Hybrid Silicon and Lithium Niobate Integrated Photonics
Dissertation
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University
By
Li Chen, M. Sc., B. Eng.
Graduate Program in Electrical and Computer Engineering
The Ohio State University
2015
Dissertation Committee:
Prof. Ronald M. Reano, Advisor
Prof. Joel Johnson
Prof. Gregory Lafyatis
Prof. Fernando L. Teixeira
© Copyright by
Li Chen
2015
Abstract
A hybrid silicon and lithium noibate (LiNbO3) material system is developed to combine the high index contrast of silicon and the second order susceptibility of lithium niobate. Ion-sliced single crystalline LiNbO3 thin film is bonded to silicon-on-insulator
(SOI) waveguides via Benzocyclobutene (BCB) as the top cladding. The LiNbO3 thin films are patterned to achieve desired size, shape and crystal orientations. Integrated electrodes are integrated to confine electric fields to the LiNbO3 thin film. Empowered by the linear electro-optic effect of LiNbO3, compact chip-scale hybrid Si/LiNbO3 integrated photonic devices are enabled on the SOI platform, including radio-frequency electric field sensors, tunable optical filters, high speed electro-optical modulators for optical interconnects, and high linearity modulators for analog optical links.
Compact and metal-free electric field sensors based on indirect bonding of z-cut ion- sliced LiNbO3 to silicon microrings are demonstrated. The demonstrated sensitivity to
-1 -1/2 electric fields is 4.5 V m Hz at 1.86 GHz. Tunable optical filters based on hybrid
Si/LiNbO3 microring resonators with integrated electrodes are also demonstrated with a tunability of 12.5 pm/V, over an order of magnitude greater than electrode-free designs.
By integrating metal thin film electrode and utilizing silicon as an optically transparent electrode, voltage induced electric fields in the LiNbO3 are enhanced. We also presented
ii low power compensation of thermal drift of resonance wavelengths in hybrid Si/LiNbO3 ring resonators. A capacitive geometry and low thermal sensitivity result in the compensation of 17 C of temperature variation using tuning powers at sub-nanowatt levels. The method establishes a route for stabilizing high quality factor resonators in chip-scale integrated photonics subject to temperature variations. Gigahertz speed hybrid
Si/LiNbO3 electro-optical microring modulators are enabled by optimizing the RC time constant of the biasing electrodes. Fabricated devices exhibit a resonance tuning of 3.3 pm/V and a small-signal electrical-to-optical 3 dB bandwidth of 5 GHz. Digital modulation with an extinction ratio greater than 3 dB is demonstrated up to 9 Gb/s.
High-speed and low tuning power chip-scale modulators that exploit the high-index contrast of silicon with the second order susceptibility of lithium niobate are envisioned.
An alternative design with x-cut LiNbO3 thin films on silicon racetrack resoantors enables compact highly linear integrated optical modulator for high spectral free dynamic range (SFDR) analog optical links. The measured third order intermodulation distortion
SFDR is 98.1 dB·Hz2/3 at 1 GHz and 87.6 dB·Hz2/3 at 10 GHz. The demonstrated SFDR is over an order of magnitude greater than silicon ring modulators based on the plasma dispersion effect, and is comparable to commercial LiNbO3 Mach-Zehnder interferometer modulators, but with a footprint three orders of magnitude smaller.
The hybrid Si/LiNbO3 photonic platform is promising for applications in optical interconnections, microwave photonics, optical computing and sensing. More broadly, empowering silicon with second-order susceptibility opens a suite of nonlinear optic applications to the chip scale.
iii
Dedication
To my parents.
iv
Acknowledgments
My PhD study has been an incredibly challenging and rewarding journey in my life. It was not only the accomplished projects and published papers, but also the great challenges and obstacles in the research that made the story memorable for the rest of my life. Here I would like to thank those who have supported and encouraged me, and helped me grow as a better researcher and a better man in the past years. Firstly, I am grateful for my parents for their everlasting love and encouragement in every stage of my life. I would like to express my sincere gratitude to my advisor, Professor Ronald M.
Reano, for his continuous support and guidance. His critical and rigorous approach toward research and science has steered me into a better researcher. This work would not have been accomplished without his help. I would also like to thank Professor Joel
Johnson, Professor Fernando L. Teixeira, and Professor Betty Anderson for their invaluable time and input while serving on my candidacy exam committee or dissertation committee.
I would like to thank my current and past colleagues in the research group: Michael
Wood, Justin Burr, Qiang Xu, Tyler Nagy, Jiahong Chen, Dr. Peng Sun, Dr. Alexander
Ruege, and Dr. Galen Hoffman, for their helpful advices, comments, and discussions. I would like to thank staff members at the Nanotech West Laboratory of the Ohio State
v
University, in particular, Aimee Price, Derek Ditmer, and Paul Steffen, for their help in the fabrication.
I also would like to acknowledge the funding support from Army Research Office
(ARO) under grant number W911NF-09-1-0073 and W911NF-12-1-0488.
vi
Vita
2008……………………………………………………. B. Eng. in Electrical Engineering Zhejiang University, China
2013…………………………………………………………M. Sc. Electrical Engineering The Ohio State University, Columbus, Ohio
2009 to present…………………………………………...….Graduate Research Associate The Ohio State University, Columbus, Ohio
Publications
Li Chen, Jonathan Nagy, and Ronald M. Reano, “Patterned Ion-Sliced Lithium Niobate for Hybrid Photonic Integration,” Opt. Mat. Express, 2015 (to be submitted)
Li Chen, Jiahong Chen, Jonathan Nagy, and Reano M. Reano, “Highly Linear Ring Modulator from Hybrid Silicon and Lithium Niobate,” Opt. Express 23, 2015 (accepted)
Li Chen, Michael G. Wood, and Ronald M. Reano, “Compensating Thermal Drift of Hybrid Silicon and Lithium Niobate Ring,” Opt. Lett. 40, 1599-1602, 2015
Li Chen, Qiang Xu, Michael G. Wood, Ronald M. Reano, “Hybrid Silicon and Lithium Niobate Electro-Optical Ring Modulator,” Optica 1, 112-118, (2014).
Qiang Xu, Li Chen, Michael G. Wood, Peng Sun, and Ronald M. Reano, “Electrically Tunable Polarization Rotation on A Silicon Chip Using Berry’s Phase,” Nat. Commun. 5:5337, (2014).
Michael G. Wood, Li Chen, Justin Bur, and Ronald M. Reano, “Optimization of Electron Beam Patterned HSQ Mask Edge Roughness for Low-Loss Silicon Waveguides,” J. Nanophotonics 8, 083098 (2014).
vii
Li Chen, Michael G. Wood, and Ronald M. Reano, “12.5 pm/V Tunable Hybrid Silicon and Lithium Niobate Optical Microring Resonator with Integrated Electrodes,” Opt. Express 21, 27003-27010 (2013).
Li Chen, and Ronald M. Reano, “Compact Electric Field sensors Based on Indirect Bonding of Lithium Niobate to Silicon Microrings,” Opt. Express 20, 4032-4038 (2012).
Fields of Study
Major Field: Electrical and Computer Engineering.
viii
Table of Contents
Abstract ...... ii
Dedication ...... iv
Acknowledgments...... v
Vita ...... vii
Table of Contents ...... ix
List of Figures ...... xiv
1. Introduction ...... 1
1.1 Background ...... 1
1.2 Silicon Photonics ...... 3
1.2.1 Silicon Modulators Based on Plasma Dispersion Effect ...... 3
1.2.2 Photonic Structures for Modulators ...... 8
1.2.3 Second Order Susceptibility on Silicon ...... 10
1.2.4 Hybrid Integration on Silicon-on-Insulator ...... 12
1.3 LiNbO3 Integrated Photonics ...... 14
1.3.1 Bulk Lithium Niobate Photonics ...... 14
ix
1.3.2 LiNbO3 Thin Film ...... 16
1.4 Hybrid Silicon and Lithium Niobate Photonics ...... 18
1.5 Contributions of This Dissertation ...... 21
1.5.1 Hybrid Integration Process ...... 23
1.5.2 Hybrid Silicon/LiNbO3 Electric Field Sensor...... 24
1.5.3 Hybrid Silicon/LiNbO3 Microring Resonator Filter ...... 25
1.5.4 Hybrid Silicon/LiNbO3 Microring Resonator Modulator ...... 26
1.6 Organization of the Dissertation ...... 27
2. Development of Fabrication Processes ...... 30
2.1 Silicon Photonic Process ...... 31
2.1.1 Low Loss Silicon Waveguide and Coupler ...... 31
2.1.2 Active Silicon Photonic Components ...... 34
2.2. Lithium Niobate Thin Film Process ...... 36
2.2.1 Ion Implantation ...... 37
2.2.2 LiNbO3 Thin Film Exfoliation ...... 42
2.3 Hybrid Silicon and LiNbO3 Integration ...... 52
2.3.1 BCB for Indirect Wafer Bonding ...... 53
2.3.2 Indirect Bonding of LiNbO3 Thin Films to Silicon ...... 57
2.4 Dry Etching of LiNbO3 ...... 62
x
2.4.1 Etching LiNbO3 with ICP-RIE ...... 63
2.4.2 Etching LiNbO3 with Focused Ion Beam Bombardment ...... 64
2.5 Chapter Conclusion and Outlook ...... 66
3. Hybrid Silicon and LiNbO3 Microring Electric Field Sensor ...... 68
3.1 Photonic Electric Field Sensors ...... 69
3.2 Device Design and Fabrication ...... 70
3.4 Device Characterization ...... 75
3.4.1 Measurement Setup and RF Measurement ...... 75
3.4.2 Sensitivity Calculation ...... 79
3.4.2 Optical Power Dependence ...... 83
3.5 Chapter Conclusion and Outlook ...... 86
4. Low Voltage Tunable Hybrid Silicon and LiNbO3 Microring Resonator ...... 87
4.1 Introduction ...... 88
4.2 Device Design ...... 90
4.3. Device Fabrication ...... 94
4.4. Measurements ...... 97
4.4.1 DC measurement ...... 97
4.4.2 Compensation of Thermal Drift ...... 100
4.5 Chapter Conclusion and Outlook ...... 104
xi
5. High Speed Hybrid Silicon and LiNbO3 Microring Modulator...... 106
5.1 Introduction ...... 107
5.2 Device Design and Fabrication ...... 108
5.2.1 Device Design ...... 108
5.2.2 Device Fabrication ...... 112
5.3 Device Characterization ...... 115
5.3.1 DC and Small-Signal High-Frequency Measurements ...... 115
5.3.2 High-Speed Digital Modulation ...... 118
5.4 Acousto-optic anomalies ...... 120
5.5 Chapter Conclusion and Outlook ...... 125
6. Highly Linear Hybrid Silicon and LiNbO3 Racetrack Modulator for Analog Optical
Links ...... 127
6.1 Introduction ...... 128
6.2 Theory ...... 129
6.3 Device Design ...... 131
6.4 Device Fabrication ...... 134
6.5 Measurement ...... 136
6.5.1 DC measurement ...... 136
6.5.1 RF measurement ...... 138
xii
6.6 Chapter Conclusion and Outlook ...... 145
7. Thesis Conclusion and Suggestions for Future Work...... 146
7.1 Thesis Conclusion ...... 146
7.2 Suggestions for Future Work ...... 147
7.2.1 Suppression of Acoustic-Optical Resonances ...... 147
7.2.2 Low Power Athermal Microring Modulator ...... 148
7.2.3 Hybrid Silicon and LiNbO3 MZI Modulator ...... 149
7.2.4 Improve the hybrid integration process ...... 149
7.2.5 Nonlinear Optics on Silicon ...... 150
Bibliography ...... 151
APPENDIX A: Fabrication Processes for Hybrid Silicon and LiNbO3 Microring
Modulator ...... 166
A.1 Silicon Waveguide and Slab ...... 166
A.2 Contact and Bottom Electrode ...... 168
A.3 LiNbO3 Thin Film ...... 170
A.4 Bonding Process ...... 171
A.5 Via and Top Electrode ...... 171
A.6 Cantilever Coupler ...... 173
xiii
List of Figures
Figure 1. Optical interconnects evolution chart ...... 2
Figure 2. Photonics for RF applications: (a) microwave signal processing [4], (b) fiber
wireless system [5] and (c) communication receiver [6]...... 2
Figure 3. Cross-sections of typical silicon waveguide modulator structure implementing
a. carrier accumulation, b. carrier injection, and c. carrier depletion [14]...... 5
Figure 4. (a) Carrier injection based silicon microring pin modulator [11], (b) Carrier
depletion based silicon microring reverse pn junction modulator [23]...... 6
Figure 5. Sources for second order susceptibility on the SOI platform with (a) electro-
optic polymer [56], (b) Barium titanate (BTO) [62], (c) plasma treatment [60]
and (d) strained silicon [58]...... 11
Figure 6. Hybrid silicon modulators using (a) silicon germanium [68], (b) III-V
semiconductors [76], (c) graphene [70], and electro-optic polymer with
plasmonics [72]...... 13
Figure 7. (a) A packaged commercial LiNbO3 MZI modulator; (b) the cross-sectional
view of the waveguide and the electrode ...... 15
Figure 8. A LiNbO3 microring fabricated on the LiNbO3-on-insulator platform with the
top-down view (a), and cross-section view (b) [88]...... 18
xiv
Figure 9. Schematic of a hybrid silicon and LiNbO3 microring resonator with direct
bonding [92]...... 19
Figure 10. Challenges and advantages of the hybrid silicon/LiNbO3 technology ...... 20
Figure 11. Summary of the challenges with modulation on the silicon-on-insulator
platform...... 21
Figure 12. Summary of the approach and results of the dissertation based on the hybrid
silicon and LiNbO3 platform ...... 22
Figure 13. Scanning electron micrographs of unclad silicon strip waveguides with top-
down view (a) and cross section view (b), Si waveguide cladded with PECVD
SiO2 (cross-section) (c), and cantilever coupler (d)...... 33
Figure 14. EDS of nickel silicide formed on the silicon slab of a SOI substrate. (a) SEM
of 5µm wide NiSi section, (b) EDS of Si element, (c) EDS of Ni element, and
(d) EDS of O element. EDS is measured along the 10 µm yellow line shown in
(a)...... 35
+ Figure 15. Simulated ion implantation projected range of He ion in LiNbO3. (a)
Implantation energy from 0 to 400 keV, (b) Implantation energy from 0 to 4000
keV...... 38
Figure 16. (a) LiNbO3 samples mounted to an aluminum carrier using silver paste; (b)
Implantation wafer holder; (c) Mounting the wafer holder; (d) Ion implanter. . 39
Figure 17. Measured wafer holder temperature versus implantation duration ...... 41
Figure 18. (a) Intentionally formed surface microcracks on implanted z-cut LiNbO3 wafer
before wet etching, (b) Angled view SEM image of the implanted LiNbO3
xv
sample showing the cross section, (c) HF undercut etching from the
microcracks, (d) Angled view SEM image of partially released LiNbO3 thin
films coated with 50 nm Chromium on a LiNbO3 substrate ...... 43
Figure 19. A z-cut LiNbO3 thin film released by wet etching and transferred to the
unpolished side of a silicon substrate. (a) Before annealing, (b) after annealing
at 1000oC for 30s with RTA...... 45
Figure 20. Exfoliated LiNbO3 thin films using the thermal blistering process, (a) on the
implanted z-cut sample after annealing, (b) transferred to unpolished silicon
substrate...... 47
Figure 21. (a) X cut LiNbO3 thin film fabrication flow. (a) He+ ion implantation on wafer
1, (b) PECVD SiO2 deposition on wafer 1,(c) BCB spin-coating on wafer 2, (d)
wafer bonding and annealing, (e) exfoliation of LiNbO3 thin film, (f) patterning
chromium mask, (g) dry etching of LiNbO3, (h) wet etching of PECVD SiO2, (i)
Transferring LiNbO3 to a unpolished silicon substrate...... 49
Figure 22. (a) X-cut LiNbO3 thin film bonded to a LiNbO3 handle substrate; (b) Scanning
electron micrograph of the cross-section of the thin film stack ...... 51
Figure 23. (a) Hydrofluoric acid etching of the patterned x-cut LiNbO3 thin film; (b)
Released LiNbO3 thin film transferred to the unpolished surface of a silicon
substrate ...... 52
Figure 24. Polymerization process of the DVS-BCB monomer [103] ...... 56
Figure 25. Complex refractive index of BCB measured with ellipsometer ...... 56
xvi
Figure 26. (a) Schematic of the glass micro-vacuum tip, (b) Top-down view microscope
image of LiNbO3 thin films on a polydimethylsiloxane (PDMS) substrate with a
micro-vacuum tip and a metal tip hovering over ...... 58
Figure 27. Time-temperature transformation isothermal cure diagram of BCB [103] ..... 59
Figure 28. Bonding of a LiNbO3 thin film on a bare silicon substrate coated with partially
cured BCB, (a) before hard curing, (b) after hard curing ...... 60
Figure 29. (a) Micrograph of silicon microrings with a LiNbO3 thin film bonded on the
top via BCB, (b) Tilted view SEM of two holes opened over the silicon ring
waveguides using focused ion beam...... 62
Figure 30. (a) Photoresist after development, (b) cross-section of 300 nm chromium mask
on LiNbO3 substrate after lift-off, (c) top-view of edged LiNbO3 ridge, (d)
Cross section of the etched LiNbO3 ridge...... 64
Figure 31. (a) Photonic crystal structures milled on LiNbO3 thin film, (b) microdisks
milled on LiNbO3 thin film, (c) a zoom-in view of the microdisk, (d) LiNbO3
microdisk transferred to silicon waveguide using a nano-manipulator ...... 65
Figure 32. Applications of electric field sensors. (a) Near field probe for RF circuit
diagnosis (Aaronia AG, Euscheid, Germany), (b) DC to 6 GHz RF probe for
EMC/EMI analysis (AFJ instruments, Milan, Italy), (c) SAR measurements for
biological application [107], (d) dielectric photonic receivers for
communications [108] ...... 69
Figure 33. (a) Schematic of an electric field sensor based on the indirect bonding of a
lithium niobate thin film to a silicon microring resonator. For clarity, a PECVD
xvii
SiO2 top-cladding layer is not shown. (b) SEM of the cross-section of the sensor
structure...... 70
Figure 34. Schematic of the sensing principle using a tunable microring resonator ...... 71
Figure 35. Optical electric field distributions in the hybrid Si/LiNbO3 sensor for the
quasi-TM (Ey component) and quasi-TE (Ex component) modes at 1550 nm
wavelength. Material boundaries are indicated by the white dashed lines and
the material regions are indicated in the inset...... 72
Figure 36. Fabrication process of electric field sensor: (a) Silicon strip waveguide ring
resonator patterned on SOI wafer using electron beam lithography and plasma
etch, (b) spin-coat of BCB, (c) indirect bonding of LiNbO3 thin film, (d) plasma
etch of BCB, (e) deposition of PECVD SiO2, (f) fabrication of cantilever
couplers...... 74
Figure 37. (a) Scanning electron micrograph of a SOI ring resonator with a diameter of
40 μm. (b) Zoom-in view of the coupling section with a gap of 350 nm...... 75
Figure 38. (a) Schematic of the measurement setup, (b) Top-view optical micrograph of
fabricated electric field sensor, (c) photo of the measurement setup...... 76
Figure 39. (a) Measured optical transmission of the electric field sensor; (b) Magnitude of
measured microwave VNA S21 RF scattering parameter versus CW laser
wavelength; (c) Corresponding phase of RF S21 versus CW laser wavelength. 78
Figure 40. (a) Photo of the microstrip circuit, (b) Electric field (vertical component)
distribution on the surface of the circuit at 1.86 GHz, (c) Simulated and
measured S11 spectrum of the RF circuit ...... 80
xviii
Figure 41. The vertical component of the electric field at one micron above the Si
substrate over the microstrip metal trace corner with 1 W of input power at 1.86
GHz. The circuit with white dashed line shows the possible position of the
sensor considering the measurement accuracy. The area of the simulation
domain is 2×2 mm...... 81
Figure 42. Flow diagram for calculating the sensor sensitivity ...... 83
Figure 43. Measured optical resonances with various optical power level. Optical
biastability is observed for high optical power...... 84
Figure 44. (a) Measured optical transmission for 0 dBm, -3dBm, -4.5 dBm optical power
(b) Magnitude of measured microwave VNA S21 RF scattering parameter versus
CW laser wavelength; (c) Corresponding phase of RF S21 versus CW laser
wavelength...... 85
Figure 45. Measured S21 magnitude maximum for various optical output power level ... 85
Figure 46. (a) Schematic of a tunable hybrid silicon and LiNbO3 microring resonator with
integrated electrodes. For clarity, the PECVD SiO2 top-cladding layer and
electrical contact pads are not shown. (b) Schematic of the cross-section of the
device structure along the dashed line shown in (a). Both schematics are not
drawn to scale...... 90
Figure 47. (a) BPM calculations of the optical mode power in LiNbO3 versus the
thickness of LiNbO3 for the TM mode and the TE mode in the hybrid
Si/LiNbO3 structure. (b) Calculations of the optical loss (blue) induced by the
top aluminum electrode and the voltage induced vertical electric field in LiNbO3
xix
(red) versus the PECVD silicon dioxide thickness. The LiNbO3 thin film
thickness is set to be 800 nm and the applied voltage is 1 V...... 92
Figure 48. (a) Calculated TM mode optical bending loss versus the ring radius for BCB
thickness of 0 nm, 20 nm, and 40 nm between the top of the silicon core and the
bottom of the LiNbO3 thin film. (b) Calculated optical electric field distribution
of the hybrid silicon and LiNbO3 structure for the fundamental TM mode at
1550 nm wavelength (Ez component)...... 94
Figure 49. Fabrication process of the device: (a) Silicon rib waveguide ring resonator
patterned on SOI wafer using electron beam lithography and plasma etch, (b)
+ BF2 ion implantation, (c) nickel silicidation, (d) 100 nm aluminum bottom
electrode deposition, (e) spin-coat of BCB, (f) indirect bonding of LiNbO3 thin
film, (g) plasma etch of BCB, (h) deposition of 125 nm PECVD SiO2
(illustrated on top of the LiNbO3 only for simplicity), (i) deposition of 250 nm
top aluminum electrode, (j) deposition of 900 nm PECVD SiO2, (k) patterning
of via and top aluminum pad, (l) fabrication of cantilever couplers...... 95
Figure 50. (a) Measurement setup; (b) Top-view optical micrograph of fabricated device.
...... 98
Figure 51. (a) Measured optical transmission of a single resonance as a function of
applied voltage. The measured resonance tunability is 12.5 pm/V. (b)
Measured optical transmission spectrum of two consecutive resonances...... 99
Figure 52. (a) The temperature change of the TEC as a function of applied bias, (b)
Measured resonance detuning versus temperature...... 100
xx
Figure 53. TM mode optical transmission of Si/LiNbO3 ring resonator for applied voltage
from -30 V to 50 V for a temperature increase of (a) 0 oC, (b) 5 oC, (c) 10 oC and
(d) 15 oC above ambient temperature...... 102
Figure 54. Measured TM mode resonance wavelength versus applied voltage, together
with linear fit, with temperature as parameter...... 103
Figure 55. Schematic of the hybrid silicon and LiNbO3 ring modulator...... 108
Figure 56. (a) Calculated optical TE mode distribution at 1550 nm wavelength (Ex
component) and DC voltage induced electric field vectors. (b) Calculation of
carrier induced optical loss, RC limited bandwidth, and waveguide serial
resistance (top axis) versus silicon waveguide doping concentration with P-type
dopants, for a 15 µm radius ring...... 111
Figure 57. (a) Scanning electron micrograph of the silicon microring resonator after slab
patterning and doping; (b) Top-view optical micrograph of fabricated device.
...... 114
Figure 58. Measured optical transmission of a single resonance as a function of applied
voltage...... 115
Figure 59. (a) RF S11 scattering parameter; (b) RC circuit model of the modulator...... 117
Figure 60. Electrical-to-optical modulation response...... 117
Figure 61. Measurement setup for digital characterization ...... 118
Figure 62. Measured (left column) and simulated (right column) optical eye: (a) and (b) 1
Gb/s, (c) and (d) 4.5 Gb/s, (e) and (f) 5 Gb/s, (g) and (h), 9 Gb/s. The red dashed
xxi
line in the measurement indicates the reference level for zero optical input. The
vertical scale is 500 μW per division...... 119
Figure 63. 2D numerical acoustic simulation of a simplified model. (a) The structure and
meshing before applying voltage on the aluminum electrode. (b) Distortion at
the first order shear mode resonance (c) Distortion at the first order longitudinal
mode resonance. The color bar shows the normalized amplitude of the
displacement...... 123
Figure 64. Measured optical response versus simulated normalized admittance based on
the simplified 2D model in Fig. 8 (a). S1 to S6 represents the first order shear
mode resonance and its higher order harmonics. L1 to L3 represents the first
order longitudinal mode resonance and its higher order harmonics...... 124
Figure 65. (a) Schematic of hybrid silicon and LiNbO3 racetrack modulator. For clarity,
the PECVD SiO2 top-cladding layer and contact pads are not shown. (b)
Schematic of the cross-section of the device structure along the dashed line in
(a). The crystal axes of LiNbO3 is marked...... 131
Figure 66. Calculated optical TE mode distribution at 1550 nm wavelength (Ex
component) and DC voltage-induced electric field vectors...... 133
Figure 67. Fabrication process of the device: (a) Silicon strip waveguide racetrack
resonator patterned on SOI wafer using EBL and plasma etching, (b) spin-coat,
partial curing, and etch back of BCB, (c) transferring and bonding of patterned
x-cut LiNbO3 thin film and plasma etch of BCB, (d) Deposition of 1 µm
xxii
PECVD SiO2 and patterning of via, (e) patterning of signal electrode, (f)
patterning of ground electrode and cantilever couplers...... 135
Figure 68. (a) Top-view optical micrograph of the fabricated device; (b) scanning
electron micrograph with a zoom-in view of the top electrodes...... 136
Figure 69. (a) Measured optical spectrum as a function of applied voltage; (b) linear
fitting of the resonance wavelength shift as a function of applied voltage ...... 137
Figure 70. Measured optical modulation response and the S11 magnitude ...... 139
Figure 71. Setup for the SFDR measurement ...... 140
Figure 72. RF output power of the fundamental, SHD, and IMD3 components as a
function of wavelength detuning from resonance wavelength at 100 MHz (a),
1GHz (b), 5GHz (c), and 10 GHz (d) respectively...... 141
Figure 73. RF output power of the fundamental and IMD3 components (the higher spur
of the two spurs) as a function of RF input power for the LiNbO3 MZI
modulator and the hybrid silicon and LiNbO3 racetrack modulator at 0.997
GHz. The noise floor is in 1 Hz bandwidth, limited by the RF spectrum
analyzer...... 143
Figure 74. RF output power of the fundamental and IMD3 components (the higher spur
of the two spurs) as a function of RF input power for the LiNbO3 MZI
modulator and the hybrid silicon and LiNbO3 racetrack modulator at 9.997
GHz. The noise floor is in 1 Hz bandwidth, limited by the RF spectrum
analyzer...... 144
xxiii
CHAPTER 1
Introduction
1.1 Background
Photonics technologies play a significant role in today’s information age. The applications of photonics include but not limited to optical communications, optical interconnects, signal processing and RF photonics. In the last few decades, starting from the fiber long haul optical communications, optical interconnects have gradually replaced copper interconnects due to its advantage of the bandwidth distance product [1, 2].
Photonics has been moving closer and closer to the processor and memory of computers.
