Compact titanium dioxide waveguides with high nonlinearity at telecommunication wavelengths

Dense integration of photonic integrated circuits demands waveguides simultaneously fulfilling requirements on compactness, low loss, high nonlinearity, and capabilities for mass production. In this work, titanium dioxide waveguides with a thick core of 380 nm exhibiting a compact mode size (0.43 μm) and a low loss (5.4 ± 1 dB/cm) at telecommunication wavelengths around 1550 nm have been fabricated and measured. A microring resonator having a 50 μm radius has been measured to have a loaded quality factor of 53500. Four-wave mixing experiments reveal a nonlinear parameter for the waveguides of 21-34 W m corresponding to a nonlinear index around 2.3-3.6 x 10 m/W, which results in a wavelength conversion efficiency of −36.2 dB. These performances, together with the potentially simple dispersion engineering to the fabricated waveguides by the post processes, yield a strong promise for the titanium dioxide waveguides applied in photonic integrated circuits, especially for nonlinear implementations. © 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement OCIS codes: (130.3120) Integrated optics devices; (160.3130) Integrated optics materials; (190.4380) Nonlinear optics, four-wave mixing. References and links 1. R. Nagarajan, C. H. Joyner, R. P. Schneider, J. S. Bostak, T. Butrie, A. G. Dentai, V. G. Dominic, P. W. Evans, M. Kato, M. Kauffman, D. J. H. Lambert, S. K. Mathis, A. Mathur, R. H. Miles, M. L. Mitchell, M. J. Missey, S. Murthy, A. C. Nilsson, F. H. Peters, S. C. Pennypacker, J. L. Pleumeekers, R. A. Salvatore, R. K. Schlenker, R. B. Taylor, M. F. Huan-Shang Tsai, M. F. Van Leeuwen, J. Webjorn, M. Ziari, D. Perkins, J. Singh, S. G. Grubb, M. S. Reffle, D. G. Mehuys, F. A. Kish, and D. F. Welch, “Large-scale photonic integrated circuits,” IEEE J. Sel. Top. Quantum Electron. 11(1), 50–65 (2005). 2. A. Politi, M. J. Cryan, J. G. Rarity, S. Yu, and J. L. O’Brien, “Silica-on-silicon waveguide quantum circuits,” Science 320(5876), 646–649 (2008). 3. T. J. Kippenberg, R. Holzwarth, and S. A. Diddams, “Microresonator-based optical frequency combs,” Science 332(6029), 555–559 (2011). 4. D. Culemann, A. Knuettel, and E. Voges, “Integrated optical sensor in glass for optical coherence tomography (OCT),” IEEE J. Sel. Top. Quantum Electron. 6(5), 730–734 (2000). 5. V. D. Nguyen, B. I. Akca, K. Wörhoff, R. M. de Ridder, M. Pollnau, T. G. van Leeuwen, and J. Kalkman, “Spectral domain optical coherence tomography imaging with an integrated optics spectrometer,” Opt. Lett. 36(7), 1293–1295 (2011). 6. X. Fan, I. M. White, S. I. Shopova, H. Zhu, J. D. Suter, and Y. Sun, “Sensitive optical biosensors for unlabeled targets: A review,” Anal. Chim. Acta 620(1-2), 8–26 (2008). 7. Y. Yanagase, S. Suzuki, Y. Kokubun, and S. T. Chu, “Box-like filter response and expansion of FSR by a vertically triple coupled microring resonator filter,” J. Lightwave Technol. 20(8), 1525–1529 (2002). 8. S. Chu, B. Little, W. Pan, T. Kaneko, S. Sato, and Y. Kokubun, “An eight-channel add-drop filter using vertically coupled microring resonators over a cross grid,” IEEE Photonics Technol. Lett. 11(6), 691–693 (1999). 9. B. Little, S. Chu, P. Absil, J. Hryniewicz, F. Johnson, F. Seiferth, D. Gill, V. Van, O. King, and M. Trakalo, “Very high-order microring resonator filters for WDM applications,” IEEE Photonics Technol. Lett. 16(10), 2263–2265 (2004). 10. C. Dragone, C. Edwards, and R. Kistler, “Integrated optics N*N multiplexer on silicon,” IEEE Photonics Technol. Lett. 3(10), 896–899 (1991). 