1. Introduction

In recent years, silicon photonics have seen substantial technological achievements beyond early expectations. Low loss high-index contrast silicon submicron-sized waveguides that can be fabricated by mature CMOS-compatible processes are showing promise for realistic dense photonic integrated circuit (PIC) [1] in various applications including optical communications, optical interconnects, signal processing and sensing. Although efficient coupling between the optical fiber and the submicron-sized waveguide is a key challenge due to the large modal mismatch, recent inspiring developments have shown that significantly improved coupling efficiencies using grating couplers [2–12] or adiabatic tapers and polished facets [13–16] is possible. Lateral coupling techniques with adiabatic tapers [13] have achieved highly efficient coupling between optical fibers and silicon nanophotonic waveguides with ~0.1μm² cross section area. The grating coupler approach is also attractive because of its inherent advantages over the alternative techniques: compatibility with wafer level optical testing in mass production, increased design flexibility by being able to place the optical input/output anywhere on a chip, and large alignment tolerance.

There are two main factors that have limited the coupling efficiency of grating couplers. One is the directionality (defined as the portion of the light power coupled upwards normalized with respect to the total out-coupled optical power). Different structures such as slanted gratings [4], a substrate mirror [5] or a silicon overlay [6] were demonstrated to diffract more light towards the fiber. However these approaches have their drawbacks. The fabrication of the slanted gratings or substrate mirror is not CMOS compatible and very complex. By applying an overlay on top of the device layer to enhance the directionality, a coupling efficiency of −1.6 dB has been reported [6], but the use of amorphous silicon limits the thermal budget. Another factor that limits the coupling efficiency is the mode mismatch between the fiber mode and the field profiles of diffracted light from waveguide grating. The coupling efficiency can be optimized by engineering the coupling strength of the grating structure to achieve better mode match [7–9, 12]. Most recently, a silicon apodized grating coupler was demonstrated with a coupling loss of −1.2dB for a waveguide thickness of 340 nm and a deep waveguide etch depth of 200 nm [8]. However, such apodized grating structure requires feature sizes (minimum feature size of 50 nm) that are beyond the limit of current deep UV lithography. This becomes an obstacle for the actual deployment of such device in mass production.

Here we report simulations and experimental results for a double-etched apodized waveguide grating coupler on silicon-on-insulator (SOI). By introducing the double-etched structure into the apodized grating design, the inherent tight feature sizes’ requirement to achieve high coupling efficiency is relaxed so that the current deep UV photolithography is capable of the device fabrication.

2. Device designs and principle

In order to reach high coupling efficiency between fiber and waveguide, a highly directional grating is required whilst the grating coupling strength α needs to be optimized to obtain a Gaussian-shaped field profile in order to match the fiber mode. Therefore, in our design, we first choose the silicon waveguide thickness of 340 nm but not the conventional thickness of 220 nm to improve the directionality of the grating. Furthermore, the apodized grating coupler design with engineered coupling strength for each grating period was employed to enhance the coupling efficiency.

The optical power confined in the grating may be described by P(z) = P₀exp(−2az) for uniform grating, where α is the coupling strength. In order to obtain a Gaussian-shaped output field profile G(z), which is defined as G²(z) = −dP(z)/dz, the coupling strength along z-axis is given by

The coupling strength can be varied by modifying the fill factor for each grating period. In a single etched apodized grating coupler on 340-nm-thick SOI, the optimized design requires the fill factor f (defined as the ratio of the etched groove width g to the grating period Λ) ranging from 0.08 to 0.4 to be applied along z-axis [8]. The grating period Λ satisfies the phase matching condition: qλ = Λ(n_eff - n_c sinθ), where λ is the center wavelength, neff is the effective index of the grating, nc is the refractive index of the cladding material, θ is the off-vertical tilt angle of the fiber, q is the diffraction order (equal to 1 for most grating couplers). The average effective index n_eff and the grating period would also change with f. According to Eq. (1) and the phase matching condition, the grating period Λ is determined to be gradually changing from 548 to 610 nm along z-axis with f changing from 0.08 to 0.4 for the central wavelength around 1550 nm. The etched groove width for each period is thus given by g = f Λ. Then the first few grating periods with small fill factor (~0.1) would result in that the groove widths (minimum feature size of ~50 nm) are beyond the limit of current deep UV lithography [8].

