Open Access
Add to BibSonomy
Abstract
Thin-film lithium niobate (TFLN) modulators are expected to be an ideal solution to achieve a super-wide modulation bandwidth needed by the next-generation optical communication system. To improve the performance, especially to reduce the driving voltage, we have carried out a detailed design of the TFLN push-pull modulator by calculating 2D maps of the optical losses and Vπ for different ridge waveguide depths and electrode gaps. Afterwards the modulator with travelling wave electrodes was fabricated through i-line photolithography and then characterized. The measured Vπ for a modulator with 5-mm modulation arm length is 3.5 V, corresponding to voltage-length product of 1.75 V·cm, which is the lowest among similar modulators as far as we know. And the measured electro-optic response has a 3-dB bandwidth beyond 40 GHz, which is the limitation of our measurement capability. The detailed design, fabrication and measurement results are presented.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Electro-optic modulators have always been key devices in modern optical communication systems [1,2]. Materials such as silicon, indium phosphide, and conventional bulk lithium niobate are now widely used to make commercial modulators and each of them has their own pros and cons [3–6]. However, none of them can meet all the requirements for next-generation optical communication systems simultaneously. These requirements include but not limited to: ultra-high electro-optical modulation bandwidth up to 60 GHz or more, low insertion loss, low ${\textrm{V}_\mathrm{\pi }}$, temperature insensitive, low cost, etc. Because of the limitation on traditional waveguide fabrication techniques, conventional bulk lithium niobate modulators cannot take full advantages of material’s large electro-optic coefficient [7]. As a newly proposed platform for modulators, thin film lithium niobate (TFLN) modulators have shown super performance as soon as they appear and have demonstrated great potential to be an ideal solution for the next-generation high-speed optical communication systems [8]. A very high bandwidth (> 40 GHz) and a very low ${\textrm{V}_\mathrm{\pi }}$ (1.4 V) have been achieved simultaneously [9]. In their work, the 5-mm long (In the following the length always means the modulator arm length) device has a high bandwidth of more than 100 GHz and a relatively high ${\textrm{V}_\mathrm{\pi }}$ of 4.4 V while the 20-mm long device has a relatively low bandwidth of 40 GHz and a low ${\textrm{V}_\mathrm{\pi }}$ of 1.4 V [9]. Voltage-length product is a very important metric to evaluate the performance of a Mach-Zehnder (MZ) modulator. A lower voltage-length product means shorter device and potentially higher modulation bandwidth to achieve a certain ${\textrm{V}_\mathrm{\pi }}$. As a relatively new modulator, the TFLN modulator still needs to minimize its voltage-length product to further improve its performance. E-beam lithography (EBL) is widely used to define the waveguide in similar works on TFLN modulators [9–11]. Although the resolution of photolithography is lower compared to that of EBL, the cost and time for exposure are much lower especially for devices with large size like the TFLN modulator [12–15].
In our design, different metal gap spacing and etching depth of the waveguide is simulated in order to optimize the Vπ·L and the optical loss simultaneously. In this work we demonstrate a low Vπ TFLN modulator fabricated with the standard i-line photolithography. The Vπ·L has reached 1.75 V·cm with a modulation bandwidth beyond 40 GHz. This paper is structured as follows: section 2 introduces the detailed design and simulation of the modulator; section 3 introduces the processing steps and the measurement result of the fabricated modulator; finally, a conclusion is given in section 4.
2. Design and simulation
2.1 Layout of the modulator
Figure 1 shows the microscope image of the fabricated TFLN modulator. From this image, we can see that the optical part of the modulator is a typical Mach-Zehnder interferometer (MZI) with waveguides and splitters based on multi-mode interferometers (MMI). For the optical input and output, cleaved facets are formed, and the waveguide width is tapered from 1.5 μm to 10 μm for better coupling to a lensed fiber of 5-µm waist diameter. The length of the taper part is 200 µm. The electrical part of the modulator includes a traveling wave electrode (TWE) for modulation and ground-signal-ground (GSG) pads for the high-speed microwave probes (GGB) to input and output radio frequency (RF) electrical signals [16]. The length of the modulator arm is set to be 5 mm and the size of the chip is 6 mm × 1.5 mm.