11. D. Dai, J. Wang, and Y. Shi, “Silicon mode (de)multiplexer enabling high capacity photonic networks-on-chip with a single-wavelength-carrier light,” Opt. Lett. 38(9), 1422–1424 (2013). 12. V. R. Almeida, C. A. Barrios, R. R. Panepucci, and M. Lipson, “All-optical control of light on a silicon chip,” Nature 431(7012), 1081–1084 (2004). 13. R. Salem, M. Foster, A. Turner, D. Geraghty, M. Lipson, and A. Gaeta, “Signal regeneration using low-power four-wave mixing on silicon chip,” Nat. Photonics 2(1), 35–38 (2008). 14. H. Fukuda, K. Yamada, T. Shoji, M. Takahashi, T. Tsuchizawa, T. Watanabe, J. Takahashi, and S. Itabashi, Vol. 26, No. 2 | 22 Jan 2018 | OPTICS EXPRESS 1055 #312922 https://doi.org/10.1364/OE.26.001055 Journal © 2018 Received 7 Nov 2017; revised 22 Dec 2017; accepted 5 Jan 2018; published 10 Jan 2018 “Four-wave mixing in silicon wire waveguides,” Opt. Express 13(12), 4629–4637 (2005). 15. J. Levy, A. Gondarenko, M. Foster, A. Turner-Foster, A. Gaeta, and M. Lipson, “CMOS-compatible multiplewavelength oscillator for on-chip optical interconnects,” Nat. Photonics 4(1), 37–40 (2010). 16. J. Li, H. Lee, T. Chen, and K. J. Vahala, “Low-pump-power, low-phase-noise, and microwave to millimeterwave repetition rate operation in microcombs,” Phys. Rev. Lett. 109(23), 233901 (2012). 17. P. Del’Haye, A. Schliesser, O. Arcizet, T. Wilken, R. Holzwarth, and T. J. Kippenberg, “Optical frequency comb generation from a monolithic microresonator,” Nature 450(7173), 1214–1217 (2007). 18. L. Razzari, D. Duchesne, M. Ferrera, R. Morandotti, S. Chu, B. Little, and D. Moss, “CMOS-compatible integrated optical hyper-parametric oscillator,” Nat. Photonics 4(1), 41–45 (2010). 19. M. Pu, L. Ottaviano, E. Semenova, and K. Yvind, “Efficient frequency comb generation in AlGaAs-oninsulator,” Optica 3(8), 823–826 (2016). 20. J. Leuthold, C. Koos, and W. Freude, “Nonlinear silicon photonics,” Nat. Photonics 4(8), 535–544 (2010). 21. A. R. Johnson, A. S. Mayer, A. Klenner, K. Luke, E. S. Lamb, M. R. E. Lamont, C. Joshi, Y. Okawachi, F. W. Wise, M. Lipson, U. Keller, and A. L. Gaeta, “Octave-spanning coherent supercontinuum generation in a silicon nitride waveguide,” Opt. Lett. 40(21), 5117–5120 (2015). 22. Y. Okawachi, K. Saha, J. S. Levy, Y. H. Wen, M. Lipson, and A. L. Gaeta, “Octave-spanning frequency comb generation in a silicon nitride chip,” Opt. Lett. 36(17), 3398–3400 (2011). 23. D. Moss, R. Morandotti, A. Gaeta, and M. Lipson, “New CMOS-compatible platforms based on silicon nitride and Hydex for nonlinear optics,” Nat. Photonics 7(8), 597–607 (2013). 24. M. Belt, M. Davenport, J. Bowers, and D. Blumenthal, “Ultra-low-loss Ta2O5-core/SiO2-clad planar waveguides on Si substrates,” Optica 4(5), 532–536 (2017). 25. C. Lacava, S. Stankovic, A. Z. Khokhar, T. D. Bucio, F. Y. Gardes, G. T. Reed, D. J. Richardson, and P. Petropoulos, “Si-rich silicon nitride for nonlinear signal processing applications,” Sci. Rep. 7(1), 22 (2017). 26. T. Wang, D. Ng, S. Ng, Y. Toh, A. Chee, G. Chen, Q. Wang, and D. Tan, “Supercontinuum generation in bandgap engineered, back-end CMOS compatible silicon rich nitride waveguides,” Laser Photonics Rev. 9(5), 498–506 (2015). 27. J. W. Choi, G. F. Chen, D. K. Ng, K. J. Ooi, and D. T. Tan, “Wideband nonlinear spectral broadening in ultrashort ultra silicon rich nitride waveguides,” Sci. Rep. 6(1), 27120 (2016). 28. C. J. Krückel, A. Fülöp, Z. Ye, P. A. Andrekson, and V. Torres-Company, “Optical bandgap engineering in nonlinear silicon nitride waveguides,” Opt. Express 25(13), 15370–15380 (2017). 29. C. J. Krückel, A. Fülöp, T. Klintberg, J. Bengtsson, P. A. Andrekson, and V. Torres-Company, “Linear and nonlinear characterization of low-stress high-confinement silicon-rich nitride waveguides: erratum,” Opt. Express 25(7), 7443–7444 (2017). 30. K. J. Ooi, D. K. Ng, T. Wang, A. K. Chee, S. K. Ng, Q. Wang, L. K. Ang, A. M. Agarwal, L. C. Kimerling, and D. T. Tan, “Pushing the limits of CMOS optical parametric amplifiers with USRN:Si7N3 above the two-photon absorption edge,” Nat. Commun. 8, 13878 (2017). 31. M. Mitrovic, X. Guan, H. Ji, L. K. Oxenløwe, and L. H. Frandsen, “Four-wave mixing in silicon-rich nitride waveguides,” in Frontiers in Optics, OSA Technical Digest (online) (Optical Society of America, 2015), paper


