SUPPLEMENTARY INFORMATION

Size: px
Start display at page:

Download "SUPPLEMENTARY INFORMATION"

Transcription

1 SUPPLEMENTARY INFORMATION doi:.38/nphoton..7 Supplementary Information On-chip optical isolation in monolithically integrated nonreciprocal optical resonators Lei Bi *, Juejun Hu, Peng Jiang, Dong Hun Kim, Gerald F. Dionne, Lionel C. Kimerling and C. A. Ross *. Department of Materials Science and Engineering, Massachusetts Institute of Technology Cambridge, Massachusetts 39. Department of Materials Science and Engineering University of Delaware Newark, Delaware 976 Supporting materials are provided for details of device integration and characterization process, device performance derivation, magneto-optical thin film characterization and optical loss evaluation of the isolator. nature photonics Macmillan Publishers Limited. All rights reserved.

2 supplementary information doi:.38/nphoton..7 I. DEVICE INTEGRATION AND CHARACTERIZATION PROCESS Figure S and figure S shows the process flow of the patterned silicon resonator and the characterization setup for the optical transmission and isolation measurements, which are described in the methods part of the manuscript. For isolation characterization, the permanent magnet was placed at ~5 mm to the resonator under testing, with the magnetic field perpendicular to the light propagation direction. The magnetic field across the resonator is measured by a Gauss probe to be ~±5 Oe, which is high enough to saturate the garnet films.. Spin on flowable oxide+rta Optical lithography and plasma etch Chemical etch by HF 3. Cleave and form end facets Test with applied magnetic field. PLD YIG film, RTA PLD Bi:YIG or Ce:YIG Fig. S Fabrication process flow of a nonreciprocal racetrack resonator on SOI. The silicon racetrack resonator is initially fabricated by electron beam lithography on SOI. Atomic force microscopy was measured between step and to verify the exposure of silicon resonator surface. PLD: pulsed laser deposition, RTA: rapid thermal annealing. a Alignment lens b Magnet Input fiber Sample Output fiber Fig. S Optical resonator (a) transmittance and (b) isolation measurement setup. In (b), the permanent magnet is placed at ~5 mm distance from the resonator under testing, with a perpendicular magnetic field of ~ ±,5 Oe at the resonator location. nature photonics Macmillan Publishers Limited. All rights reserved.

3 doi:.38/nphoton..7 supplementary information II THEORETICAL ANALYSIS OF DEVICE PERFORMANCE We show a detailed derivation of the resonator nonreciprocal wavelength shift, FSR and quality factor in this section. Considering the patterned racetrack resonator shown in figure 3 (a), the nonreciprocal resonance wavelength shift, the resonance wavelength of forward ( f ) and backward ( b ) propagating light satisfies: L n L n + eff eff f f N () L ( n eff neff + b neff ) L( neff + + n) + N b () b βtm n π (3) where N is an integer, n eff and n eff are the effective refractive index of TM polarized light corresponding to resonator section L and L respectively, and n and β TM are the TM mode effective index and propagation constant difference between forward and backward propagation light respectively. Considering f ~ b ~ r, we obtain: r L β TM π ( Ln g + Ln g) (4) where n g and n g are the group indices of resonator sections L and L respectively, which are defined as: The FSR can be expressed by: n g n eff neff (5) neff,l + neff,l neff,l + neff,l FSR r r,, where N N N N, + (6) Considering N ~N ~N>>, f ~ b ~ r,we have: g r FSR L n + L n g (7) Therefore from (4) and (7), FSR β π L TM (8) The quality factor Q can be calculated using a phasor summation approach in reference. For a nature photonics 3 Macmillan Publishers Limited. All rights reserved.

4 supplementary information doi:.38/nphoton..7 racetrack resonator coupling to a bus waveguide, the FWHM of the critical coupling resonant peak δ satisfies: I ' ' [ i π ( n L + n L ) + δ] 4 exp( α L) exp / I eff eff r (9) where I is the light intensity at the input side of the bus waveguide, α is the resonator loss per length defined in the manuscript, n eff and n eff are the effective index of L and L sections at FWHM ' wavelength of the resonance peak, with n neff + n eff eff / * δ. For Q>>, αl<< and δ<<, equation (9) can be simplified by Taylor expansion, with Q / δ, we obtain equation () in the r manuscript. III MAGNETO-OPTICAL THIN FILM CHARACTERIZATION We characterized the phase purity and magneto-optical properties of the polycrystalline garnet films. Figure S3 shows the X-ray diffraction and Faraday rotation of YIG film and Ce:YIG film deposited on YIG buffer layer respectively. Both films show pure polycrystalline garnet phases, with lattice constants of.38 Å and.46 Å for YIG and Ce:YIG respectively, comparable to bulk samples [S]. As a comparison, Ce:YIG deposited directly on oxidized silicon wafers shows amorphous XRD patterns. Faraday rotation values of + deg/cm and -63 deg/cm were observed in polycrystalline YIG and Ce:YIG films, smaller than the value of +6 deg/cm for YIG [S] and -33 deg/cm for Ce:YIG [S3] epitaxial films. The saturation magnetizations of YIG and Ce:YIG polycrystalline films were 3 emu/cm 3 and emu/cm 3 respectively, corresponding to 95 vol.% and 87 vol.% of crystalline garnet phases in the films. Both films show in-plane magnetization easy axis with low saturation magnetic field of ~ Oe and ~ Oe for YIG and Ce:YIG films, which is dominated by magnetic shape anisotropy. The thickness of Ce:YIG and YIG is chosen to be 8 nm and nm respectively to allow high magneto-optical and optical performance. Theoretically a thick Ce:YIG and thin YIG layer combination is ideal for a high NRPS waveguide. However, cracks generate when the film stack thickness exceeds nm due to thermal mismatch between garnets and silicon [S4]; while partially amorphous films were observed when the YIG thickness is below nm. The low critical thickness of garnet films on silicon has 4 nature photonics Macmillan Publishers Limited. All rights reserved.