Figure 2 shows the cross-sectional structure of the phase shifter area. The wafer consists of a layer of x-cut thin-film lithium niobate bonded on a SiO2 layer that grown on the high resistance silicon substrate. The thickness of the LN layer is 600 nm, the buried SiO2 layer is 2-μm thick and the cladding SiO2 layer is 800-nm thick. There is a natural push-pull effect on modulation using x-cut lithium niobate, because the refractive index change depends not only on strength but also on direction of the electric field, and in this case the refractive index changes of the two arms are just opposite to each other [17]. An i-line stepper instead of e-beam lithography (EBL) is used to define the waveguide. So, the upper width of the waveguide is set to be 1.5 μm to reduce the fabrication difficulty. According to our etching process, the sloping sidewall of the waveguide is about 73 degrees from horizontal. The etching depth and the gap between the signal and ground of the electrode are optimized in section 2.2.
2.2 Simulation, optimization and calculation
Using the COMSOL Multiphysics, we simulate the optical field of the transverse electric (TE) mode and the electrostatic field of the GSG electrode, as shown in Fig. 3 [18]. We can see how the electrical field of electrode acts on the lithium niobate, the refractive index of lithium niobate changes as a consequence of the electrical field. As shown in Fig. 4, from the electrical field (E) just in the lithium niobate, due to the large difference on dielectric constants (ε) of lithium niobate and silica, there is large attenuation on electrical field at the boundary between LN and silica. This is because the E×ε value is continuous at the boundary. So the electrical field mostly affects the LN core through the slab because of the large difference between dielectric constants of lithium niobate and silica. It’s intuitional that the optical loss caused by electrode absorption would be smaller with deeper etching depth of the ridge waveguide and wider gap between the ground and signal of the electrode, however the voltage-length product would be larger in these two cases as well. Therefore, there needs to make a trade-off between decreasing the optical loss and increasing the modulation efficiency.
Fig. 3. (a) The optical field of the transverse electric (TE) mode. (b) The electrostatic field of the GSG electrode.
To get an appropriate value for the gap and the etching depth, the optical loss and the ${\textrm{V}_\mathrm{\pi }}$ need to be calculated for a wide range of combinations of these two values. When a 1-V voltage applied on the electrode, the simulated electrostatic field is shown in Fig. 3(b). The change of refractive index $\Delta n$ can be expressed as
When the phase change equals $\mathrm{\pi}$, the voltage applied is the half-wave voltage ${\textrm{V}_\mathrm{\pi}}$. So, the voltage-length products can be expressed as
$\Gamma $ can be calculated in COMSOL by calculating integral of the electrical field and the optical field according to Eq. (2). By using Eq. (4) we can calculate the voltage-length products of the modulator versus different gap and etching depth. Furthermore, the optical loss caused by the electrode absorption can be calculated from the imaginary part of the effective index obtained by COMSOL.
To get a map of both the ${\textrm{V}_\mathrm{\pi}}\textrm{L}$ and the optical loss for different gaps and etching depths, we simulated modulators of different structures with gaps from 3 μm to 8 μm and etching depths from 50 nm to 550 nm, as shown in Fig. 5(a). Figure 5(b) shows the changes of the ${\textrm{V}_\mathrm{\pi}}$, from which we can see that the ${\textrm{V}_\mathrm{\pi}}\textrm{L}$ gets bigger as gaps and etching depth get bigger. This is because the electric field declines when gaps get bigger. And the mode confinement becomes stronger when etching depths get bigger, making the optical mode smaller and less affected by the electric field. These factors result in a bigger ${\textrm{V}_\mathrm{\pi}}\textrm{L}$. As can be seen from Fig. 5(c), the optical loss is just the opposite situation of the ${\textrm{V}_\mathrm{\pi}}\textrm{L}$. So, there is a trade-off between the low ${\textrm{V}_\mathrm{\pi}}\textrm{L}$ and the low optical loss thus we draw a 2-dimensional contour map of both the ${\textrm{V}_\mathrm{\pi}}$ and the optical loss for different gaps and etching depths, as shown in Fig. 5(d). From Fig. 5(d) we can see that there could be different points which can generate similar performance on ${\textrm{V}_\mathrm{\pi}}\textrm{L}$ and low optical loss. Generally we prefer etching less LN due to the concern on the exposed sidewall of the waveguide. There is an interface between LN and silica and is orthogonal to the Z-direction, which generates an effect similar to the Z-cut LN modulator which has the well-known DC bias drift issue [19]. With reference to this contour map, we have chosen the structure of our devices which has a 5-μm gap and a 0.3-μm etching depth, as shown in Fig. 5(d). The waveguide of this structure is single-mode. From the 2D map shown in Fig. 5(d) we know that the voltage-length product of our design is expected to be about 1.8 V·cm and the optical loss caused by metal absorption is expected to be less than 0.1 dB/cm.
