Two-dimensional grating coupler with a lowpolarization dependent loss of 0.25 dBcovering the C-bandJINGHUI ZOU, YU YU, AND XINLIANG ZHANG*Wuhan National Laboratory for Optoelectronics and School of Optical and Electronic Information, Huazhong University of Science andTechnology, Wuhan 430074, China*Corresponding author: xlzhang@mail.hust.edu.cn

Received 28 July 2016; revised 20 August 2016; accepted 23 August 2016; posted 24 August 2016 (Doc. ID 272446); published 8 September 2016

We design and demonstrate a two-dimensional gratingcoupler (2D GC) with a low polarization dependent loss(PDL) based on the silicon-on-insulator (SOI) platform.Using a grating cell consisting of five cylinders andcarefully optimizing the distances between the cylinders,a maximum PDL of 0.25 dB covering the C-band is real-ized, which is 1.25 dB better than a conventional 2D GCwith a single cylinder etching pattern fabricated on thesame SOI wafer. © 2016 Optical Society of America

OCIS codes: (130.3120) Integrated optics devices; (230.1950)

Diffraction gratings; (230.5440) Polarization-selective devices.

http://dx.doi.org/10.1364/OL.41.004206

Silicon-on-insulator (SOI) is becoming an attractive platformfor photonic integrated circuits (PICs) due to the complemen-tary metal oxide semiconductor compatibility and high refrac-tive index contrast [1,2]. Unfortunately, the subwavelengthwaveguide will cause a serious mode size mismatch with a singlemode fiber (SMF), making it challenging to realize highly ef-ficient fiber–chip coupling. Inverse taper and grating couplers(GCs) have been verified as two feasible approaches to solve thisproblem [3–6]. Compared to inverse taper couplers, GCs havea few advantages in terms of easy fabrication, flexible position,compact footprint, large alignment tolerance, and wafer-scaletest. However, the conventional one-dimensional (1D) GC hasan inherent defect of strong polarization sensitivity due to theeffective refractive index difference between transverse electric(TE) and transverse magnetic (TM) modes, making it helplessfor an input light with polarization state unknown. Thougha batch of polarization insensitive or polarization splitting1D GCs have been proposed and demonstrated [7–11], thesubsequent polarization beamsplitter and polarization rotatorare unavoidable considering that most of the silicon PICsare designed for TE mode, and it will not only complicate butalso enlarge the PICs. To overcome this dilemma, the 2D GChas been presented [12] and successfully applied to siliconintegrated coherent receivers [13,14].

A 2D GC can be briefly considered as a superposition oftwo orthogonal standard 1D GCs, and it can couple arbitrarilypolarized light into the PICs in TE mode with constantcoupling efficiency. A multitude of modified 2D GCs havebeen put forward to suppress the backreflection [15–17], re-duce the footprint [18], improve the coupling efficiency [19],and simplify the fabrication processes [20]. However, in allthese schemes, the tilted coupling schemes were used, whichwill make the 2D GC polarization dependent in some degree.This is because an input light with electric field parallel tothe symmetry axis (P-polarized) is slightly tilted out of the2D GC plane, while an input light with electric fieldperpendicular to the symmetry axis (S-polarized) lies in thesame plane as the 2D GC, inducing a coupling spectrum shift[21]. An active scheme based on a phase shifter can reduce thepolarization dependent loss (PDL) to a certain extent [21].Nevertheless, the bandwidth, structure complication, andextra power consumption limit its application. Another 2D GCwith rounded diamond-like etching pattern is also proposedto eliminate the PDL, but no descriptions on both of theprinciples and structures are specified [6].

In this Letter, we propose a 2D GC with an etching patternconsisting of five cylinders to lower the PDL. By adjustingthe distances between the cylinders, the coupling spectrum shiftis eliminated. For demonstration, the proposed 2D GC ismanufactured and a maximum PDL of 0.25 dB is measuredat the C-band, which is 1.25 dB better than a conventional2D GC with single cylinder etching shape.

