1. Introduction

Single-crystal lithium niobate thin film on a silicon dioxide layer (lithium niobate-on-insulator, LNOI) has shown great potential in integrated optics due to its outstanding electro-optical and nonlinear optical properties [1] as well as the high refractive index contrast between LN thin film and the buried oxide layer. Many interesting devices are demonstrated, such as electro-optical modulators [2,3] and second-harmonic generators [4‒6]. As a semiconductor, Si demonstrates superb advantages in the integrated electronics industry. The combination of Si and LN will merge the advantages of unparalleled electronic properties of Si and the excellent optical properties of LN. Recently, Si thin film on LN or LNOI platforms (Si-LNOI) [7,8], and LN thin film on SOI platform (LN-SOI) [9], and LNOI on SOI platform [10] have been demonstrated. In these cases, focused lens, tapered and lensed optical fibers, cantilever couplers, or directional couplers have been used in communicating with the platforms, respectively. Although these methods are useful and feasible, a compact and convenient coupler is also in need to benefit more compact and high-speed electro-optic devices.

A waveguide grating coupler which can be placed anywhere on a chip and needs no end face polishing has been considered as a good solution to coupling light between a fiber and devices fabricated on a chip. Due to the diffraction of gratings, light can be coupled out of the waveguide plane; when the output field mode is adjusted to match the fiber mode, light will couple effectively into the fiber. This coupling method is effective and allows wafer-scale testing of an optical integrated system. Waveguide grating couplers fabricated in LNOI have been reported, and the maximum CEs are between −12 dB and −7 dB [11‒14]; the gratings are fabricated by etching LN thin film. In Si-LNOI approach, a 70-nm-thick Si thin film has been deposited on LNOI surface, and strip-loaded waveguides are formed by etching of Si [8]. Once a waveguide grating coupler on Si-LNOI can be fabricated by just etching the upper Si thin film, the grating coupler will provide fabrication and characterization flexibilities to the Si-LNOI devices.

In this paper, a Si grating coupler for coupling light between single-mode fibers and Si-strip-loaded LNOI waveguides is designed, fabricated and characterized. The Si grating coupler is simulated and optimized by 2D-FDTD method. The simulated maximum CE is about −5 dB at wavelength 1550 nm. Experimentally, Si-strip-loaded waveguides on LNOI are fabricated by patterning and depositing a thin film of Si on LN thin film. FIB is used to etch the Si grating. The measured CE is about −18 dB at 1550 nm.

2. Design of the coupler

The simulation model is schematically shown in Fig. 1. Figure 1(a) and (b) were the top view and cross section view of the grating coupling system, respectively. The upper Si thin film was 50 nm, the LN thin film was 292 nm, and the underlying silicon dioxide layer was 1.95 μm in thickness. The refractive indices of Si and SiO₂ were set to be 3.48 and 1.46, and the extraordinary and ordinary refractive indices of LN thin film were set to be 2.138 and 2.2112, respectively. The Si waveguides (green) with gratings were located on the LN thin film, and single-mode fibers were on the top of the Si-LNOI platform. This coupling system included the input port and the output port. The two ports were exactly symmetrical, and so only the output port would be considered in the simulation. At the output port, a 300-μm-long tapered waveguide was used to convert the optical mode of a 3-μm-wide waveguide into a 12-μm-wide waveguide, on which the grating coupler was etched to diffract light out. A single-mode optical fiber which had a core radius of 4.5 μm and a cladding radius of 62.5 μm was used to collect the diffracted light. Light would scatter upwards and couple to the optical fiber by the Si grating structure effectively when the wavelength satisfied the phase-matching condition. Lumerical FDTD method with perfectly matched layer (PML) boundary conditions was used in the simulation and optimization. The FDTD method was an effective and accurate method for solving the Maxwell’s equations in complex geometries by discrete it both in space and in time. The simulation was divided into two parts; one for the light propagation from the 3-μm-wide waveguide to the 12-μm-wide waveguide through a taper structure, and the other for the coupling out-of-plane case from the 12-μm-wide waveguide through a grating to the fiber. The simulation showed that the optical loss from the 3-μm-wide waveguide to the 12-μm-wide waveguide by the 300-μm-long taper was very small (below 0.02 dB) and the propagation could be considered as an adiabatic process, and so the simulation could be simplified as an optimization of the coupling out-of-plane case.