Rack-to-rack optical interconnects has been employed in commercial super computers and servers, using III-V semiconductor vertical cavity surface emitting laser (VCSEL) transmitters and photodiode receivers [3]. Figure 1 shows the evolution chart of the optical interconnects from telecom to computercom in the past 30 years.
Besides optical interconnects, optics also has wide applications in the radiofrequency
(RF) domain [4, 5, 6]. The interaction of optics and RF enables complex functionalities in RF systems that are not directly possible in the RF domain, creating new opportunities for telecommunication networks. Figure 2 shows several examples of the application of photonics in RF.
1
Figure 1. Optical interconnects evolution chart
Figure 2. Photonics for RF applications: (a) microwave signal processing [4], (b) fiber wireless system [5] and (c) communication receiver [6].
2
1.2 Silicon Photonics
As optics is moving to the board-to-board level, chip-to-chip level and eventually to on-chip level, silicon photonics becomes a viable solution for future optical interconnections due to the compatibility to the CMOS processing technology, the feasibility of dense integration, and the high-bandwidth and high-density I/O capability enabled by the compact waveguide dimensions [1, 2]. Silicon integrated photonics provides high-bandwidth and low-power on-chip and intra-chip optical interconnections.
The past decade has witnessed tremendous research progress of silicon photonics, including high speed modulators and photodetectors and other components required for building the optical links and switching fabrics [7, 8]. A typical silicon photonic link requires an external laser source, a silicon modulator, a guiding waveguide and a silicon photodetector. The silicon modulator is a critical component in an optical link for converting the signal from the electrical domain to the optical domain. Besides optical interconnects, silicon photonics also finds wide applications in microwave photonics, optical signal processing, and sensing.
1.2.1 Silicon Modulators Based on Plasma Dispersion Effect
Optical modulation in silicon is mainly based on tuning the refractive index of silicon.
However, unstrained pure crystalline Si does not exhibit a linear electrooptic (Pockels) effect and the Franz–Keldysh effect and the Kerr effect are very weak [9]. As a result, electro-refraction resonance tuning mechanisms in silicon include tuning the refractive
3 index of silicon by the thermo-optic effect [10], and the plasma dispersion effect [11, 12].
Thermo-optic effect is inherently slow thereby is not suitable for high speed applications.
The plasma dispersion effect is associated with the free carrier density in a semiconductor. The refractive index and absorption loss of silicon is strongly related to the charge carriers in the silicon. Based on the Drude model, the change of refraction index and absorption in c-Si due to free electrons and free holes are described as:
2 2 2 2 * * n ( e / 8 c 0 n )[ Ne / m ce N h / m ch ] 3 2 2 3 *2 *2 (1) (e / 4 c 0 n )[ Ne / m ce e N h / m ch h ] where e is the electronic charge, ε0 is the permitivity in vacuum, n is the refractive index
of unperturbed c-Si, mce* is the conductivity effective mass of electrons, mch*is the conductivity effective mass of holes, µe is the electron mobility, and µh is the hole mobility. At the wavelength of 1550 nm, the refractive index and absorption change are quantified experimentally, and expressed as [13]:
22 18 0.8 n ne n h[8.8 10 n e 8.5 10 ( h ) ] 18 18 (2) e h8.5 10 n e 6.0 10 h
Charge density modulation can be achieved through mechanisms such as carrier injection, accumulation and depletion, as shown in Figure 3. Cross-sections of typical silicon waveguide modulator structure implementing a. carrier accumulation, b. carrier injection, and c. carrier depletion [14]. The carrier injection method has high tuning efficiency in a p-i-n structure, but the speed is limited by the carrier life-time [11]. The carrier depletion method can be much faster but the efficiency is relatively low. Devices
4 based on the carrier accumulation have least efficiency and requires more complex fabrication process [15].
Figure 3. Cross-sections of typical silicon waveguide modulator structure implementing a. carrier accumulation, b. carrier injection, and c. carrier depletion [14].
1.2.2.1 Silicon Modulators for Optical Interconnects
Silicon intensity modulators are one of the key components for on-chip and off-chip optical interconnects. High speed and low power silicon modulators have been demonstrated for both short-reach and long haul interconnects [1, 11, 14, 16]. Figure 4 shows two types of microring modulators based on carrier injection and carrier depletion effect, respectively. Carrier injection based silicon modulators exhibit very large tuning efficiency, result in compact device footprint and low operation voltage [11, 17].
However, the carrier lifetime limits the device speed [11]. Modulation speed can be increased using a pre-emphasis signal at the cost of increased power consumption and complex driving circuits [18, 19]. Carrier depletion based silicon modulators have the most potential for future on-chip silicon photonic interconnects [20]. Modulation speed as high as 50 Gb/s and low VπL value smaller than 0.5 V·cm have been achieved [21, 5
22]. While effective, plasma dispersion relies on carrier transport. Consequently, silicon refractive index is accompanied by absorption, and steady-state electrical tuning of optical resonances can consume significant power. The large insertion loss induced by the carrier absorption loss limits the integration density of silicon photonic integrated circuits supplied by single optical source.
Figure 4. (a) Carrier injection based silicon microring pin modulator [11], (b) Carrier depletion based silicon microring reverse pn junction modulator [0].
1.2.1.2 Silicon Modulators for Microwave Photonics
Development of microwave photonics has been constantly driven by the expanding broadband wireless access networks and the growth of fiber links directly to the home
[24]. Traditional microwave photonic systems rely on discrete components that are bulky, expensive, and power consuming. In contrast, integrated microwave photonics systems based on photonic integrated circuits (PIC) provides advantages in cost, size,
6 power consumption and reliability [25, 25]. Among many PIC platforms, silicon photonics has the potential for large scale photonic/electronic integration due to its large index contrast and compatibility with the mature silicon IC manufacturing [26, 27].
Silicon photonic devices with RF photonic functionalities such as filtering and arbitrary waveform generation have been demonstrated [28, 29].
However, it is challenging to realize high linearity and compact silicon modulators for high dynamic range integrated microwave photonic links [30]. Silicon modulators commonly rely on the plasma dispersion effect in pin or pn junctions that exhibit nonlinear phase change with applied voltage [13, 11]. The nonlinear tuning behavior
(phase tuning is not linear with applied voltage) is not desirable for high dynamic range analog optical link application. The nonlinearity of the carrier effects generally dominates the nonlinearity in silicon modulators [31]. Spurious free dynamic range
2/3 2/3 (SFDR) of 84 dB·Hz and 97 dB·Hz for third order intermodulation distortion (IMD3) are demonstrated for a silicon microring modulator and a silicon traveling wave Mach-
Zehnder interferometer (MZI) modulator, respectively [31, 32]. Several approaches have been proposed to increase the linearity of silicon MZI modulators, including compensating the nonlinearity of the MZI transfer function with the nonlinearity of the silicon PN junction by proper biasing scheme, and adopting a ring assisted MZI structure
[33-35]. Compared to MZI based structures, microrings allows for denser integration and lower power consumption. To date, a high linearity and compact microring modulator has not been demonstrated on the silicon photonics platform.
7
1.2.2 Photonic Structures for Modulators
Integrated photonic modulators are based on either electro-refraction or electro- absorption effect. For the absorptive modulators, the loss of a section of waveguide is modulated, resulting in intensity modulation. Usually no special photonic structures are required for absorptive modulators. In contrast, refractive modulators modulate by interfering an optical beam with itself, either in a single-pass two-beam interference as a
MZI structure or in resonance based cavities, such as microring resonator and photonic crystals. MZI modulators have wide wavelength bandwidth. However, due to the weak plasma dispersion effect, the phase shifter arms are on the order of a few millimeters and the required power consumption is on the pJ/bit level. In contrast, resonance based modulators are compact and power efficient. Silicon microdisk modulators with power consumption of 1 fJ/bit and device diameter of 4.8 µm have been demonstrated [0].
Among the various silicon photonic structures, silicon microring resonator is one of the most important components for compact and low-power optical filters [15], switches
[37], and modulators [14] for wavelength-division multiplexing (WDM) optical links in future optical interconnects. Besides, silicon microring resonators have been widely utilized for on-chip sensing applications, including temperature sensors, biochemistry sensors and so on [38, 39]. Thanks to the high index contrast of the SOI waveguide, the size of the microring resonators can be only a few microns [40]. While resonances produce high sensitivity and low power consumption, they also result in susceptiblity to ambient temperature variations due primarily to the large thermo-optic coefficient (TOC) of silicon [41]. The thermal sensitivity of a single mode silicon microring resonator (with
8 a cross-section dimension of about 450 nm by 250nm) clad with silicon dioxide is about
100 pm/K [42], making performance of the device greatly degraded even with a temperature change of 1K. Consequently, devices are not practical without thermal compensation. Thermal challenges need to be resolved in order to advance resonator based photonics for future network-on-chip computing systems.
One direction for addressing the thermal challenges focuses on passively reducing the thermal sensitivity of ring resonators, to achieve athermal operation, by incorporating materials with a negative TOC or by using compensating passive optical circuits [41-45].
Passive approaches generally lead to delocalized optical modes and large ring radii, reducing the benefit of high optical confinement in silicon waveguides, or requires larger footprint for interferometry. Although the techniques compensate for temperature drift of resonance wavelengths, appreciable detuning from target wavelengths remains inevitable due to manufacturing variation [46, 47]. As a result, active thermal compensation methods are required to stabilize resonances to a particular wavelength. Local temperature control using integrated resistive heaters or tuning via pn junctions are examples of active techniques [22, 0, 48-51].
Thermal tuning of resonance wavelengths results in large tuning range, however, the method requires additional contacts, is relatively slow, and consumes significant power
[52]. The power required to tune a free spectral range (FSR) can be reduced to a few milliwatts using undercut-etched or backside-etched waveguides at the cost of enhanced optical bistability that limits the optical power handling [48, 53, 54]. Furthermore, thermal tuning is unidirectional. If the resonance is close to and higher than the operating
9 wavelength, the resonance has to be red shifted by almost a FSR to the next order resonance. In contrast, wavelength stabilization based on carrier transport in pn junctions consumes power from microwatt to milliwatt levels [50]. For carrier injection based devices, the tuning range is, however, limited by resistive heating [11]. For carrier depletion based devices, the tuning is limited by doping concentration and junction breakdown. In both cases, the resonance lineshape is affected by carrier absorption.
While the required power consumption for the existing active approaches may be acceptable for current small-scale circuits, future VLSI photonic integrated circuits will demand power budgets that are not achievable on-chip [1, 55].
1.2.3 Second Order Susceptibility on Silicon
Tremendous research effects have been made to fuel silicon photonics with the second order nonlinearity as an important source for optical modulation and nonlinear optics applications. Approaches include integration of electro-optic polymers [56], breaking the centrosymmetry of the silicon crystal lattice by strain [57-59], plasma activation [60], and epitaxial growth of ferroelectric oxide material [61, 62]. The χ(2) effect on the SOI platform opens avenues for varieties of functionalities on a silicon chip, including optical modulation, RF electric field sensing [63], difference frequency generation of terahertz signal [60], and second harmonic generation of mid-IR light source [64]. The second order susceptibility on silicon enables high speed and low power silicon modulators potentially with much higher linearity and low insertion loss.
10
The χ(2) effect on the SOI platform has enabled low-power high-speed modulators [56,
58, 62]. However, it remains a challenge to achieve a platform that yields large χ(2) coefficient, high thermal stability, and low optical loss. Strained silicon has a relatively weak and non-unifrom electro-optic effect that is strongly dependent on the device geometry [59]. Reported values of VπL for strained silicon modulators are on the order of 100 V-cm, resulting in large device footprint, high operating voltage and high power consumption.
Figure 5. Sources for second order susceptibility on the SOI platform with (a) electro- optic polymer [56], (b) Barium titanate (BTO) [62], (c) plasma treatment [60] and (d) strained silicon [58].
11
On the other hand, hybrid silicon and polymer slot waveguide modulator enables very low operating voltage, thanks to the high electric field intensity in the slot region, and the large electro-optic effect in electro-optic polymers. However, electro-optic polymers suffer from thermal stability issue. The electro-optic properties decay at elevated manufacturing and operating temperatures above 120o C [65]. Barium titanate oxide
(BTO) material epitaxially grown on silicon possesses large electro-optic coefficient and high thermal stability. However, the large dielectric constant of BTO (>1000) results in large device capacitance, limiting the device bandwidth. In addition, residual oxygen vacancies in the BTO thin film generate significant optical loss [62].
1.2.4 Hybrid Integration on Silicon-on-Insulator
Hybrid material systems consisting of silicon and other materials have been extensively studied to improve the versatility and functionality of silicon photonics. The materials include polymers, III-V semiconductors, germanium, ferroelectric oxides, metal oxides, metals, and graphene and so on [66-73]. The lack of direct bandgap and second order susceptibility in silicon are the two major underlying reasons driving the research trend. Hybrid silicon photonic devices such as lasers, detectors, modulators, amplifiers, attenuators, tunable filters, and polarization diversity devices have been demonstrated
[74-78].
Hybrid silicon modulators also have been achieved taking advantage of the Pockel’s effect or electro-absorption effect of the integrated materials. Besides the hybrid silicon/polymer and hybrid silicon/BTO modulators discussed in the last section, hybrid
12 silicon electro-absorption modulators have been demonstrated using germanium, silicon germanium, III-V semiconductors, vanadium dioxide, and graphene, with some examples shown in Figure 6 [68, 70, 73, 76]. The hybrid electro-absorption modulators are realized by tuning the absorption of integrated materials. For light coupling from silicon to the integrated materials, a hybrid waveguide consist of silicon core and integrated materials in optical proximity to the silicon becomes the most popular design. Alternatively, waveguides with a core of the integrated materials can be coupled to a silicon waveguide using a waveguide coupler.
Figure 6. Hybrid silicon modulators using (a) silicon germanium [68], (b) III-V semiconductors [76], (c) graphene [70], and electro-optic polymer with plasmonics [72].
13
Hybrid silicon modulators based on the electro-absorption effect can be compact and very fast [79]. The electro-absorption modulators depend on the loss of the waveguide to produce on-off modulation [80]. As a result, the phase information is lost thereby electro- absorption modulators are not suitable for phase modulation. The phase modulation is critical for resonance and interference based tunable optical structures, such as microring resonator and Mach-Zehnder interferometers for application in wavelength reconfigurable optical devices (filters, multiplexer, and de-multiplexer) and phase modulators. In particular, electro-absorption modulators are not suitable for coherent communications that requires both the intensity and phase information of the light carrier.
1.3 LiNbO3 Integrated Photonics
1.3.1 Bulk Lithium Niobate Photonics
Decades before the emergence of the silicon photonics, lithium niobate (LiNbO3) had become an important wave-guiding material for integrated optics [81]. LiNbO3 is a widely exploited synthesized crystal material that does not exist in nature. It has a trigonal crystal structure and has excellent physical properties, including large electro- optic, pyroelectric, piezoelectric, and photoelastic coefficients [82, 83]. In addition, it has low optical absorption loss (~ 0.2 dB/cm ) at a wide wavelength span from around 0.4 micron to 4 micron. As a result, LiNbO3 finds wide applications in optical communication industry, including optical modulators, second-harmonic generators, Q- switches, beam deflectors, acoustic-optical devices, holographic data processing devices, and others. Due to the trigonal crystal structure, LiNbO3 is an anisotropic material. The
14 material is generally grown along either the x or z axis corresponding to the crystallographic a and c axes. In addition, LiNbO3 is birefringent with a refractive index difference of ~ 0.1 between the ordinary and extra-ordinary refractive indices. LiNbO3 integrated optic devices are usually polarization dependent.
Figure 7. (a) A packaged commercial LiNbO3 MZI modulator; (b) the cross-sectional view of the waveguide and the electrode
-1 -1 LiNbO3 has large electro-optic coefficients (r33 = 31 pm V , r31 = 8 pm V at 1550 nm wavelength in bulk LiNbO3). LiNbO3 integrated modulators have been key components for optical communications and signal processing applications. Lithium niobate (LiNbO3) guided-wave electro-optic modulators satisfy bandwidth, linearity, and chirp requirements in fiber-optic transmission systems [83]. Figure 7 shows a commercial LiNbO3 Mach-Zehnder interferometer (MZI) modulator based on diffused waveguide on a bulk LiNbO3 substrate. Traditional LiNbO3 waveguides are formed on the surface of a bulk wafer by Ti-diffusion or proton exchange to induce local increase of the refractive index in the core region. The index contrast between the core and cladding is so small (~ 0.005) that compact integrated optic structures such as sharp bends and 15 small-radii microring resonators cannot be realized using traditional LiNbO3 waveguides.
LiNbO3 integrated amplitude modulators are usually based on the MZI structure that are a few centimeters long. The potential for dense integration is limited.
1.3.2 LiNbO3 Thin Film
LiNbO3 thin film provides many advantages over LiNbO3 bulk wafers. Single crystalline LiNbO3 thin film (a few microns thick) can be produced in several methods.
It can be obtained by bonding a wafer to another handle wafer and polishing and thinning the top wafer. However, un-uniform thin film thickness and thin film cracking are often resulted from this method. Uniform submicrometer thick crack-free LiNbO3 thin film is almost impossible to obtain. Alternatively, thin films of single crystalline LiNbO3 with a thickness from sub-micrometer to around 10 μm can be exfoliated from the bulk substrate
+ using the crystal ion slicing process [85, 86]. LiNbO3 bulk wafer is implanted with He ions with a heavy dose (~ 4× 1016 /cm2) to form a damaged layer below the top surface.
The depth of the damaged layer is controlled by the ion implantation energy. The
LiNbO3 thin film can be separated from the substrate by hydrofluoric acid etching or thermal treatment. The thin film size obtained without a handle wafer is limited since micrometer thickness LiNbO3 thin films are very fragile. A LiNbO3-on-insulator system consisting of LiNbO3 thin films bonded to a LiNbO3 substrate with a silicon dioxide or benzocyclobutene (BCB) intermediate bonding layer has been proposed to produce wafer-scale high quality LiNbO3 thin films [87, 88]. The technology is similar to the
“smart cut” technology for fabricating silicon-on-insulator wafers. By bonding the
16 implanted LiNbO3 wafer to another LiNbO3 handle wafer as a mechanical support, full 3- inch wafer LiNbO3 thin films can be achieved. Alternatively, the implanted wafer can be bonded to a handle wafer with a thermal expansion coefficient close to that of LiNbO3, such as a quartz handle wafer [89]. Single crystalline ion-sliced thin films of LiNbO3 are attractive for compact integrated devices because of the high optical confinement and the superior optical properties of LiNbO3. The refractive index contrast between LiNbO3 and the surrounding cladding material (BCB or SiO2) is around 0.7, much greater than that of the Ti-diffused bulk LiNbO3 waveguide. Structuring of the thin film into planar waveguide devices yields compact and high density integrated optics. Various integrated waveguiding devices have been demonstrated based on the LiNbO3-on-insulator material system, including MZI modulator [89], microring tunable filter [88], submicrometer waveguide, photonic crystals, periodically poled LiNbO3 waveguides for second harmonic generation [87] and so on. Plasma etching of LiNbO3 is, however, challenging due to the redeposition of LiF compound in the etching process. Figure 8 presents a microring resonator fabricated on the LiNbO3-on-insulator material system. The demonstrated tunability of the microring resonator is only ~ 1 pm/V, requiring a few tens of volts or even over a hundred volts for practical switching applications.
LiNbO3 thin films have been integrated on silicon substrates by wafer bonding and polishing, or by ion slicing. Mode confinement is achieved by patterning the LiNbO3 thin film via dry etching [90] or selective oxidation of refractory metals on top of the
LiNbO3 thin film [91]. The silicon functions as a mechanical handle and support, and does not involve with guiding of the optical waveguide mode. As a result, the
17 demonstrated integration schemes do not combine the LiNbO3 photonics and the silicon photonics effectively.
Figure 8. A LiNbO3 microring fabricated on the LiNbO3-on-insulator platform with the top-down view (a), and cross-section view (b) [88].
1.4 Hybrid Silicon and Lithium Niobate Photonics
Recently, a hybrid material system consisting of both SOI waveguide and LiNbO3 thin film has been introduced. The LiNbO3 thin film serves a portion of the cladding mode of the hybrid silicon and LiNbO3 waveguide. A compact optical filter is demonstrated by direct bonding of LiNbO3 thin film to a silicon microring resonator [92]. The device has a tunability of 0.6 pm/V, limiting its potential for low voltage applications. Generally, direct bonding of two materials involves the concatenation of two smooth and clean material interfaces without the use of an intermediate layer. Direct bonding typically requires very flat surfaces, demanding process technology, and specialized equipment.
Also, direct bonding usually requires a high bonding temperature (~ 600oC or above) to
18 achieve desirable bonding strength. In addition, the demonstrated photonic structure with most area of the LiNbO3 thin film suspended in the air is not compatible with further microfabrication process. A more reliable bonding process is required to achieve a better bonding quality and compatibility with other microfabrication process.
Figure 9. Schematic of a hybrid silicon and LiNbO3 microring resonator with direct bonding [92].
19
Figure 10. Challenges and advantages of the hybrid silicon/LiNbO3 technology
Figure 10 demonstrates the advantages and challenges of the hybrid silicon and
LiNbO3 material system. The hybrid silicon/LiNbO3 material system combines the high index contrast and low loss of silicon waveguides with functionalities provided by the second order susceptibility of LiNbO3. Compact hybrid silicon/LiNbO3 photonic devices utilizing the linear electro-optic effect of LiNbO3 are enabled on the platform, including radio frequency electric field sensors, tunable optical filters, optical switches, and high speed optical modulators for applications in optical interconnects and analog optical links. The linear electro-optic effect available in the hybrid system provides functionalities that are not present in the traditional SOI platform, such as electric field detection and modulation with high linearity. On the other hand, the challenges of the hybrid system involves with developing a reliable integration process, and achieving high modulation efficiency and speed. 20
1.5 Contributions of This Dissertation
The overall goal of the dissertation is to address the challenges with modulation on silicon, presented in Figure 11 in terms of material, structure and device performance.
Unstrained crystalline Si does not exhibit a linear electrooptic (Pockels) effect so that modulation in silicon relies on the thermo-optic effect, which is slow and power- inefficient, and the plasma dispersion effect based on PN junction. Modulators based on the plasma dispersion effect have low linearity, large insertion loss with the phase modulation accompanied by loss modulation. In addition, the modulation speed is limited by charge mobility or charge recombination times.
Figure 11. Summary of the challenges with modulation on the silicon-on-insulator platform. 21
Our approach for addressing the challenge is to combine the second order susceptibility of LiNbO3 with the high-index contrast of silicon in a hybrid Si/LiNbO3 material system, as shown in Figure 12. The hybrid silicon and LiNbO3 photonic platform enables modulators with fast speed, high linearity, low power consumption, and low insertion loss, without modulation of the absorption loss. A method to stabilize ring resonance by applying a DC temperature-compensating voltage to a hybrid silicon/LiNbO3 ring resonator is proposed. The hybrid material system also provides new functionalities on the silicon platform, including electric field sensing and nonlinear optics.
Figure 12. Summary of the approach and results of the dissertation based on the hybrid silicon and LiNbO3 platform
22
This dissertation contains four main themes: (1) the demonstration of a reliable fabrication process to integrate ion-sliced LiNbO3 thin films on SOI based on indirect bonding; (2) the demonstration of an RF electric field sensor using a hybrid silicon/LiNbO3 microring resonator; (3) the demonstration of tunable hybrid silicon/LiNbO3 microring resonators with integrated electrodes that function as low voltage tunable optical filter; (4) the demonstration of high speed and high linearity hybrid Si/LiNbO3 electro-optical modulators for applications in optical interconnects and high spectral free dynamic range analog links.
1.5.1 Hybrid Integration Process
The purpose of this research is to experimentally demonstrate a silicon and LiNbO3 hybrid integration process based on indirect bonding using adhesive bonding material.
The goals of the work are as follows:
(1) Develop processes to obtain ion-sliced LiNbO3 thin films. Obtain LiNbO3 thin films
with controlled size, shape and orientations of the crystal axes. Increase fabrication
yield and design flexibility.
(2) Develop a reliable indirect bonding process to integrate LiNbO3 thin films on silicon
waveguides that is compatible with other complex microfabrication processes.
(3) Develop a full flow of CMOS compatible microfabrication process including
patterning of waveguides, couplers, vias, contacts, electrodes and so on.
LiNbO3 thin films are obtained using the crystal ion-slicing process. Three methods to fabricate the LiNbO3 thin films are introduced. LiNbO3 thin films with controlled size, shape and crystal orientations are obtained. DVS-BCB (divinylsiloxane-vis- 23 benzocyclobutene, also referred as BCB), a commercially available thermosetting polymer produced by the Dow Chemicals, is chosen as the adhesive for the indirect bonding process. BCB is spin-coated on the SOI substrate to planarize the waveguide topology. Individual LiNbO3 thin film can be transferred and bonded on the silicon waveguides using a micro-vacuum tip or a fiber tip. A full flow of microfabrication process is developed to fabricate complex hybrid Si/LiNbO3 integrated photonic devices.
1.5.2 Hybrid Silicon/LiNbO3 Electric Field Sensor
The purpose of this research is to experimentally demonstrate compact RF electric field sensors using hybrid silicon/LiNbO3 microring resonators. The goals of the work are as follows:
(1) Design and fabricate hybrid silicon and LiNbO3 sensor devices with high sensitivity,
high spatial resolution, large bandwidth and low invasiveness.
(2) Demonstrate the sensor by detecting the fringing field of a microstrip RF circuit.
(3) Determine the device sensitivity by simulating the electric fields of the microstrip
circuit using finite element method simulation tools.
An electric field sensor is designed and fabricated by utilizing a 600 nm thick z-cut
LiNbO3 ion-sliced thin film as the top cladding of a silicon microring resonator bonded using BCB as an intermediate layer. The optical transverse magnetic (TM) mode is used to access the r33 electro-optic coefficient in the LiNbO3. Operation of the device as an electric field sensor is demonstrated by detecting the fringing fields from a microwave frequency microstrip circuit. The electric field intensity of the mirostrip circuit is simulated to determine the sensitivity of the sensor. 24
1.5.3 Hybrid Silicon/LiNbO3 Microring Resonator Filter
The purpose of this research is to create low voltage tunable filters based on hybrid silicon/LiNbO3 microring resonators with integrated electrodes. Development of device involves with the design, simulation, fabrication and characterization. The goals of the work are as follows:
(1) Demonstrate the hybrid integration process is compatible with other complex
micro-fabrication process, including patterning of the integrated electrode, and
the fiber-chip cantilever couplers.
(2) Demonstrate tunable hybrid Si/LiNbO3 microring resonator with z-cut LiNbO3
thin films as low voltage tunable optical filters.
(3) Demonstrate a resonance stabilization scheme by applying a DC temperature-
compensating voltage to the hybrid Si/LiNbO3 microring resonator
It is demonstrated that the hybrid silicon/LiNbO3 material system can survive complex microfabrication process without degrading the bonding quality and the thin film integrity. For designs based on z-cut LiNbO3 thin films, integrated electrodes are patterned below and above the LiNbO3 thin films to confine electric fields locally. The silicon waveguide functions as both an optical guiding medium as well as an electrically conducting medium. As a result, the electric field is confined between the top electrode and the silicon core, enhancing the electric field intensity in the LiNbO3 thin film and enabling low voltage tuning. The demonstrated tunability of the filter is over an order of magnitude greater than the metal electrode-free hybrid silicon/LiNbO3 microring resonator design and the LiNbO3 thin film microring resonator. Furthermore, the
25 demonstrated tunability is comparable to state-of-the-art tunable silicon microring resonators based on reverse biased PN junctions and hybrid silicon-polymer slot waveguide microring resonators. Resonance tuning of the hybrid silicon/LiNbO3 microring resonator is linear for both forward and reserve bias with little effect on the quality factor and extinction ratio. A DC temperature-compensating voltage can be applied to the electrode to compensate the resonance drift due to thermal fluctuation.
