Silicon nitride is demonstrated as a high performance and cost-effective solution for dense integrated photonic circuits in the visible spectrum. Experimental results for nanophotonic waveguides fabricated in a standard CMOS pilot line with losses below 0.71dB/cm in an aqueous environment and 0.51dB/cm with silicon dioxide cladding are reported. Design and characterization of waveguide bends, grating couplers and multimode interference couplers (MMI) at a wavelength of 660 nm are presented. The index contrast of this technology enables high integration densities with insertion losses below 0.05 dB per 90° bend for radii as small as 35 µm. By a proper design of the buried oxide layer thickness, grating couplers with efficiencies above 38% for the TE polarization have been obtained.
©2013 Optical Society of America
Silicon nitride (SiN) is a promising platform for integrated photonics applications. This technology meets the key features that led to the success of the extensively investigated and more mature silicon-on-insulator based integrated photonics technology [1,2]. First, a refractive index contrast between the core and the oxide cladding (Δn ~0.5) that allows a tight confinement of the light and enables a high density of integration. Second, its compatibility with standard complementary-metal-oxide semiconductor (CMOS) technology, that takes advantage of the ongoing evolution in the microelectronics industry and permits a feasible cost reduction at large fabrication scales. Furthermore, silicon nitride has some advantages that make it compete favorably with silicon for some applications. The refractive index contrast is not as high as with silicon, thus significantly reducing scattering losses, and two photon absorption at near-infrared wavelengths is negligible; thereby low losses and resonators with high Q factors have been demonstrated [3–5]. At the same time, the lower index contrast makes the resulting devices more tolerant to fabrication imperfections. Finally, due to its nonlinear properties (parametric amplification [6–9], broadband supercontiniuum generation ) and its transparency in the near-infrared and visible spectrum, silicon nitride opens a wide range of new possibilities for CMOS-compatible integrated photonics applications. In particular, this technology is a leading candidate to become the standard solution for biosensing applications in the visible and near-infrared regime [11–13], where its low loss and low sensitivity to thermal variations play a crucial role.
Here we address by design and experimental validation three important problems for the development of silicon nitride integrated photonics at visible wavelengths. First, we present a waveguide with an ultrathin film layer that enables a high interaction of the field with the top cladding [11–14], while also allowing a high level of integration. Silicon nitride grating couplers have been reported in the infrared [15, 16] and recently in the near-infrared . Here, we demonstrate an ultrathin silicon nitride grating for coupling a free-space laser diode beam at visible wavelengths. Finally, we present multimode interference couplers as an effective solution for splitting the light inside the photonic circuit.
2. Designs and simulations
2.1 Waveguide design
The cross-section of the interconnection waveguides consists of a 100 nm thick silicon nitride (SiN) core layer (nSiN ≈1.87 as determined by ellipsometry) above a silicon dioxide (SiO2) bottom cladding (nSiO2 ≈1.45) and a silicon substrate. Waveguides covered with SiO2 and water (H2O) top claddings (nH2O ≈1.33) have been studied. In both cases the waveguide is fully etched down to the bottom oxide layer. We have used the FIMMWAVE (Photon Design) mode solver for the waveguide design. A width of 700 nm has been chosen in order to obtain single-mode behavior (wavelengths higher than 580 nm) for both choices of top cladding materials. The ultrathin silicon nitride film enables a strong interaction of the evanescent field with an aqueous cladding that makes it especially suitable for sensing applications. In order to assess the influence of the film thickness, we have analyzed the percentage of mode power that is contained in the entire surrounding water cladding as well as in a 50 nm thick water region in the immediate proximity of the waveguide core. Results corresponding to a wavelength of 660 nm are plotted in Fig. 1. Maximum values are achieved in both cases for waveguide thicknesses close to 100 nm. For thicker waveguides the mode field is more confined inside the waveguide core, reducing the evanescent field. In the case of a waveguide film thinner than 90 nm, the effective index of the waveguides approaches that of the bottom oxide cladding so that a disproportionally large amount of the mode field is contained within the latter, thereby reducing the fraction of the field contained inside the top water cladding.