5 doi:.38/nphoton..7 supplementary information been a limiting factor for conventional devices leading to a large required device length to achieve optical isolation. Using a nonreciprocal resonator, we demonstrated high optical isolation ratio can be achieved in these monolithically integrated thin films with a small device footprint. CeYFe5O (Ce:YIG) Ce:YIG/YIG/Si Fig. S3 X-ray diffraction spectrum of (a) YIG and (b) Ce:YIG polycrystalline films on oxidized silicon. A control sample without YIG buffer layer is also shown in (b). Room temperature Faraday rotation hysteresis at 55 nm wavelength for (c) YIG and (d) Ce:YIG films are also shown. IV LOSS EVALUATION We evaluated the optical loss contribution of the magneto-optical waveguide section, reciprocal silicon waveguide section and junction loss to the overall device insertion loss by experiment and simulation. The overall loss of the resonator is firstly calculated following equation [S5]: πr α FSR Q where Q in is the intrinsic quality factor (Q in Q at critical coupling). Using a measured TM mode intrinsic quality factor of,, the total loss is 58 db/cm. A control sample of silicon racetrack resonators with the same dimensions and SiO top cladding was fabricated. The optical transmittance spectrum of this device in near infrared wavelength range is shown in figure S4. The inset shows a TM mode resonance in L nature photonics 5 Macmillan Publishers Limited. All rights reserved.

6 supplementary information doi:.38/nphoton..7 peak. The quality factor of TM mode resonance is Q~3, with ~ db extinction ratio, from which we calculated the silicon waveguide loss to be ~3 db/cm. According to α ( αl + αl + α junction ) / L, the loss contribution from the reciprocal resonator section is.47 db/cm. The junction loss between L and L sections of the resonator is estimated by the eigenmode expansion method using FIMMPROP software to be.5 db/junction [S6]. Using equation (6-7), we calculated the distributed junction loss contribution at both facets to be 4.4 db/cm. Comparing with the measured Ce:YIG sample total loss coefficient of 58 db/cm, the magneto-optical waveguide section contributed 5 db/cm to the total optical loss in the resonator. The loss may partly due to the fabrication process during annealing and etching of the resonator top cladding. The Ce:YIG/YIG film was observed to show a loss value of 9 cm - in our recent report 6. Considering optical loss contribution of the Ce:YIG/YIG film, silicon waveguide and junction losses, we obtained a resonator quality factor of ~7, at critical coupling condition, which corresponds to an insertion loss of 5-6 db for the isolator device. Therefore fabrication induced loss is estimated to be 4-5 db, which can be optimized in future studies. Fig. S4 Transmission spectrum of a silicon racetrack resonator with SiO top cladding. The inset shows the TM resonance peak at near critical coupling condition References [S] Sekijima, T., Fujii, T., Wakino, K. & Okada, M. Optical Faraday rotator using Ce-substituted fibrous YIG single crystal grown by floating-zone method with YAG laser heating. IEEE Trans. Microwave Theory Tech., 47, (999). 6 nature photonics Macmillan Publishers Limited. All rights reserved.

7 doi:.38/nphoton..7 supplementary information [S] Dillon, J. F. Origin and Uses of the Faraday Rotation in Magnetic Crystals J. Appl. Phys., 39, 9-99 (968). [S3] Shintaku, T., Tate, A. & Mino, S. Ce-substituted yttrium iron garnet films prepared on Gd3ScGa3O garnet substrates by sputter epitaxy. Appl. Phys. Lett., 7, (997). [S4] Boudiara, T. et. al. Magneto-optical properties of yttrium iron garnet (YIG) thin films elaborated by radio frequency sputtering, J. Mag. Mag. Mat., 84, (4). [S5] Hu, J., Planar Chalcogenide Glass Materials and Devices, PhD. Thesis, (Massachusetts Institute of Technology, 9). [S6] Integrated Optics Software FIMMWAVE 4.5, Photon Design, Oxford, U.K. [Online]. Available: nature photonics 7 Macmillan Publishers Limited. All rights reserved.