Nonreciprocal racetrack resonator based on vacuum-annealed magnetooptical cerium- substituted yttrium iron garnet

Vacuum annealed polycrystalline cerium substituted yttrium iron garnet (CeYIG) films deposited by radio frequency magnetron sputtering on non-garnet substrates were used in nonreciprocal racetrack resonators. CeYIG annealed at 800°C for 30 min provided a large Faraday rotation angle, close to the single crystal value. Crystallinity, magnetic properties, refractive indices and absorption coefficients were measured. The resonant transmission peak of the racetrack resonator covered with CeYIG was nonreciprocally shifted by applying an in-plane magnetic field. ©2014 Optical Society of America OCIS codes: (160.3820) Magneto-optical materials; (230.3240) Isolators. References and links 1. L. Bi, J. Hu, P. Jiang, D. H. Kim, G. F. Dionne, L. C. Kimerling, and C. A. Ross, “On-chip optical isolation in monolithically integrated non-reciprocal optical resonators,” Nat. Photonics 5(12), 758–762 (2011). 2. S. Ghosh, S. Keyvaninia, Y. Shirato, T. Mizumoto, G. Roelkens, and R. Baets, “Optical isolator for TE polarized light realized by adhesive bonding of Ce:YIG on Silicon-on-Insulator waveguide circuits,” IEEE Photon. J. 5(3), 6601108 (2013). 3. S. Ghosh, S. Keyvavinia, W. Van Roy, T. Mizumoto, G. Roelkens, and R. Baets, “Ce:YIG/Silicon-on-Insulator waveguide optical isolator realized by adhesive bonding,” Opt. Express 20(2), 1839–1848 (2012). 4. P. Pintus, T. Ming-Chun, and J. E. Bowers, “Design of magneto-optical ring isolator on SOI based on the finiteelement method,” IEEE Photon. Technol. Lett. 23(22), 1670–1672 (2011). 5. Q. Wang, Z. Ouyang, and Q. Liu, “Multiport photonic crystal circulators created by cascading magneto-optical cavities,” J. Opt. Soc. Am. B 28(4), 703–708 (2011). 6. K. Yayoi, K. Tobinaga, Y. Kaneko, A. V. Baryshev, and M. Inoue, “Optical waveguide circulators based on two-dimensional magnetophotonic crystals: Numerical simulation for structure simplification and experimental verification,” J. Appl. Phys. 109(7), 07B750 (2011). 7. T. Goto, R. Isogai, and M. Inoue, “Para-magnetoand electro-optic microcavities for blue wavelength modulation,” Opt. Express 21(17), 19648–19656 (2013). 8. J. W. Pritchard, M. Mina, and R. J. Weber, “Improved switching for magneto-optic fiber-based technologies,” IEEE Trans. Magn. 48(11), 3772–3775 (2012). 9. Y. Arakawa, T. Nakamura, Y. Urino, and T. Fujita, “Silicon photonics for next generation system integration platform,” IEEE Commun. Mag. 51(3), 72–77 (2013). 10. M. Gomi, H. Furuyama, and M. Abe, “Strong magneto-optical enhancement in highly Ce-substituted iron garnet films prepared by sputtering,” J. Appl. Phys. 70(11), 7065 (1991). 11. B. Stadler, K. Vaccaro, P. Yip, J. Lorenzo, L. Yi-Qun, and M. Cherif, “Integration of magneto-optical garnet films by metal-organic chemical vapor deposition,” IEEE Trans. Magn. 38(3), 1564–1567 (2002). 12. T. Shintaku, A. Tate, and S. Mino, “Ce-substituted yttrium iron garnet films prepared on Gd3Sc2Ga3O12 garnet substrates by sputter epitaxy,” Appl. Phys. Lett. 71(12), 1640–1642 (1997). 13. T. Goto, M. C. Onbaşlı, and C. A. Ross, “Magneto-optical properties of cerium substituted yttrium iron garnet films with reduced thermal budget for monolithic photonic integrated circuits,” Opt. Express 20(27), 28507– 28517 (2012). 14. T. Goto, Y. Eto, K. Kobayashi, Y. Haga, M. Inoue, and C. A. Ross, “Vacuum annealed cerium-substituted yttrium iron garnet films on non-garnet substrates for integrated optical circuits,” J. Appl. Phys. 113(17), 17A939 (2013). #214609 $15.00 USD Received 23 Jun 2014; revised 21 Jul 2014; accepted 22 Jul 2014; published 29 Jul 2014 (C) 2014 OSA 11 August 2014 | Vol. 22, No. 16 | DOI:10.1364/OE.22.019047 | OPTICS EXPRESS 19047 15. K. Sato, Hikari-to-jiki (Asakura-shoten, Tokyo, 2007). 16. B. Stadler and A. Gopinath, “Magneto-optical garnet films made by reactive sputtering,” IEEE Trans. Magn. 36(6), 3957–3961 (2000). 17. S. Mino, A. Tate, T. Uno, T. Shintaku, and A. Shibukawa, “Properties of Ce-substituted yttrium iron garnet film containing indium prepared by RF-sputtering,” Jpn. J. Appl. Phys. 32(Part 2, No. 7B), L994–L996 (1993).


