1. Introduction

Lithium niobate thin film (lithium niobate on insulator, LNOI) can be a useful optical integration platform due to its high refractive index contrast between the lithium niobate thin film and the buried oxide layer, with excellent nonlinear and electro-optical properties comparable to those of bulk material [1–3]. Recently, many high-quality optical devices on LNOI are successfully demonstrated, such as low-loss waveguides [4,5], electro-optic modulators [6], second harmonic generation [7], and micro-ring/disk resonators [8–13]. They are fabricated on LNOIs by various methods, such as proton exchange [4], strip-loaded [14], femtosecond laser micromachining [12], and dry or wet etching [5‒12]. In most cases UV lithography/electron beam lithography (EBL) with following dry etching (Ar + or reactive ion etching) are used for the definition and fabrication of the structures [5‒9]. The propagating loss of waveguide on LNOI fabricated by proton exchange, EBL followed by dry etching, and strip loading is reported to be 0.2 dB/cm, 0.4 dB/cm, and 5.8 dB/cm, respectively [4,5,14]. The optical quality factor featured by micro-disk resonators are reported to be 10⁵~10⁶,[8-10] and the half-wave electro-optic efficiency featured by modulator is reported to be 1.8 V·cm and date rates up to 40 Gbps [6].

The coupling between LNOI integrated optical devices and standard single-mode fibers is an important and challenge topic due to the large mode mismatch between LNOI devices and single mode fibers. The widely-used method is end-fire coupling, where the beam is focused into the edge of a waveguide [15]. To improve the mode mismatch between the fiber and the LNOI devices, many structures are used, such as lensed optical fibers [5,7], tapered optical fibers [8-12], tapered waveguides or inverted taper waveguides. These structures are either to shrink the fiber mode sizes or to expand the waveguide mode sizes for mode matching. The constraint of the above approaches is the high requirement to the alignment tolerance and/or facet polishing. It is meaningful to have a surface coupler that could be placed anywhere on a chip rather than at the edge.

Grating couplers fabricated on a waveguide can couple light in or out between a fiber and the waveguide. The grating coupler has the advantage that it can be placed anywhere on a chip and allows wafer-scale testing of optical integrated system. When the light is coupled from a waveguide to a fiber, the light will not only radiates upwards to the fiber, but also radiates downwards into the substrate. The ratio between the power coupled upwards and downwards (up/down ratio) which is important on the coupling efficiency can be further improved. There are several methods to increase the up/down ratio. First, the use of an upper cladding layer on the thin film to increase the transmittance and reduce the reflectivity between the thin film and the upper cladding layer [16]. The most common upper material is silicon dioxide. Second, the thickness of buried oxide layer (D) below thin film can be carefully chosen. The power radiated upwards/downwards is a periodical function of D due to the interference effect, and so the upwards radiated power can be enhanced by a proper choice of D [17]. Third, the use of a bottom reflector can improve the coupling efficiency. A bottom reflector can reflect most power radiated towards substrate, thus it can improve the coupling efficiency obviously [18].

Recently, grating couplers on LNOI were successfully demonstrated with coupling efficiencies of approximately −4 dB (in simulation) and −12 dB, or −10 dB (in experiments) [19,20]. The gratings were fabricated either with an upper cladding layer or with a more appropriate buried oxide layer, but there is still some power lost to the substrate. To further improve the coupling efficiency, a bottom reflector can be added in the LNOI.

In this paper, grating couplers with a metal bottom reflector in Z-cut LNOI were designed, fabricated and characterized. By simulations, the optimized grating parameters with a metal bottom reflector were obtained, and the maximum coupling efficiency was about −2 dB. Experimentally, grating structures were etched in LNOI fabricated by ion implantation and bonding technologies [21-23]. Compared with the grating couplers demonstrated [19,20], a different fabrication method called FIB, which was an advanced micro/nano processing technology for film deposition and maskless etching, allowing a flexible, direct writing of structures with a nanometer resolution [24], was used to form the waveguide and gratings. A maximum coupling efficiency of −9.1 dB and −6.9 dB for grating without and with a metal bottom reflector was measured, respectively. Fabrication error sensitivity of etch depth was experimentally investigated and the discrepancy in the simulation and experiment was also discussed.

2. Design of the coupler

The simulation model is schematically shown in Fig. 1. From bottom to top, layers were substrate, metal layer (10 nm Cr/100 nm Au/30 nm Cr), silicon oxide layer, LN, air, and optical fiber, respectively. The thickness of LN (H) and buried oxide layer used in simulations was 480 nm and 2.23 μm, respectively. Two-dimensional (2D) FDTD simulation method with perfectly matched layer (PML) boundary conditions was applied to the designing and optimization of the grating couplers. The transverse electric mode (TE) source was injected in and transmitted along the waveguide, the fiber was fixed and tilted at a small angle (θ ≈8°) with respect to the direction normal to the waveguide surface to avoid a high second-order reflection at the waveguide-grating interface. Because coupling efficiencies in the input case (coupling the light from the fiber to the waveguide) and the output case (coupling the light from the waveguide to the fiber) were the same, only the output case would be simulated.