The vertical asymmetry introduced by the water cladding also increases the bend losses in comparison to SiO2-clad waveguides. In order to reach an acceptable level of integration, waveguide bends with radii of at least 35 µm and SiO2 top cladding are proposed for compact light distribution networks (0.027 dB/90° in simulation). In the case of water cladding, the bending radius should be increased above 60 µm for similar bending losses. Simulation results showed that transition losses between SiO2- and H2O-clad waveguides can be easily maintained below 0.2 dB by tapering up the width of the waveguide to 2 µm in the transition region.
2.2 Grating Coupler design
A grating has been designed to couple light at a fixed wavelength (660 nm) from a free-space beam source (modeled as a Gaussian) focused on the surface of the chip. The schematic 2D lateral view of the grating coupler along the propagation axis is depicted in Fig. 2. In the design process, the reciprocal problem has been simulated, exciting through the waveguide and analyzing the field diffracted from the grating. For these 2D simulations we used the commercial software FullWave (RSoft) based on the finite-difference time-domain (FDTD) method. In order to facilitate the integration of the chip in future systems, we targeted a coupling angle as close as possible to normal incidence (θ = 0°) without compromising specifications excessively. Coupling to an exactly vertical direction leads to high reflections due to second order Bragg diffraction. Solutions have been proposed to suppress these reflections [18,19], but at the expenses of a more demanding fabrication process due to smaller feature sizes. In order to maintain a high yield, we have decided to slightly increase the angle. The pitch of the grating (Λ = 420 nm) has been designed to couple to a first order diffraction at an angle θc = −3.44° (in the SiO2 cladding, θair = −5°) determined by20], and maintain a good tolerance to fabrication deviations (CD-Control) we have chosen an equal length for the grating teeth and the grooves (50% duty cycle).
The coupling efficiency (ρ), defined as the percentage of the incident power that is coupled from the free space Gaussian beam into the waveguide, is determined by two main factors (Eq. (2)): the directionality (ηdir) and the mode overlap (ηov).
The directionality (ηdir) is defined here in the reciprocal problem as the fraction of power that is radiated to the top relative to the incident power launched into the input waveguide of the grating (as such, it already takes into account back-reflections into the waveguide as well as transmission due to the finite length of the grating: 88 grates). Two main alternatives have been demonstrated to increase this parameter: First, the introduction of an asymmetry in the vertical direction by means of a thicker core layer and a partial etch step [21,22]. Second, the inclusion of a higher reflectance bottom interface such as a gold mirror or a DBR mirror that redirects the light diffracted to the bottom cladding [23,24].
The mode mismatch between the beam diffracted by the grating coupler and the coupled Gaussian beam can be calculated with the overlap integral given by Eq. (3)20]. For a grating coupler with a given Ld, the theoretical maximum overlap value is achieved when the input Gaussian beam has an optimal full-width at half-maximum (FWHMopt) verifying the relation: . Grating apodization could enhance the overlap value up to near 100%, but it would also increase the complexity of the fabrication process .
Tolerance to spatial and angular misalignment of the coupled Gaussian beam relative to the grating coupler has been a primary target during the design phase. However, spatial and angular alignment tolerances exhibit an inverse relation and a compromise had to be made. They are also directly related to the size of the beam diffracted by the grating: the smaller the diffracted beam, the more difficult it is to spatially align to it, but given the wider distribution of power in wave-vector space, the easier the angular alignment. Conversely, larger beam sizes relax the spatial alignment tolerance, but increase the demands on the angular positioning accuracy. Table 1 summarizes the 1.25 dB and 3 dB excess loss angular and spatial alignment tolerances for grating couplers with different values of Ld when they are excited with the respective optimal input FWHM Gaussian (FWHMopt).