Hence, in order to eliminate such undesirably tight requirement for minimum feature sizes, we propose the double-etched apodized waveguide grating coupler as shown in Fig. 1 . The coupling strength can also be engineered by the etch depth. By introducing a shallower etch depth on the first few grating periods, much larger fill factor (or feature size) is required to match the same coupling strength. The weak coupling strength requirement for the first few periods of the apodized grating enables the possibility of employing shallower etch depth without much influence on the directionality since there is not intensive light to be coupled at this region. In our design, the minimum feature size is released to be 135 nm by determining proper shallow etch depth e_s. The design procedure is summarized as follows. Firstly, the deeper main etch depth e_d is determined by the condition to optimize the directionality of the waveguide grating coupler. We performed FDTD simulations to derive the directionality with different etching depth of grooves for 340-nm-thick silicon SOI by assuming a uniform grating with f of 0.5 and Λ of 610 nm. The results are detailed in Ref [8]. We achieved the optimized etch depth e_d of 190 – 200 nm, which can apply to the case of either air cladding or oxide cladding. Secondly, the grating period Λ and fill factor f for each period were carefully designed to vary α in order to realize Gaussian-like output fields profile with desired angle θ. The designed coupling strength gradually increases along z-axis. We remark that the shallow etch depth should be chosen carefully to minimize the effect on the directionality and optimize the fill factor (or groove width). We set the minimum feature size of 135 nm for the grating design so that the deep UV photolithography is capable of producing the grating coupler to meet the mass production criteria. Based on this design rule, we then chose the shallow etch depth e_s of 90 nm, which gave the best coupling efficiency based on FDTD simulations when the grating groove width closest to the waveguide is 135 nm. In order to mitigate the effect on directionality, the number of shallow etch period was determined when the grating groove for the next period can be larger than 135 nm if the deep etch depth is applied. The proposed grating coupler has 23 periods including 4 periods with shallow etch depth and 19 periods with deep etch depth. The total length is about 13 μm.

 

Fig. 1 (a) Schematic of the double-etched diffractive apodized waveguide grating coupler. (b) Cross-sectional view of the double-etched apodized grating coupler.

Download Full Size | PDF

3. Simulation results

We employed commercial two-dimensional (2D) finite-difference time-domain (FDTD) numerical simulation tool (RSoft) for such grating structures. A Gaussian waveform with 1/e full width of 10.4 μm was employed to represent the single mode fiber input. Coupling efficiency between the fiber and waveguide was calculated using the power in the waveguide mode coming out from the front end of the grating coupler. During simulations, the fiber position which is the horizontal distance from the center of fiber core to the front-end of the grating coupler was adjusted to maximize the coupler efficiency. Normally in the fabrication of the grating coupler, deviations from the designed parameters is expected. The influence of all grating groove widths variation (Δg) was thus numerically studied by assuming the same Λ for each grating period. The simulated spectra with Δg of −10, 0, and 10 nm are shown in Fig. 2 . A minimum coupling loss of −0.97 dB (corresponding to 80% coupling efficiency) are predicted. The 1-dB and 3-dB bandwidth are 31 and 53 nm respectively. The grating reflects ~1% of the light back into the waveguide. The coupling loss deviation is within 0.15 dB with the variations of +/− 10 nm as shown in Fig. 2. Smaller groove width results in a reduction of the effective index of the grating and therfore a wavelength red-shift of the peak position, according to the phase matching condition. Inset shows the simulated coupling loss as a function of Δg. The coupling efficiency drops with positive Δg (larger groove widths).

 

Fig. 2 Simulated coupling loss for the double-etched apodized waveguide grating coupler as a function of wavelength with groove widths variation Δg of −10 (circles), 0 (squares), and 10 nm (triangles). (i) Inset shows the simulated minimum coupling loss as a function of groove widths variation Δg.

Download Full Size | PDF

The influence of the grating etch depths variation was also numerically studied as shown in Fig. 3 . The simulations were done by varying one etch depth while fixing the other. The deep etch depth e_d variation has more impact on the coupling efficiency, while the double-etched grating coupler shows much larger fabrication tolerance to the shallow etch depth e_s. Especially when e_s is within 50 – 120 nm range, only less than 0.3 dB coupling loss difference is expected. The coupling efficiency drops more rapidly with deeper etch depth for both shallow etch depth e_s and deep etch depth e_d.

 

Fig. 3 Simulated coupling loss for the double-etched apodized grating coupler as a function of etch depth by changing shallow etch depth (squares) and deep etch depth (circles).

