Thin-film lithium niobate (TFLN) with electro-optic effects has been investigated as an optical modulator for high-speed optical communications.1^–16 Although such optical modulators use infrared laser light, the material features of lithium niobate (LN) can allow the propagation of visible light; therefore, TFLN has potential applications in visible light applications, such as laser beam scanning (LBS)17^,18 for higher speed, which achieves higher resolution and lower power consumption than conventional direct laser modulation.

By increasing the color-change speed of a red, green, and blue (RGB)-coupled laser beam, a higher-resolution image can be obtained by combining it with a faster-speed microelectromechanical system mirror for LBS. Furthermore, because the laser beam output can be eliminated or enhanced by a Mach–Zehnder interferometer (MZI) in an analog manner for RGB light, the injection current to the laser diode (LD) can be assigned a constant value and decreased compared with the required bias current for direct laser modulation.

However, few studies have been reported on visible light applications for TFLN,14^–16 and the essential building blocks for visible light are required. We have successfully confirmed red (λ=638nm), green (λ=520nm), and blue (λ=473nm) visible light modulation by MZI fabricated by sputter-deposited TFLN.19^,20 Using the MZI, the strength of each color can be controlled by changing the voltage. To obtain a full-color laser beam for an active photonic integrated circuit (Active-PIC), modulated red, green, and blue laser lights must be coupled into a single beam by an RGB multiplexer. Active-PIC is a device that integrates an LN modulator and RGB multiplexer. It enables intensity modulation of each RGB color and multiplexing. In the future, Active-PIC can be voltage controlled by an LN modulator instead of LD current direct modulation; therefore, lower power consumption and higher resolution can be expected. For this purpose, a TFLN-based RGB multiplexer is essential for LBS; however, this has never been demonstrated, and the RGB multiplexer has been limited to glass-based PICs.21^–26

In this study, we fabricated an RGB multiplexer using TFLN and successfully confirmed its operation. This study aimed to demonstrate the operation of the RGB multiplexer, which is a passive device, as a first step to make the Active-PIC. Furthermore, TFLN was fabricated by sputter deposition, whereas conventionally, bulk-LN adhesion to the substrate is used. Sputter-deposited TFLN has a significant advantage in mass production.19^,20

To fabricate an RGB multiplexer using TFLN, we selected a multimode interferometer (MMI). MMIs have been widely used in filters, splitters, sensors, and couplers for infrared laser light waveguides because of their small size, low loss, wide bandwidth, and ease of fabrication.27^–29 We expand this capability to an RGB multiplexer. Because of the large refractive index difference between the LiNbO3 and the cladding, such as SiO2, the directional coupler method typically results in long device lengths on the order of several centimeters. Conversely, the MMI method can reduce the device length to the order of mm, and mature lithography techniques, such as KrF or ArF excimer lasers, can be used to shorten the device length and reduce losses. Because this study aimed to demonstrate the RGB multiplexer, unconstrained design values for the waveguide and gap widths were used in the i-line lithography process.

We designed an RGB multiplexer according to the self-imaging theory27^–29 using FIMMWAVE (Photon Design Ltd.). Figure 1 shows a schematic of the RGB multiplexer comprising two MMIs; their parameters are listed in Table 1. In this design, the red and blue beams were coupled at the first MMI multiplexer, whereas the green/red and blue beams were coupled at the second MMI multiplexer. Self-imaging theory was applied to minimize coupling losses. The refractive indices of the ordinary ray (no) in LiNbO3 are 2.38, 2.35, and 2.31 for wavelengths of 473, 520, and 638 nm, respectively. Similarly, the refractive indices of the extraordinary ray (ne) are 2.28, 2.25, and 2.21 for wavelengths of 473, 520, and 638 nm, respectively. These values were obtained from bulk-LN literature. Generally, LiNbO3 is inherently birefringent; thus, the RGB multiplexer is polarization dependent. Because sputter-deposited TFLN has a c-axis orientation perpendicular to the film plane, the design utilizes a transverse magnetic (TM) mode input for compatibility with the modulator. Silicon dioxide (SiO2) was used as the cladding layer because of its refractive index of 1.44, which is lower than that of LiNbO3. To adapt the RGB multiplexer to a visible LN modulator, we designed it with a ridge structure. The waveguide thickness (TLN) and slab thickness (Tslab) were set to 0.7 and 0.1μm, respectively. W1 and WArm are the width of the MMI and the arm connected to the MMI, respectively, both of which are the top base of the trapezoid. In each cross-section, Wbottom is the value of the bottom of the trapezoid, which is 0.2μm larger than Wtop. WGap is the distance between the two arm waveguides at the top of the trapezoid. The output efficiency was optimized by adjusting the length of the MMI waveguide. The design was optimized to minimize both the loss and length of the MMI for RGB wavelengths. Considering low losses, a tapered waveguide is introduced in the RGB multiplexer for mode matching with the single-mode waveguide.