 

Fig. 1 Schematic structure of the proposed grating coupler.

Download Full Size | PDF

The grating period, the etch depth and the filling factor were optimized to achieve the best coupling efficiency. For a uniform grating, it was reported that the coupling efficiency would reached the maximum when filling factor was 0.5 [25], which was consistent with our optimized result. The simulated maximum coupling efficiency at 1550 nm was achieved with a grating period of 960 nm, an etch depth of 285 nm, and a filling factor of 0.5. The coupling efficiency at 1550 nm was about −2 dB. When there was no reflector, the coupling efficiency was −3 dB at 1550 nm with more than 30% power lost to the substrate. The diffracted field radiated by grating is shown in Fig. 2. From Fig. 2(a), some power was lost into the substrate, but when a metal bottom reflector was added, the power transmitted to the substrate was reflected by the metal bottom reflector (Fig. 2(b)), which resulted in an increase in the up/down ratios, thus an increase in the coupling efficiency.

 

Fig. 2 E Field plot of the LNOI grating coupler (a) without and (b) with a metal bottom reflector in Z-cut LNOI for TE polarization. H = 480 nm, D = 2.23 μm, period = 960 nm, etch depth = 285 nm, filling factor = 0.5.

Download Full Size | PDF

3. Experiment and simulation results

3.1 Fabrication

The fabrication process of LNOI wafer with a metal bottom reflector is schematically shown in Fig. 3(a). A Z-cut LiNbO₃ wafer A with a diameter of 3 inch was implanted by He-ions with energy of 250 keV and a dose of 4 × 10¹⁶ ions/cm². The He-ions stayed in a certain depth which determined by the ion energy and the ion dose, formed an amorphous layer approximately 860 nm in LiNbO₃ substrate. This process could form an amorphous layer with a certain depth in LiNbO₃ substrate. Another Z-cut LiNbO₃ wafer B was deposited with a 10 nm Cr/100 nm Au/30 nm Cr metal layer and an oxide layer by electron beam evaporation and plasma enhanced chemical vapor deposition (PECVD), respectively. In the PECVD deposition process of SiO₂ layer, the reactive gases were SiH₄ (400 sccm, 2.5%) and N₂O (700 sccm). The deposition was at a temperature of 300°C and a pressure of 10³ mtorr. Wafer B was annealed to drive off the gases trapped in the oxide layer then sent to a chemical mechanical polishing (CMP) process. This CMP process could reduce the surface roughness and improve the surface flatness to satisfy the subsequently direct wafer bonding. Wafer A and wafer B were directly bonded at room temperature and then heated up to 220°C for 8 h to improve the bonding strength, and the wafer A would split along the He stayed amorphous layer simultaneously, formed a LNOI wafer with a bottom reflector. The wafer was then annealed at 400°C for 2 h to further the bonding strength, and a CMP process was used to reduce the surface roughness of the LN thin film. The LNOI wafer fabrication was performed at the research center of Nanoln. The fabrication process of LNOI wafer without a metal bottom reflector was the same as above, except for lacking of the metal deposition. The photograph of a LNOI wafer with a metal bottom reflector is shown in Fig. 3(b).

 

Fig. 3 Fabrication procedure (a) and paragraph (b) of a 3-inch LNOI wafer with a metal bottom reflector.

Download Full Size | PDF

The LN thin film with (without) a metal reflector had a thickness of 480 nm (480 nm), and the SiO₂ layer had a thickness of 2.23 μm (1.96 μm). FIB was used to shape the waveguide and gratings (Fig. 4(a)). A 150 nm-thickness gold was sputtered on the LNOI chip by sputtering coating. The gold layer sputtered was to dissipate the charges during the FIB etching. Two 3 μm-width strips with a distance of 12 μm were etched to form the strip waveguide. The FIB beam current in waveguide and grating etching were approximately 900 PA and 280 PA, respectively. The fine structure could be formed by a smaller beam current. Finally, the chip was soaked in the gold corrosion liquid to remove the residual gold on the surface. The scanning view of grating is shown in Fig. 4(b). The waveguide was formed between the etched two 3 μm-width strips and the gating was located on both ends of waveguides. The grating had 17 teeth and 18 grooves, each teeth and groove were 12 μm in length. An enlarged figure of the dotted box region is display in Fig. 4(c). The interface between the silicon dioxide layer and LN could be seen clearly from the enlarged figure and the partially-etched grooves and clear teeth boundaries could also be seen.

 

Fig. 4 Grating coupler in Z-cut LNOI for TE polarization. (a) Fabrication procedure (b) top view of coupler fabricated by FIB (c) the enlarged figure of the dotted box region in (b).

