We demonstrate aluminum nitride (AlN) on sapphire as a novel platform for integrated optics. High-confinement AlN microring resonators are realized by adopting a partially etched (pedestal) waveguide to relax the required etching selectivity for exact pattern transfer. A wide taper is employed at the chip end facets to ensure a low fiber-to-chip coupling loss of ~2.8 dB/facet for both transverse-electric (TE) and transverse-magnetic (TM) modes. Furthermore, the intrinsic quality factors (Qint) recorded with a high-resolution linewidth measurement are up to ~2.5 and 1.9 million at telecom band for fundamental TE00 and TM00 modes, corresponding to a low intracavity propagation loss of ~0.14 and 0.2 dB/cm as well as high resonant buildup of 473 and 327, respectively. Such high-Q AlN-on-sapphire microresonators are believed to be very promising for on-chip nonlinear optics.
© 2017 Optical Society of America
Integrated high quality factor (Q) microresonators with compact size and superb stability as well as pronounced resonant buildup are appealing for a myriad of nonlinear optics applications, such as on-chip Kerr frequency comb generation . Despite the significant success of silicon photonics in implementing complex optical functionalities , the relatively narrow bandgap (~1.1 eV) of silicon makes it susceptible to two-photon absorption and thereby ineligible for high-power and high-Q operations at near-infrared (NIR) wavelength . To overcome this challenge, alternative wide bandgap materials, such as silicon nitride (Si3N4) , aluminum nitride (AlN) [5, 6] and diamond , have proved to be promising candidates for integrated microresonator-based nonlinear photonic devices, wherein a large resonant buildup is crucial to reducing the required pump levels. For this purpose, high-Q microresonators with a small footprint are preferred, which can be realized by adopting a thick waveguide core to enable high optical confinement so as to mitigate the scattering loss at the clad-core interfaces . However, due to the high internal stress in Si3N4 film prepared by low pressure chemical vapor deposition (LPCVD), it normally requires a complex fabrication process to avoid crack formation [8, 9]. Meanwhile, the Q factors of sputtered AlN microring resonators are limited [5, 6], which may be a consequence of optical scattering and absorption at the grain interfaces.
On the other hand, epitaxial AlN is expected to exhibit superior optical properties thanks to the improved crystalline quality and reduced grain size . With a large bandgap (~6.2 eV at 300 K) as well as significant quadratic and cubic susceptibilities, AlN is attractive for broadband nonlinear optics . In addition, AlN grown on sapphire (nsapphire = ~1.75 at 1.55 μm) naturally forms a waveguide structure for guiding lights. To realize epitaxial AlN-based low-loss waveguides, it is essential to minimize the scattering and defect-related absorption loss at AlN/sapphire interface due to lattice mismatch . High-confinement waveguide with negligible mode intensity at clad-core interfaces should be an effective solution to this issue. Currently, high-quality AlN films with desired thickness are accessible via epitaxial growth technique, such as metal-organic vapor phase epitaxy (MOVPE)  and molecular beam epitaxy (MBE) . However, the high bond energy of AlN normally results in a relatively low dry etching rate , which hampers exact nano-scale pattern transfer into thick AlN film.
In this paper, we report the realization of high-confinement AlN-on-sapphire microring resonators with the intrinsic Q factor (Qint) up to ~2.5 million. A partially etched (pedestal) waveguide is adopted to relax the required etching selectivity for exact pattern transfer, which also enhances the waveguide-to-microring coupling strength and thereby expands the desired gap size for critical coupling. Meanwhile, a wide taper is employed at the chip end facets to permit a low fiber-to-chip coupling loss for both transverse-magnetic (TM) and transverse-electric (TE) modes. In addition, detailed design, fabrication, and characterization processes of the high-Q AlN microrings are presented.
