Lithium niobate (LN), as a key material in the photonic industry, exhibits a strong electro-optic (EO) effect, a large nonlinear optical coefficient, and chemical stability. Traditional LN waveguides are usually prepared through titanium (Ti) diffusion or proton exchange. Ti diffusion exchange causes an increase of 0.001-0.04 in the refractive index, which depends on Ti density, diffusion time, and temperature. Proton exchange can give rise to a change in the refractive index of up to 0.12. These waveguides have the disadvantage of low refractive index contrast between waveguide and cladding, which leads to weak optical constraints, a large mode area, and a millimeter-level bending radius. This is the major limitation of conventional bulk LN for broad applications with a large amount and a small size of LN chips. Therefore, it is necessary to develop a method with a low-cost wafer and high-index-contrast waveguide, which is also the main objective of this work.

This study mainly presents the theory and simulation. The LN material is chemically inert, and therefore, it can hardly be etched without any pre-treatment. In addition, mechanical processing of the LN material is also difficult to proceed due to its high hardness and wear resistance. It is worth noting that during the proton exchange reaction, ions diffuse into the LN substrate to exchange ions, which leads to the occurrence of phase transition and structural defects, and thus, the proton exchange region can be easily etched. Therefore, we present a process scheme combining proton exchange technology with etching technology. The LN substrate is submerged in the proton source at a high temperature for a long time to assure that the depth of the proton exchange should be equal to the required height of the waveguide at least. After the proton exchange, samples are etched by methods such as wet etching and plasma etching, and the waveguide is retained. The feasibility of the process is verified by simulations. Moreover, the proton exchange depth, etching width, and sidewall angle are changed to optimize the waveguide width of the bulk LN platform.

The experimental result demonstrates that the waveguide prepared with 1% diluted melt at 300 ℃ for 72 h can increase n_e by 0.08 at 1550 nm. We initially set the Δne to 0.08. The effective mode area under different diffusion depths is shown in Fig. 2. The minimum area is about 14.5 μm² when the diffusion depth is 2.4 μm. The etched waveguide shows its superiority in reducing width (Fig. 3), and the effective mode area in the waveguide can be reduced to 6.7 μm² by etching. Although the waveguide is decreased to 4.2 μm, it is still large compared to that of other material platforms. We can further reduce the waveguide size by increasing Δne. Δne of LN after proton exchange is set to 0.1 at 1550 nm according to theoretical research. With the growth of the diffusion depth, the effective mode area also changes (Fig. 4). The minimum effective mode area is about 7 μm² when the diffusion depth is 2 μm. To verify the advantage of the fabrication process mentioned above, we set the height of waveguides to 2.8 μm. Under the same temperature, the proton exchange depth we set can be realized by proton exchange time expansion. Considering the non-vertical sidewalls produced by the etching of LN, the angle of the tilted sidewall is 85°. When the width of the waveguide is 2.4 μm, the waveguide has the strongest ability to confine the light field (Fig. 6). Compared with the case of an unetched waveguide, increased etching can effectively reduce the width of the waveguide. In addition, as many parameters can be optimized, we use particle swarm optimization (PSO) to design the waveguide size reversely. A set of these parameters that can realize the strongest confinement of the optical field is selected with the assistance of PSO. Here, the figure of merit (FOM) is defined as the effective mode area. The optimization is conducted via Lumerical Mode Solutions. The simulations show that the lowest waveguide width is 2.5 μm, and the bending radius is reduced to the level of hundreds of microns, which greatly lessens the size of bulk LN devices. Wavelength-division multiplexing (WDM) devices are one of the key components for optical communications. As a planar waveguide component based on optical integration technology, an arrayed waveguide grating (AWG) has the advantages of high integration and low loss compared with the traditional dielectric filter. Hence, it is widely used in the optical interconnection of data centers. On the basis of the proposed process scheme, an AWG with a center wavelength of 1550 nm, four channels, and channel spacing of 400 GHz is designed. The footprint of the device is 850 μm×620 μm, and the transmission loss of the AWG is about 6 dB, as shown in Fig. 10. Since the waveguide spacing is set to be larger than 4 μm to avoid the crosstalk from the adjacent waveguides, light cannot enter the arrayed waveguides from the input coupler completely, which causes loss (Fig. 10). The crosstalk between adjacent channels is all lower than 22 dB, which further verifies the feasibility of this scheme.

In this paper, a method for fabricating AWGs over the bulk LN substrate is presented. The design of a four-channel CWDM AWG is investigated with a reduced cost due to the use of bulk LN, which proves that this process can reduce the size of the optical devices on the bulk LN platform.