Introduction
Demands for non-reciprocal phase shifters or polarization rotators, e.g., isolators [1][2][3][4], circulators [5,6], or modulators [7,8] are increasing with the rapid growth of integrated photonics [6,9].Magnetooptical (MO) garnets have been used to achieve nonreciprocal isolation with low optical losses from visible to near IR wavelengths, but their integration onto non-garnet substrates, e.g., Si, Si-on-insulator (SOI), or silica, is still challenging because of the mismatch of the thermal expansion and lattice constants between the films and substrates.In particular, cerium substituted yttrium iron garnet (Ce x Y 3-x Fe 5 O 12 or CeYIG) with 1 Ce per formula unit (Ce 1 Y 2 Fe 5 O 12 ) or more has been difficult to fabricate on nongarnet substrates because of the formation of undesirable non-garnet phases [10,11] even though a large Faraday rotation (FR) angle (-3300 deg/cm) and low absorption (-9.7 dB/cm) at a wavelength of λ = 1550 nm were demonstrated in single crystalline CeYIG on garnet substrates [12].
Recently, we demonstrated two methods to fabricate CeYIG on non-garnet substrates: two-step deposition using pulsed laser deposition (PLD) [13] and sputtering with thermal annealing in vacuum [14].The two-step deposition uses a thin YIG buffer layer as a seed layer for epitaxial growth of polycrystalline CeYIG on non-garnet substrates [1], and was used in a demonstration of an integrated isolator based on a racetrack resonator.On the other hand, the CeYIG fabricated by sputtering on silicon with thermal annealing in vacuum showed larger FR (-3000 deg/cm) at λ = 1550 nm [14], but its nonreciprocal shift as a isolator has not been reported.In this paper, we optimized the fabrication condition of CeYIG on Si and SOI substrates.The structural characteristics, the FR, refractive index, extinction coefficient, and figure of merit (FOM) of polycrystalline CeYIG are shown as a function of annealing conditions.Finally, we demonstrate a non-reciprocal shift of transmissivity in a racetrack resonator using optimized CeYIG.

CeYIG preparation
The 170 nm thick films were deposited on double-side polished SOI substrates by radio frequency sputtering using the same conditions as [14], from a sintered Ce 1 Y 2.5 Fe 5 O x target in 10 mTorr Ar at 0.3 nm/min onto substrates at room temperature.The base pressure was 1.0 × 10 −7 Torr.A vacuum anneal was carried out at a pressure of 5 Pa as shown in [14].The temperature and time of the thermal anneal was varied.Composition of the annealed films was measured by energy dispersive x-ray spectrometer (EDS) and normalized to the iron atomic fraction.Transmissivity was measured with a spectrophotometer, using specimens of CeYIG on Si substrates to suppress the oscillations in the transmission spectra generated by the SOI substrates.FR loops were measured by oscillating polarizer method [15] with a laser diode at a wavelength of λ = 1550 nm.Magnetic hysteresis loops were measured by vibrating sample magnetometry (VSM).Structural properties were characterized with atomic force microscopy (AFM) and x-ray diffraction (XRD) with CuKα radiation at 0.1541 nm wavelength.All measurements were carried out at ambient temperature (25°C).Figure 1 shows diffraction patterns for a set of CeYIG samples deposited on SOI substrates.The composition was Ce 0.97 ± 0.011 Y 1.96 ± 0.083 Fe 5 O 12.2 ± 0.030 (the value after ± is the standard deviation).Pure garnet phase was obtained from the samples annealed at 700 and 800°C, Fig. 1(a).The films annealed at more than 900°C showed YFe 2 O 4 and Y 2 O 3 peaks which lead to large optical absorption.A garnet phase was not found in the samples annealed below 600°C.In Fig. 1(b), the annealing time was varied from 1 to 120 minutes at 800°C.