The RF properties of the coplanar waveguide (CPW) electrode was simulated through HFSS [20]. By changing the thickness of the upper cladding SiO2 layer and the width of the signal metal, we finally get the characteristic impedance of 47.5 Ω, which is closed to the standard 50 Ω, and the RF phase index matched to the calculated optical group index of 2.27, as shown in Fig. 6(a). The electrical S-parameters are also simulated with the S21 and S11 parameters shown in Fig. 6(b). From the electrical S11 parameter we can see that the reflection of the microwave electrical signal is very small under our designed characteristic impedance. In the interaction region, the width of the signal electrode is 16.8 µm. In the probing region, the width of the signal electrode is 50 µm and the electrode gap is 28 µm. The length of the electrode taper from prob to arm is 50 µm. We can also estimate the EO-response of the modulator through the electrical S21 parameter when the microwave phase index is perfectly matched to the optical group index. In an electro-optic modulator, the final 3-dB electro-optical modulation bandwidth will be reached at the frequency where the driving signal decreases by 6.41 dB [21]. So, the reflection of the RF signal is expected to be low and the transmission of the RF signal is expected to be high in the range of 0 to 60 GHz.
Fig. 5. Variables of the modulator structure (a); 2-dimensional colormap plot of the ${\textrm{V}_\mathrm{\pi}}$ (b) and optical loss (c) versus different gaps and etching depths; (d) 2-dimensional contour map of both ${\textrm{V}_\mathrm{\pi}}$ and optical loss versus different gaps and etching depths, when the arm length is 5 mm.
Fig. 6. (a) The simulated microwave phase index and characteristic impedance of the modulator. (b) The simulated S-parameters of the modulator.
The theoretical bandwidth of the electro-optic response can be calculated by:
3. Fabrication and measurement
3.1 Device fabrication
The fabrication process is schematically shown in Fig. 7. An i-line stepper was used for all the patterning. First, a 200-nm-thick Cr was evaporated on LN by electron beam evaporation (EBE). Then a 2-μm-thick SPR photoresist was used to define the waveguide pattern. Inductively coupled plasma (ICP) etcher was used to transfer the photoresist pattern to the Cr mask. After removing the photoresist, the Cr mask was used to etch the LN layer with an optimized Ar-based ICP dry etching process. ICP power of 500 W and RIE power of 90 W were applied in the dry etching process, while the argon flow of 60 sccm and the chamber pressure of 3.6 mTorr was set. The etching rate is 50 nm/s. Figure 8 shows the scanning electron microscope (SEM) image of the etched waveguide’s cross-sectional view and top view. Then the Cr mask was removed by dechroming solution and an 800-nm-thick silica was deposited on the waveguide by plasma enhanced chemical vapor deposition (PECVD). The silica was etched by ICP dry etching except for where the metal crosses the waveguide before the electrode pattern was defined on a PMGI and SPR photoresist. This guarantees that the metal would not cause additional loss. Then a 900-nm-thick Ti/Au was evaporated by electron beam evaporation (EBE) and lifted off to form the electrode. Then another 800-nm-thick silica is deposited on the wafer. Silica on the pad was removed so that the RF probe can contact the electrode. The wafer was finally thinned and cleaved for test.
3.2 Measurement of the device
The measurement set-up is shown in Fig. 9. A continuous wave laser is used to input a light of 1550 nm wavelength through a lensed fiber and through a polarization controller to ensure TE polarization input into the modulator. After modulated by the modulator, the output light is collected by another lensed fiber and amplified by an EDFA before detected by a high-speed photodetector. A vector network analyzer (VNA) was used to provide the RF modulation signal and receive the RF signal from the high-speed photodetector to measure the EO-response.
A 0.7 dB/cm propagation loss was measured through a ring resonator structure of 100-μm radius without metal [22]. Figure 10 shows the structure and the normalized transmission curve. The propagation loss is mainly caused by the roughness of the shallow-etched sidewall. Through fitting to it the waveguide loss was extracted. Then a light of 2-dBm optical power was input to the modulator and to the adjacent accompanying straight waveguides, which is a test waveguide with the same length as the device and near the device. An optical power of -6 to -8 dBm was detected at the output port of the adjacent accompanying straight waveguides, which correspond to a coupling loss of 3.8 dB/facet to 4.8 dB/facet. Finally, an optical power of ∼-6 dBm was detected at one output port of the modulator when its power was maximized through adjusting the phase, which means that the additional loss caused by metal was negligible. The coupling loss is thus estimated to be about 3.8 dB/facet to 4.8 dB/facet. Under DC bias we can get the voltage power curve of the modulator shown in Fig. 11(a) from which we can see that the ${\textrm{V}_\mathrm{\pi}}$ is 3.5 V, corresponding to a voltage-length product of 1.75 V·cm. And the extinction ratio is measured to be 17 dB. Under the RF signal, the TWE output is connected to a 50-Ω terminator and the measured EO-response is shown in Fig. 11(b). The 3-dB bandwidth of the EO-response of our modulator is more than 40 GHz. We can see from the trend of the curve that the 3-dB bandwidth is actually far higher than the range of our measurement system which is 40 GHz.