2. AlN film preparation and characterization
In our experiment, a 1.2-μm-thick AlN was prepared to allow a high confinement for the fabricated microrings. The AlN was grown on a c-plane (0001) sapphire by a homemade low-pressure MOVPE system with trimethylaluminum (TMA) and ammonia (NH3) as the precursors . A 30-nm-thick low-temperature (LT) AlN buffer layer was first deposited in a continuous gas flow mode at 600°C with a V/III ratio of about 3000. Then, 1.2 μm high-temperature (HT) AlN was formed with a V/III ratio of 1000 upon elevating the temperature to 1200°C.
The X-ray diffraction (XRD) rocking curve measurement reveals a full-width at half-maximum (FWHM) linewidth of ~47 and 552 arcsec along  and  orientations, respectively, indicating an excellent crystalline quality of our AlN sample. This is also confirmed by the narrow linewidth (3.5 cm−1) of E2 (high) phonon in AlN backscattering Raman spectrum. The surface morphology is assessed by atomic force microscopy (AFM), exhibiting a low root-mean-square (RMS) roughness of ~0.33 nm over a 10 × 10 μm2 area. The refractive index measured with an ellipsometer is ~2.1 at 1.55 μm.
3. Device design and fabrication
3.1. Device design
A dimension with the width of 3.5 μm and the radius of 100 μm is adopted for our microring resonator, so as to minimize the mode overlap with the waveguide sidewall. To facilitate the pattern transfer into 1.2 μm AlN, a pedestal structure with 0.4 μm unetched AlN layer at the waveguide bottom is employed. Meanwhile, to minimize the undesirable coupling into higher-order modes within the multi-mode microring resonator, the width of the bus waveguide is optimized to ensure superior phase matching with fundamental modes . Figure 1(a) illustrates the calculated coupling Q factors (QC) for different modes within the microring based on the below expression :Fig. 1(a), three sets of transverse mode families can be identified for our adopted device structure (outer radius: 100 μm; cross section of 3.5 × 1.2 μm2) by finite element method (FEM) simulations. Nevertheless, it is noted that the coupling strength for second-order TE20 and TM20 modes is significantly reduced with increased bus waveguide width. When the bus waveguide width is increased to 1.47 μm, the coupling into fundamental TE00 and TM00 modes exceeds that to first-order TE10 and TM10 modes, along with negligible coupling into TE20 and TM20 modes. Such an optimized waveguide width not only reduces the number of excited higher-order modes within the multi-mode microring, but also ensures single-mode operation in the bus waveguide. The QC values for completely and partially etched waveguide are illustrated in Fig. 1(b). It is evident that the pedestal structure leads to an enhanced waveguide-to-microring coupling [16, 17], which in turn expands the gap size for critical coupling (QC = Qint), thereby easing the fabrication process as well as reducing the excess parasitic loss .
In general, higher-order modes inside the microring are more susceptible to the sidewall roughness (i.e., increased scattering loss) and experience a relatively large radiation loss (particularly in our pedestal structure) due to waveguide bending, thereby exhibiting a relatively lower Qint. Consequently, by adopting the optimized bus waveguide width, the value QC/Qint for fundamental modes is always smaller than that for higher-order modes. This property is particularly helpful in distinguishing fundamental modes from higher-order modes within the multi-mode microring. As depicted in Fig. 1(c), if both fundamental and high-order modes are under-coupled, i.e. QC/Qint > 1, a smaller QC/Qint value for fundamental modes indicates a higher on-resonance extinction ration (ER). On the other hand, when both fundamental and higher-order modes are over-coupled, i.e. QC/Qint < 1, the fundamental modes would have a lower ER. Additionally, when the two sets of modes are in different coupling regimes, i.e. an over-coupled fundamental mode and an under-coupled higher-order mode, they can be readily resolved from the microring phase response.
To improve fiber-to-chip coupling, the frequently used inverse nanotapers in silicon photonics  are not suitable in our case. As shown in the inset of Fig. 1(d), the TE mode field is localized in the pedestal AlN layer when the taper width is reduced to 0.5 μm, and the resultant mode profile is not large enough for efficient fiber coupling. In contrast, when a wide taper of 4 μm is employed, a better mode overlap can be ensured with our lensed fibers of 3.5 μm mode field diameter. As shown in Fig. 1(d), the coupling loss for both TE00 and TM00 modes is as low as ~2.2 dB/facet when the bus waveguide is laterally tapered to 4 μm. Meanwhile, special care is paid to the selection of taper width so as to avoid undesirable fundamental to higher-order mode conversion in the tapering process .