CeYIG properties
Figures 2(a) and 2(b) show the corresponding FR loops of the samples annealed at various conditions.The FR was measured through the film thickness which is the magnetic hard axis, so the FR did not saturate until the field reached 2-3 kOe.There is a clear dependence of the FR loop on both the temperature and time.Although the XRD results were similar after 700°C and 800°C annealing, the 800°C anneal produced a much higher FR.The sample annealed at 800°C for 30 minutes showed the largest value in this study, around -3.0 × 10 3 deg/cm, which is 90% of the value reported for single crystal CeYIG [12].
It is not known why there is an optimum anneal time, but excessive annealing could produce unwanted phases or roughness of the film, or changes in stoichiometry [16].The influence of annealing on the magnetic properties was observed in the VSM hysteresis loops, Fig. 3.The saturation magnetization Ms was ~118 emu cm −3 , similar to single crystal values [17].The dominant magnetic anisotropy in these films is the shape anisotropy and the hard-axis (out-of-plane) saturation field H s , obtained from extrapolating the out-of-plane loops to saturation, was close to the calculated shape anisotropy field of 4πMs = 1.48 kOe.For the 800°C, 30 min sample the coercivity H c for in-plane and out-ofplane loops was 45 and 102 Oe respectively.The remenant magnetization M r for in-plane and out-of-plane loops was 75 and 6.3 emu cm −3 respectively.Figure 4(a) shows the transmission spectra of CeYIG on Si substrate annealed at 800°C for various times.The transmissivity was increased because of the anti-reflection effect of introducing the lower refractive index garnet on the higher refractive index Si substrates.This is not the case for the silica substrates.To obtain the refractive index and absorption, the transmissivity of CeYIG on a silica substrate was measured, Fig. 4(b).The silica substrate is highly transmissive, enabling the optical index to be found from the fringes.The simulation software (SCOUT) [14], based on the Fresnel equation, was used to derive the refractive index and extinction coefficient of CeYIG annealed at 800°C for 30 min, Fig. 5.These extinction coefficients are one order larger than the single crystal value, 5.5 dB/cm at 1550 nm wavelength [12].This was probably caused by the surface cracking of CeYIG shown in Fig. 6.The large difference of the thermal expansion coefficients between the films and substrates led to mismatch strain and cracks.In contrast, in an uncracked area of 0.2 × 0.2 μm, the grain size and roughness were ~10 nm and ~0.4 nm rms respectively.We expect the optical losses to decrease if cracks are avoided, e.g. by patterning the CeYIG to relieve strain.The FOM defined as FR/absorption coefficient for the best film in this study was (-3.0 × 10 3 deg/cm)/(-62.3dB/cm) = 48 deg/dB, close to previous work [1,14].Fig. 6.Surface topography of Si/CeYIG obtained by atomic force microscopy (AFM).Anneal time was varied from 1 to 60 minutes, and two scan sizes, 0.2 × 0.2 μm 2 and 10 × 10 μm 2 were observed.