Due to the small thickness of the silicon nitride film and the index contrast of the technology, only a weak scattering strength (large Ld) can be achieved with a partial etch. As a consequence, the resulting gratings are rather long and the angular alignment tolerance very stringent for a matching Gaussian beam. A full etch can contribute to alleviate these problems by maximizing the coupling strength. At the same time, the fabrication complexity is reduced, since the grating can be defined with the same etch step as the rest of the structures and the underlying SiO2 cladding can be used as an etch stop layer. Furthermore, an over-etch is optically irrelevant if the upper cladding of the grating coupler also consists in SiO2.
Finally, the optimal thicknesses of the bottom (dox bot) and top (dox top) SiO2 cladding layers have been determined. Both directionality and decay length are directly related to these parameters. Figures 3(a) and 3(b) show the simulated periodic dependence of directionality and decay length on dox bot and dox top, where the period in both axes is given by the half-wavelength of the light in the oxide (~227 nm). First, the bottom SiO2 layer thickness has been designed to achieve constructive interference between the wave reflected at the substrate interface and the field that is directly radiated to the top cladding. An inverse relation between decay length and directionality as a function of dox bot can be clearly distinguished in Fig. 3. Destructive interference in the top cladding leads to the light reflected by the bottom substrate to constructively couple back into the grating coupler. Due to interference, light scattered out of the grating coupler then also preferentially radiates down towards the partial reflector at the substrate interface. Thus, destructive interference in the top cladding effectively adds an additional guiding mechanism that confines the light to the chip surface for a longer distance, increasing Ld. Therefore, a proper design of the bottom SiO2 layer thickness does not only increase the grating efficiency due to a higher ηdir but also contributes to reduce the decay length. The nominal design value for the dox bot has been fixed to 1.88 µm.
Conversely, there is also a direct relation between decay length and directionality as a function of the top SiO2 layer thickness. In this case, a destructive interference between the wave that is reflected at the air-oxide interface and the wave that is directly radiated to the bottom cladding not only creates a guiding mechanism that enlarges Ld, but also reduces the amount of power that is transmitted through the substrate and hence increases the directionality. As a consequence, in this case there exists a trade-off between directionality and decay length in the design of dox top. However, the dependence of these performance metrics on dox top is much weaker than with dox bot since the reflectance of the SiO2-air interface is much lower than the reflectance of the SiO2-Si interface. In order to achieve a maximum coupling efficiency, the nominal value of dox top has been chosen to be 1.23 µm. Assuming a maximum fabrication deviation in the bottom oxide thickness of ± 40 nm, and an uncontrolled top oxide thickness, the expected value of the decay length is in the range 17 µm – 25 µm and the directionality in the range 0.51 – 0.68. The optimal input Gaussian FWHMopt should be in the range 13.5 µm – 19.9 µm for a high overlap value. Figure 4 shows a subset of the data shown in Figs. 3(a) and 3(b) that exemplifies clearly the correlation and anti-correlation of the directionality and the decay length when the top and bottom SiO2 thicknesses are respectively being varied.
In order to achieve a good mode matching in the transverse direction, the width of the grating has been set to 32 µm (13 µm optimal input Gaussian FWHMopt for high overlap with the ground mode). A 2 mm long taper is adiabatically converting the transverse dimension of the in-plane waveguide to the standard interconnection waveguide width (700 nm). Considering an expected input Gaussian with a FWHM of 11.5µm, a theoretical maximum grating efficiency of 50.5% can be achieved (see Fig. 3(c)). With variations of ± 40 nm in the bottom oxide thickness and an uncontrolled top oxide thickness, the worst case grating efficiency remains above 35%.
Although a very small incident angle can solve the problem of high reflections back into the waveguide, a reduced bandwidth is predicted (less than 20 nm) due to the occurrence of the Bragg condition at a nearby wavelength. This filtering characteristic could be favorable in some sensing applications where only a resonance at a specific wavelength is targeted. For applications with broader bandwidth requirements, an angle higher than 10° should be preferred . Figure 3(d) shows the simulated efficiency at different wavelengths with the incident beam angle assumed to be tracking the grating diffraction angle. A steep decay in the grating efficiency can be distinguished at around 625 nm due to the second order Bragg reflection. A symmetric window of operation is possible at 590 nm for coupling angles near 5°. An even higher efficiency is expected at 590 nm (55%) due to a higher directionality as a result of a higher reflectance of the silicon substrate and a higher scattering strength of the grates at smaller wavelengths.