3.2. Device fabrication
For AlN microring fabrication, a SiNx mask is deposited on the AlN-on-sapphire wafer by plasma enhanced chemical vapor deposition (PECVD), and the microring (outer radius: 100 μm; cross section: 3.5 × 1.2 μm2) as well as the associated bus waveguide are subsequently defined by an electron beam lithography (EBL) system (JEOL JBX-9300FS) with ZEP520A resist. Since the wafer is electrical insulating, a 10 nm Ti is sputtered on top of the e-beam resist prior to EBL writing to avoid any surface charge-up effects and then removed afterwards. The pattern is first transferred to SiNx mask by reactive ion etching (RIE) with a selectivity of ~1:1. Then, dry etching of AlN is carried out with Cl2/BCl3/Ar-based chemistries in an inductively coupled plasma (ICP) system (Sentech SI 500). During the etching process, Cl2/BCl3/Ar flow rate, ICP power, bias voltage and chamber pressure are 40/8/5 sccm, 500 W, −200 V and 0.3 Pa, respectively. These parameters are optimized to ensure a relatively high etching rate of ~211 nm/min as well as a smooth etched sidewall, as identified by the scanning electron microscopy (SEM) image in Fig. 2(a). Meanwhile, a reasonably high selectivity of ~3:1 between AlN and SiNx is recorded. Following the etching process, the integrated microring resonators are embedded in 3 μm PECVD SiO2 for protection, and no additional annealing treatment is performed.
To facilitate the cleavage of AlN-on-sapphire devices, the bus waveguide is defined perpendicular to the primary flat a-plane of sapphire substrate, which is parallel to the natural cleavage facet m-plane of AlN . Figure 2(b) shows the cross section of the input and output waveguides cleaved with a dicing machine after thinning the sapphire down to 150 μm and subsequent scribing with a high-power pulsed ultra-violet (UV) laser . It is noted that wrinkle-like features in the cleaved facet can be clearly identified, which are attributed to the slight misalignment between the crystallographic orientations of AlN and sapphire . Meanwhile, a pedestal waveguide with 418 nm unetched AlN layer at the ridge bottom is fabricated to balance the required etching selectivity among the e-beam resist, SiNx mask and 1.2 μm AlN for exact EBL pattern transfer. As a by-product, a reduced sidewall scattering loss induced by dry etching is expected, since the AlN film is partially etched.
4. Device characterization
4.1. Experimental setup
To characterize the performance of fabricated devices, a pump-probe experimental setup is adopted, as schematically illustrated in Fig. 3. For transmission spectrum measurement, an external cavity tunable laser source (TLS, Santec TSL-510) is employed as the light source, and lensed fibers are used to couple light into and out of the chip. Meanwhile, the polarization is selected by a fiber polarization controller (FPC), while the output is monitored by an optical power meter.
To determine the high Q factors, an additional optical path (dotted-line encircled part) is introduced into Fig. 3 for accurate resonance linewidth measurements. By fine tuning the oscillating wavelength of Santec TSL-510 (TLS1) through piezoelectric-varied cavity length, a scanning step below 0.1 pm can be ensured. Then, the frequency is calibrated by beating the light from TLS1 with another tunable laser source (TLS2) at a high-speed photodetector (PD), which is connected to an electrical spectrum analyzer (ESA). Meanwhile, an optical spectrum analyzer (OSA) is used to monitor the relative wavelength of TLS1 and TLS2 (not shown in Fig. 3). To avoid thermal effect induced distortion of resonance peaks in the high-Q microrings, a 5/95 coupler is used to allow only 5% power entering the chip. The 3-dB linewidth of the beat note signal is measured to be ~2.5 MHz, indicating the capability to resolve ultra-high Q factors up to ~70 million at 1.55 μm.