Figure 2 shows a color map indicating the intensity distribution of the propagation mode through the RGB multiplexer when red, green, and blue lights are input. The RGB light propagates via multimode interference at the MMI, which is indicated by white dotted lines. Figure 3(a) illustrates the multiplexer loss when L2 is maintained at a certain value and L1 is varied. Similarly, Fig. 3(b) shows the multiplexer loss when L1 is fixed at a certain value and L2 is varied. Therefore, the propagation efficiency was optimized by adjusting the lengths of the two MMI multiplexers. L1 and L2 were finally set to 620 and 1680μm, respectively, and the expected losses of the RGB multiplexer were 3.8 dB (R), 2.1 dB (G), and 3.1 dB (B).

Using the design simulated in the previous section, the device was fabricated using the process described previously.19^,20 Instead of the conventional bulk-LN adhesion process, radio frequency sputter deposition was used to fabricate the RGB multiplexer. The LiNbO3 target was sputtered with Ar-ion gas, whereas O2 gas was introduced into the sputtering chamber. An Al2O3 sapphire (001) substrate was heated to 650°C during sputter deposition to obtain TFLN with good crystallinity. A ridge-shaped waveguide of the TFLN was fabricated using physical Ar-ion etching. Sputtering was used to deposit SiO2 as a cladding layer.

Figure 4 shows the measurement system for the sputter-deposited TFLN RGB multiplexer. The wavelengths of the visible light lasers for the red, green, and blue colors were 638, 520, and 473 nm, respectively. To ensure stable operation, the LDs were maintained at room temperature using a temperature controller. The laser light intensity was controlled using a current controller. The laser lights were coupled from the three-channel fiber arrays into the sample, and the output from the RGB multiplexer was observed using a 100× objective lens. The image from the objective lens was acquired on a personal computer using a neutral density filter and a camera. When the optical power intensity was measured, the objective lens was replaced with a lensed fiber, the three-channel fiber arrays were replaced using a focuser with a polarizer, and the polarization state on the input side was TM mode. The RGB multiplexer loss was the insertion loss (IL) of each RGB multiplexer minus the IL of the straight line. Thus, the coupling and reflection losses at the end faces were neglected.

Figure 5 depicts a cross-sectional image of the RGB multiplexer, which was obtained using a scanning electron microscope (SEM). The parameters measured using the cross-sectional images were equivalent to the designed values listed in Table 1. The sidewalls of the LN waveguide appear slight asymmetry. Nonetheless, its impact is small, and IL is <1% for symmetrical or asymmetrical waveguides. The asymmetry of the LN waveguide may result from the difference between the inner and outer circumference during ion milling.

The measurement system described in Sec. 3 was used to extract the output image and optical power. Figure 6 shows the laser power dependence of the injection current on the LDs for red, green, and blue colors. Thus, the intensity of each color was modulated by changing the intensity of the LD current. Figure 7 illustrates the output images of two-color and three-color multiplexing. Magenta, which is a combination of red and blue, is successfully generated, as shown in Fig. 7(a). Similarly, yellow, which is a combination of red and green, is generated, as shown in Fig. 7(b), whereas cyan, a combination of green and blue, is generated, as shown in Fig. 7(c). Thus, we successfully produced magenta, cyan, and yellow, which are the two intermediate colors of the three primary colors of light. Finally, white light was successfully obtained by combining all three primary colors of light, as shown in Fig. 7(d). Figure 8 presents a chromaticity diagram, which shows that by changing the current in each of the three LDs, all colors could be generated, not just the intermediate colors mentioned above.

Thus, the RGB multiplexer with sputter-deposited TFLN successfully generated all colors, indicating that our device performed sufficiently well as a light source for displays such as LBS. The experimentally confirmed losses of each RGB multiplexer were 4.8 dB (red), 3.7 dB (green), and 7.0 dB (blue), which will be discussed in detail in Sec. 5.