Introduction
Efficient photonic integrated circuits (PICs) operating at typical telecommunication wavelengths around 1550 nm are becoming increasingly favorable in many applications ranging from ultra-dense data processing [ [15][16][17][18][19]. When talking about efficiency, compactness and low loss are the two basic requirements. Additionally, large nonlinearity and small nonlinear loss, i.e. energyeffectiveness, are usually also required for the waveguides comprising the devices. Moreover, the materials and fabrication processes for the PICs are highly preferred to be compatible with the complementary metal-oxide-semiconductor (CMOS) technologies matured by the electronics industry to exploit manufacturing scalability. Unfortunately, the past decades have witnessed very few material platforms for PICs meeting simultaneously the above-mentioned requirements. As to CMOS-compatible photonic waveguides, biggest hopes are anchored to silicon waveguides due to silicon's large refractive index and its capability of integration with silicon-based electronic integrated circuits (ICs). However, because of a small electronic bandgap (1.12 eV) of silicon, twophoton absorption (TPA) and the associated free carrier absorption (FCA) dramatically limit the nonlinear performances of silicon waveguides around 1550 nm [20]. As an alternative, stoichiometric silicon nitride (Si 3 N 4 ) waveguides have been investigated and applied to achieve key nonlinear devices, including octave supercontinuum generators [21] and frequency comb generators [22], thanks to small propagation losses (0.7 dB/cm) and a large bandgap (5 eV). However, the small refractive index of Si 3 N 4 (~2.0) yields a large footprint of the photonic devices based on Si 3 N 4 waveguides, and the high tensile stress inside thick Si 3 N 4 layer may create cracks in the film limiting the ability for dispersion engineering [23]. Resembling Si 3 N 4 waveguides, newly reported CMOS-compatible tantalum pentoxide (Ta 2 O 5 ) waveguides also encounter the problem of large dimensions due to a similarly small refractive index of Ta 2 O 5 (2.05) in spite of an ultra-low propagation loss [24]. Most recently, silicon-rich nitride (SRN) waveguides have attracted much attention since SRN film can be fabricated with low stress, while possessing a large bandgap as well as having a large refractive index [25][26][27][28][29][30][31], by tuning the reactive gas ratios to enrich the silicon concentration. Whereas it is a quite promising platform for PICs, research on SRN waveguides is still in its infancy and some of the deposited SRN material exhibit quite large material losses (7 dB/cm [26]).
In the work presented here, compact TiO 2 waveguides are fabricated and measured, showing a strong confinement to light and a loss of 4-7 dB/cm at telecom wavelengths. A fabricated TiO 2 MRR having a ring radius as small as 50 μm is found to exhibit a loaded Qvalue of 53500. Four-wave mixing measurements have been carried out on the waveguides, from which nonlinear parameters of 21-34 W −1 m −1 are extracted and a wavelength conversion efficiency of −36.2 dB is achieved.