The top view of the single grating cell for the proposed 2DGC is illustrated in Fig. 1. The etching part is formed by fivecylinders with same radius of R. One of them is centered in thegrating cell, and the other four are distributed along the P- andS-axes with distances of dp and d s, respectively. As analyzed anddemonstrated in [20], the S-polarized coupling spectrum has aredshift compared to the P-polarized coupling spectrum, whichis the major origin of the PDL. Thus, we can relief the PDL bymoving the S-polarized spectrum back. According to theexpression of the peak wavelength λ of a GC [22]:

λ � �neff − ncladding sin θ�P: (1)

4206 Vol. 41, No. 18 / September 15 2016 / Optics Letters Letter

0146-9592/16/184206-04 Journal © 2016 Optical Society of America

λ depends on the effect refractive index neff of the GC if clad-ding refractive index ncladding, coupling angle θ, and GC pitch Premain constant. So we can blueshift the S-polarized couplingspectrum by lowering its neff , and it can be realized by enlargingthe distance d s between the cylinders lying along the S-axis.

A common SOI wafer with 220 nm top silicon layer and2 μm buried oxide layer is used in our design. To suppressthe backreflection, a coupling angle of θ � 14° is adopted.For fabrication convenience, an etching depth of e � 70 nmis employed, enabling the 2D GC to be manufactured withthe same processes of 1D GC [23]. The cylinder radius isselected to be R � 110 nm, easily fulfilling the boundary ofdeep-ultraviolet lithography. Then, dp, d s, and P are the re-maining parameters to be optimized. Commercial 3D Finite-Difference Time-Domain (FDTD) software from LumericalSolutions, Inc. [24] is utilized for the optimization. A focusingscheme is introduced [18] to compact the footprint, and thelayout is defined as the following equation:�

neffffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�x − L�2 � y2

p�

ffiffi2

p2 sin θ��x − L� � y� � mλ

neffffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix2 � �y − L�2

p�

ffiffi2

p2

sin θ�x � �y − L�� � nλ; (2)

where m and n are both integers, and L is the focusing distance.Each solution (x, y) indicates the center location of agrating cell.

We first fixed dp to 250 nm, and a sweep of d s is performed.As shown in Fig. 2, the peak wavelength for S-polarized inputlight decreases as d s increases, and it moves back to 1550 nmwhen d s increases to 360 nm. The P- and S-polarized couplingspectra are shown in Fig. 3. The PDL defined as the absolutevalue of difference between P- and S-polarized coupling spectrais illustrated as Fig. 4. A maximum PDL of 0.20 dB locates at1547 nm covering a broad span from 1530 to 1570 nm. Forcomparison, a conventional 2D GC with an etching pattern of

a single cylinder is also explored at the same time with an etch-ing depth of 70 nm, cylinder radius of 200 nm, and pitch of612 nm. The calculated coupling spectra are also shown inFig. 3, and a redshift of 11 nm is observed for the S-polarizedcoupling spectrum, causing a large PDL. As shown in Fig. 4,the PDL is dramatically increased beyond 1.25 dB at 1570 nm,which is 1.05 dB larger than the proposed 2D GC. As for thebandwidth, the conventional 2D GC merely can work at anarrow band from 1550 to 1555 nm while the proposed2D GC can work from 1530 to 1570 nm if a PDL boundaryof 0.20 dB is requested. With the same method, 2D GCs withdp � 210, 230, and 270 nm are also explored with the corre-sponding optimum d s of 340, 350, and 370 nm, respectively.The PDLs are illustrated in Fig. 4. Compared with the conven-tional 2D GC, all of the proposed 2D GCs show an over-whelmingly superior performance in PDL, especially for the2D GC with dp � 270 nm. However, we should give consid-eration to another key character of a 2D GC, namely thecoupling efficiency. Although the 2D GC with dp � 270 nmhas a 0.06 dB better PDL than that with dp � 250 nm, it has a0.22 dB larger coupling loss. Taking both PDL and couplingefficiency into consideration, we select the parameter group ofdp � 250 nm and d s � 360 nm in this Letter.

The fabrication tolerance of the proposed 2D GC is alsoexplored. In the fabrication processes, the actual etching depthcannot be completely the same as designed, and the impact ofthe variations of etching depth is calculated as shown in Fig. 5.

Fig. 1. Top view of the single grating cell for the proposed 2D GC.

Fig. 2. Peak wavelength for S-polarized input light as a functionof d s.

Fig. 3. P- and S-polarized coupling spectra of the proposed andconventional 2D GCs.

Fig. 4. PDLs of conventional 2D GC and proposed 2D GCs withdifferent parameters.