 

Fig. 1 Schematic diagram of the proposed grating couplers (a) top view (b) cross section view.

Download Full Size | PDF

The optimization of coupling was performed at a fiber angle of 8° and a filling factor of 0.5 to avoid the high second-order reflection and obtain a maximum coupling strength at wavelength 1550 nm [15]. The CEs for different periods and etch depths are shown in Fig. 2. In Fig. 2, the CEs decreased with the increasing of grating period when etch depth varied from 10 nm to 40 nm; when the etch depth was 50 nm, the CEs clustered and achieved maximums. This might be caused by the interference and the diffraction effects of grating, which related to grating period, and etch depth. The grating period and etch depth influenced the angle of outgoing beams and the electric field intensity, which decided the outgoing electric field distribution. A bigger deviation from the designed angle (8°) and lower electric field intensity resulted in a lower CE. Each etch depth related to an optimized period which made the CE maximum, and the optimized periods varied with etch depths. For example, when the etch depth was 20 nm and 50 nm, the optimized period was 840 nm and 880 nm, respectively. The maximum period deviations shown in Fig. 2 for etch depth of 20 nm and 50 nm were 80 nm and 40 nm, respectively. A smaller deviation of period resulted in a smaller decease in the optimized CE, and so the CEs clustered much closer when etch depth was 50 nm. When the period and etch depth fulfilled the phase-matching condition, the CE would reach the optimized value. For this structure, the optimized CE (about −5 dB at wavelength 1550 nm) was at the period of 880 nm and the etch depth between 44 nm to 50 nm. Etch depth of 50 nm which equaled to the thickness of Si was selected in consideration of keeping a high CE and reducing the fabrication complexity.

 

Fig. 2 Simulated CE of the Si grating couplers on Si-LNOI. (The thickness of Si thin film was 50 nm, the filling factor was 0.5, fiber angle θ was 8°, and the wavelength was 1550 nm).

Download Full Size | PDF

3. Experiment result

3.1 Fabrication and measurement

The detailed fabrication process is shown in Fig. 3. The LN thin film had a thickness of 292 nm, and the SiO₂ layer had a thickness of 1.95 μm. First, a positive photoresist mask was patterned via photolithography. Second, a thin film of Si was deposited by magnetic sputtering technology. The thickness of this Si thin films was 50 nm measured by a step profiler. Third, a lift-off process was taken to remove the photoresist, and so Si-strip-loaded LN waveguides were formed. FIB was used to etch gratings on the Si strip waveguide, and the beam current of FIB was 230 pA. A grating with 18 teeth and 19 grooves was formed. The size of the grating zone was about 12-μm-wide and 17.5-μm-long. The Si strip was fully etched.

 

Fig. 3 Fabrication procedure of the Si-strip-loaded LNOI.

Download Full Size | PDF

Figure 4(a) is the optical microscope photograph of the sample before the etching of gratings. A 12-μm-wide Si waveguide, a Si tapered waveguide, a 3-μm-wide Si waveguide and the LN thin film were indicated. The FIB photography of Si grating with and without Au layer on the surface was shown in Fig. 4(b) and (c), respectively. The surface roughness in Fig. 4(b) was mainly caused by the uneven Au layer which was deposited on the sample surface to dissipate the charges during the FIB etching. The measured period (about 870 nm) and filling factor (about 0.6) which were deviated from the optimized values (880 nm, 0.5) would introduce about 2 dB declination of CE at wavelength 1550 nm and about 30 nm peak wavelength shift.

 

Fig. 4 Si-strip-loaded waveguides and grating. (a) Optical microscope photograph of the Si- strip-loaded waveguide (b) FIB photograph of grating with and (c) without Au deposited.

Download Full Size | PDF

The performance of the coupler was evaluated by the experimental setup of the coupling measurement, which had been described in [13]. The coupling system included three parts, as shown schematically in Fig. 1(a): a standard single-mode fiber which tilted above the input grating as the light input source, grating couplers and waveguides on Si-LNOI, and another standard single-mode fiber which tilted above the output grating as the output light collector. A tunable laser source (Santec TSL-210) with a wavelength range of 1420–1580 nm was used as the light source. An InGaAs detector was connected to the output fiber to measure the transmitted power of the fiber-grating-waveguide-grating-fiber system. The normalized transmitted power T was related to the CE of the input grating, the propagation efficiency η of the waveguide, and the CE of the output grating. Due to the consistence of the two ports, the input CE was considered the same as the output CE. When the dB unit was used, the following equation was fulfilled: T = 2CE + η. T could be obtained directly by the optical measurement, and η could be deduced by the optical measurements of waveguides with different length.