Waveguide fabrication and linear characterization
Starting with a thermally grown oxide (~2.4 μm) on a silicon wafer, we deposit the TiO 2 film by using a reaction between oxygen and sputtered titanium metal in an argon/oxygen (Ar 2 /O 2 ) ion beam sputtering machine (Oxford, Ionfab 300). The gas flows used are 9 sccm and 2 sccm for Ar 2 and O 2 , respectively, and the sputtering power is 875 W. During the deposition, the silicon wafer is kept at room temperature (20 °C) by using a helium cooling gas, which is supposed to produce an amorphous TiO 2 film according to the previous literature [44]. After 8 hours of deposition, ellipsometry reveals a TiO 2 film thickness of 380 nm. The measured refractive index (n) and the extinction coefficient (k), fitted by the CodyLorentz model, of the deposited film are shown in Fig. 1(a) a as a function of wavelength. At 1550 nm, the refractive index is 2.31 and the film is without significant loss (k<10 −5 ) from 570 nm to 1600 nm, which is derived from a large bandgap of ~3.4 eV. Figure 1(b) shows an atomic force microscope (AFM) image of the deposited TiO 2 film and reveals an average surface roughness, R a , of ~0.4 nm of our amorphous TiO 2 film. TiO 2 waveguides are fabricated in the deposited film. Firstly, electron-beam (e-beam) lithography is used to pattern the waveguides in an e-beam positive resist (ZEP520A) deposited on the TiO 2 film. By utilizing a metal evaporation and lift-off process, the patterns are subsequently transferred to a 55 nm thick aluminum (Al) film to act as an etching mask for the TiO 2 . The TiO 2 is dry-etched by using a gas mixture of tetrafluoromethane (CF 4 , 20 sccm) and hydrogen (H 2 , 10 sccm) gasses. The residual Al mask is stripped from the TiO 2 waveguides and no cladding is covered on the top. Figure 2(a) shows a scanning electron microscope (SEM) image of the cross section of a fabricated TiO 2 waveguide. The shape of the waveguide with height h = 380 nm is trapezoidal with base width w 1 and top width w 2 resulting in sloped sidewalls with angles θ = 73°. The waveguides are slightly over etched by h in = 75 nm into the buried oxide layer. Figures 2(b) and 2(c) show the calculated field distributions of the fundamental transverse-electric (TE) mode and the transverse-magnetic (TM) mode, respectively, of a TiO 2 waveguide with w 1 = 1150 nm. Here, an eigenmode solver, (Mode Solution from Lumerical Solutions) is used and the wavelength is 1550 nm. In the figures, the effective index (n eff ), the effective mode area (A eff ) and the power ratio confined in the TiO 2 (Γ TiO2 ) are also given. Due to a relatively large thickness of the TiO 2 , both the TE-and TM-modes are tightly confined in the TiO 2 with the TE mode having the higher confinement (Γ TiO2 = 81%). The strong confinement will not only allow for compact footprints of the TiO 2 photonic devices in general but will also enhance the nonlinear performances of the TiO 2 waveguides. Using a cut-back method, we have measured the propagation loss of the fabricated waveguides and the coupling losses to a tapered and lensed fiber. Figure 3(a) shows the normalized transmission measured at 1550 nm for different lengths of waveguides having w 1 = 1150 nm for the TM-(red triangles) and the TE-modes (blue dots). Also shown are measurements for TE-polarized light in a waveguide having w 1 = 1630 nm (green squares). Linear fittings for each set of measurements are also given. All the measured transmissions are normalized to a fiber-to-fiber measurement to exclude the influence of the setup. For a given width and length, 4 or 5 identical waveguides are fabricated and measured to estimate the uncertainty and process variations. The propagation loss α for the TE mode in the waveguide with w 1 = 1150 nm is found to be 5.4 ± 1.0 dB/cm, which is larger than the loss of the TM mode (4.1 ± 0.9 dB/cm) in the same waveguides since the TM mode has less overlap with the slightly rough sidewalls. The wider waveguides with w 1 = 1630 nm generally has a lower propagation loss (5.0 ± 1.6 dB/cm) when the light is TE-polarized due to a lower overlap with the sidewalls in the wider waveguides. Figure 3(b) shows the propagation loss and the coupling loss as a function of wavelengths ranging from 1500 nm to 1580 nm for the TE mode in a waveguide with w 1 = 1150 nm. Generally, the propagation loss is 4-7 dB/cm and the coupling loss is ~8 dB per facet (8.0 ± 0.6 dB at 1550 nm). The large coupling loss is due to a strong confinement of the waveguide compared to the lensed fiber, and can be reduced by introducing an inverse waveguide taper scheme to more efficiently couple light from the lensed fibers to the waveguides. Microring resonators (MRRs) with a compact footprint and a high quality factor are key components in passive photonic integrated circuits and for several nonlinear applications like high-efficiency frequency comb generation [19]. The high index contrast between the TiO 2 core and its surroundings (i.e. silica and air) and the large thickness of our waveguides produce a strong confinement of light propagating in the waveguide and facilitate compact TiO 2 MRRs. To verify this, we have fabricated and characterized MRRs having waveguide widths w 1 = 1150 nm and bending radii R = 50 μm but designed with different coupling gaps between the ring and the bus waveguide. Figure 4(a) shows the measured and normalized spectrum for TE-polarized light