Letter Vol. 41, No. 18 / September 15 2016 / Optics Letters 4207

A depth variation of 10 nm will introduce ∼13 nm wavelengthshift while the PDL stays below 0.30 dB.

In order to validate our design, the proposed 2D GC is fab-ricated using electron beam lithography (EBL) and inductivelycoupled plasma (ICP) etching processes. The optical micro-graph of fabricated 2D GCs is shown in Fig. 6(a). And the2D GC is characterized through fiber-to-fiber transmissionmeasurement using a back-to-back configuration [18]. Theinput light from a broadband optical source is first coupledand focused into the subsequent single mode rib waveguidesby the input 2D GC. After transmitting through the connect-ing waveguides, the light is coupled into the output SMF withan identical 2D GC and measured using an optical spectrumanalyzer. The scanning electron microscope (SEM) pictures ofthe proposed 2D GC are illustrated in Fig. 6(b). At the sametime, a focusing conventional 2D CG with 200 nm radiusand 612 nm pitch is fabricated on the same wafer as shownin Fig. 6(c). Besides, TE and TM 1D GCs are also fabricatedto identify the P- and S-polarization states of the input light.

To obtain a P- or S-polarized input light, the polarizationcontroller before the input SMF is turned to minimize thetransmission of the TE or TM 1D GC.

The measured coupling spectra and PDLs of the proposedand conventional 2D GCs are illustrated in Fig. 7. For theproposed 2D GC, the peak wavelengths of P- and S-polarizedcoupling spectra are both located at 1548 nm with losses of5.0 and 5.2 dB. As for the conventional 2D GC, the P- andS-polarized coupling spectra have a wavelength difference of12 nm with peak losses of 4.8 and 4.4 dB. A great ameliorationin PDL can be easily observed as shown in Fig. 7. The proposed2D GC has a PDL under 0.25 dB across a wavelength range of40 nm while the conventional 2D GC has a maximum value of1.50 dB at the same band. The experimental results are wellin agreement with the simulation results shown in Figs. 3and 4 except for a loss degradation, which may be causedby the imperfectly smooth edge induced in the etching process.Though the loss of the proposed 2D GC is 0.2 dB higher thanthe conventional 2D GC, it is at the forefront of 2D GCs thathave been presented based on a 220 nm SOI wafer withoutbottom reflector [12,15,17,18]. The coupling efficiency canbe improved by a further optimization of the structure param-eters including etching depth and cylinder radius, which arenot swept here considering the time-consuming 3D FDTDsimulation and fabrication convenience. Also, a bottom metalreflector can dramatically improve the coupling efficiency atthe expense of fabrication complication [25,26].

In conclusion, a 2D GC with grating cell consisting of fivecylinders is proposed to alleviate the PDL for a tilted couplingscheme. By adjusting the distances between the cylinders, thecoupling spectrum shift of P- and S-polarized input light iseliminated and the PDL is greatly decreased. For demonstra-tion, the proposed 2D GC is fabricated based on a 220 nmSOI wafer using EBL and ICP etching processes. A maximumPDL of 0.25 dB is measured at the C-band, which is 1.25 dBbetter than the conventional 2D GC fabricated on thesame wafer.

Funding. National Basic Research Program of China(2011CB301704); National Natural Science Fund forDistinguished Yong Scholars (61125501); NSFC MajorInternational Joint Research Project (61320106016);

Fig. 5. Impact of the variations of etching depth.

Fig. 6. (a) Optical micrograph of back-to-back cascaded 2D GCs.(b) SEM pictures of proposed 2D GC. (c) SEM pictures of conven-tional 2D GC.

Fig. 7. Measured coupling spectra and PDLs of the proposed andconventional 2D GCs.

4208 Vol. 41, No. 18 / September 15 2016 / Optics Letters Letter

National Natural Science Foundation of China (NSFC)(61275072, 61475050); New Century Excellent TalentProject in Ministry of Education of China (13-0240);Foundation for Innovative Research Groups of the NaturalScience Foundation of Hubei Province (2014CFA004).

Acknowledgment. We thank He Zhang, engineer in theCenter of Micro-Fabrication and Characterization of WuhanNational Laboratory for Optoelectronics, for the support inthe EBL and ICP etching processes.