Racetrack resonator preparation
An optical isolator based on a racetrack resonator with optimized CeYIG as cladding was fabricated by electron beam lithography (EBL).In this design one straight side of the racetrack is clad with MO material which is magnetized perpendicular to the track to produce a nonreciprocal phase shift in the circulating light.The device structure was based on Ref [1], and also its process was similar, but several steps were improved, e.g., the SiO 2 cladding layer was formed by chemical vapor deposition (CVD) and etched by reactive ion etching (RIE) in CHF 3 gas instead of by a wet processes, which improved the control over the deposition and etching rate.The base pressure was 1.0 × 10 −5 Torr, forward power was 200 W, and the etching rate was 1.0 nm/sec.A 1 μm thick SiO 2 was deposited onto the sample by plasma-enhanced CVD (PlasmaTherm) with a deposition rate of 28.9 nm/min, Fig. 7(f).The substrate temperature was 250°C, base pressure was 10 mTorr, and working pressure was 500 mTorr.The introduced gasses were 150 sccm 5% SiH 4 in He and 600 sccm N 2 O.The input power was 25 W.
The quality of the racetrack resonator without MO cladding was evaluated by a transmission measurement performed on a Newport Auto Align workstation using an optical vector analyzer (LUNA Technologies OVA-5000) with a built-in tunable laser.The laser light was coupled into and out of the devices with tapered lens-tip fibers to improve coupling efficiency.The transmitted light was analyzed with an optical vector analyzer (LUNA Technologies).The sample was placed onto a thermostat stage kept at 25°C to decrease the effects of the temperature-dependent index of silicon.Figure 8 shows the transmission spectra of the Si racetrack resonator.The TE mode undergoes greater attenuation than the TM mode because it is more sensitive to sidewall roughness, even though the cross-sectional dimensions of the waveguide suggest it should propagate the TE mode better, according to Lumerical MODE simulations.The Q factor was 3.4 × 10 4 and free spectral range (FSR) was 0.82 nm.Although the FSR was comparable to the reported values (0.90 nm) in [1], the Q factor was one order smaller than the value 1.3 × 10 5 in [1].To deposit CeYIG only in a 20 × 200 μm region on the racetrack, a window in the SiO 2 cladding layer was etched using a positive resist.The 1 μm thick positive resist (ZEP520A) was spin coated for 1 minute at 5000 rpm followed by baking at 180°C for 3 minutes on a hot plate.The resist was exposed by EBL under the conditions described above except the current was 2 nA, aperture 120 μm, and dose time 0.2 μsec, Fig. 7(g).To form the window at the appropriate part of the racetrack, two cross-shaped alignment marks were formed on first 3 μm thick SiO 2 layer at the edge of the SOI substrate, and covered with polyimide tapes during SiO 2 deposition and the ZEP coating so they would be visible.The pattern was developed with o-xylene (ZED-WN) for 1 minute at room temperature (RT) without sonication followed by rinsing with IPA for 1 minute at RT and N 2 blowing.The SiO 2 was etched by RIE in 200 mTorr (20 sccm) CHF 3 gas with a substrate temperature of 200°C, base pressure of 1 μTorr, and input power of 100 W, Fig. 7(h).The etching rate for the resist and SiO 2 were both ~21 nm/min.The top image of the sample is shown in Fig. 7(i).The presence of residual SiO 2 (from the HSQ) beside the waveguide was confirmed by comparing the profiles in Figs.7(e) and 7(j).80 nm thick CeYIG was deposited on this sample and annealed with the above conditions, Figs.7(k) and 7(l).A cross sectional image of the waveguide covered with MO material was obtained by focused ion beam (FIB) cutting and scanning electron microscope (SEM), Fig. 7(l).Only the CeYIG on the top of the waveguide was in direct contact with the waveguide because the side of the Si waveguide was covered with residual SiO 2 and was decoupled from the MO layer.Cracks in the CeYIG were not observed on the Si waveguide, but grain growth was observed and might lead to optical losses.This result suggests that it will be useful to optimize the fabrication conditions of the CeYIG in the geometry used in the actual device.

Nonreciprocal shift measurement
The nonreciprocal shift of the resonant peak in the transmission spectra was measured with the waveguide characterization system described above.A magnetic field of -1.1, -0.7, 0.0, + 0.7, and + 1.1 kOe was applied in-plane, perpendicular to the waveguide, Figs. 9.In Fig. 9(a), the resonant peak for the TM mode was shifted as a function of applied magnetic field at 25°C, but the TE mode was not shifted.The shift of the resonant wavelength Δλ was 16.2 pm, similar to the previous report of ~18.4 pm [1].The TE mode was not shifted when the magnetic field was reversed which confirmed that the device exhibited the expected nonreciprocal magnetooptical response only for TM, Fig. 10.The isolation ratio was only 1 dB based on the maximum field values applied.The resonant peak was too shallow to calculate the Q factor, but if we subtract the background loss (~-27.5 dB), the quasi Q factor is 3.3 × 10 4 , compared to 1.1 × 10 4 in [1].The rough surface decreased the quality factor of this device.