Download Full Size | PDF

3.2 Measurement

The experimental setup of the coupling measurement is shown in Fig. 5. Two single mode fibers were used to couple the power in and out from the gratings. The fibers were mounted on two precision 3-axis translation stages, and were adjusted close to the grating facet observed by an optical microscope. The translation stages were adjusted by monitoring the output power at 1550 nm. During the coupling efficiency versus wavelength measurement, the air gap between the fiber facet and the grating was kept constant, and the angles between the fibers axis and the vertical axis were fixed at 8 degrees by two optical fiber clamps. A tunable laser source (Santec TSL‒210) with a range of 1500 nm‒1600 nm was connected to a single mode fiber as the input port, and a paddle fiber polarization controller was used in the input fiber to select a transverse electric (TE) mode which would be injected to the input grating. The output fiber was fixed with one end on the output grating and the other end connected to an optical power meter to collect the transmitted power T. The ridge waveguide had a length of 400 μm and the loss was ignored. The coupling efficiency of one coupler could be calculated from the measured transmitted power T, as the input and output coupling parameters were the same and they would yield the same coupling efficiency: η² = T, So $\text{η}=\sqrt{\text{T}}$ [12].

 

Fig. 5 Optical setup for the transmission measurement of the grating couplers.

Download Full Size | PDF

3.3 Results and discussions

In the experiment, grating couplers with a 200-μm-long waveguide were fabricated and tested, and there was no obvious measured coupling efficiency difference with the couplers which had a 400-μm-long waveguide, and so the propagation loss of the waveguide was neglected. For couplers on LNOI without a bottom reflector, the maximum transmission T measured by the power meter was approximately −18.2 dB (1.5%) at 1562 nm with an etch depth of 285 nm. Thus, the maximum coupling efficiency η for one coupler was −9.1 dB (12.3%). For couplers on LNOI with a metal reflector, the maximum coupling efficiency was −6.9 dB (20.6%) at 1550 nm with an etch depth of 285 nm. The results are shown in Fig. 6.

 

Fig. 6 Measured and simulated coupling efficiencies for different etch depths in Z-cut LNOI for TE polarization. For gratings in LNOI with a metal bottom reflector, parameters are: H = 480 nm, D = 2.23 μm, period = 960 nm, etch depth = 285 nm, filling factor = 0.5; For grating in LNOI without a metal bottom reflector, parameters are: H = 480 nm, D = 1.96 μm, period = 960 nm, etch depth = 285 nm, filling factor = 0.5.

Download Full Size | PDF

In Fig. 6, the measured coupling efficiencies for different etch depths (245 nm and 265 nm) are also shown. For a deviation of 20 nm from the optimal etch depth (285 nm), the maximum coupling efficiency decreased by approximately 0.6 dB, and the peak wavelength shifted. This phenomenon was consistent with the simulations, which were shown in Fig. 6. In addition, the coupling performances when etch depths were bigger than 285 nm were also studied by simulation. When the etch depth was chosen to 305 nm, the coupling efficiency decreased by 0.3 dB at 1550 nm, and when the etch depth was 325 nm, the coupling efficiency decreased by 1.2 dB at 1550 nm. Etch depth played an important role on the interference conditions, thus had an important influence on the up/down ratio. Compared the simulation and experimental results, the lines (coupling efficiency with respect to wavelength) had similar shapes, but they had discrepancies in the absolute values. These discrepancies were attributed to fabrication and setup errors, such as period, duty cycle and fiber tilted angle, which had considerable influences on the coupling efficiency. In the FIB fabrication process, the fabrication error such as instability of the emission current might result in some deviations in the designed grating parameters. For example, a 30 nm deviation in the grating periodicity would result in about 2 dB decrease in the maximum coupling efficiency. Furthermore, there were some deviations in the measurement setup, such as the fiber angle and position. Finally, the propagation loss in the waveguide would result in a decreased measurement coupling efficiency.

4. Conclusion

In conclusions, fiber-to-chip grating couplers on LNOI with a metal bottom reflector were designed, fabricated, and characterized. Firstly, LNOI wafers were fabricated by ion implantation and bonding technologies, which could produce single-crystalline LiNbO₃ thin films; Secondly, waveguide and gratings were structured on LNOIs by FIB, which could simplify the grating fabrication process for saving the processing of mask and lithograph. Lastly, coupling efficiency with respect to wavelength from 1500 nm to 1600 nm was measured. A maximum coupling efficiency of −9.1 dB and −6.9 dB for grating without and with a metal bottom reflector was obtained, respectively. In addition, the fabrication error sensitivity of etch depth was experimentally investigated, and for a deviation of 20 nm from the optimal etch depth, the decreased coupling efficiency was approximately 0.6 dB. To further improve the coupling efficiency, some structures were under consideration, such as chirped and apodized grating coupler [26], and photonic crystals which might be a promising solution with precisely controlled fabrication tolerance [27-29].