1.5.4 Hybrid Silicon/LiNbO3 Microring Resonator Modulator
(1) Demonstrate low power and high speed hybrid Si/LiNbO3 microring resonator
electro-optical modulator with z-cut LiNbO3 thin films for optical interconnects.
(2) Demonstrate high linearity and high speed hybrid silicon/LiNbO3 racetrack
resonator electro-optical modulator with x-cut LiNbO3 thin films for high spectral
free dynamic range analog optical links.
Two types of hybrid Si/LiNbO3 modulators are demonstrated. The first type adopts z- cut LiNbO3 thin films and the silicon transparent conductor. The device speed is limited by the RC time constant of the biasing circuit. By reducing the resistance of the silicon transparent conductor and the capacitance between the electrodes, the device speed can be increased to multi-gigahertz for high speed optical modulation. A hybrid silicon and
LiNbO3 electro-optical microring resonator modulator operating at gigahertz frequencies is presented experimentally. The modulator design is based on tradeoffs between electrical frequency response and optical loss. High frequency scattering parameters are used to extract an RC circuit model for the modulator. The small-signal electrical-to-
26 optical 3 dB bandwidth is measured to be 5 GHz. Digital modulation with an extinction ratio greater than 3 dB is demonstrated up to 9 Gb/s. High-speed and low tuning power chip-scale optical modulators that exploit the high-index contrast of silicon with the second order susceptibility of LiNbO3 are envisioned.
An alternative modulator design is based on x-cut LiNbO3 thin films. A hybrid
Si/LiNbO3 racetrack modulator using patterned x-cut LiNbO3 thin films and a co-planar electrode design is demonstrated for high SFDR analog optical links. The measured
SFDR is 98.1 dB·Hz2/3 at 1 GHz and 87.6 dB·Hz2/3 at 10 GHz for the third order intermodulation distortion. The demonstrated SFDR is over an order of magnitude greater than silicon ring modulators based on the plasma dispersion effect, and is comparable to commercial LiNbO3 MZI modulators. High-speed and low tuning power chip-scale optical modulators that exploit the high-index contrast of silicon with the second order susceptibility of LiNbO3 are envisioned.
1.6 Organization of the Dissertation
Chapter 2 of this dissertation demonstrates the fabrication process and supporting technologies for creating the hybrid silicon and LiNbO3 material system and devices.
Four main parts are included: silicon photonic process, LiNbO3 thin film process, hybrid integration process, and dry etching process of LiNbO3. The demonstrated processes lay the foundation for the hybrid silicon and LiNbO3 integrated photonic technology. In
Chapter 3, a hybrid silicon and LiNbO3 microring electrical field sensor is demonstrated for detecting the fringe electric fields of a RF microstrip circuit. Thin film of LiNbO3 is
27 bonded to silicon microring waveguides as a portion of the cladding. The RF electric field is converted to optical intensity modulation via the linear electro-optical effect of
LiNbO3, which can be detected by a photodetector. Sensitivity of the sensor is estimated by calculating the electric field intensity of the RF circuit using finite element method simulation tools. Chapter 4 demonstrates low voltage tunable hybrid Si/LiNbO3 microring resonator with integrated electrode. An optically transparent and electrically conductive silicon bottom electrode is adopted to increase the electric fields in LiNbO3 and the tunability of the device. The design, fabrication, and measurement results of the device are presented. In chapter 5, a high speed Si/LiNbO3 microring electro-optical modulator is demonstrated for optical interconnects application. The RC time constant of the biasing circuit is reduced to enable gigahertz speed operation by patterning the silicon slab and lightly doping the silicon transparent conductor. DC tuning, small-signal and digital modulation of the device are characterized. Chapter 6 presents a high linearity hybrid Si/LiNbO3 racetrack electro-optical modulator for analog optical links. The device utilizes patterned x-cut LiNbO3 thin films and top coplanar metal electrodes to achieve large bandwidth, large tunability and reduced acoustic-optical distortion in modulation response. The design achieves high device linearity by avoiding the nonlinearity from pn-junctions and the plasma dispersion effect. It is demonstrated that the SFDR performance of the device is comparable to commercial LiNbO3 MZI modulators, and is over an order of magnitude greater than silicon microring modulators based on the plasma dispersion effect. Future chip-scale modulators that exploit the second order susceptibility of LiNbO3 on the silicon-on-insulator platform are envisioned
28 for analog optical links with high linearity. Chapter 7 concludes the dissertation and suggests possible future works.
29
CHAPTER 2
Development of Fabrication Processes
The hybrid silicon and LiNbO3 integrated photonic material platform depends crucially on the development of the microfabrication process. The fabrication process requires not only the standard CMOS fabrication techniques for forming silicon photonic circuits but also novel techniques for LiNbO3 material processing and hybrid integration. For rapid photonic circuit prototype in a university research environment, the process depends heavily on the aligned direct write electron-beam lithography (EBL) that is capable of precise definition of the critical device dimensions and accurate alignment of multiple lithography layers. Low loss silicon waveguides are enabled by the optimization of hydrogen silsequioxane (HSQ) mask edge roughness. Low resistance metal contact on silicon is formed for active photonic devices by ion implantation, nickel silicidation, and aluminum deposition. On the LiNbO3 side, three fabrication methods are developed for obtaining ion-sliced LiNbO3 thin films. LiNbO3 thin films with controlled size, shape and crystal axes orientation are achieved. With the basic building blocks of silicon photonic circuit and ion-sliced LiNbO3, a hybrid Si/LiNbO3 material system is developed based on indirect bonding using Benzocyclobutene (BCB). BCB is spin-coated on silicon waveguides and partially cured before bonding. LiNbO3 thin films are transferred 30 and bonded to silicon waveguides based on a pick-and-place process using a micro- vacuum tip or a fiber tip on a probe station. The high bonding strength, high temperature stability, and strong chemical resistance of BCB allow the devices to go through further complex fabrication process including formation of the cladding, via, electrode and fiber- waveguide couplers without degrading the bonding quality or cracking the LiNbO3 thin films. The development of the fabrication process ensures complex hybrid Si/LiNbO3 photonic devices can be successfully fabricated as designed.
2.1 Silicon Photonic Process
Silicon is the substrate material of the complementary metal-oxide-semiconductor
(CMOS) technology. Silicon photonics can take advantage of the well-developed and cost-efficient CMOS design, fabrication and testing resources. An in-house fabrication process is developed for patterning passive and active silicon photonic circuits. Low loss silicon waveguide and coupler, high quality microring resonators, microring modulators based on pin junction are enabled.
2.1.1 Low Loss Silicon Waveguide and Coupler
Silicon waveguide and fiber-waveguide coupler are the most fundamental photonic circuit components. High-index contrast silicon waveguides and inverse width waveguide tapers typically have cross-sectional dimensions of a few hundred nanometers or smaller. As the critical dimensions of integrated device reduce, reliable and stable fabrication process is required to achieve desired device performance. More importantly,
31 the sidewall roughness of silicon waveguide should be reduced to achieve low optical propagation loss by optimizing the electron beam lithography and the dry etching process. Cantilever couplers and taper fibers with well-matched mode profiles are used for broadband and low loss fiber-waveguide coupling [94, 95].
Hydrogen Silsequioxane (HSQ) is a negative-tone electron beam capable of producing features smaller than 10 nm. HSQ is used as a mask for patterning low optical loss silicon waveguides. Since mask roughness is transferred to the waveguide by plasma etching, minimizing mask edge roughness is critical for producing low-loss waveguides.
A multiparameter fabrication study is carried out to reduce the line edge roughness (LER) of HSQ patterned with a Leica EBPG-5000 EBL tool. Reduced mask roughness was achieved for 50°C pre-exposure baking for 40 min, 5000 μC/cm2 dose with an electron beam spot size of more than twice as large as the electron beam step size, development in
25% tetramethylammonium hydroxide for 1 min and post development baking with rapid thermal annealing in an O2 ambient at 1000°C for 1 min. Silicon waveguide is etched with inductively coupled plasma reactive ion etching (ICP-RIE) using Cl2 and O2 chemistry. Silicon strip waveguides (450 nm by 250 nm) patterned with the optimized mask have root-mean-square sidewall roughness of 2.1 nm with a correlation length of 94 nm, as measured by three-dimensional atomic force microscopy [93]. Figure 13 presents the unclad silicon strip waveguide with HSQ mask remaining on the top in (a) and (b).
The waveguides are cladded with 1 micron thick plasma enhanced chemical vapor deposition (PECVD) silicon dioxide, as shown in Figure 13(c).
32
Figure 13. Scanning electron micrographs of unclad silicon strip waveguides with top- down view (a) and cross section view (b), Si waveguide cladded with PECVD SiO2 (cross-section) (c), and cantilever coupler (d).
Cantilever couplers are fabricated on two ends of the waveguides, enabling broadband and low loss light coupling to silicon photonic integrated waveguides on an entire chip surface without the need for cleaving or dicing the chip [94, 95]. Cut-back method using waveguides with various lengths is adopted to measure the waveguide propagation loss and the coupler loss simultaneously. Tapered fibers with a diameter of 1.5 µm at the fiber end are used to couple with the cantilever couplers. The taper fibers are fabricated by etching standard SMF28 silica optical fiber with concentrated hydrofluoric (HF) acid.
Measured optical propagation losses for the waveguides with optimized sidewall roughness were below 2.5 and 2.8 dB/cm across the telecommunications C-band for the
33
TM and TE modes, respectively [93]. The coupling loss between the cantilever coupler and the fiber varies with the silicon waveguide taper and cantilever coupler dimensions, including the length and profile of the silicon waveguide taper, and the length and width of the coupler. Adopting an optimal design, coupling loss can be reduced to as low as
0.62 dB per connection for TE mode and 0.5 dB per connection for TM mode across the
C band [94].
2.1.2 Active Silicon Photonic Components
The process demonstrated in this section is mainly developed for silicon modulators based on pn junctions. The process involves with ion-implantation, RTA annealing, contact formation with nickel silicide, via formation, and metal electrode deposition.
Aligned direct write EBL using alignment markers allows multi-layer lithography with accurate alignment between layers, enabling complex fabrication process with more than
10 mask layers. Silicon pin modulators are demonstrated based on the process.
The patterning process mainly relies on EBL using the PMMA and Copoly P(MMA-
MAA 8.5)EL 11 positive resists. A single layer PMMA resist can be used as a mask for ion-implantation and via etching, while a double layer PMMA/MMA resist can be adopted for metal lift-off process. Rectangular or square metal markers or topology markers are patterned as alignment markers for layer-to-layer mask alignment. The edge length of the markers typically varies from 5 µm to 20 µm. Materials for metal markers include Au, Pt, and Ti/W. Metal markers are easily oxidized and corrupted in the high temperature RTA process in O2 ambient even covered with SiO2 or Si3N4 protection
34 materials. Silicon markers, usually formed at the same step for patterning silicon waveguides, is not affected by the RTA process, and does not introduce metal containments that become deep bandgap traps for silicon. Silicon markers are difficult to auto-identified by the EBL tool, so that their center coordinates are identified manually by the EBL operator.
Figure 14. EDS of nickel silicide formed on the silicon slab of a SOI substrate. (a) SEM of 5µm wide NiSi section, (b) EDS of Si element, (c) EDS of Ni element, and (d) EDS of O element. EDS is measured along the 10 µm yellow line shown in (a).
Typically metal-silicon contacts are formed on a silicon slab layer with a thickness of
+ around 50 nm. Using a PMMA mask, BF2 dopants are implanted to the silicon slab at
45 keV with a dose of 3×1015 cm-2 for P type contact. For N type contact, P+ dopants are implanted at 33 keV with a dose of 2×1015 cm-2. After activation of the dopants by RTA at 850oC for 2 min, 25 nm of nickel is deposited on the contact window and RTA is performed at 500oC for 1 minute to form nickel silicide (NiSi). Unreacted nickel is
35 removed by wet etching using piranha solution. The successful formation of NiSi is verified by an energy-dispersive X-ray spectroscopy (EDS) tool. Figure 14 shows the
EDS of NiSi after piranha etching. The Si and Ni signals on the NiSi clearly demonstrate successful formation of the NiSi. For a pin ring diode with a radius of 10 micron, the contact resistance is as low as 25 Ω for P type contact, and 10 Ω for N type contact. The contact resistance cab be much higher if the Si surface is not clean before Ni deposition, or if the silicon doping level is lower than expected.
Silicon pin micoring modulators are fabricated. After formation of the silicon strip ring waveguides, the basis for the pin diode is formed through two ion implantation steps
+ + of BF2 and P ions. Next, low-loss electrical connections are made by forming NiSi contact vias and Al pads. The fabricated device has a turn-on voltage of 0.8V and a large
DC tunability of 1.1 nm/V at the optimal voltage bias. The small-signal electro-optical modulation bandwidth is 2.3 GHz.
2.2. Lithium Niobate Thin Film Process
+ LiNbO3 thin films are produced using the crystal ion slicing process using He ions. Ion implantation creates a buried damaged layer beneath wafer surface with the depth controlled by the implantation energy. The damaged layer facilitates exfoliation of
LiNbO3 thin films either by chemical etching or mechanical splitting. Three methods are developed to obtain LiNbO3 thin films. LiNbO3 thin films with uncontrolled and inconsistent size and shape can be produced using wet etching or thermal blistering process after implantation. Alternatively, large area LiNbO3 thin film is first obtained on
36 a LiNbO3 substrate by wafer bonding, and LiNbO3 thin films are patterned by dry etching and released by wet etching using a sacrificial SiO2 layer introduced in the wafer bonding process, resulting in controlled size, shape, and crystal axis orientation.
2.2.1 Ion Implantation
Helium is chosen as the implantation species due to its small atomic mass that results in deeply buried damage layer and small crystal damage above the center damage layer.
Two dominant mechanisms for energy loss determine the distribution of lattice damage and the implantation profile in the LiNbO3 crystal [96]. At high energies, energy loss is dominated by electronic scattering that energy loss per unit trajectory length is proportional to square root of the ion energy along its trajectory. This process at high energy generates little crystal lattice damage. At low energies, energy loss is mostly due to Rutherford scattering with the host nuclei and is inversely proportional to the square of the ion energy, which ensures that the majority of the ions are stopped in a relatively narrow layer under the sample surface. The center of the implantation damage layer
+ (projected range) for He ion in LiNbO3 crystal is calculated using Transport-of-ions-in- matter (TRIM) calculation [96]. Figure 15 presents the simulation result for 0 to 400 keV range and 0 to 4000 keV range. An implantation energy of 342 keV and 3.8 MeV is required to obtain 1 µm and 10 µm depth, respectively. It is worth to mention that most of the ion implanters from the industry vendors only provide low to medium implantation energy below 420 keV.
37
+ Figure 15. Simulated ion implantation projected range of He ion in LiNbO3. (a) Implantation energy from 0 to 400 keV, (b) Implantation energy from 0 to 4000 keV.
The He+ ion implantation was performed in the Michigan Ion Beam Laboratory Three- inch lithium niobate wafers (z-cut or x-cut) were cut into smaller pieces with various size from one cm2 to a few cm2 using a dicing saw. The lithium niobate samples were cleaned with RCA1 solution, acetone and IPA and mounted to a 4” aluminum carrier using silver paste, as shown in Figure 16(a). The +z side is facing up for z-cut samples, and the +x size is facing up for x-cut samples. After mounting, the samples on the aluminum carrier were baked on a hotplate at 150oC for at least one hour to dry the silver paste and strengthen the bond. The LiNbO3 samples were mounted about 5mm away from the edge of the carrier so that it can fit within the slot on the copper wafer holder for ion implantation, as shown in Figure 16(b). Three pieces of doubled sided copper tape were applied to the copper wafer holder to ensure good thermal contact and uniform heating. The aluminum carrier with the LiNbO3 samples was then mounted to the copper holder, which was loaded into the chamber of the ion implanter with a tilted angle of 6o as shown in Figure 16(c) and Figure 16(d). The tilted angle prevents channeling that 38 affects the formation of a sharply localized implantation profile. The chamber was pumped down for about 30min to a pressure of on the order of 10-6 Torr before the ion beam is set up.
Figure 16. (a) LiNbO3 samples mounted to an aluminum carrier using silver paste; (b) Implantation wafer holder; (c) Mounting the wafer holder; (d) Ion implanter.
The staff of the Ion Beam Lab was operating the ion implanter. The implantation energy is set to 195 keV for 700 nm implantation depth, 342 keV for 1 micron and 380 keV for 1.1 micron. It took around 50 minutes to set up the implantation energy and current condition before the implantation can be started. The implantation area is a 5- 39 inch square that covers the 4-inch wafer holder. The current density started at a low level of 0.05 µA/cm2, and gradually increased to 0.25 µA/cm2 in 30 minutes. The beam current density was maintained at 0.25 µA/cm2 throughout the process before ramping down. The total beam current is around 39 µA, which is close to the maximum beam current of the implanter. The current fluctuates sometimes so the staff checked and readjusted the current every once a while. The dose required for z cut wafer is 4×1016
He/cm2. The implanted charge per unit area is Q = 4×1016 × 1.6×10-19 = 0.0064 C/cm2.
With a current density of 0.25 µA/cm2, the implantation time is 0.0064 C/0.25 µA =
25600 s or around seven hours. The charge per unit area can be control by various parameters. The ions are generated as pulses, which are measured using a current integrator. The dose is calculated by multiplication of the current density per pulse and the pulse number. The pulse was set to have an average charge density of 0.5 ×10-6
C/cm2 and a pulse period of two second so that the total pulse number required was
12800. The wafer in the chamber was occasionally checked through a glass window with the aid of a flash light to make sure the samples did not fall off the holder.
It is critical to avoid over heating of the wafer during the implantation that could cause thermal shock of the wafer surface leading to cracking or blistering. The silver paste between the LiNbO3 and the aluminum carrier and the copper tape between the aluminum carrier and the copper wafer holder ensure good thermal conduction and fast heat dissipation from the LiNbO3 sample to the wafer holder. We have successfully achieved crack-free implanted LiNbO3 dies bonded to the carrier with silver paste through an external implanted service vendor. We also tried mounting a full 3-inch wafer directly on
40 the wafer holder without any heat dissipation enhancement procedure. The whole wafer surface were blistered after the implantation.
Figure 17. Measured wafer holder temperature versus implantation duration
In the Michigan Ion Beam Lab, the wafer holder temperature can be monitored using an integrated thermal coupler on the backside of the holder. The measured temperature was close to the actual wafer temperature after the steady state was reached. The temperature was recorded every 10 minute during the ion implantation process as shown in Figure 17. Note that the actual wafer temperature was higher than the monitored wafer holder temperature. The temperature increased from room temperature to around 80 oC within the first 150 minute, and was maintained below 85 oC throughout the implantation process. After implantation for 430 minute, the beam was turned off, and wafer holder
41 was cooled in the chamber with the vacuum on for 100 minutes before unloading. With the careful heat management, crack-free implanted LiNO3 samples with desired dose were obtained.
2.2.2 LiNbO3 Thin Film Exfoliation
The ion implantation process provides a damaged crystal layer underneath the surface of a LiNbO3 wafer where the host nuclei are significantly dislodged. The LiNbO3 layer above the damaged layer can be separated from the substrate by breaking the damaged layer. Three methods are developed to exfoliate LiNbO3 thin film from the bulk wafer after implantation, including wet etching, thermal blistering, and patterned thin film releasing process.
2.2.2.1 Wet Etching Process
The wet etching process was first proposed by Levy for fabricating 10 µm thick LiNbO3 thin films [85]. LiNbO3 thin film is separated from the bulk wafer by etching the implantation-generated damaged layer using hydrofluoric acid (HF). HF etches LiNbO3 fast in the lateral direction and slowly in the vertical direction in the damaged layer. As a result, the high etching selectivity allows LiNbO3 thin films to be released from the bulk.
The lateral etching speed and the selectivity can be dramatically increased by a post- annealing process [97]. Before wet etching, the sample is annealed by rapid thermal annealing. During the annealing, He+ ions are filled in the implantation-generated microvoids so that the internal pressure in the damaged sacrificial layer is increased. The
42 thermal evolution of internal voids and internal pressure buildup can increase the lateral undercut etching speed by more than 100 times.
Figure 18. (a) Intentionally formed surface microcracks on implanted z-cut LiNbO3 wafer before wet etching, (b) Angled view SEM image of the implanted LiNbO3 sample showing the cross section, (c) HF undercut etching from the microcracks, (d) Angled view SEM image of partially released LiNbO3 thin films coated with 50 nm Chromium on a LiNbO3 substrate
It is found that similar methods can be applied to submicron thick z-cut LiNbO3 thin films as well. To facilitate the undercut etching, vertical surface microcracks are intentionally formed on the wafer surface, providing local etching entries across the 43 wafer surface for the chemical. The microcracks can be generated either by introducing large non-uniformly distributed stress to the wafer during the ion implantation process by tightly clamping the LiNbO3 wafer, or rapidly changing the temperature of an non- cracked implanted wafer.
The microcracks are formed along the crystal planes of the LiNbO3 crystal, as shown in Figure 18(a). The depth of the microcrack is a few microns, deeper than the buried damaged layer, as shown in Figure 18(b). As a result, wet etching of the damaged layer is initiated from the microcracks. Before wet etching, the implanted LiNbO3 sample (195 keV energy, 700 nm target thickness) is annealed at 250oC for 30s. The annealed z-cut
LiNbO3 sample is etched for a few minutes to a few tens of minutes in 5% hydrofluoric acid made of one part of 49% HF and nine part of DI water at room temperature to exfoliate the thin films. The undercut etching rate is around a few µm min-1. Figure
18(c) shows the sample surface during the undercut etching process, with the white arrows indicating the etching direction. The exfoliated thin films have areal shape of strips and triangles with the edge length ranging from several tens of micrometers to several hundred micrometers. Released LiNbO3 thin films are slightly curled due to the residual stress on the –z side of the films. Figure 18(d) shows a SEM image of partially released 700 nm thick z-cut LiNbO3 thin films coated with 50 nm chromium to prevent charging in the SEM imaging. The large thin film stress of chromium causes LiNbO3 thin films to bend out of plane.
The released LiNbO3 thin film can be transferred to another LiNbO3 or silicon substrate for further process. After annealing on a hotplate at 250oC for one hour or
44
o annealed by RTA at 1000 C for 30s, curled LiNbO3 thin films are flattened, as shown in
Figure 19. Flat LiNbO3 thin films are desirable for hybrid silicon and LiNbO3 integration process in which LiNbO3 thin films are bonded to silicon waveguides by indirect bonding. Curling of LiNbO3 thin films results in poor bonding quality and unbonded area.
Figure 19. A z-cut LiNbO3 thin film released by wet etching and transferred to the unpolished side of a silicon substrate. (a) Before annealing, (b) after annealing at 1000oC for 30s with RTA.
The wet etching process is also performed on implanted x-cut LiNbO3 wafer. In contrast to z-cut LiNbO3, no obvious HF undercut is observed after etching in 5% HF solution for a few hours at room temperature. It was also showed by Schrempel that the
-1 undercut etching rate for implanted x-cut LiNbO3 is on the order of 100 nm min using
3.7% HF at 40 oC [98]. As a result, it is time-consuming to process large area x-cut
LiNbO3 thin films using the wet etching process. The different etching behaviors
45 between z-cut and x-cut LiNbO3 indicate that etching of the damaged layer is anisotropic.
For bulk LiNbO3 without implantation, etching rate using HF is anisotropic, with the largest etching rate in –z direction. Etching rate in x, y and +z direction is extremely slow. The large undercut etching rate for implanted z-cut LiNbO3 is a result of fast etching of the –z side of the LiNbO3 crystal through the damaged layer. Etching occurs in both vertical and lateral directions in the damaged layer for z-cut LiNbO3, effectively increasing the undercut etching rate. In contrast, access to the –z side of LiNbO3 is in the horizontal direction for the x-cut LiNbO3, so that no fast vertical etching occurs in the damaged layer.
2.2.2.2 Thermal Blistering Process
LiNbO3 thin films can also be obtained using a thermal blistering process. An implanted
LiNbO3 sample is annealed at an elevated temperature by oven/hotplate baking or rapid thermal annealing. No surface coating or bonded supporting substrate is used. During the annealing, helium bubbles are formed and internal stress builds up in the damaged layer, leading to blistering of the LiNbO3 thin films [92]. The threshold temperature depends on the ion implantation energy which determines the thickness of the LiNbO3
o thin film. Typically 700 nm thick LiNbO3 thin film requires annealing at above 200 C,
o and 1 µm thick LiNbO3 thin film requires 250 C and above for around 30s in a rapid thermal annealing tool. Figure 20(a) shows a piece of blistered z-cut LiNbO3 thin film staying on top of the implanted LiNbO3 host wafer. The LiNbO3 wafer can be put on top of a silicon receiver substrate manually with the implanted side facing down. LiNbO3
46 thin films are blistered and transferred to the silicon substrate after thermal annealing, as shown in Figure 20.
Figure 20. Exfoliated LiNbO3 thin films using the thermal blistering process, (a) on the implanted z-cut sample after annealing, (b) transferred to unpolished silicon substrate. LiNbO3 thin films produced with the thermal blistering process have irregular size and shape.
The largest achievable thin film dimensions is around 200 µm in length. The size is large enough for compact hybrid silicon and LiNbO3 microring devices but the scalability is limited. In contrast to the wet etching process, the thermal blistering process works for both z-cut and x-cut LiNbO3.
2.2.2.3 Patterned LiNbO3 Thin Film Releasing Process
The LiNbO3 thin film process based on wet etching or thermal blistering results in thin films with uncontrolled size, shape, and unknown orientation of the crystal axes. In addition, the process is relatively time-consuming with lateral undercut etching rates of 47
-1 -1 several µm min for z-cut LiNbO3 and only ~100 nm min for x-cut LiNbO3. The lack of control of the ion-sliced LiNbO3 limits the fabrication yield, device size, and design flexibility of the hybrid silicon and LiNbO3 integrated photonic devices.
In this section, a process to obtain patterned ion-sliced LiNbO3 is demonstrated, as shown in Figure 21. The process can be applied to LiNbO3 wafers with different cut
+ orientations. As an example, an x-cut LiNbO3 sample is implanted with He ions with an implantation energy of 380 keV and a fluence of 3.5×1016 ions cm-2. After implantation, the LiNbO3 sample is cleaned using RCA1 solution and clad with 1 µm of PECVD SiO2.
The PECVD SiO2 layer will serve as a sacrificial layer to release the ion-sliced LiNbO3 later in the process. A second x-cut LiNbO3 handle wafer is cleaned with RCA1 and coated with adhesion promoter (Dow Chemical AP3000) and BCB (Dow Chemical
CYCLOTENE 3022-35). The implanted wafer is coated with adhesion promoter and bonded to the LiNbO3 handle wafer via the BCB film. The crystal z axes of the wafers are aligned to each other under a microscope. For LiNbO3 dies, the samples are cut with the edge along the z or y axis of the wafer so that they can be aligned. The bonded wafer pair is heated gradually to 300 °C in an oven in an N2 ambient. First the oven temperature ramps to 200 °C from room temperature in one hour, holds for one hour, and increases by 10 °C in 5 minutes and holds for one hour until 250 °C. During the ramping process, the bonding is strengthened and the BCB is fully cured. Finally the wafer pair is heated to 300 °C in 2 hours and held for one hour. As a result, 1.1 µm thick ion-sliced
LiNbO3 is exfoliated from the implanted wafer and transferred to the handle wafer.