2.3 1x4 Multimode Interference device
Splitter designs based on directional couplers and multimode interference couplers (MMI) for the distribution of light inside the chip have been considered. However, multimode interference couplers have demonstrated to be a more reliable option due to their lower sensitivity to fabrication deviations in the film thickness, waveguide width and refractive index. In this work, we opted for a 1x4 MMI in order to realize a compact power distribution network. A comparison between 1x4 and the more common 1x2 MMIs in terms of fabrication tolerance and bandwidth can be found in . The schematic top-view of the symmetric 1x4 MMI design is depicted in Fig. 5. Input and output waveguides have been widened up to 1.5 µm (WA) along a 10 µm taper (Ltaper) in order to reduce excitation of the higher order modes in the multimode region and consequently achieve a better fabrication tolerance . To prevent coupling between output ports, a center-to-center separation of 2.5 µm has been chosen (gap = 1 µm). Simulation results obtained with FIMMPROP (Photon Design) show an optimal multimode section length (LMMI) of 57.3 µm for a 9.7 µm wide (WMMI) MMI with a top oxide layer cladding.Table 2 summarizes the simulated results of the sensitivity analysis for fabrication parameters varied within the indicated ranges (the MMI length is expected to be set by design). The most critical parameter is the width of the multimode region, but even a maximum deviation of ± 0.2 µm will maintain a high efficiency (>91%) and a low imbalance between output ports (<2.5%). The explored parameter ranges are well within the capability of optical lithography, so that a high fabrication yield is expected.
3. Experimental results
The different structures were fabricated on 200 mm silicon wafers in a standard CMOS pilot line. Plasma enhanced vapor deposition and DUV optical lithography were used in the definition of the different layers and waveguides. For the optical characterization, a laser diode beam at a wavelength of 660 nm was focused with beam shaping optics onto the surface of the chip with a FWHM of 11.5 µm. The samples were mounted on an XYZ-stage for spatial alignment and a mirror stage with two degrees of freedom was used to adjust the incident angles of the laser beam. All the functional devices are connected at the input and output ports to identical grating couplers. At the output grating, the light is again coupled out and the signal is redirected to a CCD camera. After background subtraction of the camera image, the detected signal is integrated in a region around the output grating coupler in order to determine the total out-coupled power. The set-up has been characterized by focusing the laser beam on top of a chrome sample of known reflectance. This reference measurement has allowed us to normalize the output relative to the incident power. Additionally, the polarization was fixed to maximize the coupling power (linear TE polarization).
Characterization of the propagation losses was performed with the fabrication and measurement of waveguides of different lengths (1, 2.5, 5, 10 and 49.8 mm) with the cut-back method. Waveguides with different widths (0.7, 1.5 and 2 µm) were also included to evaluate the impact of etch roughness at the side-walls of the waveguides. Finally, to evaluate the influence of the top cladding we have fabricated wafers cladded with SiO2 and wafers without top oxide cladding immersed into water. By performing an exponential fit of the experimental data, propagation losses below 0.51 dB/cm have been obtained for top SiO2 cladding and 0.71 dB/cm for water cladded waveguides (see Table 3). SiO2-clad waveguides exhibit a slightly better performance. The small difference in the measured attenuation between waveguides of different widths indicates that material absorption losses and/or roughness at the top and bottom of the SiN film play a significant role in the overall propagation losses.
Meandering waveguides with different curvature radii and number of 90° bends (16, 32, 64, 96, 128 and 144) were fabricated and measured for the characterization of the bend loss. All the measured structures were covered with SiO2 cladding and the width was fixed to 700 nm. Figure 6 shows the measured excess output loss as a function of the number of 90° bends. Low insertion loss of only 0.05 ± 0.01 dB per 90° bend has been obtained for a 35 µm radius, enabling a high level of integration. This result is slightly higher than the value obtained by simulations (0.027 dB per 90° bend).