3.2 Results and discussions

To determine the CE of the grating coupler and the propagation loss of Si-strip-loaded waveguide, three groups of photonic structures containing grating couplers and waveguides with different size were etched in Si strips. The photonic structures are schematically shown in Fig. 5(a). Structure A had a waveguide length of 0.5 mm and a width of 12 μm. Structure B had a waveguide length of 1 mm and a width of 12 μm. Structure C contained two 300-μm-long tapered waveguides, two 100-μm-long and 12 μm-wide strip waveguides, and a 200-μm-long and 3-μm-wide strip waveguide. The identical gratings were etched in both ends of each structure (A, B, C). In Fig. 5(b), the measured T of A, B, C is shown. The black, blue and red lines were the measured T of A, B, and C, respectively. It could be seen that the three lines had a similar trend but a decreased T from A to C. At 1550 nm wavelength, the T was −38 dB, −40.4 dB, and −41.4 dB, respectively. The maximum T was at 1500 nm wavelength. Comparing the T of A and B, the measured CE of each grating coupler and propagation loss of the waveguide could be determined. The measured CE for one grating coupler was −18 dB at wavelength 1550 nm, and the propagation loss of the 12-μm-wide Si-strip-loaded waveguide was about 48 dB/cm. The optical absorption and scattering caused by the large number of defects, such as dislocations, vacancies, stacking faults, and the inactive impurities might be responsible for the propagation loss of the Si-strip-LNOI waveguide [16]. Comparing the T of B and C, the extra propagation loss caused by Si tapered waveguides and 3-μm-wide strip waveguide could also been determined, which was about 1 dB.

 

Fig. 5 Schematic and normalized transmission T of Si-strip-loaded waveguide grating couplers. Structure A had a waveguide length of 0.5 mm and a width of 12 μm. Structure B had a waveguide length of 1 mm and a width of 12 μm. Structure C contained two 300-μm-long tapered waveguides, two 100-μm-long and 12-μm-wide strip waveguides, and a 200-μm-long and 3-μm-wide strip waveguide. The identical gratings (red strips) were etched in both ends of each waveguide (A, B, C).

Download Full Size | PDF

The measured CE of Si grating coupler is shown in Fig. 6. The measured CE was −18 dB at wavelength 1550 nm, and the maximum CE was −13 dB at wavelength 1500 nm. The simulated CE according to the fabricated grating structure (Fig. 4) which had a period of 870 nm and a filling factor of 0.6 is shown in Fig. 6. It could be seen that the simulated CE was about −7 dB and −6 dB at wavelength 1550 nm and 1500 nm, respectively; the shape of the curve was similar to the measured one, but CE was different. The discrepancy of CE might be from the setting of refractive index of Si. In the simulation, the refractive index of Si was 3.48, but the refractive index of Si thin film was estimated to be 2.7 at wavelength 1539 nm by prism coupling measurement. The simulated CE of grating coupler with a period of 870 nm, a filling factor of 0.6 and a Si refractive index of 2.7 is also shown in Fig. 6. The simulated CE was −13 dB at wavelength 1550 nm, and the maximum CE was −9 dB at 1500 nm, which were closer to the measurement values.

 

Fig. 6 Simulated and measured CEs of Si grating coupler on Si-LNOI. (The thickness of Si thin film was 50 nm, the thickness of LN thin film was 292 nm, the period was 870 nm, the filling factor was 0.6, and the fiber angle θ was 8°)

Download Full Size | PDF

4. Conclusion

In conclusion, Si grating couplers on LNOI waveguides were fabricated and characterized. The Si layer was deposited on LNOI by magnetron sputtering technology, and Si-strip-loaded LNOI waveguides were fabricated by lift-off process. The Si-strip-loaded LNOI waveguide was then etched by FIB to fully etch the Si to form the grating structure. CE of −18 dB was measured at wavelength 1550 nm. In addition, the propagation loss of the 12-μm-wide waveguide and the tapered waveguide were also measured and estimated. Si grating coupler which was fabricated by etching the upper Si thin film on LNOI would provide testing and characterization flexibilities to the Si-LNOI devices.