Figure 22(a) shows a 0.85 cm by 1.2 cm x-cut ion-sliced LiNbO3 bonded to a handle
48 wafer after annealing. A thin film stack of ion-sliced LiNbO3, PECVD SiO2 and BCB is formed on the surface of the handle wafer. The surface of the transferred x-cut ion-sliced
LiNbO3 is measured with atomic force microscopy, showing a peak to peak roughness over 100 nm. In contrast, z-cut ion-sliced LiNbO3 bonded to a z-cut LiNbO3 wafer using the same fabrication process results in a root mean square roughness of 6 nm. The difference is attributed to the anisotropicity of the crystal properties [82]. The surface roughness can be reduced to sub-nanometer level by chemical mechanical polishing [86].
Figure 21. (a) X cut LiNbO3 thin film fabrication flow. (a) He+ ion implantation on wafer 1, (b) PECVD SiO2 deposition on wafer 1,(c) BCB spin-coating on wafer 2, (d) wafer bonding and annealing, (e) exfoliation of LiNbO3 thin film, (f) patterning chromium mask, (g) dry etching of LiNbO3, (h) wet etching of PECVD SiO2, (i) Transferring LiNbO3 to a unpolished silicon substrate.
49
The areal shape of the ion-sliced LiNbO3 is defined by a Cr mask patterned by EBL and plasma etching. A 320 nm thick Cr layer is first deposited on the surface. The cross- section of the thin film stack is shown in Figure 22(b). The sample is then deposited with
300 nm PECVD SiO2 and coated with 660 nm polymethyl methacrylate (PMMA) resist.
The PMMA resist is patterned by EBL and serves as a mask to etch the PECVD SiO2 layer by inductively coupled plasma reactive ion etching (ICP-RIE) using CHF3 chemistry. The patterns are in rectangular shapes with edge sizes in the range of hundreds of µm. The spacing between each rectangle pattern is 5 µm. After removing the PMMA resist, the Cr mask is formed using the PECVD SiO2 mask by ICP-RIE using
Cl2 and O2 chemistry, as shown in Figure 21(f). The ion-sliced LiNbO3 is then etched through to the BCB using the Cr mask by ICP-RIE using Ar and CHF3 chemistry. The etching rate of LiNbO3 is 10 nm/min and the etching selectivity to the Cr mask is 5:1.
Five micron wide trenches are formed between each ion-sliced LiNbO3 pattern.
After dry etching, the chromium mask is removed with CR-7S chromium etchant.
The sample is then etched in 5% hydrofluoric acid to undercut the PECVD SiO2 layer underneath the ion-sliced LiNbO3 from the etched trenches between patterns. The horizontal undercut etching rate is approximately 5 µm/min. Figure 23(a) shows the undercut after etching for 4 minutes. Before ion-sliced LiNbO3 is fully released, the
LiNbO3 sample is flipped to an unpolished silicon substrate and the etching is resumed with both substrates in the etchant. After etching for around 12 min, the ion-sliced
LiNbO3 patterns break away from the SiO2 pillar. The samples are then rinsed in DI water and dried on a hotplate at 80 °C. Released ion-sliced LiNbO3 patterns are
50 transferred to the unpolished silicon substrate. Figure 23(b) shows patterned ion-sliced
LiNbO3 transferred to silicon surface with edge lengths of 100 µm × 150 µm and 100
µm×450 µm. The maximum pattern size achieved is 100 µm × 2 mm. The z axis of the crystal is along the short edges of the patterns. One corner of the patterns is etched so that the crystal axes can be identified. Rapid thermal annealing (RTA) is then performed at 1000 °C for 30 s to repair the crystal lattice and restore the electro-optical properties of the ion-sliced LiNbO3 [86].
Figure 22. (a) X-cut LiNbO3 thin film bonded to a LiNbO3 handle substrate; (b) Scanning electron micrograph of the cross-section of the thin film stack. The black line in the BCB portion is a result of charging in SEM imaging.
Freestanding LiNbO3 thin films on the silicon substrate can be transferred to a silicon-on-insulator substrate using a pick-and-place process. The demonstrated material platform also opens routes for other membrane transfer techniques that allow multiple
LiNbO3 patterns to be transferred at the same time, such as the transfer printing process using an elastomer stamp [99, 100]. 51
Figure 23. (a) Hydrofluoric acid etching of the patterned x-cut LiNbO3 thin film; (b) Released LiNbO3 thin film transferred to the unpolished surface of a silicon substrate
2.3 Hybrid Silicon and LiNbO3 Integration
A hybrid silicon and LiNbO3 material system is created using the silicon waveguides and the LiNbO3 thin films fabricated with the process described above. It is critical to achieve a reliable process that integrates micrometer thick LiNbO3 thin films on the silicon-on-insulator platform. The challenges are fourfold. First, the physical dimensions of the LiNbO3 thin films are so small that handling and manipulation of the thin films are challenging. Second, nanometer accuracy is required for the relative position of the silicon waveguides and the LiNbO3 thin films as photonic performances is highly sensitive to the waveguide dimensions. Third, low temperature process and careful stress management are demanded for avoiding cracking of the LiNbO3 thin films induced by the large thermal expansion ratio discrepancy between the silicon and LiNbO3 [82].
Fourth, the integration process needs to be compatible with other complex microfabrication processes for creating hybrid Si/LiNbO3 devices with advanced functionalities.
52
In this work, an integration process is developed based on indirect bonding of LiNbO3 thin films to silicon waveguides via benzocyclobutene. The indirect bonding process satisfies the strict structural and thermal requirements of the hybrid material system, and is proved to be a reliable and repeatable process for advanced hybrid Si/LiNbO3 integrated photonics.
2.3.1 BCB for Indirect Wafer Bonding
Wafer bonding refers to a processing technology that brings polished, flat and clean wafers into contact to achieve strong bond between them either by direct bonding or by using an intermediate layer. Generally, direct bonding of two materials involves the concatenation of two smooth and clean material interfaces without the use of an intermediate layer. Direct bonding typically requires very flat surfaces, demanding process technology, and specialized equipment. An alternative to direct bonding is indirect bonding. Indirect bonding of two materials involves the use of an intermediate layer such as polymer, spin-on-glass, or metal [101]. It is a robust and low temperature technique that is relatively insensitive to surface topography [102].
Three types of adhesive materials have been used for adhesive bonding, thermoplastic, elastomeric and thermosetting materials. Both thermoplastic and elastomeric materials have low thermal stability, thus the post-bonding thermal budget is low. In contrast, thermosetting adhesives are more attractive for their higher thermal stability.
53
DVS-BCB (divinylsiloxane-vis-benzocyclobutene, also referred as BCB), a commercially available thermosetting polymer produced by the Dow Chemicals, has been widely used as wafer bonding adhesive and low dielectric constant material in the semiconductor industry [102]. BCB has a high bonding strength, a high thermal stability, a high chemical resistance, and a low dielectric constant. Various material properties are listed in Table 1[103]. BCB creates void-free and very strong wafer bonds with excellent chemical resistance and thermal stability. BCB has low volume shrinkage, shrinking by less than 5% without releasing significant amounts of by-products during the bonding process. The glass transition temperature of BCB is as high as 350oC. After thermal- setting curing at 250oC for one hour, BCB material properties stay stable up to 300oC
[103]. Figure 24 shows the polymerization process of the DVS-BCB monomer [103].
BCB possess a superior capability of planarization of wafer surface topology, and is commonly used for wafer topology planarization and wafer-to-wafer bonding of wafers with surface topologies. BCB is transparent at the telecommunication wavelengths, and the refractive index is as low as 1.535 at 1.55 µm. Finally, BCB provides fabrication flexibility. BCB can be applied to the wafer by spin coating with its thickness controlled by spinning speed and solution concentration. Also, BCB can be etched by dry etching, for example by inductively coupled plasma reactive ion etching (ICP-RIE).
B-staged DVS-BCB solutions are commercialized by Dow Chemicals as the
CYCLOTENE 3022 product series. The BCB thickness depends on the concentration of the solution, which is over one micron for the commercially available products.
Therefore, a diluted DVS-BCB solution is formulated by adding mesitylene solvent to
54
CYCLOTENE 3022-35. BCB thickness as thin as 50 nm can be obtained and the thickness can be fine-tuned by controlling the spin-coating speed. Submicrometer thick
BCB has been adopted as the bonding material for the hybrid III/V and silicon photonics
[103].
Table 1. Various properties of BCB [103]
Thin BCB is coated on silicon substrate and cured at 250oC for one hour, yielding a thickness of ~ 104 nm. The complex refractive index (n and k) of BCB is analyzed using a J. A. Woollam VUV-UVASE spectroscopic ellipsometer over the wavelength range 55 from 140 nm to 2480 nm. As shown in Figure 25, BCB has an absorption peak in the UV range. The imaginary refractive index k is smaller than 1×10-6 above 500 nm, indicating a negligible optical absorption loss.
Figure 24. Polymerization process of the DVS-BCB monomer [103]
Figure 25. Complex refractive index of BCB measured with ellipsometer
56
2.3.2 Indirect Bonding of LiNbO3 Thin Films to Silicon
LiNbO3 thin film is bonded to silicon waveguides as the top cladding using BCB as an intermediate layer. This section describes the details of the integration bonding process.
2.3.2.1 Handling and Manipulation of LiNbO3 Thin Films
Free standing LiNbO3 thin films obtained using the methods described above are small and fragile so that handling and manipulation of the thin films are challenging. A pick- and-place process is developed to transfer the free-standing LiNbO3 thin films from an unpolished silicon substrate to the silicon waveguide. Handling of LiNbO3 thin films using a fiber tip based on static electric force has been demonstrated to transfer a LiNbO3 thin film microring to a GaN waveguide [104]. This process is sensitive to the static charge conditions of the two materials. Also the electric force is typically not strong enough for picking up large thin films. In this work, LiNbO3 thin films prepared on unpolished silicon substrate are picked up, transferred, and bonded to the silicon waveguides using a micro-vacuum tip on a probe station. The micro-vacuum glass tip with a hose diameter of 25 µm at the end is connected to a vacuum source using a plastic tube, as shown in Figure 26(a). Compared to other pick-and-place tools based on static electric force, such as a fiber tip, the micro-vacuum tip provides reliable and accurate manipulation of larger size thin films. The micro-vacuum tip is fixed to a probe station under a microscope for micrometer accuracy positioning. Figure 26(b) shows the micro- vacuum tip and a metal tip hovering over some LiNbO3 thin films. LiNbO3 thin films are sucked to the bottom of the tip and transferred to the area directly above silicon waveguides. When the vacuum is turned off, LiNbO3 thin films are released from the
57 micro-vacuum tip, dropping to the top of the silicon waveguides. Pressure is applied to the films using the micro-vacuum tip as well to complete the pre-annealing bonding process.
Figure 26. (a) Schematic of the glass micro-vacuum tip, (b) Top-down view microscope image of LiNbO3 thin films on a polydimethylsiloxane (PDMS) substrate with a micro- vacuum tip and a metal tip hovering over
2.3.2.1 Bonding using Partially Cured BCB
For the hybrid silicon and LiNbO3 devices, LiNbO3 thin films are transferred and bonded to silicon waveguides coated with BCB. It is ideal to have the LiNbO3 thin film bonded directly above the silicon waveguide without no gap. A small BCB gap between the
LiNbO3 and the silicon waveguide dramatically affects the performance of the device, especially for TM mode with strong optical field confined in the BCB gap. Unlike bonding between bulk substrates, no external pressure is applied on the LiNbO3 thin film during the curing process. It is found that without pre-curing the BCB layer before
58 bonding, BCB reflows during the hard curing, resulting in non-uniform BCB thickness around and under the LiNbO3 thin film. As a result, the degree of polymerization of the
BCB layer prior to bonding is critical for obtaining desirable hybrid silicon and LiNbO3 waveguide structures.
Figure 27. Time-temperature transformation isothermal cure diagram of BCB [103]
Figure 27 shows the time and temperature dependent curing process of BCB from a liquid state to a solid state. Such a diagram is obtained by measuring the times for events to occur during isothermal cure at different temperature, including the monitoring of the degree of polymerization. Complete polymerization is typically done by annealing at
250oC for 1 hour. Prior to bonding, if the polymerization is too high, the BCB layer is 59 not tacky enough and the bonding may fail. If the polymerization is too low, the BCB layer reflows in the curing. It is found partially curing the BCB layer at a moderate temperature with a proper time prior to bonding gives high bonding strength and little
BCB reflowing [106]. A thin layer of BCB (~50 nm) is spin-coated on a bare silicon
o substrate, and partially cured at 190 C for 15 minute. A 700nm LiNbO3 thin film is transferred to the top of the BCB, as shown in Figure 28(a).
Figure 28. Bonding of a LiNbO3 thin film on a bare silicon substrate coated with partially cured BCB, (a) before hard curing, (b) after hard curing
Prior to curing, optical interference patterns are observed, indicating the LiNbO3 thin film does not contact the BCB layer uniformly. The sample is then cured at 250oC for an hour to fully cure BCB. As shown in Figure 28(b), the interference patterns disappear after hard curing, indicating the LiNbO3 thin film is bonded to silicon with uniform thickness BCB layer in between. The results show the hard curing process enables
60 seamless and uniform bonding of the thin film even with imperfect initial bonding of the
LiNbO3 thin film to silicon using partially cured BCB. Cross-sectional inspection with scanning electron microscopy and optical thin film thickness measurement reveal the
BCB thickness does not have noticeable change before and after the hard curing process.
2.3.2.1 Bonding LiNbO3 Thin Film to Silicon Waveguide
In contrast to a flat bare silicon substrate, the topology of silicon waveguides on a SOI substrate needs to be planarized. Typical strip silicon waveguide has a width of 450 nm and a height of 250 nm. A thin BCB layer is first spin-coated, and baked at 150 oC for one hour in the oven prior to partially curing at 190oC for 15 minutes, resulting a BCB thickness of ~ 350 nm. The BCB is then etched back to 250 nm with ICP-RIE using CF4 and O2 chemistry. A companion BCB sample on bare silicon substrate is processed at the same time to monitor the BCB thickness for accurate thickness control. LiNbO3 thin films are then transferred and bonded to the BCB coated silicon waveguides. The time period between partial curing and the hard curing can be as long as 24 hours without degrading the final bonding quality. Figure 29 shows microscope images of hybrid silicon and LiNbO3 microrings after hard curing along with a SEM image of milled holes using focused ion beam (FIB) to reveal the cross-sectional structure. The platinum shown in the SEM image is coated to prevent charging during the FIB milling of the two holes. It shows that BCB allows for excellent planarization of the silicon waveguide and high bonding quality without cracking the LiNbO3 thin film. The BCB gap between the
LiNbO3 and silicon waveguide cam be maintained below 30 nm. The BCB etch back
61 process enables optimal planarization of the waveguide topology and minimization of the
BCB gap between LiNbO3 and silicon.
Figure 29. (a) Micrograph of silicon microrings with a LiNbO3 thin film bonded on the top via BCB, (b) Tilted view SEM of two holes opened over the silicon ring waveguides using focused ion beam.
2.4 Dry Etching of LiNbO3
Dry etching of LiNbO3 is essential for creating patterned lithium niboate thin films and
LiNbO3 thin film photonic waveguides. LiNbO3 is well known to be difficult to etch either by wet etching or dry etching due to its inert material properties. ICP-RIE and focused ion beam (FIB) etching of LiNbO3 are demonstrated in this section. Dry etching of LiNbO3 mainly relies on physical bombardment of reactive ions or focused ion beams.
Etching LiNbO3 with ICP-RIE is generally slow compared to other materials, such as
+ silicon. Etching LiNbO3 with FIB, such as Ga ion beams can result in efficient local etching.
62
2.4.1 Etching LiNbO3 with ICP-RIE
Reactive ion etching (RIE) tools is widely used in microfabrication in the semiconductor industry. Inductively coupled plasma RIE (ICP-RIE) is able to produce high density plasma with separate RF and ICP generator to provide separate control over ion energy and ion density, enabling high etch rate and high process flexibility. Bulk z-cut LiNbO3 substrate patterned with chromium mask is etched with ICP-RIE using Ar and CHF3 chemistry.
Chromium hard mask is patterned on top of a LiNbO3 substrate using photolithograph and lift-off process using the AZ5214E photoresist. Electron-beam lithography using
PMMA mask can also be employed to pattern the metal mask. After development, 300 nm chromium is deposited and lifted off, with the cross section shown in Figure 30(b).
The sample is etched in a PlasmaTherm SLR770 ICP-RIE tool. The chamber temperature is maintained at 12 oC and the wafer backside helium cooling is turned on.
The chemistry combination is 25 sccm of Ar and 25 sccm of CHF3 with a pressure of
12.5 mTorr. The DC bias is set to 270 V and the ICP power is set to 250 W. The etching is paused for one minute after every 10 minute of etching to avoid over-heating of the sample. Figure 30(c) and Figure 30(d) show the top-down view and cross-sectional view of the etched LiNbO3 ridge structure after etching for 70 minutes. The etching depth is
900nm. The etching rate is 12.8 nm min-1 and the etching selectivity over the chromium mask is around 5:1. Micro-masking effect near the mask results in two “feet” on two sides of the ridge. The sidewall is relatively smooth with a sidewall angle of 73oC.
63
Figure 30. (a) Photoresist after development, (b) cross-section of 300 nm chromium mask on LiNbO3 substrate after lift-off, (c) top-view of edged LiNbO3 ridge, (d) Cross section of the etched LiNbO3 ridge
2.4.2 Etching LiNbO3 with Focused Ion Beam Bombardment
FIB milling of LiNbO3 is performed using a FEI Helios Nanolab dual focused ion beam and SEM system. 700 nm thick LiNbO3 thin films are fabricated using the wet etching method described in section 2.2.2.1. Fifty nanometer thick chromium is deposited on the top to avoid the charging effect during milling. The stress of the chromium causes
+ LiNbO3 thin films to bend out of plane. The Ga ion beam energy is set to 30 kV. Ion beam size increases as the ion current increases. Larger beam current results in larger etching rate at a cost of worse pattern resolution.
64
Figure 31. (a) Photonic crystal structures milled on LiNbO3 thin film, (b) microdisks milled on LiNbO3 thin film, (c) a zoom-in view of the microdisk, (d) LiNbO3 microdisk transferred to silicon waveguide using a nano-manipulator
Photonic crystal and microdisk structures are milled on the LiNbO3 thin films with a beam current of 0.92 nA and a beam spot size of 44 nm, as shown in Figure 31. For forming the microdisk, a 500 nm wide ring structure with a diameter of 30.25 micron is milled through the LiNbO3 thin films. The milling time for a 30 micron diameter disk is
2 minutes 47 seconds using the 0.92 nA beam. The sidewall of the disk is very smooth compared to the ICP-RIE results. The disks can be picked up and transferred to silicon waveguides using a nano-manipulator installed in the FIB tool, as shown in Figure 31(d).
65
2.5 Chapter Conclusion and Outlook
In this chapter, details of microfabrication techniques required to realize the hybrid silicon and LiNbO3 integrated photonic platform are discussed. The process development includes a full flow of CMOS compatible fabrication process for passive and active silicon photonic structures, LiNbO3 thin film fabrication and LiNbO3 micro-structuring by dry etching, and integration of LiNbO3 thin films to the SOI platform using pick-and- place and indirect bonding processes. For multiple-layer patterning, direct write electro- beam lithography is developed to allow for accurate layer-to-layer alignment. Silicon passive components including low loss silicon waveguides and couples, and active components including low resistance contact, metal electrode, and doped waveguides are enabled. On the LiNbO3 side, three fabrication methods are developed for obtain ion- sliced LiNbO3 thin films. LiNbO3 thin films with controlled size, shape and crystal axes orientation are achieved. A pick-and-place process is adopted to transfer and bond free- standing LiNbO3 thin films to silicon waveguides using BCB. The indirect bonding process satisfies the strict structural and thermal requirements of the hybrid material system, and is compatible with other complex microfabrication steps, proved to be a reliable and repeatable process for advanced hybrid Si/LiNbO3 integrated photonics. Dry etching of LiNbO3 using ICP-RIE and focused ion beam milling is also demonstrated.
The fabrication techniques presented in this chapter is a fundamental for further works in this dissertation. In Chapter 3, an electrode free hybrid silicon and LiNbO3 microring is demonstrated as an electric field sensor. By integrating silicon transparent electrode and metal thin film electrodes to the structure, tunable filters and high speed modulators
66 can be fabricated, shown in Chapter 4 and Chapter 5, respectively. In Chapter 6, an alternative design using patterned x-cut LiNbO3 thin film and a co-planar electrode configuration is shown.
67
CHAPTER 3
Hybrid Silicon and LiNbO3 Microring Electric Field Sensor
This chapter presents a compact photonic electric-field sensor based on a hybrid silicon and lithium niobate microring resonator. An electric field sensor based on the indirect bonding of submicrometer thin films of lithium niobate to silicon microring resonators is presented using benzocyclobutene as an intermediate bonding layer. The hybrid material system combines the electro-optic functionality of lithium niobate with the high-index contrast of silicon waveguides, enabling compact and metal-free electric field sensors. A sensor is designed and fabricated using ion-sliced z-cut lithium niobate as the top cladding of a 20 m radius silicon microring resonator. The optical quasi transverse magnetic mode is used to access the largest electro-optic coefficient in the lithium niobate. Optical characterization of the hybrid device results in a measured loaded quality factor of 13,000 in the infrared. Operation of the device as an electric field sensor is demonstrated by detecting the fringing fields from a microstrip electrical circuit operating at 1.86 GHz. The demonstrated sensitivity to electric fields is deducted by simulating the electric field of the microstrip circuit using three-dimensional finite element method simulation.
68
3.1 Photonic Electric Field Sensors
Figure 32. Applications of electric field sensors. (a) Near field probe for RF circuit diagnosis (Aaronia AG, Euscheid, Germany), (b) DC to 6 GHz RF probe for EMC/EMI analysis (AFJ instruments, Milan, Italy), (c) SAR measurements for biological application [107], (d) dielectric photonic receivers for communications [108]
Advances in electric field sensors are important for a host of applications including electromagnetic compatibility (EMC) measurements, high-frequency electronic circuit diagnostics, medical equipment field monitoring, radio-frequency reception, and high power microwave detection, as shown in Figure 32 [107-111112]. In the case of electric field sensors based on the linear electro-optic (Pockels) effect, a high-frequency (DC to 69
THz) electric field modifies the indices of refraction of an electro-optic medium, resulting in high-speed modulation of an optical carrier signal. Electric field sensors based on optical technology are advantageous, compared to electronic technology, because they can be metal free, compact, and broadband [112]. The use of metal free sensors minimizes invasiveness. Furthermore, optical technology is amenable to realizing high spatial resolution sensor arrays and signal routing can utilize fiber optics.
3.2 Device Design and Fabrication
Figure 33. (a) Schematic of an electric field sensor based on the indirect bonding of a lithium niobate thin film to a silicon microring resonator. For clarity, a PECVD SiO2 top-cladding layer is not shown. (b) SEM of the cross-section of the sensor structure.
A schematic of the hybrid Si/LiNbO3 electric field sensor is shown in Figure 33(a). The sensor consists of a bus coupled SOI strip waveguide ring resonator and a thin film of
LiNbO3 which serves as a portion of the top cladding. The silicon waveguide core width is 450 nm and the height is 250 nm. The LiNbO3 thin film is 600 nm thick and is bonded to the silicon resonator via BCB. The whole sensor is covered by one micrometer thick 70 plasma enhanced chemical vapor deposition (PECVD) SiO2. Figure 33(b) shows a scanning electron micrograph (SEM) of the cross-section of the sensor structure. A ring radius of 20 m is chosen to avoid large bending losses.
Figure 34. Schematic of the sensing principle using a tunable microring resonator
An optical carrier signal propagating as a guided wave in the SOI bus waveguide is evanescently coupled into the ring resonator via a 375 nm wide coupling gap. A portion of the guided-mode is within the LiNbO3 thin film. For optical wavelengths near the resonance wavelengths of the ring resonator, the optical transmission is sensitive to modulations in the effective index of the guided-mode in the ring resonator. The use of a high-Q resonator is desirable because the sensitivity depends on the slope of the optical transmission versus wavelength. When the sensor is immersed within a high-frequency electric field, the electric field can be detected because it modifies the refractive indices in the LiNbO3 thin film via the electro-optic effect. Consequently, the effective index of
71 the guided-mode in the ring resonator is modulated, resulting in an intensity modulation of the optical carrier. The output light is detected by a photodetector to convert the optical signal to electric signal. Figure 34 shows an illustration for the sensing principle.
Figure 35. Optical electric field distributions in the hybrid Si/LiNbO3 sensor for the quasi-TM (Ey component) and quasi-TE (Ex component) modes at 1550 nm wavelength. Material boundaries are indicated by the white dashed lines and the material regions are indicated in the inset.
The optical polarization of the guided mode is chosen to maximize the fraction of the optical mode that overlaps with the LiNbO3. Maximum overlap optimizes the change in the effective index of the optical mode that occurs for a change in the refractive indices of the LiNbO3. Figure 35 shows the cross-section electric field distributions, for the quasi-
TE and quasi-TM modes at 1550 nm optical wavelength, calculated using the beam propagation method (BPM). The commercial software package Rsoft Beamprop is used for the simulation. The grid size is set to 10 nm for x, y and z directions, and the iterative method is used for solving the mode. The effective index calculates to 2.33 for TM mode
72 and 2.58 for TE mode. Since the TM mode is less confined, a larger fraction of the optical mode is in the LiNbO3 for the TM mode than the TE mode. In addition, the TM mode accesses the r33 electro-optic coefficient of the LiNbO3, whereas the TE mode accesses the r13 electro-optic coefficient. Therefore, the optical TM mode is chosen in the design. Compared to fabricated devices, the simulations shown in Figure 35 neglect several tens of nanometers of BCB that resides on the top in pasurface of the silicon core.
The presence of BCB between the top of the silicon core and the bottom of the LiNbO3 thin film results in a reduction of the TM mode electric field inside the LiNbO3 due to the electromagnetic boundary conditions. Therefore, in order to maximize the fraction of the optical mode in the LiNbO3, it is important to minimize the BCB thin film thickness that is directly on top of the silicon core.
The fabrication process is shown in Figure 36. The process begins with a SOI wafer.
The thickness of the silicon device layer is 250 nm, the thickness of the buried oxide
(BOX) is 1 m, and the thickness of the silicon substrate is thinned to 210 m. Silicon strip waveguides forming a silicon microring and a bus waveguide with inverse width tapers are defined in HSQ resist using electron-beam lithography and inductively coupled plasma reactive ion etching (ICP-RIE) using Cl2/O2 chemistry. The cross-sectional width and height of the silicon strip waveguides are 450 nm and 250 nm, respectively. The radius of the ring resonator is 20 m and the coupling gap is 375 nm, as shown in Figure
37.
Thin films of LiNbO3 are obtained from a bulk single crystal z-cut LiNbO3 wafer using helium ion implantation and thermal treatment [92]. The implantation energy is
73
195 keV and the fluence is 41016 ions cm-2. After implantation, the wafer is annealed using rapid thermal annealing (RTA) and then etched in hydrofluoric acid (HF) solution
[97]. Non-uniform stress produces 700 nm thick LiNbO3 thin films in the areal shape of strips and triangles whose edges are formed along crystal planes. The edge length ranges from several tens of micrometers to several hundred micrometers. The surface roughness of the implanted side of the LiNbO3 thin film is reduced to 4 nm by ICP etching with Ar chemistry resulting in a final film thickness of 600 nm.