Loop structures with input and output gratings connected by a waveguide were defined and measured for characterization of the coupling efficiency (see Fig. 7(a)). According to the design presented in section 2.1, the period of the fabricated grating coupler is 420 nm and the duty cycle 50%. The scattering section has 88 grates. The gratings have a width of 32 µm and are tapered down to a width of 700 nm (taper length = 2 mm) for connection with the single mode waveguides. A maximum of 26% of the incident power has been measured at the output for the optimum coupling angle. Figure 7(b) shows the fraction of the input power measured at the output as a function of the incident angle. The structures present an angular alignment tolerance of ± 0.4° for 3 dΒ excess losses, this value is consistent with the analysis presented in Table 1, and confirms the expected stringent angular tolerances of the proposed grating, due to the large size of the diffracted beam. The power measured at the output is related to the incident power and the coupling efficiency (ρ) through the expression. Note that the overlap occurs only once in the expression, since the light at the output grating coupler is directly imaged. Hence, the coupling efficiency can be calculated by measuring the fraction of the incident power at the output if either the overlap value or the directionality is known. In a worst-case analysis, the maximum theoretical directionality of the grating coupler with an ideally placed silicon substrate and top cladding thickness can be considered (0.68). This ηdir value gives us a lower limit of the grating efficiency of at least 38%.
Additionally, CCD camera images of the diffracted beam have been analyzed to estimate the decay length, the transversal FWHM and the overlap values. Figures 8(b) and 8(c) show the radiated beam profiles, respectively along the propagation axis (z axis) and the transverse axis (x axis).
A decay length of 17.4 µm and a FWHM in the transverse direction of 13 µm were obtained. The overlap with the Gaussian beam estimated from the power distribution obtained from the CCD image at the output grating coupler is 0.75. Based on this overlap value, the grating efficiency would be 44%. However, it is important to point out that only the power distribution and not a possible phase front distortion can be captured by the camera. Thus these 44% are an upper bound of the coupling efficiency. Simulation results also showed that the reflected power measured at the input grating (8.4%) is consistent with the assumed directionality. The experimentally determined insertion efficiency, in the range 38% to 44%, is well in line with the theoretically predicted 35% worst case taking into account the tolerances of the fabrication process and the theoretically predicted best case of 50.5% corresponding to a perfectly fabricated structure.
For the characterization of the MMIs different devices with variations in the multimode section length have been fabricated (see Fig. 9(a)). The four output ports of the MMI were routed to output grating couplers located on both sides of the input grating coupler in order to minimize biasing effects of the imaging system used to characterize the structure. Additional insertion losses due to waveguides and bends are negligible. All four output signals were captured in the CCD camera image and analyzed to compute the total MMI efficiency and imbalance.
Figure 9(b) shows the measured imbalance and efficiency of the MMIs as a function of the multimode section length. For the calculation of the efficiency, the insertion loss of a reference grating coupler loop has been subtracted. The MMIs with the chosen design length (57.3 µm) present an efficiency higher than 93% and an imbalance level below 10%. Figure 9(b) also compares simulations and measurements of the MMI efficiency and imbalance as a function of the length. A good agreement has been found in the efficiency, but the experimental results show a higher imbalance than in simulations. This discrepancy between measured and expected imbalance has been attributed to small differences in the performance of the output gratings and stray light, which translates small deviations of the position of the input beam into imbalance between outputs (the input grating was routed in line with the MMI).
We have experimentally demonstrated that silicon nitride is a highly efficient and standard CMOS compatible technology platform for dense photonic integrated applications in the visible spectrum. Designs of waveguide bends, grating couplers and multimode interference couplers at a wavelength of 660 nm have been presented. The ultrathin film layer of the waveguides enables a high interaction of the field with the top cladding and makes it especially suitable for sensing applications. Low waveguide propagation losses (below 0.71 dB/cm and 0.51 dB/cm, respectively with water cladding and SiO2 cladding) and bend losses (0.05 dB/90° bend for 35 µm radius and SiO2 cladding) have been measured. Light coupling has been addressed by a grating design with measured coupling efficiencies of more than 38%. We have also shown that compact multimode interference couplers are an efficient solution for the light distribution network inside the chip.