in an MRR with a coupling gap of 300 nm. From the spectrum, a group index n g of 2.568 can be extracted from a free spectral range (FSR) of 2.982 nm and agrees well with the theoretical value of 2.536 calculated by using n g = λ 2 /(2π•R•FSR), where λ = 1550.894 nm. A zoom-in around the resonance at 1550.894 nm is shown in Fig. 4(b) where a Lorentzian curve (green line) is used to fit the measured spectrum (blue dot). From the fitted curve, a full-width at half-maximum (FWHM) of 29 pm is extracted resulting in a loaded Q-factor (Q load ) of ~53500. With the coupling gap of 300 nm, the MRR is under-coupled and the intrinsic Q-factor Q int can be calculated as ~73500 according to [37], where T 0 ( = 0.207) is the transmission normalized to the maximum value of the fitting curves at the resonant wavelength (λ 0 = 1550.894 nm). Knowing the radius R and the FSR of the MRR, the waveguide loss of the ring α ring can be calculated by α ring = 2πn g /(Q int •λ 0 ) giving a loss of 6.1 dB/cm, which is slightly higher than the measured loss of the straight waveguide (5.4 ± 1.0 dB/cm) due to a larger mode overlap with the sidewalls in the bent waveguide.  Figure 5 shows the experimental setup for performing four-wave mixing (FWM) experiments in the fabricated straight TiO 2 waveguides. The pump and probe signals from two continuous wave (CW) lasers are separately polarized by a polarization controller (PC), amplified by a erbium-doped fiber amplifier (EDFA), filtered by a 1-nm band-pass filter, and then combined in a ratio 10:1, respectively, by using a 10 dB fiber coupler. After passing through a polarization beam splitter (PBS) and another PC, the two signals are injected into the TiO 2 waveguides by using a tapered and lensed fiber. Here, the PBS is used to switch between the TE and TM polarizations. The output light from the waveguides is collected by another tapered and lensed fiber and analyzed by an optical signal analyzer (OSA). In order to monitor the input and output powers of the waveguides, two power meters are connected via two 20 dB fiber couplers positioned before and after the waveguides. During the measurements, we fix the wavelengths of the pump and the probe to 1550.1 nm and 1551.3 nm, respectively. We have measured the output FWM spectra for the TE mode and the TM mode in a TiO 2 waveguide with the width w 1 = 1150 nm and the length L = 1.1 cm when the power of the input probe signal is fixed at 12.7 dBm but varying the power of input pump signal. Figure 6(a) shows the spectra for the maximum input pump power used (27.8 dBm/27.6 dBm) for the TE/TM (blue/red) mode, respectively. From the spectra, one can extract the FWM conversion efficiencies (CEs) of −36.2 dB and −37.2 dB for the TE mode and the TM mode, respectively, with the CE defined as the ratio of the output idler power to the output probe signal power. Furthermore, the nonlinear parameter γ can be calculated by [45],

Nonlinear characterization
where P pump is the coupled pump power being the input pump power minus the fiberwaveguide coupling loss and L eff is the effective interaction length of the pump and probe signal in the waveguide defined as (1-e -αL )/α, where L is the length of the waveguide with propagation loss α. We reasonably assume that γ is real since no TPA is present at the wavelengths in focus due to a large measured bandgap (3.4 eV). Moreover, we assume that η = 1 since the phase mismatch is negligible for the small spectral separation between the pump and the probe signals. Figure 6(b) shows the calculated CEs with respect to the different coupled pump powers for the TE (blue square) mode and the TM (red circle) mode along with quadratic fittings (dashed lines) when using the maximum value of the coupling losses, i.e. the highest nonlinear efficiency. The quadratic fittings are in very good agreement with the measured values which supports the assumption of neglecting TPA effects. The highest value of the nonlinear parameter γ can be extracted to 34 W −1 m −1 and 19 W −1 m −1 using the maximum values of the propagation loss in the calculations for the TE mode and the TM mode, respectively. The measured nonlinearity of the TE mode is almost twice as large as that of the TM mode since the TE mode is more confined in the TiO 2 core. The lowest value of γ can be calculated by using the minimum values of the coupling loss and the propagation loss in the calculations resulting in nonlinear parameters γ of 21 W −1 m −1 and 14 W −1 m −1 for the TE mode and the TM mode, respectively. Knowing the waveguide dimensions and the calculated nonlinear parameters, it is easy to calculate the effective nonlinear refractive index (n 2,eff ) of the TiO 2 waveguide as n 2,eff = (γλA eff )/(2π). Using the parameters of the TE mode and the lowest and the highest values of γ, n 2,eff is calculated to be 2.3-3.6 x 10 −18 m 2 /W. Although both the TiO 2 core and the buried silica layer contribute to the nonlinearity of the waveguides, the contribution of the silica layer is smaller than 0.2% due to a low refractive index (1.444), a low power ratio in this layer (Γ silica~1 1%) as well as a quite low nonlinear refractive index of silica (~0.028 x 10 −18 m 2 /W [46]). Consequently, the calculated n 2,eff is almost equal to the nonlinear refractive index of the deposited TiO 2 material, which is one order of magnitude larger than that of Si 3 N 4 and approximately half the value of that of silicon.