Figure 36. Fabrication process of electric field sensor: (a) Silicon strip waveguide ring resonator patterned on SOI wafer using electron beam lithography and plasma etch, (b) spin-coat of BCB, (c) indirect bonding of LiNbO3 thin film, (d) plasma etch of BCB, (e) deposition of PECVD SiO2, (f) fabrication of cantilever couplers.
Next, BCB is spin-coated from solution onto the silicon strip waveguides. LiNbO3 thin films are transferred to the top of the SOI ring resonators using a fiber tip on a probe station. After transfer, the sample is annealed to 300 C to cure the BCB and to partially 74 recover the r33 coefficient [88], resulting in a BCB layer approximately 20 nm thick between the LiNbO3 and the silicon microring. The LiNbO3 thin film does not crack after annealing despite the large mismatch in thermal expansion coefficients between silicon and LiNbO3. BCB that is not covered by the LiNbO3 thin film is etched using
CF4/O2 ICP-RIE. A 1 μm thick SiO2 top cladding is deposited by PECVD. Finally, cantilever couplers are fabricated for fiber coupling to the bus waveguide [95]. The bus waveguide, including the cantilever couplers, is 640 m long.
Figure 37. (a) Scanning electron micrograph of a SOI ring resonator with a diameter of 40 μm. (b) Zoom-in view of the coupling section with a gap of 350 nm.
3.4 Device Characterization
3.4.1 Measurement Setup and RF Measurement
The electric field sensor is demonstrated in the laboratory by detecting fringing electric fields from a radio-frequency (RF) microstrip circuit. At a constant RF power, the RF electric field varies spatially above the circuit. By positioning the sensor at various
75 points above the circuit, a field map of the electric field distribution can be obtained. The microstrip circuit is an RF coupled line resonator operating on resonance at 1.86 GHz with an RF return loss of 7.4 dB. The RF frequency of the circuit is well within the RF bandwidth of the sensor which is limited by the photon lifetime of the ring resonator.
Figure 38. (a) Schematic of the measurement setup, (b) Top-view optical micrograph of fabricated electric field sensor, (c) photo of the measurement setup.
The experimental setup is shown in Figure 38 along with a top-view optical micrograph of the fabricated electric field sensor. In the demonstration, the chip is placed on the surface of the RF microstrip circuit in a region where the fringing electric field is a maximum. Port 1 of a calibrated microwave vector network analyzer (VNA) drives the 76
RF microstrip circuit with 1 mW of RF power. An infrared continuous-wave laser source is connected to a polarization controller which outputs linearly polarized quasi-TM light with cross-polarization rejection ratio of more than 17 dB. Tapered optical fibers with tip diameter approximately equal to 2 μm are butt-coupled to the input and output cantilever couplers of the electric field sensor.
Fringing electric fields from the microstrip circuit modify the effective index of the optical mode in the electric field sensor, producing an intensity modulation on the optical beam. The modulated lightwave is out-coupled via optical fiber which is terminated in a high-speed photodiode with responsivity equal to 0.9 A/W and transimpedance gain equal to 1000 V/A, resulting in a photoreceiver conversion gain of 900 V/W. The demodulated RF signal is fed into port 2 of the VNA. The RF power entering port 2 is proportional to the square of the magnitude of the vertical component of the RF electric field in the LiNbO3 portion of the sensor, weighted by the optical mode intensity distribution. Furthermore, the RF power entering port 2 is proportional to the square of the optical power and the square of the slope of the optical transmission versus wavelength. The microwave VNA is set to measure the port 1 to port 2 S21 scattering
2 parameter. The square of the magnitude of S21, |S21| , is the RF power delivered to port 2 normalized by the RF power available from port 1. The VNA operates with an intermediate frequency (IF) bandwidth of 10 Hz and 20 averages.
The measured optical transmission of the electric field sensor is shown in Figure 39(a) for 1.3 dBm of input optical power near an optical resonance at 1526.274 nm. The full width half maximum (FWHM) is 118 pm, the free spectral range (FSR) is 5.05 nm, and
77 the extinction ratio is 13.7 dB. The total optical insertion loss is 5.3 dB. The insertion loss is from 4 dB of fiber-to-waveguide coupling loss and 1.3 dB of waveguide transmission loss. The loaded quality factor is 13,000 and the finesse is 43. The measured optical resonance has a slight asymmetric red shift due to the relatively large thermo-optic nonlinearity in silicon [113]. Figure 39(b), andFigure 39(c) show the corresponding values of the microwave VNA S21 scattering parameter versus CW laser wavelength. The magnitude of S21, denoted |S21|, peaks at approximately -104 dB on the red and blue sides of the optical resonance.
Figure 39. (a) Measured optical transmission of the electric field sensor; (b) Magnitude of measured microwave VNA S21 RF scattering parameter versus CW laser wavelength; (c) Corresponding phase of RF S21 versus CW laser wavelength.
78
Since the thermal nonlinearity is a slow effect compared to the electro-optic effect, the peak RF S21 magnitude remains equal on both sides of the optical resonance. At the resonance wavelength, the |S21| is more than 10 dB less than the peak values. The |S21| data is consistent with the steep slopes of the optical transmission near the optical resonance. The phase of S21 exhibits a phase change of approximately 180 degrees as the optical wavelength is swept through the resonance, consistent with the sign change in the slope of the optical transmission.
The VNA noise, shown as the dashed red curve in Figure 39(b), is obtained by terminating port 1 of the VNA with a matched 50 Ω load and also connecting the laser directly into the photodiode. As the laser wavelength is swept in time, the VNA noise is recorded. A noise floor of -115 dB is calculated by fitting a Gaussian distribution to the linear magnitude of the noise data. The noise floor is defined as the sum of the mean value of the linear magnitude of the noise and three times its standard deviation.
3.4.2 Sensitivity Calculation
As shown in Figure 39(b), there is an 11 dB difference between the peak |S21| and the noise floor. Numerical modeling of the electric field sensor together with the RF microwave circuit is conducted, using the finite element method, to estimate the RF electric field sensitivity. Since the microwave circuit and the optical thin films composing the sensor differ in spatial dimensions by 4 to 5 orders of magnitude, the GHz varying electromagnetic field in the optical thin films are considered to be quasi-static.
Therefore, an approximate two-step modeling procedure is utilized. First, a three-
79 dimensional time-harmonic RF simulation is conducted involving the RF microstrip circuit and the silicon substrate without the optical thin films (ie. without the BOX, BCB, patterned silicon waveguide core, LiNbO3, and PECVD oxide).
Figure 40. (a) Photo of the microstrip circuit, (b) Electric field (vertical component) distribution on the surface of the circuit at 1.86 GHz, (c) Simulated and measured S11 spectrum of the RF circuit
A commercial software package Ansoft High Frequency Structure Simulator (HFSS)
13.0 is used for the simulation. The HFSS employs the vector finite element method that divides the full problem space into thousands of smaller regions and represents the field in each element with a local function [114-119]. A single tetrahedron is treated as a basic element with the collection of tetrahedral referred to as the finite element mesh. To produce the optimal mesh, HFSS uses an interactive process in which the mesh is automatically refined in critical regions. A maximum delta S (the magnitude of the change of the S-parameters between consecutive passes) of 0.02 is set for
80 the simulation. HFSS starts the simulation with initial mesh, and gradually refines the mesh size after each pass until the convergence criteria is met. At some critical areas, the mesh size can be defined to not larger than a certain value so as to solve the fields more accurately.
Figure 40(a) shows the photo of the microstrip circuit. Figure 40(c) shows the simulated and measured S11 spectrum of the RF circuit. The RF resonance frequency from the simulation is 1.86 GHz and the RF return loss is 6.8 dB, in agreement with the measurements of the RF circuit. At the resonance, the distribution of the vertical component of the electric field on the surface of the circuit is shown in Figure 40(b).
Figure 41. The vertical component of the electric field at one micron above the Si substrate over the microstrip metal trace corner with 1 W of input power at 1.86 GHz. The circuit with white dashed line shows the possible position of the sensor considering the measurement accuracy. The area of the simulation domain is 2×2 mm.
81
As shown in Figure 41, the boundaries of the microstrip metal trace is shown as black dashed line. The position of the sensor measured by the microscope is marked as the white triangle. The area inside the white dash circle is the area where the actual sensor position could possibly be, considering the measurement accuracy tolerance of ~400 micron. The maximum vertical component of the electric field inside the white circuit is then used as an approximate input source field in a static two-dimensional simulation involving only the optical thin films. The static 2D simulation is conducted with a commercial software package COMSOL 3.5.
The process to calculate the device sensitivity is illustrated in Figure 42. The small
(<3%) effect of the optical thin films on the value of the input source field is neglected.
From the simulation, the vertical component of the electric field in the LiNbO3 portion of the sensor, weighted by the optical mode intensity distribution, is found to be 50.8 V/m for an RF input power of 1 mW. From Figure 39(b), reduction of the RF input power by
11 dB produces a signal-to-noise ratio of one. The corresponding RF electric field scales from 50.8 V/m to 14.3 V/m. Figure 42 shows the flow diagram for obtaining the sensor sensitivity. Based on our system bandwidth of 10 Hz, the demonstrated sensitivity to electric fields is 4.5 V m-1 Hz-1/2. The sensitivity depends on several factors specific to our experiments. Most notably, the RF signal power entering port 2 of the VNA scales quadratically with input optical power, slope of the resonator optical transmission versus wavelength, photoreceiver conversion gain, and r33 in the electro-optic medium.
82
Figure 42. Flow diagram for calculating the sensor sensitivity
3.4.2 Optical Power Dependence
As the optical input power increases, the resonance wavelength red-shifts and optical bistability is observed with off-resonance optical output power higher than 2 dBm, as shown in Figure 43. The optical bistability is induced by thermal-optical nonlinearity of silicon [113]. The optical power dependent measurement is performed with the measurement setup shown in Figure 38.
83
Figure 43. Measured optical resonances with various optical power level. Optical biastability is observed for high optical power.
The S21 as a function of optical wavelength is measured as the optical power level is varied. Figure 44 shows the results for three output optical power level, -4.5 dBm, -3 dBm and 0 dBm. The nominated optical power level indicates the maximum off- resonance output power. Below the bistability optical power threshold, the maximum S21 magnitudes on two sides of the resonance remain the same. Since the thermal nonlinearity is a slow effect compared to the electro-optic effect, the peak RF S21 magnitude remains equal on two sides of the optical resonance. The phase of S21 exhibits a phase change of approximately 180 degrees as the optical wavelength is swept through the resonance for various optical power levels. The S21 magnitude maximum scales quadratically with the optical power level, shown in Figure 45. An increase of 9 dB in optical power results in 18 dB increase in S21 magnitude.
84
Figure 44. (a) Measured optical transmission for 0 dBm, -3dBm, -4.5 dBm optical power (b) Magnitude of measured microwave VNA S21 RF scattering parameter versus CW laser wavelength; (c) Corresponding phase of RF S21 versus CW laser wavelength.
Figure 45. Measured S21 magnitude maximum for various optical output power level
85
3.5 Chapter Conclusion and Outlook
A compact and metal-free hybrid silicon and lithium niobate microring electric field sensor is designed, fabricated, and characterized. The hybrid device combines an ion- sliced lithium niobate thin film with a silicon-on-insulator strip waveguide ring resonator using BCB as an intermediate bonding layer. The sensor has a ring radius of 20 μm and measured loaded quality factor of 13,000 at infrared wavelengths for the quasi-TM mode.
The sensor is demonstrated by detecting the fringing electric fields from a microwave circuit operating at 1.86 GHz using a test and measurement setup that incorporates a
VNA. The dependence of the magnitude and phase of VNA S21 scattering parameter on the optical wavelength is consistent with electro-optic modulation from the lithium niobate portion of the electric field sensor. Future work involves optimizing the radius of the ring resonator to reduce the footprint of the sensor. Reduction in footprint enables dense integration of multiple electric field sensors for high spatial resolution field mapping of free-space electromagnetic fields.
86
CHAPTER 4
Low Voltage Tunable Hybrid Silicon and LiNbO3 Microring Resonator
In this chapter a silicon microring resonator with a lithium niobate top cladding and integrated tuning electrodes is presented. Submicrometer thin films of z-cut lithium niobate are bonded to silicon microring resonators via benzocyclobutene. Integrated electrodes are incorporated to confine voltage controlled electric fields within the lithium niobate. The electrode design utilizes thin film metal electrodes and an optically transparent electrode wherein the silicon waveguide core serves as both an optical waveguide medium and as a conductive electrode medium. The hybrid material system combines the electro-optic functionality of lithium niobate with the high-index contrast of silicon waveguides, enabling compact low tuning voltage microring resonators. Optical characterization of fabricated devices results in a measured loaded quality factor of
11,500 and a free spectral range of 7.15 nm in the infrared. The demonstrated tunability is 12.5 pm/V, which is over an order of magnitude greater than electrode-free designs. A scheme to achieve low-power resonance stabilization by applying a temperature- compensating DC voltage to the hybrid ring resonator is proposed. Fabricated devices demonstrate a tuning range of 1 nm with a voltage range from -30 V to 50 V. A
87 capacitive geometry and low thermal sensitivity result in the compensation of 17 C of temperature variation using tuning powers at sub-nanowatt levels. The method establishes a route for stabilizing high quality factor resonators in chip-scale integrated photonics subject to temperature variations.
4.1 Introduction
Voltage controlled silicon microring resonators enable electrically tunable optical filters [114], switches [121], and modulators [14] in compact integrated photonics for applications such as on-chip wavelength-division multiplexing (WDM) optical networks.
Resonance tuning of a ring resonator can be achieved by changing the effective index of the optical mode in the ring waveguide. Electro-refraction based resonance tuning mechanisms in silicon microring resonators include tuning the refractive index of silicon by the thermo-optic effect and the plasma dispersion effect [10-12]. Alternatively, modifying the refractive index of integrated materials, such as polymers [15] and III-V semiconductors [74] in optical proximity to the silicon can also be exploited for tuning.
Due to inversion symmetry, unstrained bulk silicon does not exhibit second-order susceptibility. Therefore, hybrid material systems consisting of silicon and other materials have been extensively studied to enhance the versatility and functionality of silicon photonics [66-68].
Recently, a hybrid silicon-on-insulation (SOI) and LiNbO3 material system has been introduced to combine the benefits of high optical confinement in silicon waveguides and the second order susceptibility of LiNbO3 [92, 122]. In the hybrid system, a thin film of
88
LiNbO3 is bonded to the top of a silicon waveguide to serve as a portion of the top cladding of an optical waveguide mode by direct [92] or indirect [122] bonding.
Modulation of optical resonances at DC and RF has been demonstrated in the hybrid system. However, no design to date includes integrated electrodes, limiting the tunability
(shift of the optical resonance with unit voltage) and their applications in integrated photonics. By incorporating integrated electrodes in the silicon/LiNbO3 material system, voltage induced electric fields are locally confined to the LiNbO3 thin film thereby enabling large tunability. In contrast to the polymer-on-silicon platform that requires in- device poling of electro-optic polymers [88, 123], the hybrid silicon/LiNbO3 system has immediate access to the second order susceptibility of the LiNbO3 thin film. The use of
BCB for indirect bonding in the hybrid silicon/LiNbO3 system takes advantage of its high thermal stability, having a glass transition temperature of 350 C [106].
In this chapter, a low tuning voltage hybrid silicon and LiNbO3 optical microring resonator with integrated electrodes is presented. The device is fabricated by indirectly bonding an 800 nm thick z-cut LiNbO3 ion-sliced thin film to serve as a portion of the top cladding of a 15 m radius silicon microring resonator using BCB. The design utilizes thin film metal electrodes and an optically transparent electrode wherein the silicon waveguide core serves as both an optical waveguide medium and as a conductive electrode medium. The demonstrated tunability from fabricated devices is 12.5 pm/V, which is over an order of magnitude greater than electrode-free designs. The large tunability enables a scheme for low-power resonance stabilization by applying a temperature-compensating DC voltage.
89
4.2 Device Design
A schematic of the tunable silicon/LiNbO3 microring resonator is shown in Figure 46(a).
The resonator consists of a silicon rib waveguide microring and a LiNbO3 thin film bonded to the silicon resonator as a portion of the top cladding via BCB. The silicon rib waveguides are 500 nm wide with a 70 nm slab thickness and 180 nm rib height. A bottom metal electrode is formed around and exterior to the microring on the silicon slab to provide an electric path to the silicon core. A top metal electrode is aligned to the microring on top of the LiNbO3 thin film and a SiO2 buffer layer that is deposited by plasma enhanced chemical vapor deposition (PECVD). Figure 46(b) shows a schematic of the cross-section of the device structure.
Figure 46. (a) Schematic of a tunable hybrid silicon and LiNbO3 microring resonator with integrated electrodes. For clarity, the PECVD SiO2 top-cladding layer and electrical contact pads are not shown. (b) Schematic of the cross-section of the device structure along the dashed line shown in (a). Both schematics are not drawn to scale.
The LiNbO3 thin film and the PECVD SiO2 buffer layer serve as the top cladding. On the cross-section of the microring waveguide, a portion of the optical guided-mode is 90 within the LiNbO3 thin film. Figure 47(a) shows the fraction of the optical mode power within the LiNbO3 thin film as a function of the thickness of the LiNbO3 thin film for the transverse electric (TE) and transverse magnetic (TM) modes, at 1550 nm optical wavelength, calculated using the semi-vector beam propagation method (BPM). In the simulation, the PECVD SiO2 layer below the aluminum electrode is set to be 1 µm thick.
The refractive indices of silicon, SiO2, LiNbO3 and BCB are set to 3.48, 1.44, 2.14 and
1.54, respectively, in the simulation. The maximum fraction of the optical mode power in the LiNbO3 approaches 42% for the TM mode and 11% for the TE mode as the
LiNbO3 thickness approaches 1 µm . The larger optical mode field overlap with the
LiNbO3 for the TM mode is desirable for greater tuning efficiency. Moreover, the TM mode accesses the r33 electro-optic coefficient in LiNbO3, whereas the TE mode accesses
-1 -1 the r13 electro-optic coefficient (r33 = 31 pm V , r31 = 8 pm V in bulk LiNbO3).
Therefore, the device is designed for TM mode. Furthermore, the fraction of the optical mode power in the LiNbO3 for the TM mode is marginally improved for LiNbO3 thicknesses greater than 600 nm. The thickness of the LiNbO3 is therefore chosen to be
800 nm.
The use of the silicon microring as a transparent conductor minimizes the dielectric layer thickness required to isolate the optical mode from the electrodes [124]. As a result, the voltage induced electric field inside the LiNbO3 cladding layer is enhanced thereby enabling a larger change in the refractive index of the LiNbO3 via the linear electro-optic effect. The PECVD SiO2 cladding layer thickness significantly influences the electric field intensity in the LiNbO3 (z = 29.1) thin film since the permittivity of
91
LiNbO3 is much higher than that of PECVD SiO2 (relative permittivity r = 4.2 [125]).
The red curve in Figure 47(b) shows the vertical (perpendicular to the surface of the substrate) electric field intensity (Ez) of the applied DC field in the LiNbO3 thin film as a function of the PECVD SiO2 thickness simulated using the finite element method (FEM) electro-static solver (COMSOL). The electric field distribution within LiNbO3 is not uniform due to the rib topology of the silicon waveguide. Therefore, the electric fields are evaluated at the center of the LiNbO3 thin film directly above the center of the silicon waveguide. For an 800 nm thick LiNbO3 thin film and an applied voltage of 1 V between the top electrode and the silicon core, Ez in the LiNbO3 thin film decreases from 1.15
V/µm to 0.45 V/µm as the PECVD SiO2 layer increases from 0 nm to 300 nm.
Figure 47. (a) BPM calculations of the optical mode power in LiNbO3 versus the thickness of LiNbO3 for the TM mode and the TE mode in the hybrid Si/LiNbO3 structure. (b) Calculations of the optical loss (blue) induced by the top aluminum electrode and the voltage induced vertical electric field in LiNbO3 (red) versus the PECVD silicon dioxide thickness. The LiNbO3 thin film thickness is set to be 800 nm and the applied voltage is 1 V.
92
A thinner PECVD SiO2 layer is desirable for achieving higher electric field in LiNbO3.
Conversely, reducing the PECVD SiO2 thickness leads to optical loss from the optical mode interacting with the top aluminum electrode. The blue curve in Figure 47(b) shows the TM mode optical loss caused by the aluminum electrode at 1550 nm wavelength, as a function of the PECVD SiO2 thickness, calculated by BPM (Rsoft Beamprop). In the simulation, the imaginary part of the refractive index of aluminum is set to 16 [126]. The thickness of the PECVD SiO2 layer is chosen to be 125 nm. Relatively large electric fields in the LiNbO3 and metal-induced optical loss less than 0.2 dB/cm are expected.
Compared to fabricated devices, the simulations shown in Figure 47 neglect several tens of nanometers of BCB that resides on the top surface of the silicon core. We have previously demonstrated that the BCB layer between the top of the silicon core and the bottom of the LiNbO3 thin film can be as thin as approximately 20~30 nm [122]. The presence of the BCB layer affects both the optical mode overlap in the LiNbO3 thin film and the bending loss of the microring resonator. The optical mode power decreases in the
LiNbO3 as the thickness of the BCB on top of the silicon core increases. The result is lower tuning efficiency. Figure 48(a) shows the TM mode optical bending loss as a function of ring radius for BCB thicknesses of 0 nm, 20 nm, and 40 nm. As the BCB thickness increases, the optical mode becomes less confined and the bending loss increases. Considering fabrication tolerances, the ring radius is chosen to be 15 µm. For a BCB layer of 20 nm on top of the silicon core, the bending loss for a 15 µm radius ring is 0.02 dB/cm. Figure 48(b) shows the cross-section electric field distribution of the final design for the fundamental TM mode at 1550 nm optical wavelength calculated using
93
BPM. Material boundaries are indicated by the white dashed lines and the material regions are indicated in the insets. The BCB thickness is 0 nm in the simulation. The effective index is calculated to be 2.3.
Figure 48. (a) Calculated TM mode optical bending loss versus the ring radius for BCB thickness of 0 nm, 20 nm, and 40 nm between the top of the silicon core and the bottom of the LiNbO3 thin film. (b) Calculated optical electric field distribution of the hybrid silicon and LiNbO3 structure for the fundamental TM mode at 1550 nm wavelength (Ez component).
4.3. Device Fabrication
The fabrication process is shown schematically in Figure 49. The process begins with a silicon-on-insulator (SOI) wafer with a p-type background doping of 1015 cm-3, a buried oxide (BOX) thickness of 1 m, and a silicon device layer thickness of 250 nm. Silicon rib microring waveguides and bus waveguides with inverse width tapers are defined in hydrogen silsesquioxane (HSQ) resist using electron-beam lithography (EBL) and
94 inductively coupled plasma reactive ion etching (ICP-RIE) using Cl2/O2 chemistry. The silicon rib waveguides are 500 nm wide with a 70 nm slab thickness and 180 nm rib height. The radii of the ring resonators are 15 m and the coupling gaps are 350 nm.
Figure 49. Fabrication process of the device: (a) Silicon rib waveguide ring resonator + patterned on SOI wafer using electron beam lithography and plasma etch, (b) BF2 ion implantation, (c) nickel silicidation, (d) 100 nm aluminum bottom electrode deposition, (e) spin-coat of BCB, (f) indirect bonding of LiNbO3 thin film, (g) plasma etch of BCB, (h) deposition of 125 nm PECVD SiO2 (illustrated on top of the LiNbO3 only for simplicity), (i) deposition of 250 nm top aluminum electrode, (j) deposition of 900 nm PECVD SiO2, (k) patterning of via and top aluminum pad, (l) fabrication of cantilever couplers.
95
The bottom electrode fabrication process consists of ion implantation, nickel silicidation, and aluminum evaporation, as shown in Figure 49(b)-Figure 49(d). The
+ silicon slab area around and exterior to the microring is doped with BF2 ions with a fluence of 2×1015 cm-2, resulting in a doping level of 8×1019 cm-3, for contact with the bottom metal electrode. Twenty five nanometers of nickel silicide is formed in the doped region before 100 nm of aluminum is deposited in order to reduce contact resistance.
Thin films of LiNbO3 are obtained from a bulk single crystal z-cut LiNbO3 wafer using helium ion implantation, thermal treatment, and chemical etch [0]. The wafer is implanted on the +z side of the wafer with an implantation energy of 342 keV and a fluence of 41016 ions cm-2. After implantation, the wafer is annealed using rapid thermal annealing (RTA) at 300 C and then etched in 10% hydrofluoric acid (HF) solution. Non-uniform stress produces 900 nm thick LiNbO3 thin films in the areal shape of strips and triangles whose edges are formed along crystal planes. The exfoliated thin films are directly bonded to an unpolished silicon carrier and annealed by RTA at 1000
C for 30s to repair the crystal lattice and restore the electro-optic properties [127]. The surface roughness of the exfoliated side of the LiNbO3 thin film is reduced to 4 nm by
ICP etching with Ar chemistry resulting in a final film thickness of 800 nm. Next, BCB is spin-coated from solution onto the silicon rib waveguides and bottom electrode as shown in Figure 49(e). LiNbO3 thin films are transferred to the top of the SOI ring resonators with the +z side facing down using a fiber tip on a probe station. After transfer, the sample is annealed to 250 C to cure the BCB. The LiNbO3 thin film does not crack after annealing despite the large mismatch in thermal expansion coefficients between silicon
96 and LiNbO3. BCB that is not covered by the LiNbO3 thin film is etched via ICP-RIE with CF4/O2 chemistry as shown in Figure 49(g). A 125 nm thick PECVD SiO2 layer is then blanket deposited to provide optical isolation between the optical mode and the top metal electrode. A 250 nm thick top aluminum electrode is patterned on top of the
PECVD SiO2 isolation layer. A final 900 nm thick PECVD SiO2 film is deposited as a capping layer. Electrical interconnects through the SiO2 capping layer are etched and aluminum contact pads are patterned to allow access to the top and bottom electrodes.
Finally, cantilever couplers are patterned for fiber-to-chip optical coupling [94, 95].
Throughout the fabrication process, EBL with alignment markers is used to generate multi-layer patterns with layer-to-layer misalignment less than 100 nm. Positive tone resist poly-methyl methacrylate (PMMA) and negative tone resist HSQ are used to generate the masks.
4.4. Measurements
4.4.1 DC measurement
The ability to tune the hybrid microring resonator is characterized by optical transmission measurements. The experimental setup is shown in Figure 50 along with a top-view optical micrograph of the fabricated device. An infrared continuous-wave laser source is connected to a polarization controller which outputs linearly polarized TM mode light.
Tapered optical fibers with tip diameters of approximately 2 μm are butt-coupled to the input and output cantilever couplers of the device. The output light is detected using a photodetector and measured by a power meter. Bias voltage is applied to the device
97 through the integrated electrodes using a voltage source and DC probes. DC bias is applied to the top aluminum electrode and the bottom aluminum electrode is grounded.
The measured transmission near 1552 nm as a function of applied voltage is given in
Figure 51(a). By changing the DC bias between -10 V and 10 V, the resonance position linearly shifts by 250 pm between 1552.195 nm and 1551.945 nm. The measured shift in resonance position yields a tunability of 12.5 pm/V. The blueshift (redshift) of the resonance with positive (negative) applied voltage is consistent with the orientation of the
LiNbO3 c-axis.