This work was supported by the European Research Council (ERC FP7/2011-2016 No. 279770) and the European Union’s Seventh Framework Programme (CIG FP7/2011-2015 No. 293767). The authors would like to acknowledge Sean Cheng and Allen Timothy Chang from the Taiwan Semiconductor Manufacturing Company (TSMC).
References and links
1. G. Masini, G. Capellini, J. Witzens, and C. Gunn, “High-speed, monolithic CMOS receivers at 1550 nm with Ge on Si waveguide photodetectors,” Proc. 20th Lasers and Electro-Optics Soc.(LEOS), 848-849 (2007).
2. D. Feng, B. J. Luff, and M. Asghari, “Recent advances in manufactured silicon photonics integrated circuits,” Proc. SPIE 8265, 826507, 826507-9 (2012). [CrossRef]
3. N. Daldosso, M. Melchiorri, F. Riboli, M. Girardini, G. Pucker, M. Crivellari, P. Bellutti, A. Lui, and L. Pavesi, “Comparison among various Si3N4 waveguide geometries grown within a CMOS fabrication pilot line,” J. Lightwave Technol. 22(7), 1734–1740 (2004). [CrossRef]
4. A. Gorin, A. Jaouad, E. Grondin, V. Aimez, and P. Charette, “Fabrication of silicon nitride waveguides for visible-light using PECVD: a study of the effect of plasma frequency on optical properties,” Opt. Express 16(18), 13509–13516 (2008). [CrossRef] [PubMed]
6. J. S. Levy, A. Gondarenko, M. A. Foster, A. C. Turner-Foster, A. L. Gaeta, and M. Lipson, “CMOS-compatible multiple-wavelength oscillator for on-chip optical interconnects,” Nat. Photonics 4(1), 37–40 (2010). [CrossRef]
7. L. Razzari, D. Duchesne, M. Ferrera, R. Morandotti, S. Chu, B. E. Little, and D. J. Moss, “CMOS-compatible integrated optical hyper-parametric oscillator,” Nat. Photonics 4(1), 41–45 (2010). [CrossRef]
8. F. Ferdous, H. Miao, D. E. Leaird, K. Srinivasan, J. Wang, L. Chen, L. T. Varghese, and A. M. Weiner, “Spectral line-by-line pulse shaping of on-chip microresonator frequency combs,” Nat. Photonics 5(12), 770–776 (2011). [CrossRef]
9. J. Pfeifle, C. Weimann, F. Bach, J. Riemensberger, K. Hartinger, D. Hillerkuss, M. Jordan, R. Holtzwarth, T. J. Kippenberg, J. Leuthold, W. Freude, and C. Koos, “Microresonator-Based Optical Frequency Combs for High-Bitrate WDM Data Transmission,” Optical Fiber Communication Conference, Los Angeles, USA, Mar. 4 (2012). [CrossRef]
10. R. Halir, Y. Okawachi, J. S. Levy, M. A. Foster, M. Lipson, and A. L. Gaeta, “Ultrabroadband supercontinuum generation in a CMOS-compatible platform,” Opt. Lett. 37(10), 1685–1687 (2012). [CrossRef] [PubMed]
11. I. Goykhman, B. Desiatov, and U. Levy, “Ultrathin silicon nitride microring resonator for biophotonic applications at 970 nm wavelength,” Appl. Phys. Lett. 97(8), 081108 (2010). [CrossRef]
12. G. Voirin, D. Gehringer, O. M. Parriaux, and B. A. Usievich, “Si3N4/SiO2/Si waveguide grating for fluorescent biosensors,” Proc. SPIE 3620, 109–116 (1999). [CrossRef]