Discussion
The nonlinear index n 2 (2.3-3.6 However, to the best of our knowledge, our results successfully verify for the first time the high nonlinear index of TiO 2 used for realizing compact waveguides. TiO 2 waveguides have already been demonstrated for spectral broadening, but the extracted n 2 (0.16 x 10 −18 m 2 /W) is relatively small [34]. Furthermore, due to a larger waveguide height of 380 nm, our waveguides confine 81% of power in the TiO 2 core and correspondingly enhance the nonlinearity of the waveguide to 21-34 W −1 m −1 . This is an order of magnitude larger than previously demonstrated waveguides confining ~60% of the power in a 250nm-high TiO 2 core [34]. Anomalous dispersion is usually needed to achieve key nonlinear processes like supercontinuum generation [21,51]. Having a thicker waveguide core, i.e. having a larger mode confinement in the core, may simplify the dispersion engineering. Figure 7 shows the calculated dispersion D as a function of wavelengths for the TE mode in a TiO 2 waveguide with w 1 = 1150 nm and thicknesses of 250 nm (green line), 380 nm (red line) and 400 nm (blue line). From the curves, one can find that the dispersion of the thinner waveguide (250 nm) is away from the anomalous dispersion regime around 1550 nm. In contrast, the 380nmthick TiO 2 waveguide is close and can easily be tuned to have anomalous dispersion (D>0) in a broad wavelength region by increasing slightly the thickness of the waveguide to 400 nm. This re-depositing of 20 nm TiO 2 on the fabricated waveguides can be achieved by using either sputtering or atomic layer deposition (ALD). The latter can provide an even more accurate control of the re-deposition and can also be used to smoothen the waveguide sidewalls to decrease the propagation loss [35].

Conclusion
In conclusion, we have fabricated compact TiO 2 waveguides and characterized their linear and nonlinear performances. The waveguides possess a strong confinement of light due to a large refractive index, a thick film of the deposited titanium dioxide and also a full etching for a stripe configuration. Such a strong confinement not only benefits the compactness of the devices but also plays an important role for achieving a high nonlinearity. The contemporary fabrication processes produce a propagation loss of 5.4 ± 1 dB/cm and 4.1 ± 0.9 dB/cm for the TE mode and the TM mode, respectively, which is comparable with that in the previous literatures but may be further decreased by optimizing the processes, including using a resist mask instead of the rough metal mask and tuning the gas ratio in the etching to achieve a more chemical (i.e. softly) etching. Microring resonators have been fabricated and measured to have a moderately large loaded Q (53500) in a compact footprint. Four-wave mixing experiments have been carried out in the TiO 2 waveguides and a conversion efficiency of −36.2 dB can be achieved without the presence of detrimental TPA effects. From the FWM experiments, the nonlinear parameters of the fabricated TiO 2 waveguide and the nonlinear index of the TiO 2 material are calculated to 21-34 W −1 m −1 and 2.3-3.6 x 10 −18 m 2 /W, respectively. These results, especially the high nonlinearity, pave a promising way for TiO 2 waveguides to be applied in photonic integrated circuits simultaneously having good linear and nonlinear performances and retaining CMOS-compatibility in the fabrication.