Fig. 11. (a) The voltage power curve of the modulator under DC bias. (b) The measured and calculated EO-response curve.
3.3 Comparison of TFLN modulators in the literature
Table 1 shows recent works on TFLN MZ modulators. They have similar structures and working principle. Among all these modulators, our modulators fabricated through the standard i-line photolithography has shown the lowest voltage-length products. While the 3-dB EO-response bandwidth of our 5-mm-long device is more than 40 GHz, which is also reasonably good. Therefore, our modulator has achieved a good balance between modulation voltage and bandwidth.
Table 1. Performance comparison of TFLN modulators
4. Conclusion
In summary, we report the design optimization of a shallowly etched TFLN modulator. A 0.3-μm-depth rib on 0.3-μm-thickness slab and a 5-μm-width gap between signal and ground are chosen through the 2D contour maps calculated by us. Then we fabricated the modulator through standard photolithography and ICP dry etching process. Finally, we measured the modulator with the half-wave voltage measured to be 3.5 V, corresponding to a voltage-length product of 1.75 V·cm. And the 3-dB EO-response bandwidth is measured to be more than 40 GHz. The waveguide propagation loss was measured to be 0.7 dB/cm. All these measured results match well with the calculated results, proving that the simulation has a good guidance on improving the modulator performance. Compared to other works on TFLN modulators, our results are leading-edge especially the voltage-length product. What’s more, our fabrication with photolithography is fast to fabricate, which can reduce the cost of the device. All these will help reducing the cost of ultra-high-performance commercial modulators in the future.
Funding
National Key Research and Development Program of China (2019YFB2203304); National Natural Science Foundation of China (61861136001).
Disclosures
The authors declare no conflicts of interest.
References
1. D. A. B. Miller, “Attojoule Optoelectronics for Low-Energy Information Processing and Communications,” J. Lightwave Technol. 35(3), 346–396 (2017). [CrossRef]
2. G. T. Reed, G. Mashanovich, F. Y. Gardes, and D. J. Thomson, “Silicon optical modulators,” Nat. Photonics 4(8), 518–526 (2010). [CrossRef]
3. S. S. Azadeh, F. Merget, S. Romero-Garcia, A. Moscoso-Martir, N. von den Driesch, J. Muller, S. Mantl, D. Buca, and J. Witzens, “Low V(pi) Silicon photonics modulators with highly linear epitaxially grown phase shifters,” Opt. Express 23(18), 23526–23550 (2015). [CrossRef]
4. Y. Ogiso, J. Ozaki, Y. Ueda, N. Kashio, N. Kikuchi, E. Yamada, H. Tanobe, S. Kanazawa, H. Yamazaki, Y. Ohiso, T. Fujii, and M. Kohtoku, “Over 67 GHz Bandwidth and 1.5 V Vπ InP-Based Optical IQ Modulator With n-i-p-n Heterostructure,” J. Lightwave Technol. 35(8), 1450–1455 (2017). [CrossRef]
5. S. Wolf, H. Zwickel, C. Kieninger, M. Lauermann, W. Hartmann, Y. Kutuvantavida, W. Freude, S. Randel, and C. Koos, “Coherent modulation up to 100 GBd 16QAM using silicon-organic hybrid (SOH) devices,” Opt. Express 26(1), 220–232 (2018). [CrossRef]
6. “100G/400G LN Modulator.”, retrieved https://www.fujitsu.com/jp/group/foc/en/products/optical-devices/100gln/.
7. A. Boes, B. Corcoran, L. Chang, J. Bowers, and A. Mitchell, “Status and Potential of Lithium Niobate on Insulator (LNOI) for Photonic Integrated Circuits,” Laser & Photonics Reviews 12(2018).