4.2. Results discussion
Figures 4(a) and 4(b) depict the transmission spectra for TE- and TM-polarized incident lights, respectively. The off-resonance insertion loss is as low as ~5.6 dB, which includes the reflection loss at the waveguide facets, thus in good accordance with the calculated coupling loss of ~2.2 dB/facet in Fig. 1(d). It is noted that only two sets of TE and TM mode families are excited in the microring. The absence of second-order TE20 and TM20 modes is due to the optimized bus waveguide width, as illustrated Fig. 1(a). The microring phase response measurement  reveals that the excited modes are slightly under-coupled around 1.55 μm. Thus, the fundamental modes should exhibit higher on-resonance ERs as analyzed in Section 3.1. The recorded ERs for TE00 and TM00 modes are above 20 dB for the gap size of 0.7 and 0.6 μm, respectively, indicating nearly-critical coupling.
Figures 4(c) and 4(d) illustrate the resonance linewidth measurements. It is obvious that the resonance lineshape deviates from the symmetric Lorentzian shape as a result of Fabry-Pérot interference induced by the waveguide end facet reflection. Thus, a Fano-like fit is employed for accurate linewidth extraction . The FWHM linewidth (Δ fFWHM) for TE00 and TM00 modes is measured to be ~135 and 180 MHz, corresponding to a loaded Q factor (QL) of ~1.4 and 1 million, respectively. As the microring is slightly under-coupled from the phase response measurement, Qint and QC can then be determined by the following formulas: and 1/QC = 1/QL − 1/Qint, where T0 represents the normalized on-resonance transmission value . The extracted Qint is ~2.5 and 1.9 million for TE00 and TM00 modes, respectively, and their intracavity propagation loss (αring) is estimated to be ~0.14 and 0.2 dB/cm based on the equation: αring = 10 · log10e · f0/(Qint · FSR · Rring) (f0, FSR and Rring being the resonance frequency, free spectrum range and radius of the microring, respectively) . To our knowledge, the achieved Qint in this work is 3 times higher than the best value (Qint = 0.8 million) reported for sputtered AlN microrings . It should be mentioned that the extracted QC values for TE00 and TM00 modes are in good agreement with our simulation results in Fig. 1(b) at gap sizes of 0.7 and 0.6 μm, respectively. To confirm our analysis in Section 3.1, the Q factors for TE10 and TM10 modes are also recorded in our experiment (e.g., Qint = 1.3 million for TM10 mode in Fig. 4(b) around 1552.634 nm), which indeed exhibits a relatively lower Qint than that for fundamental modes. Additionally, the intracavity power buildup within the microring, which is of great significance for nonlinear optics applications, is determined by :Eq. (2), compact microrings with a small radius (i.e. a large FSR) and critical coupling are favorable for obtaining a high resonant buildup. For our fabricated AlN-on-sapphire microring, the estimated intracavity power buildup is ~473 and 327 for TE00 and TM00 modes, respectively, making it promising to enable on-chip nonlinear optical process.
The simulated TE00 and TM00 modal profiles are depicted in the inserts of Figs. 4(c) and 4(d), in which localized fields with high confinement are clearly identified. As a result, the Q factor of the AlN microring is expected to be less susceptible to the lossy SiO2 cladding. This is confirmed by the fact that high-Q performance is demonstrated without any annealing treatment after the device fabrication. To unveil the possibility of further improvement in Q factors, the gap-filling property of PECVD SiO2 is investigated. As illustrated in Fig. 5(a), a large void occurs at the waveguide coupling region when the gap size is 0.56 μm, which results in excess loss due to mode mismatch around the coupling section. When the gap size is increased to ~1.1 μm, an improved gap filling is observed in Fig. 5(b). However, expanding the gap size would lead to a less efficient waveguide-to-microring coupling. To circumvent this issue, a promising approach is to clad the device with conformal LPCVD SiO2 followed by PECVD to fill the remaining part . In addition, spin-on glass such as hydrogen silsesquioxane (HSQ) can serve as a low-cost alternative, since it exhibits superior gap-filling capability as well as sufficient thickness, and is in analogous to SiO2 after an optimized curing process .