As described in Sec. 4, we successfully fabricated and demonstrated an RGB multiplexer using sputter-deposited TFLN.

The RGB multiplexer had a device length of ∼2.3mm, which was smaller than or similar to that used for a directional coupler21^–23 and MMI.24^,25

There are two major limitations in this study that could be addressed in future research. First, the study focused on sputter-deposited TFLN to realize wider circulation with large market sizes.19^,20 However, bulk-LN is expected to exhibit better crystal structure than is sputter-deposited TFLN, and this limitation must be addressed. The full-width at half maximum of the rocking curve for the LN (006) peak of the sputter-deposited TFLN used in the RGB multiplexer was 0.49 deg. This value is larger than that of the bulk-LN and requires improvement because of its possible effect on the loss of the RGB multiplexer. In fact, at blue wavelengths (λ=473nm), the loss of the RGB multiplexer fabricated using sputter-deposited TFLN was 7 to 10 dB, which is larger than the simulated value. By contrast, the loss of the RGB multiplexer fabricated using the bulk-LN was 3 to 4 dB, which was almost the same as the simulated value. Careful control of the sputtering conditions is important to achieve better crystalline TFLN and thereby improve the propagation loss and lower the device IL. Second, the etching of TFLN waveguides is crucial for achieving TFLN waveguides with smooth surfaces. At shorter wavelengths, Rayleigh scattering has a larger effect on the propagation loss owing to the sidewall roughness of the waveguide. In particular, sidewall roughness must be less than 1 nm for visible light devices such as RGB multiplexers.16 This process development may further improve the propagation loss and reduce the device IL.

Subsequently, we discussed the difference between calculated and measured RGB multiplexer losses at blue wavelength (473 nm). The calculated blue loss value (3.1 dB) assumed that the LiNbO3 material itself had no scattering and absorption losses in the device design. Actually, however, the propagation loss of the sputtered film measured with a prism coupler (Metricon model 2010/M) was 4.3dB/cm at 446 nm, suggesting that the crystallinity of the sputtered film was still poor and that scattering or absorption losses existed in this LiNbO3 material itself. Considering this in the calculation, the blue multiplexer loss slightly increased to 4.1 dB. However, this does not explain the measured blue multiplexer loss (7.0 dB). Next, we considered the effect of the sidewall roughness of the waveguide. The root mean square roughness, Rq, of the sidewall of the waveguide evaluated by atomic force microscopy was ∼6nm. This value was used to simulate the propagation loss of the waveguide by Ansys Lumerical, which resulted in a propagation loss of 22dB/cm. This value is significantly larger than that of the sputter film (4.3dB/cm). Therefore, we conclude that the roughness of the waveguide sidewalls dominantly worsens the blue multiplexer loss.

The values evaluated by the prism coupler of the sputtered film itself (FWHM=0.49deg) are 4.3dB/cm at 446 nm and 1.6dB/cm at 633 nm. However, the values, measured by the prism coupler of the bulk LN wafer, are <1dB/cm at 446 and 633 nm. This suggests that propagation loss correlates with crystallinity, and crystallinity improvement of sputtered films is necessary. However, it is possible to reduce the loss of blue light because the crystallinity of sputter-deposited TFLN can be improved by optimizing the initial sputtering process.

Regarding the device-to-device variation in optical performance, there is ∼10% of wafer-to-wafer crystallinity variation and several % of in-plane wafer distribution due to crystallinity and fabrication of the LN film. It is possible that these may have affected the loss, and improvement is necessary.

We fabricated an RGB multiplexer using TFLN and experimentally demonstrated its operation. Three different laser lights—red, green, and blue—were successfully coupled as a single laser beam by the RGB multiplexer. Thus, an active PIC was realized using TFLN, which allows high-resolution LBS with higher speed and lower power consumption compared with direct laser modulation. Furthermore, sputter-deposited TFLN is advantageous for large-volume mass production.

Atsushi Shimura is a research engineer at the Advanced Products Development Center, Technology and Intellectual Property HQ, TDK Corporation. He received his BS degree in electronics engineering in 2011 and his MS degree in 2013, both from Tohoku University, Japan, and joined TDK in 2013. He has been engaged in research and development of spintronics devices, optical modulators, and optical devices.

Biographies of the other authors are not available.