Conclusion
A nonreciprocal peak shift was demonstrated in a racetrack resonator made using sputtered and vacuum-annealed CeYIG cladding.The CeYIG deposition was optimized on Si substrates to produce Faraday rotation and saturation magnetization that were close to bulk values after a 30 min anneal at 800°C.The process modifications and the simplified growth process of the CeYIG produced similar nonreciprocal phase shift of the resonator, though lower isolation ratio, than that shown in previous work where the CeYIG was made by pulsed laser deposition.This report provides a further step toward realizing integrated magnetooptical devices on non-garnet substrates.

Fig. 2 .
Fig. 2. FR loops of the samples of Si substrate/CeYIG (170 nm) annealed (a) at various temperatures for 60 minutes and (b) at 800°C for various times.Insets show the saturation FR dependence on annealing temperature and time.

Fig. 3 .
Fig. 3. (a) In-plane and (b) out-of-plane magnetic properties of the samples Si substrate/CeYIG (170 nm) annealed at 800°C for various minutes.Insets show the enlarged portions of these data, respectively around zero.

Fig. 5 .
Fig. 5. (a) Refractive index and (b) extinction coefficient of Si/CeYIG annealed at 800°C for 30 minutes.The error bars are determined by the thickness deviation of the CeYIG.

Fig. 7 .
Fig. 7. Process for formation of optical isolator with CeYIG cladding annealed in vacuum.

Figure 7
Figure7shows the process flow.The 80 nm thick negative-contrast hydrogen silsesquioxane (HSQ) resist (Dow Corning, XR-1541, 6%) was spin coated on SOI wafers (Soitec) for 1 minute at 5000 rpm, Fig.7(a).The SOI substrates were composed of 0.5 mm thick Si/3 μm thick SiO 2 /250 nm thick Si.The coated substrates were exposed by electron beam lithography (Elionix, ELS-F125) with acceleration voltage of 125 kV, current of 10 nA, aperture of 120 μm, 6 × 10 4 pixels in a 600 × 600 μm 2 area, at a dose time of 0.20 μsec, Fig. 7(b).The exposed width of the waveguide was 450 nm.The curvature radius of the racetrack was 45 μm.The length of the straight line portion of the racetrack was 200 μm.After development with tetra-methyl-ammonium hydroxide (TMAH, MF-CD-36) at 25°C with sonication for 3 minutes and a rinse with deionized water, Fig. 7(c), the uncovered Si area was etched by RIE (PlasmaTherm SLR) in 10 mTorr HBr gas pressure, Figs.7(d) and 7(e).The base pressure was 1.0 × 10 −5 Torr, forward power was 200 W, and the etching rate was 1.0 nm/sec.A 1 μm thick SiO 2 was deposited onto the sample by plasma-enhanced CVD (PlasmaTherm) with a deposition rate of 28.9 nm/min, Fig.7(f).The substrate temperature was 250°C, base pressure was 10 mTorr, and working pressure was 500 mTorr.The introduced gasses were 150 sccm 5% SiH 4 in He and 600 sccm N 2 O.The input power was 25 W.The quality of the racetrack resonator without MO cladding was evaluated by a transmission measurement performed on a Newport Auto Align workstation using an optical vector analyzer (LUNA Technologies OVA-5000) with a built-in tunable laser.The laser light was coupled into and out of the devices with tapered lens-tip fibers to improve coupling efficiency.The transmitted light was analyzed with an optical vector analyzer (LUNA Technologies).The sample was placed onto a thermostat stage kept at 25°C to decrease the effects of the temperature-dependent index of silicon.Figure8shows the transmission spectra of the Si racetrack resonator.The TE mode undergoes greater attenuation than the TM mode because it is more sensitive to sidewall roughness, even though the cross-sectional dimensions of the waveguide suggest it should propagate the TE mode better, according to Lumerical MODE simulations.The Q factor was 3.4 × 10 4 and free spectral range (FSR) was 0.82 nm.Although the FSR was comparable to the reported values (0.90 nm) in[1], the Q factor was one order smaller than the value 1.3 × 10 5 in[1].

Fig. 8 .
Fig. 8. Transmission spectra of the waveguide without CeYIG cladding at a temperature of 25°C.Solid and dashed lines show the TM and TE mode, respectively.

Fig. 9 .
Fig. 9. Nonreciprocal phase shift of the (a) TM and (b) TE modes as functions of magnetic field H in kOe.

Fig. 10 .
Fig. 10.The peak shifts of TE and TM mode vs. applied magnetic field.