High-performance AlN-on-sapphire microring resonators are demonstrated with an optimized design and fabrication process, featuring low insertion loss and high Q factor for both TE and TM modes, respectively. A low intracavity propagation loss of 0.14 dB/cm and a high power buildup of 473 are recorded, which, to our knowledge, is the best result for AlN-based microrings so far. With the mature epitaxial growth techniques, high quality AlN-on-sapphire film is accessible, and can be an attractive platform for on-chip nonlinear optical process, such as Raman lasing  and Kerr frequency comb formation . Additionally, thanks to its large bandgap and excellent crystalline quality, AlN-on-sapphire is also appealing for integrated optics at UV regime .
This work was supported in part by National Basic Research Program of China (2014CB340002); National Natural Science Foundation of China (61210014, 61621064, 61574082, 51561165012); High Technology Research and Development Program of China (2015AA017101); Tsinghua University Initiative Scientific Research Program (20131089364, 20161080068, 20161080062); and Open Fund of State Key Laboratory on Integrated Optoelectronics (IOSKL2014KF09).
References and links
1. J. S. Levy, A. Gondarenko, M. A. Foster, A. C. T.-Foster, A. L. Gaeta, and M. Lipson, “CMOS-compatible multiple-wavelength oscillator for on-chip optical interconnects,”Nat. Photonics 4,37–40 (2010). [CrossRef]
2. B. Jalali and S. Fathpour, “Silicon photonics,”J. Lightwave Technol. 24(12),4600–4615 (2006). [CrossRef]
4. J. F. Bauters, M. J. R. Heck, D. John, D. X. Dai, M.-C. Tien, J. S. Barton, A. Leinse, R. G. Heideman, D. J. Blumenthal, and J. E. Bowers, “Ultra-low-loss high-aspect-ratio Si3N4 waveguides,”Opt. Express 19(4),3163–3174 (2011). [CrossRef] [PubMed]
5. W. H. P. Pernice, C. Xiong, and H. X. Tang, “High Q micro-ring resonators fabricated from polycrystalline aluminum nitride films for near infrared and visible photonics,”Opt. Express 20(11),12261–12269 (2012). [CrossRef] [PubMed]
7. B. J. M. Hausmann, I. B. Bulu, P. B. Deotare, M. McCutcheon, V. Venkataraman, M. L. Markham, D. J. Twitchen, and M. Lončar, “Integrated high-quality factor optical resonators in diamond,”Nano Lett. 13,1898–1902 (2013). [CrossRef] [PubMed]
9. M. H. P. Pfeiffer, A. Kordts, V. Brasch, M. Zervas, M. Geiselmann, J. D. Jost, and T. J. Kippenberg, “Photonic Damascene process for integrated high-Q microresonator based nonlinear photonics,”Optica 3(1),20–25 (2016). [CrossRef]
10. A. Soltani, A. Stolz, J. Charrier, M. Mattalah, J.-C. Gerbedoen, H. A. Barkad, V. Mortet, M. Rousseau, N. Bourzgui, A. BenMoussa, and J.-C. De Jaeger, “Dispersion properties and low infrared optical losses in epitaxial AlN on sapphire substrate in the visible and infrared range,”J. Appl. Phys. 115,163515 (2014). [CrossRef]
11. H. Jung, R. Stoll, X. Guo, D. Fischer, and H. X. Tang, “Green, red, and IR frequency comb line generation from single IR pump in AlN microring resonator,”Optica 1(6),396–399 (2014). [CrossRef]
12. A. W. Bruch, C. Xiong, B. Leung, M. Poot, J. Han, and H. X. Tang, “Broadband nanophotonic waveguides and resonators based on epitaxial GaN thin films,”Appl. Phys. Lett. 107,141113 (2015). [CrossRef]