Download Full Size | PDF

4. Device fabrication and characterization

The double-etched apodized waveguide grating coupler was fabricated on a commercial 8-inch 340-nm-thick SOI wafer with 2-μm-thick buried oxide layer using CMOS-compatible technology. Firstly, a 70-nm-thick oxide layer was deposited as hard mask. The grating structure was defined by deep UV photolithography and transferred onto the hard mask layer by dry etching. Secondly, a shallow dry etching was performed through silicon device layer with etch depth e_s. After that, the first four periods of the grating coupler was covered by the patterned photoresist using photolithography. Then another silicon dry etching process with etch depth (e_d - e_s) was performed in order to reach the target deep etch depth e_d for other grating periods. Finally, the waveguide was patterned and etched with a remained slab height of ~100 nm. We remark that our design has inherent large alignment tolerance with double-etched process because the misalignment specification of the deep UV photolithography facility is within ± 60 nm, which is far less than the grating slit width of ~400 nm.

Figure 4(a) shows the top-view scanning electron micrograph (SEM) of the fabricated double-etched waveguide grating coupler on SOI. Figure 4(b) and 4(c) show the zoom-in view SEM images of the device. We can clearly observe the double etched structure as shown in Fig. 4(c). The measured shallow etch depth e_s is ~85 nm and the deep etch depth e_d is ~210 nm. The fill factor and the period are purposely designed to achieve engineered coupling strength for each grating period (shown in Fig. 4(a)). The minimum feature size is ~135 nm.

 

Fig. 4 (a) Top-view scanning electron micrograph (SEM) of the fabricated grating coupler. (b), (c) Zoom-in view SEM images of the device.

Download Full Size | PDF

We then characterized the fabricated double-etched grating couplers by measuring the insertion loss of a waveguide with grating couplers at both ends using cleaved single mode fibers tilted at 10° as optical input and output. A pair of adiabatic linear tapers, each with a length of 500 μm, is used to connect the 10-μm-wide grating region and a 3-μm-wide waveguide section with a length of 550 μm. Another pair of 200-μm-long adiabatic tapers is then used to connect the 3-μm-wide waveguide to a 500-nm-wide waveguide. The total distance between the two grating couplers is 2.9 mm. A tunable laser (Agilent 8164B) was employed as the input light source. The resolution is set to be 0.02 nm. A polarization controller was used to adjust the input light in TE-polarization for the waveguide grating coupler. We remark that there is no index-matching liquid applied between the optical fiber and waveguide grating coupler.

The measured coupling loss for double-etched grating coupler as a function of wavelength is shown in Fig. 5 assuming that the waveguide and tapers are lossless and that the input and output grating coupler have the same coupling loss. The measurement results were normalized with the link loss without including the possible extra loss from the cleaved input and output fibers and connectors. The minimum coupling loss is −1.5 dB (corresponding to 70.8% coupling efficiency) with a 3-dB bandwidth of 54 nm. We note that the fabricated etch depth e_d of ~210 nm is not the optimal value to achieve the maximum coupling efficiency as illustrated in Fig. 3. We then simulated the coupling loss of the grating coupler by assuming shallow etch depth of 85 nm and deep etch depth of 210 nm without considering loss, fabrication errors and so on. The simulation suggests −1.3 dB minimum coupling loss with 53 nm 3-dB bandwidth. The measurement matched with simulation results quite well.

 

Fig. 5 Measured coupling loss for the double-etched apodized grating coupler as a function of wavelength with groove widths variation Δg of −10, 0, and 10 nm.

Download Full Size | PDF

We also varied the widths of grating grooves in the mask layout to investigate the fabrication tolerance. Figure 5 shows the measured coupling loss with different Δg (−10, 0, and 10 nm). The coupling loss deviation remains less than 0.3 dB despite the +/− 10 nm variations in actual groove widths. The measurement results are also consistent with the simulations as shown in Fig. 2.

5. Conclusion

A high efficiency double-etched fiber-to-waveguide grating coupler with engineered periods, which was fabricated by CMOS-compatible technology using deep UV photolithography on SOI platform, was reported. The measured coupling loss is −1.5 dB, which is corresponding to 70.8% coupling efficiency. The 3-dB bandwidth is 54 nm. This result is the highest coupling efficiency demonstrated to date for waveguide grating couplers fabricated in a CMOS line using deep UV photolithography. Measurements show the double-etched grating coupler with engineered periods is tolerant to large fluctuations in groove width in fabrication process. Further improvement may be obtained by optimizing the etch depth.