Figure 50. (a) Measurement setup; (b) Top-view optical micrograph of fabricated device.
The optical transmission of two consecutive resonances with varying DC bias is shown in Figure 51(b). The measured quality factor for both resonances is approximately
11,500 and the free spectral range (FSR) is 7.15 nm, indicating a group index of 3.57.
The ripple between resonances is due to Fabry-Pérot fringes created between the input and output fiber-to-chip coupling facets. Resonance tunability for the resonance near
98
1559 nm is again measured to be 12.5 pm/V. For the measured rings, 10.8 V applied bias is required for tuning of 135 pm which is equal to the full-width-half-maximum (FWHM) of the measured resonance. The demonstrated tunability is equivalent to a VπL value of
2.63 V cm.
Figure 51. (a) Measured optical transmission of a single resonance as a function of applied voltage. The measured resonance tunability is 12.5 pm/V. (b) Measured optical transmission spectrum of two consecutive resonances.
The hybrid silicon/LiNbO3 microring device with integrated electrodes is attractive for tunable on-chip filters, switches and modulators. The tunability demonstrated in this paper is over an order of magnitude greater than the metal electrode-free hybrid silicon/LiNbO3 microring resonator design discussed in reference 16 (~0.6 pm/V) and the
LiNbO3 thin film microring resonator discussed in reference 12 (~1 pm/V). Furthermore, the demonstrated tunability is comparable to state-of-the-art tunable silicon microring resonators based on reverse biased PN junctions (~20 pm/V at the optimal bias condition) 99
[128, 129] and hybrid silicon-polymer slot waveguide microring resonators (16.5 pm/V)
[15]. Finally, resonance tuning of the hybrid microring resonator is linear for both forward and reserve bias with little effect on the quality factor and extinction ratio.
4.4.2 Compensation of Thermal Drift
The temperature sensitivity of the Si/LiNbO3 ring resonator is characterized and compared to a silicon-on-insulator (SOI) ring resonator with similar radius [130]. The SOI ring is clad with
PECVD SiO2 and has the same core dimensions of the silicon in the Si/LiNbO3 device. The device temperature is controlled by a thermoelectric cooler (TEC). The TEC can be either heated or cooled depending on the polarity of the voltage. No temperature feedback is provided. With the temperature control, the resonance wavelength of a hybrid silicon and LiNbO3 ring and a silicon ring is measured as a function of the temperature change, as shown in Figure 52(b).
Figure 52. (a) The temperature change of the TEC as a function of applied bias, (b) Measured resonance detuning versus temperature.
100
The initial ambient temperature is 22 C. The temperature sensitivity of the SOI ring is 103 pm/C, comparable to typical results in the literature [43]. In comparison, the temperature sensitivity of the Si/LiNbO3 ring is 58 pm/C for TM polarization and 87 pm/C for TE polarization. The lower temperature sensitivity is a result of the negative TOC of BCB (-1.5×10-4
-6 -5 /C) and smaller TOC of LiNbO3 (dno/dT = 3.3×10 /C and dne/dT = 3.7 ×10 /C) compared to silicon (1.86×10-4 /C) at room temperature [131, 83, 43]. The lower sensitivity allows for the compensation of a larger temperature range with the same applied voltage.
The compensation of resonance wavelength shift by temperature variations is achieved by applying a DC voltage to the electrodes. AC switching or modulating voltages can be superimposed on the temperature compensating DC voltage using the same electrodes.
15 The resistivity of bulk LiNbO3 crystal at room temperature is approximately 1×10 Ω∙cm
[83] so DC current flow is negligible. Ideally, zero steady state tuning power is drawn in the absence of leakage current. The DC voltage can have either positive or negative polarity, enabling bidirectional temperature compensation.
The TM optical transmission of the Si/LiNbO3 ring is shown for various temperatures in Figure 53 with voltage as parameter. The quality factor and free spectral range are
11,500 and 7.15 nm, respectively. The resonance is blue or red shifted, depending on the polarity of the voltage. A blue shift of resonance frequency is observed for increasingly positive voltage, indicating a decrease in refractive index, consistent with the orientation of the applied electric field and the z-axis of the LiNbO3. Nominally, the extinction ratio of the resonance should not change with voltage and temperature. At a given temperature, the observed variation of extinction ratio with voltage is attributed to Fabry-
101
Pérot fringes created between the two input/output fiber-to-chip coupling facets. At different temperatures, the fiber-to-chip coupling is slightly modified. The slab and tapered waveguide combination in the fiber-to-chip coupler functions as a TM to TE mode converter [132]. Variations in fiber-to-chip coupling produce changes in mode conversion efficiency. Since the TE mode power is all-pass for the TM ring resonance,
TE power variation produces changes in the extinction ratio for the TM output. The extinction ratio variations can be minimized by optimizing the coupler design in the presence of the slab [94, 95,132].
Figure 53. TM mode optical transmission of Si/LiNbO3 ring resonator for applied voltage from -30 V to 50 V for a temperature increase of (a) 0 oC, (b) 5 oC, (c) 10 oC and (d) 15 oC above ambient temperature.
102
As shown in Figure 54, the resonance red shifts by 0.9 nm for a temperature increase of 15 C.
The device tunability is in the range of 12.5 pm/V. The resonance is tuned by 1 nm for a voltage sweep from -30 V to 50 V, allowing for thermal compensation against a temperature fluctuation of up to 17 C. Feedback control can be applied to stabilize the resonance to a given wavelength
[50, 51]. The biasing current is measured with a Keithly 2400 source meter. The measured DC current is below the measurement noise floor of 20 pA over the applied voltage range, indicating the static power consumption is below 1 nW for a bias of 50 V.
Figure 54. Measured TM mode resonance wavelength versus applied voltage, together with linear fit, with temperature as parameter.
The amount of voltage that can be applied to the hybrid Si/LiNbO3 device is limited by the coercive field (21 kV/mm in bulk LiNbO3) and the breakdown fields of LiNbO3 (70 kV/mm)
103 and PECVD SiO2 which is process dependent [133, 134]. A decrease of tunability is observed as the voltage is increased above +45 V, suggesting possible DC drift and material degradation at high electric fields [84]. Lower voltage and larger range of temperature compensation can be achieved by optimizing both the passive athermal design and the active compensation, taking advantage of the negative TOC of BCB that confines a portion of the optical mode.
4.5 Chapter Conclusion and Outlook
A low tuning voltage hybrid silicon and LiNbO3 microring resonator with integrated electrodes is designed, fabricated, and characterized. The hybrid device combines an ion- sliced LiNbO3 thin film with an SOI rib waveguide ring resonator using BCB as an intermediate bonding layer. The tunability of the resonator is enhanced by optimizing the electrode design to increase the electric field intensity in the thin film. The electrode design utilizes thin film metal electrodes and an optically transparent electrode wherein the silicon waveguide core serves as both an optical waveguide medium and as a conductive electrode medium. A tunability of 12.5 pm/V is demonstrated, which is over an order of magnitude greater than electrode-free designs. The large tunability of the hybrid silicon and LiNbO3 device is attractive for compact integrated photonics. Future work involves device optimization for high electrical frequency operation. In addtion, we present the compensation of thermal drift of resonance wavelengths in a Si/LiNbO3 ring resonator by tuning the refractive index of the LiNbO3 via the linear electro-optic effect.
A tuning range of 1 nm with sub-nanowatt static power consumption is demonstrated, enabling compensation of a 17 C temperature variation. In contrast to traditional tuning
104 mechanisms based on resistive heating and carrier transport, the tuning in the Si/LiNbO3 structure draws considerably lower current. The concept presented in this paper can be applied to other ferroelectric materials and electro-optic polymers on silicon-on-insulator.
The approach paves a route for stabilizing resonators in integrated photonics subject to temperature variations
105
CHAPTER 5
High Speed Hybrid Silicon and LiNbO3 Microring Modulator
This chapter presents the experimental demonstration of a hybrid silicon and LiNbO3 electro-optical microring resonator modulator operating at gigahertz frequencies. The device consists of a 15 μm radius silicon ring resonator and a 1 μm thick z-cut LiNbO3 ion-sliced thin film bonded together by benzocyclobutene (BCB). The modulator design is based on tradeoffs between electrical frequency response and optical loss. Fabricated devices operating in the transverse-electric (TE) optical mode exhibit an optical loaded quality factor of 14,000 and a resonance tuning of 3.3 pm/V. High frequency scattering parameters are used to extract an RC circuit model for the modulator. The small-signal electrical-to-optical 3 dB bandwidth is measured to be 5 GHz. Digital modulation with an extinction ratio greater than 3 dB is demonstrated up to 9 Gb/s. Details of design, fabrication, and characterization are conveyed. High-speed and low tuning power chip- scale optical modulators that exploit the high-index contrast of silicon with the second order susceptibility of LiNbO3 are envisioned.
106
5.1 Introduction
Optical modulators are fundamental components for communications and interconnects.
Lithium niobate (LiNbO3) guided-wave electro-optic modulators satisfy bandwidth, linearity, and chirp requirements in fiber-optic transmission systems [83]. Modulators based on diffused waveguides in bulk LiNbO3 substrates are, however, relatively large.
The potential for dense integration is limited.
Modulators based on silicon-on-insulator (SOI) are of interest for CMOS compatible chip-scale modulation [14]. Applications include short-reach optical interconnects [1], long-haul optical communications [16], analog optical links [32], and reconfigurable optical filters [133]. While the large refractive index and low optical absorption of silicon make it an attractive medium for passive waveguiding in the telecommunications wavelength range, unstrained crystalline silicon does not exhibit a linear electro-optic effect. Consequently, silicon optical modulators rely on alternative mechanisms such as the plasma dispersion effect to achieve electro-optical modulation [11, 129].
Single crystalline LiNbO3 thin film provides large χ(2) coefficient with high material stability and low optical loss. A hybrid silicon and LiNbO3 material system consisting of silicon waveguide ring resonators bonded to ion-sliced LiNbO3 thin films has been introduced to combine the dense integration of silicon photonics with the second order susceptibility of LiNbO3 [92, 122, 136]. In the hybrid system, a thin film of LiNbO3 is bonded to the top of a silicon waveguide to serve as a portion of the top cladding of an optical waveguide mode. The advantages are threefold. First, the silicon waveguide can serve as an optically transparent electrode to enhance voltage induced electric fields in
107 the LiNbO3. Second, large shifts of optical resonance with applied voltage are enabled in the absence of significant absorption. Third, steady state DC power consumption for tuning of optical resonance frequencies can potentially be very low due to the capacitive geometry. Tunable filters and radio-frequency electric field sensors have been demonstrated, however, a high-speed electro-optical ring modulator has not been reported in the literature to date.
5.2 Device Design and Fabrication
5.2.1 Device Design
Figure 55. Schematic of the hybrid silicon and LiNbO3 ring modulator.
A schematic of the hybrid silicon and LiNbO3 modulator is shown in Figure 55. The cross-section is through the center of the ring resonator. The device consists of a 15 m radius silicon rib waveguide ring and a one micrometer thick z-cut ion-sliced LiNbO3 thin film bonded via BCB [122]. The rib waveguides are 500 nm wide with a 45 nm slab 108 thickness and 205 nm rib height. The silicon slab is patterned so that it exists only around and exterior to the ring. The silicon core and surrounding silicon slab layer are doped to function as a transparent conductor with reduced series resistance [136]. A voltage applied between the top electrode and the bottom electrode produces an electric field confined between the top electrode and the silicon waveguide core. The electric field interacts with the portion of the optical mode in the LiNbO3 cladding, modifying the mode effective index by the linear electro-optic effect. As a result, the optical transmission response of the ring resonator is modulated.
The 1550 nm wavelength TE optical mode distribution is shown in Figure 56(a) calculated by the beam propagation method. The refractive indices of silicon, SiO2,
LiNbO3 and BCB are set to 3.48, 1.44, 2.21 and 1.54, respectively, in the simulation.
The effective index of the optical mode is 2.66. The optical mode and the electric field vectors overlap in the LiNbO3. The fraction of the optical mode power in the LiNbO3 is
-1 11%. The TE mode accesses the r31 electro-optic coefficient in LiNbO3 (r31 = 8 pm V in bulk LiNbO3) and takes advantage of the nearly vertical voltage induced electric field in the LiNbO3. Also shown is the voltage induced electric field (yellow vectors) from a DC voltage applied between the top electrode and the silicon transparent conductor, using material permittivities in finite element method calculations.
The effective index change of the waveguide mode due to local change of the refractive index of LiNbO3 can be expressed as [137]:
n n (3) eff LiNbO3
109
2 Eopt(rˆ) drˆ n g LiNbO3 2 (4) nLiNbO E (rˆ) drˆ 3 opt where Γ is the confinement factor, neff is the effective index of the mode, ng is the group index, n is the refractive index of LiNbO3, ε is the material permittivity, E (rˆ) is the LiNbO3 opt optical electric field distribution. The integral in the numerator is taken over the LiNbO3 region, and the integral in the denominator is taken over the entire cross section of the optical mode. According to the Pockels effect,
1 n (rˆ) r n3 E(rˆ) (5) LiNbO3 13 LiNbO3 2 where r13 is the electro-optic coefficient of lithium niobate, and E(rˆ) is the vertical component of the electrical field in the lithium niobate. Due to the topology of the silicon waveguide, is not uniformed in the LiNbO3. Combining equation (3) (4) and
(5), the effective index change is expressed as:
2 E (rˆ) E(rˆ) drˆ opt 1 2 LiNbO3 neff r13ng n LiNbO3 2 2 E (rˆ) drˆ (6) opt
For a microring resonator, the resonance wavelength change δλ can be expressed as:
n n eff (7) g The optical mode and the voltage induced electric field are simulated with the beam propagation method and the finite element method, respectively. The group index and
110 the confinement factor are calculated to be 4 and 0.13 respectively. Assuming r13 equals to 8 pm/V and a voltage of 1V is applied, δneff is calculated to be 1.05e-5, and the resonance wavelength tuning is calculated to be 4.08 pm at 1550 nm wavelength, yielding a tunability of 4.08 pm/V.
Figure 56. (a) Calculated optical TE mode distribution at 1550 nm wavelength (Ex component) and DC voltage induced electric field vectors. (b) Calculation of carrier induced optical loss, RC limited bandwidth, and waveguide serial resistance (top axis) versus silicon waveguide doping concentration with P-type dopants, for a 15 µm radius ring.
Due to the fast response time of the electro-optic effect in LiNbO3, the device speed is limited by the RC time constant of the device and the photon lifetime in the ring resonator [14]. The serial resistance of the biasing circuit is mainly from the resistance of the silicon transparent conductor and the contact resistance. To reduce the resistance, the silicon waveguide is blanket implanted with P-type dopants at a light dose, followed by a
P-type heavy dose implantation on the silicon slab for forming the contact. The heavily- doped region is 300 nm away from the silicon core to avoid excessive optical loss. The 111 waveguide serial resistance as a function of the doping concentration for a 15 µm radius ring is shown on the top axis of Figure 56(b), assuming uniform carrier concentration in the lightly-doped waveguide core and slab.
The area of the silicon slab is minimized to reduce both the resistance and the capacitance. The capacitance is extracted from the stored electric energy in the structure shown in Figure 56(a). The parasitic capacitance, CP, and the device capacitance, CJ, in the electrical path through the lightly-doped waveguide core and slab are calculated to be
0.2 fF/µm and 0.21 fF/µm, respectively. The blue curve in Figure 56(b) shows the calculated RC limited bandwidth for a 15 µm radius ring, assuming a practical additional serial resistance of 100 Ω in the circuit, including the metal-semiconductor contact resistance and the pad resistance. The modeling neglects the current path through the buried silicon dioxide and the silicon substrate as the response time is much longer. The calculated RC limited bandwidth can be much faster than 40GHz, showing a potential for the hybrid silicon and LiNbO3 modulator to achieve high speed modulation. For a moderate optical quality factor of around ~10000 with a photon lifetime bandwidth of around 20 GHz, the device speed is mainly limited by the photon lifetime. While a lower resistance and higher bandwidth are enabled by doping the waveguide, waveguide optical propagation loss from the carriers is also induced, as shown by the red curve in Figure
56(b).
5.2.2 Device Fabrication
112
The fabrication process begins with a SOI wafer with a silicon device layer thickness of
250 nm and a buried SiO2 layer thickness of 1 μm. The silicon waveguides are patterned with HSQ resist using electron beam lithography and ICP-RIE to obtain a rib waveguide geometry with a slab thickness of 45 nm [93]. The coupling gap between the bus waveguides and the ring waveguides are 200 nm. The slab is patterned with a second
HSQ resist layer and ICP-RIE. After patterning, the slab only exists around and exterior to the ring waveguides and on one side of the bus waveguides. After patterning the slab regions, the HSQ mask over the waveguide and the slab is removed, and 20 nm of plasma-enhanced chemical vapor deposition (PECVD) SiO2 is deposited. The sample is
+ 13 -2 then blanket implanted with 45 keV BF2 ions at a fluence of 1.3×10 ion cm to lightly dope the silicon core. After implantation, the sample is annealed using rapid thermal annealing (RTA) to activate and drive in the dopants.
+ 15 -2 The slab is then heavily doped with 45 keV BF2 ions at a fluence of 3×10 cm .
Figure 57(a) shows the top view scanning electron micrograph of the device after this step. Nickel silicide is formed in the heavily doped region before 100 nm of aluminum is deposited as the bottom electrode. A 350 nm thick BCB layer is then spin-coated and etched back to 250 nm thickness to allow for planarization of the waveguide topology.
+ To obtain LiNbO3 thin films, a z-cut LiNbO3 wafer is implanted with He ions with an implantation energy of 342 keV and a fluence of 41016 ions cm-2. After implantation, the wafer is flip bonded to an unpolished silicon wafer and baked on a hotplate at 300 oC for 5 minutes. The heating causes blistering of the implanted LiNbO3 wafer surface due to the aggregation of helium bubbles [88]. One micron thick LiNbO3 thin films are
113 exfoliated from the bulk and transferred to the silicon wafer. The exfoliated thin films on the silicon wafer are annealed then by RTA at 1000 C for 30s to repair the crystal lattice and restore the electro-optic properties. The surface roughness of the exfoliated side of the LiNbO3 thin film is 6 nm. LiNbO3 thin films are transferred and bonded to the Si microring resonators using a micro vacuum tip on a probe station. After transfer, the sample is annealed to cure the BCB. Residual BCB not covered by the LiNbO3 thin film is etched via ICP-RIE with CF4/O2 chemistry. One micron thick PECVD SiO2 film is deposited as a capping layer. The SiO2 film over the ring resonator is removed to form a
40 micron diameter via. A 300 nm thick top aluminum electrode, connected to a ground- signal-ground radio-frequency (RF) pad, is patterned on top of the LiNbO3 thin film.
Finally, cantilever couplers are patterned for fiber-to-chip optical coupling [94, 95]. A top-view optical micrograph of the fabricated device is shown in Figure 57(b).
Figure 57. (a) Scanning electron micrograph of the silicon microring resonator after slab patterning and doping; (b) Top-view optical micrograph of fabricated device.
114
5.3 Device Characterization
5.3.1 DC and Small-Signal High-Frequency Measurements
Figure 58. Measured optical transmission of a single resonance as a function of applied voltage.
Optical transmission measurements are performed to characterize the electrical tuning of optical resonances. TE mode light from a tunable infrared continuous-wave laser source is coupled through the input and output fiber-to-chip cantilever couplers of the modulator. DC voltage is applied to the top electrode of the modulator while the bottom electrode is grounded. Figure 4 shows the measured TE mode spectrum as a function of the applied DC voltage. The total optical insertion loss is 4.3 dB. The insertion loss is from 3 dB of fiber-to-waveguide coupling loss and 1.3 dB of waveguide transmission loss. The decrease of the optical transmission minimum with voltage is attributed to
Fabry-Pérot fringes that are present due to fiber-to-chip coupling. The resonance shift is
115
66 pm for a change in the DC bias from -10 V to 10 V, indicating 3.3 pm/V tuning, which equivalent to a VL of 9.1 V-cm. The measured quality factor is 14,000, the full-width half-maximum is 13.7 GHz, and the group index is 4.0. At 1551.856 nm, the optical transmission intensity varies by 5.2 dB with a -5 V to 5 V voltage swing, as indicated by the dashed arrow in Figure 58.
The RF scattering parameter, S11, is measured with a 20 GHz vector network analyzer
(VNA) operating in a 50 system. Figure 59(a) shows the measured S11 magnitude and phase data from 40MHz to 20 GHz, which is the frequency range available in the VNA.
The magnitude of S11 drops sharply to -2 dB from 40 MHz to 2 GHz, and then gradually rolls off to -4.5 dB at 20 GHz. An RC circuit model extracted from the measured S11 is shown in Figure 59(b) [129]. Parameter RC denotes the resistance of the metal electrodes.
Parameter CP represents the parasitic capacitance between the metal electrodes through the top dielectric materials and the air. Parameters CJ and RS model the electrical path through the silicon transparent electrode and the LiNbO3 layer. Parameters COX and RSi model the electrical path through the buried oxide and the silicon substrate. The value of
CJ is in good agreement with finite element method simulations. The relatively high value of RS indicates a lower than expected doping concentration in the waveguide and a higher than expected contact resistance [138]. The relatively large CP value is mainly attributed to the parasitic capacitance of the electric pads. The sharp drop at the lower frequency band is attributed to the RF coupling through the substrate.
The small-signal electrical-to-optical modulation response from 40MHz to 12 GHz is shown in Figure 60. The modulated optical signal is detected with a photodetector with a
116
3 dB bandwidth of 12 GHz. Also shown in Figure 60 is the frequency response of the voltage across CJ in the circuit model shown in Figure 59(b) [129]. The measured 3 dB optical response bandwidth of the modulator is approximately 5 GHz. While the general trend of the optical responses match the circuit model, the optical response contains additional dips and peaks across the entire frequency band.
Figure 59. (a) RF S11 scattering parameter; (b) RC circuit model of the modulator.
Figure 60. Electrical-to-optical modulation response.
117
5.3.2 High-Speed Digital Modulation
A schematic for the high-speed digital modulation characterization is shown in Figure 61
[129]. A pulse pattern generator (PPG) outputs a 231-1 pseudo-random-bit-stream data stream (PRBS31) with a voltage swing from -1 V to 1V. The peak-to-peak amplitude of the signal is amplified to 5 V (-2.5V to 2.5 V) with a modulator driver amplifier. Due to the microwave reflection from the capacitive device, the voltage swing across the modulator electrode is doubled to 10 V at DC [129]. The output light from the modulator is passed through a fiber amplifier and an optical passband filter with 0.5 nm bandwidth to amplify the signal and suppress the amplified spontaneous emission noise. The amplified optical signal is attenuated and connected to a 30 GHz optical module on a digital communication analyzer (DCA) synchronized to the clock of the PPG for generating optical eye diagrams. The optical bias wavelength is tuned to maximize the extinction ratio (ER).
Figure 61. Measurement setup for digital characterization
118
Figure 62. Measured (left column) and simulated (right column) optical eye: (a) and (b) 1 Gb/s, (c) and (d) 4.5 Gb/s, (e) and (f) 5 Gb/s, (g) and (h), 9 Gb/s. The red dashed line in the measurement indicates the reference level for zero optical input. The vertical scale is 500 μW per division.
The left column of Figure 62 shows measured optical eye diagrams at 1 Gb/s, 4.5
Gb/s, 5 Gb/s, and 9 Gb/s, with extinction ratios of 4.7 dB, 4.5 dB, 4 dB and 3 dB, respectively. The right column of Figure 62 shows simulated eye diagrams at the corresponding bit rates with ideal PRBS driving signals. The modeling incorporates the measured frequency response in Figure 60 and the optical resonance spectrum of the 119 modulator in Figure 58. The extinction ratio and shape of the simulated optical eyes agree well with the measurement results. In particular, the amplitude ringing for the 1
Gb/s eye and the partial eye closing on the left part of the 5 Gb/s eye closely match the simulation.
The total energy consumption of the modulator is estimated to be 4.4 pJ/bit at 9 Gb/s with a 10 V swing. The energy per bit encompasses 0.4 pJ/bit from CJ, 1.8 pJ/bit from
CP, and 2.2 pJ/bit from COX. The energy consumption on CJ is only 9% of the total energy consumption. The majority of the energy consumption is attributed to charging and discharging of the large area electrical pads for test and measurement. Energy consumption can be reduced by optimizing device capacitances. Operating in the optical
TM mode also reduces the energy per bit due to the increase in tunability from a larger
-1 electro-optic coefficient (r33 = 31 pm V ) and greater overlap between RF and optical fields. Steady state DC power consumption for tuning of optical resonance frequencies is ideally zero in the capacitive device.
5.4 Acousto-optic anomalies
The resonances on the modulation response spectrum are detrimental for both digital and analog applications. These resonances are a result of acousto-opitc resonances. It is well known that the modulation response of the bulk LiNbO3 Mach-Zenhder interferometer (MZI) modulators are affected by acousto-optic anomalies [139-141].
Due to the piezoelectricity of LiNbO3, the electrodes can function as an acoustic transducer that launches surface acoustic waves and bulk acoustic waves. The acoustic
120 waves change the refractive index of LiNbO3 by the elasto-optic effect. At the resonant frequencies of the acoustic modes, the elasto-optic refractive index change may dominate the modulation response. Modulation response ripples can exceed 10 dB at acoustic resonance frequencies, which are usually at less than 1 GHz [140]. The impact of acousto-optic effects on electro-optic modulation are also observed in III-V material and polymer electroopitc modulators. Depending on the waveguide and electrode design, the resonances can occur at the gigahertz frequency range [142].
Unlike the bulk LiNbO3 MZI modulator, the hybrid silicon and LiNbO3 modulator can be treated as a lossy thin film bulk acoustic wave resonator [143]. The electric fields between the top and bottom electrode modulate the refractive index of LiNbO3 via the electro-optic effects. At the same time, the electric fields also excite acoustic waves through the converse piezoelectric effect of LiNbO3. The acoustic waves propagating in the vertical direction are confined in the LiNbO3 thin film, resulting in thickness excitation mode acoustic resonances. With the 1 μm thickness of the LiNbO3 thin film, the fundamental resonance and its harmonics occur in the gigahertz frequency range. The acoustic energy leaks into the silicon substrate and scatter at the rough silicon bottom surface. Despite the relative weak acoustic resonances for the hybrid modulator compared to thin film bulk acoustic wave resonators design with high-quality acoustic reflectors, the acousto-optic effects are still large enough to generate significant distortion in the electro-optical response by the elasto-optic effect.
Based on the piezoelectric properties of the z-cut LiNbO3 thin film, vertical electrical fields excite the thickness longitudinal mode acoustic resonances, and horizontal
121 electrical fields (in plane) excite the thickness shear mode acoustic resonances [83]. Due to the asymmetry on two sides of the LiNbO3 thin film, both even and odd higher order harmonics of the fundamental resonances can be excited. The resonance frequencies of the higher order resonances can deviate from the harmonics of the fundamental frequencies, depending on the electrode designs. Spurious modes are also excited due to the topology and finite dimensions of the electrodes. Besides the primary modes, acoustic resonances at much lower frequency can also occur, corresponding to acoustic waves travelling parallel to surfaces of the electrodes and bouncing off the edges of the
LiNbO3 thin film.