REFERENCES

1. R. Soref, IEEE J. Sel. Top. Quantum Electron. 12, 1678 (2006).2. T. Baehr-Jones, T. Pinguet, P. L. Guo-Qiang, S. Danziger, D. Prather,

and M. Hochberg, Nat. Photonics 6, 206 (2012).3. G. Roelkens, P. Dumon, W. Bogaerts, D. Van Thourhout, and R.

Baets, IEEE Photon. Technol. Lett. 17, 2613 (2005).4. J. Cardenas, C. B. Poitras, K. Luke, L. Luo, P. A. Morton, and M.

Lipson, IEEE Photon. Technol. Lett. 26, 2380 (2014).5. D. Taillaert, W. Bogaerts, P. Bienstman, T. F. Krauss, P. van Daele, I.

Moerman, S. Verstuyft, K. De Mesel, and R. Baets, IEEE J. QuantumElectron. 38, 949 (2002).

6. A. Mekis, S. Gloeckner, G. Masini, A. Narasimha, T. Pinguet, S. Sahni,and P. De Dobbelaere, IEEE J. Sel. Top. Quantum Electron. 17, 597(2011).

7. S. Shao and Y. Wang, Opt. Lett. 35, 1834 (2010).8. X. Chen and H. K. Tsang, Opt. Lett. 36, 796 (2011).9. W. S. Zaoui, A. Kunze, W. Vogel, and M. Berroth, IEEE Photon.

Technol. Lett. 25, 1395 (2013).

10. Z. Cheng and H. K. Tsang, Opt. Lett. 39, 2206 (2014).11. J. H. Song, F. E. Doany, A. K. Medhin, N. Dupuis, B. G. Lee, and F. R.

Libsch, Opt. Lett. 40, 3941 (2015).12. D. Taillaert, C. Harold, P. I. Borel, L. H. Frandsen, R. M. De La Rue,

and R. Baets, IEEE Photon. Technol. Lett. 15, 1249 (2003).13. C. R. Doerr, P. J. Winzer, Y.-K. Chen, S. Chandrasekhar, M. S.

Rasras, L. Chen, T.-Y. Liow, K.-W. Ang, and G.-Q. Lo, J. LightwaveTechnol. 28, 520 (2010).

14. C. Doerr, L. Buhl, Y. Baeyens, R. Aroca, S. Chandrasekhar, X. Liu, L.Chen, and Y.-K. Chen, IEEE Photon. Technol. Lett. 23, 762 (2011).

15. W. Bogaerts, D. Taillaert, P. Dumon, D. Van Thourhout, and R. Baets,Opt. Express 15, 1567 (2007).

16. X. Chen, C. Li, and H. K. Tsang, Opt. Commun. 283, 2146 (2010).17. S. Pathak, M. Vanslembrouck, P. Dumon, D. Van Thourhout, and W.

Bogaerts, Opt. Express 20, B493 (2012).18. F. Van Laere, W. Bogaerts, P. Dumon, G. Roelkens, D. Van

Thourhout, and R. Baets, J. Lightwave Technol. 27, 612 (2009).19. L. Carroll, D. Gerace, I. Cristiani, and L. C. Andreani, Opt. Express 22,

14769 (2014).20. J. Zou, Y. Yu, and X. Zhang, Opt. Express 23, 32490 (2015).21. R. Halir, D. Vermeulen, and G. Roelkens, IEEE Photon. Technol. Lett.

22, 389 (2010).22. Y. Ding, H. Ou, and C. Peucheret, Opt. Lett. 38, 2732 (2013).23. A. J. Lim, S. Junfeng, F. Qing, L. Chao, T. Xiaoguang, D. Ning, C. Kok

Kiong, R. P. C. Tern, and L. Tsung-Yang, IEEE J. Sel. Top. QuantumElectron. 20, 405 (2014).

24. http://www.lumerical.com/tcad‑products/fdtd/.25. W. S. Zaoui, M. F. Rosa, W. Vogel, M. Berroth, J. Butschke, and F.

Letzkus, Opt. Express 20, B238 (2012).26. W. S. Zaoui, A. Kunze, W. Vogel, M. Berroth, J. Butschke, F. Letzkus,

and J. Burghartz, Opt. Express 22, 1277 (2014).

Letter Vol. 41, No. 18 / September 15 2016 / Optics Letters 4209