8. G. Poberaj, H. Hu, W. Sohler, and P. Günter, “Lithium niobate on insulator (LNOI) for micro-photonic devices,” Laser Photonics Rev. 6(4), 488–503 (2012). [CrossRef]
9. C. Wang, M. Zhang, X. Chen, M. Bertrand, A. Shams-Ansari, S. Chandrasekhar, P. Winzer, and M. Loncar, “Integrated lithium niobate electro-optic modulators operating at CMOS-compatible voltages,” Nature 562(7725), 101–104 (2018). [CrossRef]
10. M. Xu, W. Chen, M. He, X. Wen, Z. Ruan, L. Chen, L. Liu, S. Yu, and X. Cai, “Lithium Niobate Michelson Interferometer Modulator on Silicon-On-Insulator Platform,” arXiv e-prints, arXiv:1905.01623 (2019).
11. M. He, M. Xu, Y. Ren, J. Jian, Z. Ruan, Y. Xu, S. Gao, S. Sun, X. Wen, L. Zhou, L. Liu, C. Guo, H. Chen, S. Yu, L. Liu, and X. Cai, “High-performance hybrid silicon and lithium niobate Mach–Zehnder modulators for 100 Gbit s−1 and beyond,” Nat. Photonics 13(5), 359–364 (2019). [CrossRef]
12. K. Luke, P. Kharel, C. Reimer, L. He, M. Loncar, and M. Zhang, “Wafer-scale low-loss lithium niobate photonic integrated circuits,” Opt. Express 28(17), 24452–24458 (2020). [CrossRef]
13. P. O. Weigel, J. Zhao, K. Fang, H. Al-Rubaye, D. Trotter, D. Hood, J. Mudrick, C. Dallo, A. T. Pomerene, A. L. Starbuck, C. T. DeRose, A. L. Lentine, G. Rebeiz, and S. Mookherjea, “Bonded thin film lithium niobate modulator on a silicon photonics platform exceeding 100 GHz 3-dB electrical modulation bandwidth,” Opt. Express 26(18), 23728–23739 (2018). [CrossRef]
14. A. J. Mercante, P. Yao, S. Shi, G. Schneider, J. Murakowski, and D. W. Prather, “110 GHz CMOS compatible thin film LiNbO3 modulator on silicon,” Opt. Express 24(14), 15590–15595 (2016). [CrossRef]
15. S. Jin, L. Xu, H. Zhang, and Y. Li, “LiNbO3Thin-Film Modulators Using Silicon Nitride Surface Ridge Waveguides,” IEEE Photonics Technol. Lett. 28(7), 736–739 (2016). [CrossRef]
16. “GGB Industries, Inc.”, retrieved http://www.ggb.com/.
17. L. Arizmendi, “Photonic applications of lithium niobate crystals,” Phys. Stat. Sol. (a) 201(2), 253–283 (2004). [CrossRef]
18. “COMSOL Multiphysics”, retrieved https://www.comsol.com/.
19. M. Xu, M. He, H. Zhang, J. Jian, Y. Pan, X. Liu, L. Chen, X. Meng, H. Chen, Z. Li, X. Xiao, S. Yu, S. Yu, and X. Cai, “High-performance coherent optical modulators based on thin-film lithium niobate platform,” Nat. Commun. 11(1), 3911 (2020). [CrossRef]
20. “Ansys HFSS”, retrieved https://www.ansys.com/products/electronics/ansys-hfss.
21. J. H. den Besten, “Integration of multiwavelength lasers with fast electro-optical modulators,” (2004).
22. L.-W. Luo, G. S. Wiederhecker, J. Cardenas, C. Poitras, and M. Lipson, “High quality factor etchless silicon photonic ring resonators,” Opt. Express 19(7), 6284–6289 (2011). [CrossRef]
23. C. Wang, M. Zhang, B. Stern, M. Lipson, and M. Loncar, “Nanophotonic lithium niobate electro-optic modulators,” Opt. Express 26(2), 1547–1555 (2018). [CrossRef]
24. A. Rao, A. Patil, P. Rabiei, A. Honardoost, R. DeSalvo, A. Paolella, and S. Fathpour, “High-performance and linear thin-film lithium niobate Mach-Zehnder modulators on silicon up to 50 GHz,” Opt. Lett. 41(24), 5700–5703 (2016). [CrossRef]
25. A. N. R. Ahmed, S. Shi, A. Mercante, S. Nelan, P. Yao, and D. W. Prather, “High-efficiency lithium niobate modulator for K band operation,” APL Photonics 5(9), 091302 (2020). [CrossRef]