Two-dimensional numerical modeling of the acoustic resonances of the modulator is conducted using the finite element method. The commercial software package COMSOL
4.4 based on the finite element method, together with the MEMS module is used for the simulation. A rectangular mesh element is used and the meshing size is 100 nm in the vertical direction and 300 nm in the horizontal direction. A mesh convergence study is done to verify that the mesh fulfills mesh convergence. To give a more clear presentation of the acoustic mode and resonances behavior, a simplified model without the silicon waveguide and bottom electrode is adopted, as shown in Figure 63(a). The thickness of each material layer and the width of the aluminum electrode are the same as the fabricated device. The electrical ground is set on the bottom surface of the LiNbO3 layer.
Perfectly matching layers are placed on the left and right side of the simulation domain, and on the bottom silicon substrate [144-146]. Figure 64 shows the measured optical modulation response spectrum (as Figure 60) and the simulated admittance spectrum of
122 the structure in Figure 63(a). The admittance resonances represent the acoustic resonances [143]. The admittance is normalized to the admittance of the structure without the acoustic effects. The shear mode and longitudinal mode resonances are identified and labeled as “S” and “L” respectively, with the number indicating the order of the harmonics. Figure 63(b) and Figure 63(c) show the simulated distortion of the first order shear mode and first order longitudinal mode at the resonances (S1 and L1 in
Figure 64) with the color of the meshes representing the normalized amplitude of the displacement. The amplitude of the distortion is exaggerated for demonstration purpose.
Due to the energy trapping of the aluminum electrode, S1 (L1) mode has major horizontal (vertical) and minor vertical (horizontal) displacement components.
Figure 63. 2D numerical acoustic simulation of a simplified model. (a) The structure and meshing before applying voltage on the aluminum electrode. (b) Distortion at the first order shear mode resonance (c) Distortion at the first order longitudinal mode resonance. The color bar shows the normalized amplitude of the displacement.
As shown in Figure 64, the simulated acoustic resonance frequencies match closely to the resonances on the optical response, indicating that the ripples on the modulation 123 response are a result of the acoustic-optical resonances. Calculation of the optical response based on the acoustic simulation requires accurate resonance and distortion simulation with a 3D model, and is not covered in this dissertation.
Figure 64. Measured optical response versus simulated normalized admittance based on the simplified 2D model in Fig. 8 (a). S1 to S6 represents the first order shear mode resonance and its higher order harmonics. L1 to L3 represents the first order longitudinal mode resonance and its higher order harmonics.
The acoustic resonances can be suppressed by roughening the surface of the piezoelectric material to increase the scattering loss of the acoustic wave [144]. It has been shown that the ripples on the electro-optic response of a bulk LiNbO3 MZI modulator can be smoothen out by roughening the edges and the backside of the LiNbO3 substrate [148]. The surface roughness can be produced by dry or wet etching.
Sufficiently thick LiNbO3 avoids optical mode interaction with the rough surface.
124
Numerical simulation also shows that the amplitude of the admittance resonances of the hybrid silicon and LiNbO3 modulator can be significantly reduced as the top side surface of the LiNbO3 thin film is roughen.
5.5 Chapter Conclusion and Outlook
A hybrid silicon and LiNbO3 microring resonator modulator operating at GHz frequencies is designed, fabricated, and characterized. The hybrid modulator consists of an ion-sliced LiNbO3 thin film bonded to a silicon waveguide ring resonator via BCB.
The modulator design is based on tradeoffs between electrical frequency response and optical loss. Fabricated devices operating in the TE optical mode exhibit an optical loaded quality factor of 14,000 and a resonance tuning of 3.3 pm/V. High frequency scattering parameters are used to extract an RC circuit model for the modulator. The small-signal electrical-to-optical 3 dB bandwidth is measured to be 5 GHz. Digital modulation with an extinction ratio greater than 3 dB is demonstrated up to 9 Gb/s.
Future work involves optimizing the RC time constant, suppressing the acousto-optic resonances, and reducing the driving voltage in TE and TM mode designs to achieve higher data rates and lower energy consumption. High-speed and low tuning power chip- scale optical modulators that exploit the high-index contrast of silicon with the second order susceptibility of LiNbO3 are envisioned. More broadly, new horizons become apparent when exploiting the capability of silicon to provide submicrometer spatial confinement of light and the ability of lithium niobate to mediate second order nonlinear optical effects. Empowering silicon with second order susceptibility opens a suite of
125 nonlinear optics to the chip-scale, including second harmonic generation, difference frequency generation, optical rectification, and sum frequency generation for applications in classical and quantum information processing [60, 64, 149].
126
CHAPTER 6
Highly Linear Hybrid Silicon and LiNbO3 Racetrack Modulator for
Analog Optical Links
In this chapter, a high linearity and compact hybrid silicon and lithium niobate (LiNbO3) modulator for analog optical links is presented based on the integration of a patterned ion-sliced x-cut LiNbO3 thin film on a silicon racetrack [150]. The design enables high speed and high linearity optical modulation based on the linear electro-optic effect of
LiNbO3 on the compact silicon photonic platform. The linearity as a function of the bias wavelength and the RF frequency is studied. The third order intermodulation distortion
2/3 2/3 spurious free dynamic range (SFDR) is measured to be 98.1 dB/Hz and 87.6 dB/Hz at 1 GHz and 10 GHz, respectively. The measured SFDR is over an order of magnitude greater than the SFDR for silicon microring modulators based on the plasma dispersion effect, and is 3.4 dB higher and 1.4 dB lower than a commercial LiNbO3 MZI modulator biased at the quadrature point and measured with the same measurement system. Highly linear, high speed and compact electro-optical modulators that exploit the linear electro- optical effect of LiNbO3 on the silicon-on-insulator platform for analog optical links are envisioned.
127
6.1 Introduction
Lithium niobate MZI modulators based on diffused waveguides, on the other hand, have high linearity and high SFDR [151]. However, the large device size and the
LiNbO3 material itself limit the potential for dense integration with electronics [88].
Recently, the hybrid integration of z-cut ion-sliced LiNbO3 thin films on silicon waveguides with indirect bonding has been demonstrated, enabling compact devices such as RF electric field sensors, tunable filters, and high speed modulators that take advantage of both the high optical confinement of silicon and the second order susceptibility of LiNbO3 [92, 122, 136].
In this chapter, a high linearity and compact hybrid silicon and LiNbO3 racetrack modulator is demonstrated. The device includes a patterned 1 µm thick x-cut LiNbO3 thin film bonded to a silicon racetrack resonator. The design enables high speed and high linearity optical modulation based on the linear electro-optic effect of LiNbO3 on the compact silicon photonic platform. The linearity as a function of the bias wavelength and the RF frequency is studied. The SFDR performance is compared to a packaged all-
LiNbO3 MZI modulator biased at quadrature with the same measurement system. At the optimal bias wavelength, the device has a measured SFDR of 98.1 dB·Hz2/3 at 1 GHz and
2/3 87.6 dB·Hz at 10 GHz, which are 3.4 dB higher and 1.4 dB lower than the LiNbO3
MZI modulator, respectively. For the same optical insertion loss of 3 dB, the SFDR of the device is comparable to the LiNbO3 MZI. The demonstrated IMD3 SFDR in the hybrid silicon and LiNbO3 racetrack is over an order of magnitude greater than the silicon
2/3 ring modulators based on the plasma dispersion effect (84 dB/Hz at 1 GHz) [31].
128
6.2 Theory
We consider a general waveguide optical phase shifter under voltage control. The relationship between waveguide effective index and applied voltage is expressed by the following power series
2 3 푛푒푓푓(푉푖푛) = 푛푒0 + 훼푉푖푛 + 훽푉푖푛 + γ푉푖푛 +⋅⋅ (8) where Vin is the applied voltage, neff is the effective index, and ne0 is the effective index without applied voltage. For silicon phase shifters based on the plasma dispersion effect, the effective index changes nonlinearly with Vin, resulting in relatively large non-zero higher order terms in Eq. ((8). For phase shifters based on the linear electro-optic effect such as in a LiNbO3 MZI modulator or in the hybrid Si/LiNbO3 ring modulator, the effective index changes more linearily with Vin. A good approximation for small signals is
푛푒푓푓(푉푖푛) = 푛푒0 + 훼푉푖푛 (9)
MZI and ring modulators based on highly linear phase shifters still exhibit nonlinearity due to their nonlinear modulation transfer functions [152]. For ring resonators at the critical coupling condition, the steady state photocurrent received as a function of the wavelength and Vin is given by
1 퐼 (휆, 푉 ) = 푅 ∙ 푃 ∙ [1 − ] 표푢푡 푖푛 표푢푡,푚푎푥 4퐹2 휋 ∙ 퐿 ∙ 푛 (푉 ) (10) 1 + sin2 [ 푒푓푓 푖푛 ] 휋2 휆
129 where Iout and R are the detected photo current and the responsivity of the photodetector,
Pout,max is the maximum optical transmission power, F is the finesse of the resonator, L is the length of the resonator, and λ is the wavelength of operation [153]. Based on the nonlinearity in Vin, we express the photocurrent in terms of a power series as
2 3 퐼표푢푡(휆, 푉푖푛) = 푐0(휆) + 푐1(휆) ∙ 푉푖푛 + 푐2(휆) ∙ (푉푖푛) + 푐3(휆) ∙ (푉푖푛) + ⋯ (11)
where cn is the nth order term of the Taylor expansion.
With a two-tone input signal in the form of
푖휔1푡 ∗ −푖휔1푡 푖휔2푡 ∗ −푖휔2푡 푉푖푛 = 퐴1푒 + 퐴1푒 + 퐴2푒 + 퐴2푒 (12) spurious signals are generated at the output due to the nonlinearity. For RF frequencies much smaller than the device bandwidth, the second harmonic distortion (SHD) terms at
2ω1 and 2ω2 and the IMD3 terms at 2ω1-휔2 and 2ω2- ω1 are
2 푖2휔1푡 ∗2 −푖2휔1푡 2 푖2휔2푡 ∗2 −푖2휔2푡 SHD: 푐2(휆) ∙ (퐴1푒 + 퐴1 푒 + 퐴2푒 + 퐴2 푒 (13)
IMD3: 3푐 (휆) ∙ [퐴2퐴∗ 푒푖(2휔1−휔2)푡 + 퐴∗2퐴 푒−푖(2휔1−휔2)푡 + 퐴2퐴∗ 푒푖(2휔2−휔1)푡 3 1 2 1 2 2 1 (14) ∗2 −푖(2휔2−휔1)푡 + 퐴2 퐴1푒 ].
Equations (13) and (14) highlight that SHD and IMD3 depend on bias wavelength.
Steady state numerical calculation shows that the Lorentzian transfer function of a ring resonator has zero third order distortion with a bias point at 0.48 optical transmission and zero second order distortion with a bias point at 0.24 optical transmission [154]. The phenomenon has been demonstrated in a 1 mm radius polymer microring resonator at 30 kHz [154]. However, Distortion-free points are, however, absent in a silicon ring
130 modulator based on the plasma dispersion effect, since the linearity is mainly contributed by the pn junction instead of the Lorentzian transfer function. In contrast, the hybrid
Si/LiNbO3 platform can take advantage of these distortion free bias points since the voltage controlled phase shift is based on the second order susceptibility of the LiNbO3.
6.3 Device Design
Figure 65. (a) Schematic of hybrid silicon and LiNbO3 racetrack modulator. For clarity, the PECVD SiO2 top-cladding layer and contact pads are not shown. (b) Schematic of the cross-section of the device structure along the dashed line in (a). The crystal axes of LiNbO3 is marked.
A schematic of the hybrid silicon and LiNbO3 racetrack modulator is shown in Figure 65.
The device consists of a silicon strip waveguide racetrack resonator and a 1 µm thick x- cut ion-sliced LiNbO3 thin film bonded with BCB as the cladding [155]. The silicon strip waveguide is 550 nm wide and 170 nm thick. The racetrack has a radius of 10 µm on the
131 curves and a length of 50 µm on the straight sections. The bus waveguide to racetrack gap is 180 nm. In contrast to previous hybrid silicon and LiNbO3 devices using un- patterned z-cut LiNbO3 thin films, the modulator integrates patterned x-cut LiNbO3 thin films with controlled size, shape, and crystal axis orientation, providing flexibility and scalability in device design. The z-axis of the crystal is oriented along the edge of the
LiNbO3 thin film and is perpendicular to the straight sections of the racetrack.
Aluminum electrodes are patterned on top of the LiNbO3 thin film with a 1 µm electrode gap centered on the silicon waveguide. The electrodes are interdigitated so that the directions of the applied electric field in the two straight waveguide sections are the same. As a result, the electric field induced phase change accumulates constructively as light propagates along the racetrack. The electrode configuration allows the device to access the r33 electro-optic coefficient of the LiNbO3 along the straight waveguide
-1 sections (r33=31 pm V in bulk LiNbO3). The curved sections have weaker modulation efficiency as the applied electric field and the optical electrical field in the z direction gradually decrease from the end of the straight sections to the center of the arcs. A large ratio between the lengths of the straight section and the curve section is desired for high modulation efficiency.
Figure 66 shows the optical transverse electrical (TE) mode optical distribution in a straight waveguide section at 1550 nm calculated with the beam propagation method.
The commercial software package Rsoft Beamprop is used for the simulation. The grid size is set to 10 nm in x, y, and z direction. The mode effective index is 2.45 and the fraction of the optical mode power in LiNbO3 is 25%. Also shown is the electric field
132
(yellow vectors) from a DC voltage applied between the top metal electrodes. The high waveguide mode confinement allows the top metal electrodes to be patterned close to each other directly over LiNbO3 without inducing large optical absorption loss, enabling large electric fields in LiNbO3 and large device tunability.
Figure 66. Calculated optical TE mode distribution at 1550 nm wavelength (Ex component) and DC voltage-induced electric field vectors.
Compared to previous designs using z-cut LiNbO3 thin film and a silicon transparent conductor, this design with x-cut LiNbO3 and a co-planar top electrode configuration enables low device resistance, large bandwidth and large tunability without the need of doping and contact process [155]. The capacitance per unit length of the electrode is calculated to be 2×10-10 F/m based on finite element method simulation. With a conservative estimation of a device capacitance of 30 fF and a device resistance of 30 Ω, the RC limited bandwidth is calculated to be 66 GHz in a 50 Ω system. As a result, the
133 device speed is limited by the photon lifetime of the racetrack resonator. As the silicon core is not a portion of the biasing circuit, the influence of the silicon carrier effect is minimized and a higher linearity is expected. In addition, the design reduces acoustic resonances built up in the z orientation of the LiNbO3 thin film, resulting in improved optical modulation response with less acousto-optic distortion [139, 140, 155].
6.4 Device Fabrication
X-cut LiNbO3 thin film is obtained by indirect wafer bonding using BCB/silicon dioxide
(SiO2) double layers, patterned by electron-beam lithography and dry etching, and released from the substrate by removing the embedded SiO2 layer using wet etching, as described in Chapter 2. Figure 67 shows the fabrication details of the device.
The fabrication process starts with a silicon on insulator wafer with a 170 nm thick silicon device layer on 1 µm thick buried SiO2 layer. The silicon waveguides are patterned with hydrogen silsesquioxane resist using EBL and ICP-RIE [83]. The waveguide width is 550 nm, and the coupling gap between the bus waveguides and the racetrack waveguides is 180 nm. The sample is coated with 250 nm thick BCB using
BCB solution diluted in mesitylene solvent (Dow Chemical T1100). The sample is then baked in a N2 ambient at 150 °C for one hour to maximize the planarization of the waveguide topology with BCB, and then heated up to 190 °C for 15 minutes to partially cure the BCB. The partial curing process prevents the BCB from reflowing in the later curing process after bonding, providing high BCB thickness uniformity [139]. The BCB is etched back to a 170 nm thickness using ICP-RIE with O2 and CF4 chemistry prior to
134 bonding. Patterned LiNbO3 thin films prepared on silicon substrate are picked up, transferred, and bonded to the silicon racetrack ring waveguides using a micro-vacuum tip on a probe station. The glass micro-vacuum tip with a hose diameter of 25 µm at the end is connected to a vacuum source using a plastic tube. Compared to other pick-and- place tools based on static electric force, such as a fiber tip, the micro-vacuum tip provides reliable and accurate manipulation of larger size thin films. The z axis of the
LiNbO3 along the short thin film edge is aligned to the bus waveguide with an accuracy of ± 3o.
Figure 67. Fabrication process of the device: (a) Silicon strip waveguide racetrack resonator patterned on SOI wafer using EBL and plasma etching, (b) spin-coat, partial curing, and etch back of BCB, (c) transferring and bonding of patterned x-cut LiNbO3 thin film and plasma etch of BCB, (d) Deposition of 1 µm PECVD SiO2 and patterning of via, (e) patterning of signal electrode, (f) patterning of ground electrode and cantilever couplers.
135
After bonding, the sample is capped with a 1 µm thick PECVD SiO2 layer. A via hole is formed by removing the SiO2 film over the racetrack resonator. The thickness of
LiNbO3 in the via hole is reduced to around 900 nm in the etching. Three-hundred nanometer thick Al electrodes are patterned with a two-step lithography process to accurately control the narrow electrode gap. The signal electrode is first formed, and the ground electrode is aligned to the signal electrode in the second lithography step, forming a 1 µm wide electrode gap with an alignment error smaller than 50 nm. Finally, cantilever couplers are patterned for fiber-to-chip optical coupling [95]. A top-view optical micrograph of the fabricated device along with a scanning electron micrograph of the top electrodes are shown in Figure 68.
Figure 68. (a) Top-view optical micrograph of the fabricated device; (b) scanning electron micrograph with a zoom-in view of the top electrodes.
6.5 Measurement
6.5.1 DC measurement 136
Figure 69. (a) Measured optical spectrum as a function of applied voltage; (b) linear fitting of the resonance wavelength shift as a function of applied voltage
The hybrid Si/LiNbO3 racetrack is compared to an all-silicon racetrack. The two devices have the same silicon racetrack core with a straight section length of 50 µm and a curve radius of 10 µm. The free spectral range (FSR) is 4.05 nm and 3.74 nm and the group index is 3.65 and 3.95 for the hybrid Si/LiNbO3 racetrack and the silicon racetrack, respectively. The difference in the group index is a result of different effective index and dispersion. The loaded quality factor is 25,850 for the silicon racetrack, and drops to
15,500 for the hybrid Si/LiNbO3 racetrack. The waveguide loss in the resonator is estimated by
L / 2 e ng Q L (15) (1 e ) where Q is the loaded quality factor, α is the power loss per unit length, L is the length of the resonator, ng is the group index, and λ is the resonance wavelength [156]. The waveguide loss is calculated to be 13.5 dB/cm and 20.7 dB/cm for the silicon racetrack and the hybrid Si/LiNbO3 racetrack. The 7.2 dB/cm excess loss for the hybrid Si/LiNbO3
137 racetrack is mainly attributed to scattering loss from the rough top LiNbO3 surface, which can be reduced by polishing the LiNbO3. Other loss sources can include absorption loss from the top metal electrode and LiNbO3 crystal damage that is not completely restored from thermal annealing. Measurement of another hybrid Si/LiNbO3 racetrack with a wider electrode gap of 1.4 µm results in a loaded quality factor of 17,200 and an estimated loss of 18.7 dB/cm, indicating the metal induced loss is over 2 dB/cm.
Figure 69(a) shows the measured TE-mode optical transmission spectrum as a function of the applied DC voltage between the electrodes. The resonance wavelength blueshifts with increasingly positive voltage, indicating the refractive index is decreased as the voltage increase, which is consistent with the relative orientation of the applied electric field and the z axis of the LiNbO3 thin film. A linear fitting of the resonance wavelength shifts in Figure 69(b) shows a tunability of 5.3 pm/V. The r33 electro-optic coefficient of the LiNbO3 thin film is estimated to be 25.7 pm/V based on the measured tunability, which is 83% of the bulk value.
6.5.1 RF measurement
The RF scattering parameter, S11, is measured with a 20 GHz vector network analyzer
(VNA) operating in a 50 Ω system. As shown in Figure 70, S11 is 0 dB at DC and decreases to –3.3 dB at 20 GHz. For optical wavelength biased at -3 dB optical transmission, the small-signal electrical-to-optical modulation response is obtained from the VNA and a 25 GHz photodetector by taking 1/2 of the S21 scattering parameter in dB.
Ignoring for a moment the deep resonance at 7.5 GHz, the 3 dB modulation bandwidth is
138 approximately 15 GHz. Further device speed improvement can be achieved by decreasing the loaded quality factor of the resonator to obtain a higher photon life time limited bandwidth. With a Q factor of 5000, a modulation bandwidth of up to 40 GHz can be expected.
The strong dips on the optical modulation response are a result of acousto-optic resonances [20,25,26]. Compared to a hybrid silicon and LiNbO3 modulator using z-cut
LiNbO3, fewer acoustic-optic resonances are observed for the x-cut design. The acoustic resonances can be further suppressed by roughening the LiNbO3 thin film top surface
[157]. Alternatively, the thickness of the LiNbO3 thin film can be reduced to push the resonances to higher frequencies that exceed the bandwidth of the modulator.
Figure 70. Measured optical modulation response and the S11 magnitude
139
The SFDR measurement setup is shown in Figure 71. TE-mode infrared light from a tunable laser source is coupled into and out of the modulator via cantilever couplers. The output light is amplified with an erbium doped fiber amplifier (EDFA) and filtered prior to measurement from a 25 GHz photodetector with a conversion gain of 15V/W (New
Focus 1414) and a 40 GHz spectrum analyzer. The optical power reaching the photodetector is maintained at 1 mW for all measurements. Two RF tones with equal power are combined and launched to the modulator using a RF probe. A frequency separation of 6 MHz between the two tones centered about 100 MHz, 1GHz, 5GHz, and
10GHz is used. The fundamental, second harmonic distortion (SHD), and third order intermodulation (IMD3) terms are measured with the RF spectrum analyzer to obtain the
SFDR. The optical bias wavelength is varied on the blue side of the resonance to obtain wavelength dependent response. The noise floors are -157 dBm/Hz at 100 MHz and
1GHz, -150 dBm/Hz at 5GHz, and -148 dBm/Hz at 10 GHz, which are limited by the noise of the RF spectrum analyzer.
Figure 71. Setup for the SFDR measurement
140
Figure 72 shows the fundamental, SHD, and IMD3 components versus detuning wavelength from resonance. At large detuning, the fundamental power increases as the detuning is decreased because the slope of the resonator transfer function is increasing and the EDFA gain is increasing to maintain 1 mW total optical power at the photodetector. Within 10 pm of the resonance, the slope of the resonator transfer function decreases sharply, resulting in the sharp decrease in the fundamental power.
Figure 72. RF output power of the fundamental, SHD, and IMD3 components as a function of wavelength detuning from resonance wavelength at 100 MHz (a), 1GHz (b), 5GHz (c), and 10 GHz (d) respectively.
141
The fundamental power at -54 pm resonance detuning, which is the -3 dB optical transmission wavelength, is -49.3 dBm, -49.1 dBm, -49.3 dBm, and -51.3 dBm for 100
MHz, 1 GHz, 5 GHz, and 10 GHz, respectively, consistent with the optical modulation response in Figure 70. The fundamental power peaks at -37 dBm, -37 dBm, -39.3 dBm and -43.8 dBm for 100 MHz, 1 GHz, 5 GHz, and 10 GHz, respectively. Generally, SHD and IMD3 power increase as the wavelength is tuned closer to the resonance. In the 100
MHz case, local minima are observed for SHD and IMD3, attributed to the Lorentzian transfer function of the ring. Lower distortion occurs at the local minima. At higher frequencies, the local minima no longer appear.
To provide context for the measurements, the Si/LiNbO3 modulator is compared to a commercial LiNbO3 MZI modulator with a Vπ of 3.1 V at DC (Lucent 2623-NA). The
MZI is biased at quadrature and the optical power reaching the photodetector is again maintained at 1 mW for all measurements by adjusting the the EDFA current. As shown
2/3 in Fig. 8 and Fig. 9, the IMD3 SFDR of the LiNbO3 MZI modulator are 94.7 dB Hz
2/3 and 89 dB Hz at 1 GHz and 10 GHz, respectively. The lower SFDR at 10 GHz is primarily due to the higher noise floor of the RF spectrum analyzer at 10 GHz. For the
Si/LiNbO3 modulator, the optimal bias wavelengths for maximum IMD3 SFDR are found to be 15 pm and 40 pm on the blue side of the resonance at 1 GHz and 10 GHz, respectively. At the optimal bias, the IMD3 SFDR of the Si/LiNbO3 modulator is 98.1
2/3 2/3 dB Hz at 1 GHz, and 87.6 dB Hz at 10 GHz. The measured IMD3 SFDR is 3.4 dB higher and 1.4 dB lower than the LiNbO3 MZI at 1 GHz and 10 GHz respectively.
Furthermore, at the -3 dB bias wavelength, the IMD3 SFDR are 94.2 dB Hz2/3 at 1 GHz
142
2/3 and 86.7 dB Hz at 10 GHz, which are 0.5 dB lower and 2.3 dB lower, respectively, than the LiNbO3 MZI. For the same optical insertion loss of 3 dB, the SFDR of the
Si/LiNbO3 ring is comparable to the LiNbO3 MZI, but with a footprint three orders of magnitude smaller.
Figure 73. RF output power of the fundamental and IMD3 components (the higher spur of the two spurs) as a function of RF input power for the LiNbO3 MZI modulator and the hybrid silicon and LiNbO3 racetrack modulator at 0.997 GHz. The noise floor is in 1 Hz bandwidth, limited by the RF spectrum analyzer.
143
Figure 74. RF output power of the fundamental and IMD3 components (the higher spur of the two spurs) as a function of RF input power for the LiNbO3 MZI modulator and the hybrid silicon and LiNbO3 racetrack modulator at 9.997 GHz. The noise floor is in 1 Hz bandwidth, limited by the RF spectrum analyzer.
With proper bias the inherent third order nonlinearity of the Lorentzian transfer function for a ring resonator can be comparable to or smaller than the sinusoidal transfer function of a MZI structure. A high linearity electro-optical modulator with IMD3 SFDR performances comparable to a commercial LiNbO3 MZI modulator can be achieved on the hybrid silicon and LiNbO3 photonic platform with a device size a few hundred times smaller. The demonstrated IMD3 SFDR is over an order of magnitude greater than silicon ring modulators based on the plasma dispersion effect (84 dB·Hz2/3) [31], and is comparable to the state-of-the art silicon MZI carrier depletion modulator using differential drive (97 dB·Hz2/3) [32]. Note that the system noise floor in our measurement is limited by the noise floor of the RF spectrum analyzer. Using state-of- the-art equipment sets, a noise floor of -165 dBm/Hz has been demonstrated using the
144 same photodetector and received optical power (1 mW) [32]. The IMD3 SFDR of our
2/3 2/3 modulator can be improved to 103.4 dB·Hz at 1 GHz and 98.9 dB·Hz at 10 GHz with such a measurement system. The SFDR can be further improved by using a photodetector with higher saturation current or by implementing the modulator in a ring assisted MZI structure [34].
The SHD SFDR of the device is also characterized. SHD SFDR of 78.4 dB·Hz1/2 and
1/2 69.8 dB·Hz are measured at the optimal bias for 1 GHz and 10 GHz, respectively. The measured SHD SFDR is also over an order of magnitude greater than the silicon microring modulator based on the plasma dispersion effect (64.5 dB·Hz2/3 at 1GHz) [31].