13. J. C. Yan, J. X. Wang, P. P. Cong, L. L. Sun, N. X. Liu, Z. Liu, C. Zhao, and J. M. Li, “Improved performance of UV-LED by p-AlGaN with graded composition,”Phys. Status Solidi C 8(2),461–463 (2011). [CrossRef]
14. R. Fan, Z. B. Hao, C. Zhang, J. N. Hu, and Y. Luo, “High quality A1N with a thin interlayer grown on a sapphire substrate by plasma-assisted molecular beam epitaxy,”Chin. Phys. Lett. 27(6),068101 (2010). [CrossRef]
15. X. W. Liu, C. Z. Sun, B. Xiong, L. Niu, Z. B. Hao, Y. J. Han, and Y. Luo, “Smooth etching of epitaxially grown AlN film by Cl2/BCl3/Ar-based inductively coupled plasma,”Vacuum 116,158–162 (2015). [CrossRef]
16. E. S. Hosseini, S. Yegnanarayanan, A. H. Atabaki, M. Soltani, and A. Adibi, “High quality planar silicon nitride microdisk resonators for integrated photonics in the visible wavelength range,”Opt. Express 17(17),14543–14551 (2009). [CrossRef] [PubMed]
17. X. W. Liu, C. Z. Sun, B. Xiong, J. Wang, L. Wang, Y. J. Han, Z. B. Hao, H. T. Li, Y. Luo, J. C. Yan, T. B. Wei, Y. Zhang, and J. X. Wang, “Broadband tunable microwave photonic phase shifter with low RF power variation in a high-Q AlN microring,”Opt. Lett. 41(15),3599–3602 (2016). [CrossRef] [PubMed]
18. D. T. Spencer, J. F. Bauters, M. J. R. Heck, and J. E. Bowers, “Integrated waveguide coupled Si3N4 resonators in the ultrahigh-Q regime,”Optica 1(3),153–157 (2014). [CrossRef]
19. G. H. Ren, S. W. Chen, Y. P. Cheng, and Y. Zhai, “Study on inverse taper based mode transformer for low loss coupling between silicon wire waveguide and lensed fiber,”Opt. Commun. 284,4782–4788 (2011). [CrossRef]
21. K. Dovidenko, S. Oktyabrsky, J. Narayan, and M. Razeghi, “Aluminum nitride films on different orientations of sapphire and silicon,”J. Appl. Phys. 79(5),2439–2445 (1996). [CrossRef]
22. J.-R. V. Look, S. Einfeldt, O. Krüger, V. Hoffmann, A. Knauer, M. Weyers, P. Vogt, and M. Kneissl, “Laser scribing for facet fabrication of InGaN MQW diode lasers on sapphire substrates,”IEEE Photonics Technol. Lett. 22(6),416–418 (2010). [CrossRef]
24. M. Soltani, “Novel integrated silicon nanophotonic structures using ultra-high Q resonators,” Ph. D. thesis,Georgia Institute of Technology (2009).
25. C. W. Holzwarth, T. Barwicz, and H. I. Smith, “Optimization of hydrogen silsesquioxane for photonic applications,”J. Vac. Sci. Technol. B 25(6),2658–2661 (2007). [CrossRef]
26. X. W. Liu, C. Z. Sun, B. Xiong, J. Wang, L. Wang, Y. J. Han, Z. B. Hao, H. T. Li, Y. Luo, J. C. Yan, T. B. Wei, Y. Zhang, and J. X. Wang, “Low-threshold chip-scale aluminum nitride Raman laser,” in Proceedings of ISLC (2016), paperThD1.
27. X. W. Liu, C. Z. Sun, B. Xiong, J. Wang, L. Wang, Y. J. Han, Z. B. Hao, H. T. Li, Y. Luo, J. C. Yan, T. B. Wei, Y. Zhang, and J. X. Wang, “Broadband frequency comb generation in aluminum nitride microring resonators,” in Proceedings of ECOC (2016), pp.746–748.
28. T. Troha, M. Rigler, D. Alden, I. Bryan, W. Guo, R. Kirste, S. Mita, M. D. Gerhold, R. Collazo, Z. Sitar, and M. Zgonik, “UV second harmonic generation in AlN waveguides with modal phase matching,”Opt. Mater. Express 6(6),2014–2023 (2016). [CrossRef]