabatic splitter.

Fig. 2. The simulated normalized transmission of the splitter as a function of the wavelength.

Fig. 3. The simulated normalized transmission of the splitter as a function of the adiabatic region length $L$.

Fig. 4. The simulated normalized transmission of the splitter as a function of the narrowest gap $G$.

To highlight the advantages of the proposed $1\times3$ adiabatic splitter, we simulate the designed structure with parameters in keeping with the ones mentioned before and display the simulated normalized transmission curves of the splitter as a function of wavelength in Fig. 2. The wavelength coverage is from 1450 nm to 1700 nm. Here I–O1 in the figure represents light transmitting from I to O1, I–O2 and I–O3 can be understood in the same manner. The green line (I–O1) and the black line (I–O3) are nearly overlapped due to the complete symmetry of W1 and W3. As shown in Fig. 2, the uniformity of the three output ports is better than 0.5 dB (about 4%) over a broadband wavelength range of 250 nm (from 1450 nm to 1700 nm). The splitting effects are extremely close to $-$4.77 dB (the transmission value when light is trisected) around wavelength 1550 nm. The critical points of the proposed splitter are adiabatic evolution and sufficient coupling. Adiabatic evolution means that the adiabatic region length $L$ should be long enough, and sufficient coupling needs an appropriate $G$. To illustrate the significance of $L$ and $G$, we introduce single variables $L$ and $G$ into the proposed splitter structure separately and show the simulated normalized transmissions in Figs. 3 and 4, respectively. As we can see in Fig. 3, the variation range of $L$ is from 100 μm to 500 μm, and the transmission curves are tending towards stability when $L$ is longer than 275 μm, which means that when $G$ is 100 nm, a 275-μm-long adiabatic region is ample to realize adiabatic evolution. In Fig. 4, we make $G$ vary from 25 nm to 275 nm, and the curves distinctly indicate that 100 nm is an adequate $G$ value for sufficient coupling. The large fringes when $G$ is large are owing to the unsimultaneous tapering of velocity and coupling and the mild imbalance when $G$ is small indicates the slightly insufficient coupling. Moreover, in Fig. 4, the proposed splitter maintains a high uniformity better than 0. 5 dB (about 4%) when $G$ is altered from 50 nm to 125 nm, yielding a large fabrication tolerance.

Fig. 5. The simulated normalized transmission of the splitter as a function of the width deviation.

The preceding simulations have revealed the broadband characteristic and large fabrication tolerance of the proposed $1\times3$ adiabatic splitter, in the following we will further study the fabrication tolerance to width deviation (Fig. 5) and etching depth offset (Fig. 6) of the three waveguides composing the splitter. As shown in Figs. 5 and 6, the uniformity of the proposed splitter is better than 0.5 dB (about 4%) when the width deviation changes from $-$20 nm to 5 nm and when the etching depth offset varies from $-$25 nm to 20 nm. When the width deviation is between 5 nm to 20 nm, the splitter can still acquire a splitting ratio better than 0.6 dB (about 5%).

Fig. 6. The simulated normalized transmission of the splitter as a function of the etching depth deviation.

Fig. 7. The microscope and scanning electron microscope (SEM) views of the measured splitter.

Fig. 8. The normalized transmission of the measured splitter as a function of the wavelength.

The proposed $1\times3$ adiabatic splitter is fabricated on an SOI wafer with a 340-nm-thick top silicon film separated by a 2-μm-thick SiO$_2$ layer from the silicon substrate. Electron-beam lithography (EBL) is used to define the structures, and inductively coupled plasma (ICP) etching is then used to etch 200 nm on the top silicon layer. Figure 7 presents the microscope and scanning electron microscope (SEM) views of the tested splitter. A laser that is tunable from 1540 nm to 1620 nm is used in the measurement. A TE-polarized light beam is coupled from a single-mode fiber to the input waveguide (W2) by a high-efficiency grating coupler designed for wavelength 1550 nm. Optical power from each output waveguide is then measured. The test results are exhibited in Figs. 8 and 9.