13. H. Cai and A. W. Poon, “Optical trapping of microparticles using silicon nitride waveguide junctions and tapered-waveguide junctions on an optofluidic chip,” Lab Chip 12(19), 3803–3809 (2012). [CrossRef] [PubMed]
15. C. R. Doerr, L. Cheng, Y. K. Chen, and L. Buhl, “Wide bandwidth silicon nitride grating coupler,” IEEE Photon. Technol. Lett. 22(19), 1461 (2010). [CrossRef]
16. G. Maire, L. Vivien, G. Sattler, A. Kaźmierczak, B. Sanchez, K. B. Gylfason, A. Griol, D. Marris-Morini, E. Cassan, D. Giannone, H. Sohlström, and D. Hill, “High efficiency silicon nitride surface grating couplers,” Opt. Express 16(1), 328–333 (2008). [CrossRef] [PubMed]
17. A. Z. Subramanian, S. Selvaraja, P. Verheyen, A. Dhakal, K. Komorowska, and R. Baets, “Near-infrared grating couplers for silicon nitride photonic wires,” IEEE Photon. Technol. Lett. 24(19), 1700–1703 (2012). [CrossRef]
18. J. Witzens, A. Scherer, G. Pickrell, D. Louderback, and P. Guilfoyle, “Monolithic integration of vertical-cavity surface-emitting lasers with in-plane waveguides,” Appl. Phys. Lett. 86(10), 101105 (2005). [CrossRef]
19. G. Roelkens, D. Van Thourhout, and R. Baets, “High efficiency grating coupler between silicon-on-insulator waveguides and perfectly vertical optical fibers,” Opt. Lett. 32(11), 1495–1497 (2007). [CrossRef] [PubMed]
20. T. Tamir and S. T. Peng, “Analysis and Design of Grating Couplers,” Appl. Phys. (Berl.) 14(3), 235–254 (1977). [CrossRef]
21. D. Vermeulen, S. Selvaraja, P. Verheyen, G. Lepage, W. Bogaerts, P. Absil, D. Van Thourhout, and G. Roelkens, “High-efficiency fiber-to-chip grating couplers realized using an advanced CMOS-compatible Silicon-On-Insulator platform,” Opt. Express 18(17), 18278–18283 (2010). [CrossRef] [PubMed]
22. C. Alonso-Ramos, A. Ortega-Moñux, I. Molina-Fernández, P. Cheben, L. Zavargo-Peche, and R. Halir, “Efficient fiber-to-chip grating coupler for micrometric SOI rib waveguides,” Opt. Express 18(14), 15189–15200 (2010). [CrossRef] [PubMed]
23. F. V. Laere, G. Roelkens, M. Ayre, J. Schrauwen, D. Taillaert, D. V. Thourhout, T. F. Krauss, and R. Baets, “Compact and highly efficient grating couplers between optical fiber and nanophotonic waveguides,” J. Lightwave Technol. 25(1), 151–156 (2007). [CrossRef]
24. W. Sfar Zaoui, M. Félix Rosa, W. Vogel, M. Berroth, J. Butschke, and F. Letzkus, “High-Efficient CMOS-compatible grating couplers with backside metal mirror,” Europ. Conf. Opt. Comm. (ECOC), Amsterdam, Netherlands, Sept. 16-20 (2012).
25. R. Halir, P. Cheben, J. H. Schmid, R. Ma, D. Bedard, S. Janz, D.-X. Xu, A. Densmore, J. Lapointe, and Í. Molina-Fernández, “Continuously apodized fiber-to-chip surface grating coupler with refractive index engineered subwavelength structure,” Opt. Lett. 35(19), 3243–3245 (2010). [CrossRef] [PubMed]
26. P. A. Besse, M. Bachmann, H. Melchior, L. B. Soldano, and M. K. Smit, “Optical bandwidth and fabrication tolerances of multimode interference couplers,” J. Lightwave Technol. 12(6), 1004–1009 (1994). [CrossRef]