Fig. 9. The excess loss of the measured splitter as a function of the wavelength.

Fig. 10. The normalized transmission of the post simulation as a function of the wavelength.

As described by the parameters of waveguides and gaps shown in Fig. 7, the tested structure is not conformed to the design very well, but the results are still acceptable with a worst splitting ratio of about 1.5 dB (about 10%) as demonstrated in Fig. 8 between the wavelengths 1540 nm and 1620 nm. The measured excess loss (EL) is lower than 0.8 dB as shown in Fig. 9. A post simulation using the parameters shown in Fig. 7 is implemented and the simulated transmissions are demonstrated in Fig. 10, and the comparison of Figs. 10 and 8 indicates that the actual results meet the simulation well. In conclusion, an SOI-based $1\times3$ adiabatic splitter, with simultaneous tapering of phase velocity and coupling, has been proposed and experimentally demonstrated. The splitter shows high uniformity, low loss and large fabrication tolerance for a wide wavelength range in both simulations and experiments. In addition, we believe that this splitter, with preferable performance and acceptable dimensions, will have a promising future in SOI-based integrated optics. References The Past, Present, and Future of Silicon PhotonicsCompact beam splitters based on self-imaging phenomena in one-dimensional photonic crystal waveguidesDesigning analysis of the polarization beam splitter in two communication bands based on a gold-filled dual-core photonic crystal fiberPolarization Beam Splitter Based on a Self-Collimation Michelson Interferometer in a Silicon Photonic CrystalDesign of a Novel Polarized Beam Splitter Based on a Two-Dimensional Photonic Crystal Resonator Cavity1×N Multimode Interference Beam Splitter Design Techniques for On-Chip Optical InterconnectionsExperimental Demonstration of Two-Dimensional Multimode-Interference Optical Power Splitter1×3 Beam splitter based on self-collimation effect in two-dimensional photonic crystals1x3 beam splitter for TE polarization based on self-imaging phenomena in photonic crystal waveguidesSilicon-on-insulator-based adiabatic splitter with simultaneous tapering of velocity and couplingCoupled-mode theory for guided-wave opticsMultimode add-drop multiplexing by adiabatic linearly tapered coupling

[1]	Soref R 2006 IEEE J. Sel. Top. Quantum Electron. 12 1678
[2]	Chen B, Huang L, Li Y D, Liu C L and Liu G Z 2012 Chin. Opt. Lett. 10 111301
[3]	Zhao Y Z, Zhang Y and Li B J 2005 Chin. Opt. Lett. 3 S196
[4]	Fan Z K, Li S G, Fan Y Q, Zhang W, An G W and Bao Y J 2014 Chin. Phys. B 23 094212
[5]	Chen X Y, Qiang Z X, Zhao D Y, Wang Y F, Li H, Qiu Y S and Zhou W D 2012 Chin. Phys. Lett. 29 014210
[6]	Zhang X, Chen S Y, Liao Q H, Tianbao Y U, Liu N H and Huang Y Z 2011 Chin. Phys. Lett. 28 084201
[7]	Hosseini A, Kwong D N, Zhang Y, Subbaraman H, Xu X C and Chen R T 2011 IEEE J. Sel. Top. Quantum Electron. 17 510
[8]	Sun Y L, Jiang X Q, Yang J Y, Tang Y and Wang M H 2003 Chin. Phys. Lett. 20 2182
[9]	Wan Y, Yun M J, Xia L H and Zhao X S 2011 Optik 122 337
[10]	Zhang M, Malureanu R, Krüger A C and Kristensen M 2010 Opt. Express 18 14944
[11]	Yang J B, Xu S Z, Zhou K and Xu J 2014 Chin. Opt. Lett. 12 310501
[12]	Xing J J, Xiong K, Xu H, Li Z Y, Xiao X, Yu J Z and Yu Y D 2013 Opt. Lett. 38 2221
[13]	Yariv A 1973 IEEE J. Quantum Electron. 9 919
[14]	Greenberg M and Orenstein M 2005 Opt. Express 13 9381