6.6 Chapter Conclusion and Outlook
A hybrid silicon and LiNbO3 electro-optical racetrack modulator is designed, fabricated and characterized for analog optical link applications. Single crystalline patterned x-cut
LiNbO3 thin film is integrated on silicon waveguide by indirect bonding, enabling optical modulation fueled by the linear electro-optical effect of LiNbO3 on the compact silicon photonic platform. Spurious free dynamic range (SFDR) of 98.1 dB·Hz2/3 and 87.6
2/3 dB·Hz for IMD3 are measured at 1 GHz and 10 GHz, respectively. The measured
SFDR is over an order of magnitude greater than silicon microring modulators based on the plasma dispersion effect, and is 3.4 dB higher and 1.4 dB lower than a commercial
LiNbO3 MZI modulator biased at the quadrature point. Future chip-scale modulators that exploit the second order susceptibility of LiNbO3 on the silicon-on-insulator platform are envisioned for analog optical links with high linearity.
145
CHAPTER 7
Thesis Conclusion and Suggestions for Future Work
This chapter summarizes the entire dissertation and suggests several possible topics for future work.
7.1 Thesis Conclusion
In this dissertation a hybrid silicon and LiNbO3 integrated photonic platform is proposed for applications in on-chip optical interconnects, filtering, sensing, and analog optical links. The hybrid platform combines the high-index contrast of silicon and the second order susceptibility of LiNbO3. Details of the design, fabrication, and measurement of a series of hybrid silicon and LiNbO3 integrated photonic devices are discussed. A basic hybrid silicon and LiNbO3 photonic structure consists of ion-sliced LiNbO3 thin films bonded to silicon waveguides as the cladding via BCB. Microfabrication techniques are developed for fabricating LiNbO3 thin films, passive and active silicon photonic circuits and hybrid integration of the two materials. Compact hybrid silicon and LiNbO3 microring electric field sensors are first demonstrated for detecting the fringing field of a microstrip RF circuit. With integrated electrodes, voltage-induced electric field is 146 confined to the LiNbO3 thin films, allowing for controlled modulation and filtering functionalities of the devices. Using z-cut LiNbO3 thin films and a silicon transparent conductor design, low voltage tunable filters and high speed microring electro-optical modulators are enabled. Demonstrated tunability of 12.5 pm/V for TM mode is over an order of magnitude greater than previous electrode free designs. Gigahertz speed digital modulation is also demonstrated for optical interconnects. An equivalent circuit model of the modulator is extracted, and anomalies on the modulation frequency response is explained with acoustic optic analysis and simulation. Alternatively, by adopting patterned x-cut LiNbO3 thin film and co-planar metal electrode design, a high linearity hybrid silicon and LiNbO3 racetrack modulator is presented. The SFDR performance for
IMD3 of the device is comparable to commercial LiNbO3 MZI modulator operated at quadrature, and is over an order of magnitude greater than silicon microring modulators based on the plasma dispersion effect.
7.2 Suggestions for Future Work
7.2.1 Suppression of Acoustic-Optical Resonances
In the demonstrated hybrid silicon and LiNbO3 modulators with either z-cut or x-cut
LiNbO3, anomalies caused by acoustic-optical resonances are found on the modulation frequency response. The anomalies are detrimental for both digital and analog optical modulation of the devices, reducing the bandwidth and impairing the modulation quality.
Successful suppression of the acoustic-optical resonance will result in much smoother modulation response, extending the operation bandwidth and greatly improving the
147 modulation quality. Potential approaches may involve with parametric study of the modulation response with various electrode designs and different LiNbO3 thin film dimensions and crystal orientations to identify the effects of each individual parameter.
Three dimensional finite element simulation can be employed to study the acoustic- optical response prior to experimental studies. The key is to avoid or cancel out the acoustic-optic excitation by choosing proper crystal orientation and electrode design, or increase and absorb the acoustic energy by roughening the LiNbO3 surface and applying acoustic impedance matching materials, or push the acoustic resonance frequencies over the device bandwidth by decreasing the critical dimension of LiNbO3 thin films.
7.2.2 Low Power Athermal Microring Modulator
A great advantage of the hybrid silicon and LiNbO3 microring modulator compared to traditional silicon microring modulators is that RF and DC signals can be launched to the same electrode to enable high speed modulation and low-power and linear adjustment of the resonance at the same time. The DC biasing signal is used to compensate the resonance change caused by the temperature fluctuation. The resonance can be tuned to the blue side or the red side by applying positive or negative DC voltages, in response to a temperature increase or decrease. A feedback control loop can be adopted to lock the resonance wavelength to a desirable value. Compared to the traditional method using thermal tuning, the proposed method is much faster and power efficient. In addition, no additional heat is introduced, adding no burden to the limited silicon thermal budget. The challenge is how to increase the tuning range to compensate larger temperature change.
148
7.2.3 Hybrid Silicon and LiNbO3 MZI Modulator
The current hybrid Silicon and LiNbO3 MZI modulators are microring based modulator that have narrow operating wavelength band and are susceptibility to fabrication imperfection and temperature fluctuation. A hybrid silicon and LiNbO3 MZI modulator can have both large operating wavelength range and low temperature sensitivity. In addition, the RF bandwidth is not limited to the photon lifetime in the resonator. The x- cut LiNbO3 design demonstrated in Chapter 6 provides a promising direction for the MZI modulator. Push-pull co-planar electrodes can be employed to increase the modulation efficiency. High speed and low power hybrid silicon and LiNbO3 MZI modulator using x-cut LiNbO3 thin film is envisioned. The challenge is how to fabricate such a hybrid silicon and LiNbO3 photonic structure that is much larger than the current microring based designs, and how to achieve low voltage and high speed operation with proper electrode design.
7.2.4 Improve the hybrid integration process
The current process to integrated LiNbO3 thin films to silicon waveguides is based on a pick-and-place process, which is inherently time-consuming and lack of accurate control of the position of the LiNbO3 thin films. Particles are also more likely to be introduced in the process that degrade the bonding quality. An integration process capable of parallel transferring of the LiNbO3 thin films can dramatically improve the fabrication yields and speed, and provide design flexibility as all the LiNbO3 thin films on the same die are configured and processed at the same time. The patterned LiNbO3 thin film releasing
149 process that enables wafer-scale production of LiNbO3 thin film patterns shown in
Chapter 2 is promising for such an integration process. Potentially, the patterned LiNbO3 thin films can be released from the LiNbO3 substrate and transferred to the silicon waveguides using a “transfer printing” process with a polydimethylsiloxane (PDMS) stamp. The transfer printing process has been widely employed to transfer arrays of thin film electronic and photonic devices with high yield. The challenge is to engineer the adhesion among various materials so that thin films can be released from the host substrate and bonded to the receiver substrate. Besides the transfer printing technique, it is also possible to first flip bonded the LiNbO3 substrate with LiNbO3 thin films on it to the SOI substrate and remove the LiNbO3 substrate. The challenge is to maintain the process at a low temperature to avoid cracking of LiNbO3 due to the thermal expansion mismatch between LiNbO3 and silicon.
7.2.5 Nonlinear Optics on Silicon
High-speed and low tuning power chip-scale optical modulators are envisioned. More broadly, new horizons become apparent when exploiting the capability of silicon to provide submicrometer spatial confinement of light and the ability of lithium niobate to mediate second order nonlinear optical effects. Empowering silicon with second order susceptibility opens a suite of nonlinear optics to the chip-scale, including second harmonic generation, difference frequency generation, optical rectification, and sum frequency generation for applications in classical and quantum information processing
150
Bibliography
1. D. A. B. Miller, “Device requirements for optical interconnects to silicon chips,” Proc. IEEE. 97, 1166-1185 (2009)
2. J. W. Goodman, F. J. Leonberger, S.-Y. Kung, and R. A. Athale, “Optical interconnections for VLSI systems,” Proc. IEEE 72, 850 (1984)
3. B. Arimilli, and S. Baumgartner, "The IBM POWER7 HUB module: a terabyte interconnect switch for high-performance computer systems," Hot Chips. 22, (2010)
4. J. Capmany and D. Novak, “Microwave photonics combines two worlds”, Nat. Photon. 1, 319 (2007).
5. C. Lim, A. Nirmalathas, M. Bakaul, P. Gamage, Y. Ka-Lun Lee, Yizhuo Yang, D. Novak, and R. Waterhouse, “Fiber-wireless networks and subsystem technologies,” J. Lightwave Technol. 28(4), 390–405 (2010).
6. R. C. J. Hsu, A. Ayazi, B. Houshmand, and B. Jalali, “All-dielectric photonic- assisted radio front-end technology,” Nature Photon. 1, 535-538 (2007).
7. L. Liao, A. Liu, D. Rubin, J. Basak, Y. Chetrit, H. Nguyen, R. Cohen, N. Izhaky, and M. Paniccia, “40 Gbit/s silicon optical modulator for high-speed applications,” Electron. Lett. 43, 1196 (2007)
8. S. Assefa, F. Xia, S. W. Bedell, Y. Zhang, T. Topuria, P. M. Rice, and Y. A. Vlasov, "CMOS-integrated high-speed MSM germanium waveguide photodetector," Opt. Express.18, 4986-4999 (2010)
9. Boyd, R. W. Nonlinear Optics. Boston : Academic Press, (2002)
10. P. Dong, W. Qian, H. Liang, R. Shafiiha, N.-N. Feng, D. Feng, X. Zheng, A. V. Krishnamoorthy, and M. Asghari, “Low power and compact reconfigurable multiplexing devices based on silicon microring resonators,” Opt. Express 18, 9852–9858 (2010)
151
11. Q. Xu, B. Schmidt, S. Pradhan, and M. Lipson, “Micrometre-scale silicon electro- optic modulator,” Nature 435, 325–327 (2005)
12. F. Y. Gardes, A. Brimont, P. Sanchis, G. Rasigade, D. Marris-Morini, L. O’Faolain, F. Dong, J. M. Fedeli, P. Dumon, L. Vivien, T. F. Krauss, G. T. Reed, and J. Martí, “High-speed modulation of a compact silicon ring resonator based on a reverse- biased pn diode,” Opt. Express 17, 21986–21991 (2009)
13. R. A. Soref, and B. R. Bennett, “Electrooptical effects in silicon,” IEEE J. Quantum Electron. 23(1), 123–129 (1987).
14. G. T. Reed, G. Mashanovich, F. T. Gardes, and D. J. Thomson, “Silicon optical modulators,” Nat. Photonics 4(8), 518–526 (2010)
15. A. Liu, R. Jones, L. Liao, D. Samara-Rubio, D. Rubin, O. Cohen, R. Nicolaescu, and M. Paniccia, “A high-speed silicon optical modulator based on a metal-oxide- semiconductor capacitor,” Nature 427(6975), 615–618 (2004).
16. A. Biberman, S. Manipatruni, N. Ophir, L. Chen, M. Lipson, and K. Bergman, “First demonstration of long-haul transmission using silicon microring modulators,” Opt. Express 18(15), 15544–15552 (2010).
17. S. Manipatruni, K. Preston, L. Chen, and M. Lipson, “Ultra-low voltage, ultra-small mode volume silicon microring modulator,” Opt. Express18(17), 18235–18242 (2010)
18. S. Manipatruni, Q. Xu, B. Schmidt, J. Shakya, and M. Lipson, “High speed carrier injection 18Gb/s silicon micro-ring electro-optic modulator,” in proceedings of IEEE on the Annual Meeting of the IEEE LEOS (Institute of Electrical and Electronics Engineers, New York, 2007), pp.537–538, (2007)
19. W. M. Green, M. J. Rooks, L. Sekaric, and Y. A. Vlasov, “Ultra-compact, low RF power, 10 Gb/s silicon Mach-Zehnder modulator,” Opt. Express15(25), 17106– 17113 (2007)
20. A. Liu, L. Liao, D. Rubin, H. Nguyen, B. Ciftcioglu, Y. Chetrit, N. Izhaky, and M. Paniccia, “High-speed optical modulation based on carrier depletion in a silicon waveguide,” Opt. Express 15(2), 660–668 (2007)
21. Y. Tang, H.-W. Chen, S. Jain, J. D. Peters, U. Westergren, and J. E. Bowers, “50 Gb/s hybrid silicon travelingwave electroabsorption modulator,” Opt. Express 19(7), 5811–5816 (2011).
152
22. E. Timurdogan, C. M. Sorace-Agaskar, J. Sun, E. S. Hosseini, A. Biberman, and M. R. Watts, “An ultralow power athermal silicon modulator,” New Communication 5, 4008 (2014).
23. P. Dong, R. Shafiiha, S. Liao, H. Liang, N. N. Feng, D. Feng, G. Li, X. Zheng, A. V. Krishnamoorthy, and M. Asghari, “Wavelength-tunable silicon microring modulator,” Opt. Express 18(11), 10941–10946 (2010).
24. J. Capmany and D. Novak, “Microwave photonics combines two worlds,” Nat. Photonics 1(6), 319–330 (2007)
25. D. Marpaung, C. Roeloffzen, R. Heideman, A. Leinse, S. Sales, and J. Capmany, “Integrated microwave photonics,” Laser Photon. Rev. 7(4), 506-538 (2013)
26. B. Jalali and S. Fathpour, “Silicon Photonics,” J. Lightwave Technol. 24(12), 4600 (2006).
27. R. A. Soref, “The past, present and future of silicon photonics,” IEEE J. Sel. Top. Quantum Electron. 12(6), 1678–1687 (2006)
28. M. H. Khan, H. Shen, Y. Xuan, L. Zhao, S. Xiao, D. E. Leaird, A. M. Weiner, and M. Qi, “Ultrabroad-bandwidth arbitrary radiofrequency waveform generation with a silicon photonic chip-based spectral shaper,” Nat. Photonics 2(4), 117-122 (2010)
29. P. Dong, N.-N. Feng, D. Feng, W. Qian, H. Liang, D. C. Lee, B. J. Luff, T. Banwell, A. Agarwal, P. Toliver, R. Menendez, T. K. Woodward, and M. Asghari, “GHz-bandwidth optical filters based on high-order silicon ring resonators,” Opt. Express 18(23), 23784– 23789 (2010)
30. C. H. Cox III, E. I. Ackerman, G. E. Betts, and J. L. Prince, “Limits on the performance of RF-over-fiber links and their impact on device design,” IEEE Trans. Microwave Theory Tech. 54, 906–920 (2006)
31. A. Ayazi, T. Baehr-Jones, Y. Liu, A. E.-J. Lim, and M. Hochberg, “Linearity of silicon ring modulators for analog optical links,” Opt. Express 20(12), 13115–13122 (2012)
32. M. Streshinsky, A. Ayazi, Z. Xuan, A. E.-J. Lim, G.-Q. Lo, T. Baehr-Jones, and M. Hochberg, “Highly linear silicon traveling wave Mach-Zehnder carrier depletion modulator based on differential drive,” Opt. Express 21(3), 3818–3825 (2013).
153
33. A. Khilo, C. M. Sorace, and F. X. Kärtner, “Broadband linearized silicon modulator,” Opt. Express 19(5), 4485-4500 (2011)
34. J. Cardenas, P. A. Morton, J. B. Khurgin, A. Griffith, C. B. Poitras, K. Preston, and M. Lipson, “Linearized silicon modulator based on a ring assisted Mach Zehnder inteferometer,”Opt. Express 21(19), 22549-22557 (2013)
35. C. Xiong, D. M. Gill, J. C. Rosenberg, M. Khater, T. Barwicz, S. Assefa, S. M. Shank, C. Reinholm, E. Kiewra, J. Ellis-Monaghan, S. Kamlapurkar, A. D. Stricker, W. M. J. Green, Y. A. Vlasov, W. Haensch, “A linear push-pull silicon optical modulator,” in Frontier in Optics (FIO), Optical Society of America, (2014)
36. M. S. Dahlem, C. W. Holzwarth, A. Khilo, F. X. Kärtner, H. I. Smith, and E. P. Ippen, “Reconfigurable multichannel second-order silicon microring-resonator filterbanks for on-chip WDM systems,” Opt. Express 19(1), 306–316 (2011)
37. H. L. R. Lira, S. Manipatruni, and M. Lipson, “Broadband hitless silicon electro- optic switch for on-chip optical networks,” Opt. Express 17(25), 22271–22280 (2009)
38. K. De Vos, I. Bartolozzi, E. Schacht, P. Bienstman, and R. Baets, “Silicon-on- insulator microring resonator for sensitive and label-free biosensing,” Opt. Express 15, 7610-7615 (2007)
39. M.-S. Kwon and W. H. Steier, “Microring-resonator-based sensor measuring both the concentration and temperature of a solution,” Opt. Express 16, 9372–9377 (2008)
40. S. C. Hagness, D. Rafizadeh, S. T. Ho, and A. Taflove. "FDTD microcavity simulations: design and experimental realization of waveguide-coupled single-mode ring and whispering-gallery-mode disk resonators," J. Lightwave Technol. 15, 2154-2165, (1997)
41. B. Guha, B. B. C. Kyotoku, and M. Lipson, “CMOS-compatible athermal silicon microring resonators,” Opt. Express 18, 3487–3493 (2010)
42. B. Guha, K. Preston, and M. Lipson, “Athermal silicon microring electro-optic modulator,” Opt. Lett. 37, 2253–2255 (2012)
43. J. Teng, P. Dumon, W. Bogaerts, H. Zhang, X. Jian, X. Han, M. Zhao, G. Morthier, and R. Baets, “Athermal Silicon-on-insulator ring resonators by overlaying a polymer cladding on narrowed waveguides,” Opt. Express 17, 14627–14633 (2009)
154
44. B. Guha, and M. Lipson, “Controlling thermo-optic response in microresonators using biomaterial cantilevers,” Opt. Lett., 40(1), 103-106 (2015).
45. S. S. Djordjevic, K. Shang, B. Guan, S. T. S. Cheung, L. Liao, J. Basak, H. Liu, and S. J. B. Yoo, “CMOS-compatible, athermal silicon ring modulators clad with titanium dioxide,” Opt. Express, 21(12), 13958-13968 (2013).
46. A. V. Krishnamoorthy, X. Zheng, G. Li, J. Yao, T. Pinguet, A. Mekis, H. Thacker, I. Shubin, Y. Luo, K. Raj, and J. E. Cunningham, “Exploiting CMOS Manufacturing to Reduce Tuning Requirements for Resonant Optical Devices,’’ IEEE Photonics J., 3(3), 567-579 (2011).
47. W. A. Zortman, D. C. Trotter, and M. R. Watts, “Silicon photonics manufacturing,” Opt. Express 18(23), 23598–23607 (2010).
48. P. Dong, W. Qian, H. Liang, R. Shafiiha, D. Feng, G. Li, J. E. Cunningham, A. V. Krishnamoorthy, and M. Asghari, “Thermally tunable silicon racetrack resonators with ultralow tuning power,” Opt. Express, 18(19), 20298–20304 (2010).
49. K. Padmaraju, D. F. Logan, X. Zhu, J. J. Ackert, A. P. Knights, and K. Bergman, “Integrated thermal stabilization of a microring modulator,” Opt. Express, 21(12), 14342–14350 (2013).
50. K. Padmaraju, J. Chan, L. Chen, M. Lipson, and K. Bergman, “Thermal stabilization of a microring modulator using feedback control,” Opt. Express 20, 27999–28008 (2012)
51. C. T. DeRose, M. R. Watts, D. C. Trotter, D. L. Luck, G. N. Nielson, and R. W. Young, “Silicon microring modulator with integrated heater and temperature sensor for thermal control,” Proc. Conference on Lasers and Electro-Optics (Optical Society of America, 2010), paper CThJ3
52. K. Padmaraju and K. Bergman, “Resolving the thermal challenges for silicon microring resonator devices,” Nanophotonics, 13, 1–14 (2013).
53. J. E. Cunningham, I. Shubin, X. Zheng, T. Pinguet, A. Mekis, Y. Luo, H. Thacker, G. Li, J. Yao, K. Raj, and A. V. Krishnamoorthy, "Highly-efficient thermally-tuned resonant optical filters," Opt. Express, 18(18), 19055-19063 (2010).
54. X. Zheng, Y. Luo, G. Li, I. Shubin, H. Thacker, J. Yao, K. Raj, J. E. Cunningham, and A. V. Krishnamoorthy, "Enhanced optical bistability from self-heating due to free carrier absorption in substrate removed silicon ring modulators," Opt. Express, 20, 11478-11486 (2012). 155
55. B. E. Little, S. T. Chu, W. Pan, and Y. Kokubun, “Microring resonator arrays for VLSI photonics,” IEEE Photon. Technol. Lett. 12, 323–325 (2000).
56. M. Gould, T. Baehr-Jones, R. Ding, S. Huang, J. Luo, A. K.-Y. Jen, J.-M. Fedeli, M. Fournier, and M. Hochberg, “Silicon-polymer hybrid slot waveguide ring-resonator modulator,” Opt. Express 19, 3952–3961 (2011)
57. R. S. Jacobsen, K. N. Andersen, P. I. Borel, J. Fage-Pedersen, L. H. Frandsen, O. Hansen, M. Kristensen, A. V. Lavrinenko, G. Moulin, H. Ou, C. Peucheret, B. Zsigri, and A. Bjarklev, “Strained silicon as a new electro-optic material,” Nature 441, 199–202 (2006).
58. B. Chmielak, M. Waldow, C. Matheisen, C. Ripperda, J. Bolten, T. Wahlbrink, M. Nagel, F. Merget, and H. Kurz, “Pockels effect based fully integrated, strained silicon electro-optic modulator,” Opt. Express 19, 17212–17219 (2011).
59. B. Chmielak, C. Matheisen, C. Ripperda, J. Bolten, T. Wahlbrink, M. Waldow, and H. Kurz, “Investigation of local strain distribution and linear electro-optic effect in strained silicon waveguides,” Opt. Express 21, 25324–25332 (2013).
60. M. Wächter, C. Matheisen, M. Waldow, T. Wahlbrink, J. Bolten, M. Nagel, and H. Kurz, “Optical generation of terahertz and second-harmonic light in plasma-activated silicon nanophotonic structures,” Appl. Phys. Lett. 97, 161107 (2010).
61. S. Abel, T. Stöferle, C. Marchiori, C. Rossel, M. D. Rossell, R. Erni, D. Caimi, M. Sousa, A. Chelnokov, B. J. Offrein, and J. Fompeyrine, “A strong electro-optically active lead- free ferroelectric integrated on silicon,” Nat. Commun. 4, 1671 (2013).
62. C. Xiong, W. H. Pernice, J. H. Ngai, J. W. Reiner, D. Kumarh, F. J. Walker, C. H. Ahn, H. X. Tang, “Active silicon integrated nanophotonics: ferroelectric BaTiO3 devices,” Nano Lett. 14, 1419-1425 (2014).
63. X. Zhang, A. Hosseini, H. Subbaraman, S. Wang, Q. Zhan, J. Luo, A. K. –Y. Jen, and R. T. Chen, “Integrated photonic electromagnetic field sensor based on broadband bowtie antenna coupled silicon organic hybrid modulator,” J. Lightwave Technol., 2319152 (2014).
64. M. Cazzanelli, F. Bianco, E. Borga, G. Pucker, M. Ghulinyan, E. Degoli, E. Luppi, V. Véniard, S. Ossicini, D. Modotto, S. Wabnitz, R. Pierobon, and L. Pavesi, “Second- harmonic generation in silicon waveguides strained by silicon nitride,” Nat. Mater. 11, 148–154 (2011).
156
65. J. W. Wu, J. F. Valley, S. Ermer, E. S. Binkley, J. T. Kenney, G. F. Lipscomb, and R. Lytel, “Thermal stability of electroopitc response in poled polyimide systems,” Appl. Phys. Lett. 58, 225 (1991)
66. M. J. R. Heck, H.-W. Chen, A. W. Fang, B. R. Koch, D. Liang, H. Park, M. Sysak, and J. E. Bowers, “Hybrid Silicon Photonics for Optical Interconnects,” IEEE J. Sel. Top. Quantum Electron. 17, 333-346 (2011)
67. C. Koos, P. Vorreau, T. Vallaitis, P. Dumon, W. Bogaerts, R. Baets, B. Esembeson, I. Biaggio, T. Michinobu, F. Diederich, W. Freude, and J. Leuthold, “All-optical high-speed signal processing with silicon organic hybrid slot waveguides,” Nat. Photonics 3, 216–219 (2009)
68. J. Liu, M. Beals, A. Pomerene, S. Bernardis, R. Sun, J. Cheng, L. Kimerling, and J. Michel, “Waveguideintegrated, ultralow-energy GeSi electroabsorption modulators,” Nat. Photonics 2(7), 433–437 (2008).
69. Y. H. Kuo, Y. K. Lee, Y. Ge, S. Ren, J. E. Roth, T. I. Kamins, D. A. Miller, and J. S. Harris, “Strong quantumconfined Stark effect in germanium quantum-well structures on silicon,” Nature 437(7063), 1334–1336 (2005).
70. M. Liu, X. Yin, E. Ulin-Avila, B. Geng, T. Zentgraf, L. Ju, F. Wang, and X. Zhang, “A graphene-based broadband optical modulator,” Nature 474, 64–67 (2011)
71. Z. Lu, L. Zhao, “Nanoscale electro-optic modulators based on graphene-slot waveguides,” J. Opt. Soc. Am. B 6(6), 1490–1496 (2012)
72. A. Melikyan, L. Alloatti, A. Muslija, D. Hillerkuss, P. C. Schindler, J. Li, R. Palmer, D. Korn, S. Muehlbrandt, D. V. Thourhout, B. Chen, R. Dinu, M. Sommer, C. Koos, M. Kohl, W. Freude, and J. Leuthold, “High-speed plasmonic phase modulators,” Nat. Photonics 8(3), 229–233 (2014)
73. R. M. Briggs, I. M. Pryce, and H. A. Atwater, “Compact silicon photonic waveguide modulator based on the vanadium dioxide metal-insulator phase transition,” Opt. Express 18, 11192–11201 (2010)
74. A. W. Fang, H. Park, O. Cohen, R. Jones, M. Paniccia, J. E. Bowers, “Electrically pumped hybrid AlGaInAs-silicon evanescent laser,” Opt. Express 14, 9203-9210 (2006).
75. H. Park, Y.-H. Kuo, A. W. Fang, R. Jones, O. Cohen, M. J. Pannicia, J. E. Bowers, "A hybrid AlGaInAssilicon evanescent preamplifier and photodetector," Opt. Express 15, No. 21, (2007).
157
76. Y. H. Kuo, H. W. Chen, and J. E. Bowers, “High speed hybrid silicon evanescent electroabsorption modulator,” Opt. Express 16(13), 9936–9941 (2008).
77. D. Dai and S. He, “A silicon-based hybrid plasmonic waveguide with a metal cap for a nano-scale light confinement,” Opt. Express 17(19), 16646–16653 (2009)
78. J. N. Caspers, M. Z. Alam, and M. Mojahedi, “Compact hybrid plasmonic polarization rotator,” Opt. Lett. 37(22), 4615–4616 (2012).
79. Y. Tang, J. D. Peters, and J. E. Bowers, “Over 67 GHz bandwidth hybrid silicon electroabsorption modulator with asymmetric segmented electrode for 1.3 μm transmission,” Opt. Express 20, 11529–11535 (2012).
80. S. Srinivasan, D. Liang, M. Fiorentino, R. G. Beausoleil, J. E. Bowers, “Micro-ring resonator based electro-absorption modulators on the hybrid III-V on silicon platform,” in Group IV Photonics (GFP), 99-101 (2012)
81. M. Lawrence, “Lithium niobate integrated optics," Reports on Progress in Physics 56, 363-429, (1993)
82. R. S. Weis, and T. K. Gaylord, “Lithium niobate: summary of physical properties and crystal structure,” Appl. Phys., A Mater. Sci. Process. 37, 191–203 (1985)
83. K. K. Wong, Properties of Lithium Niobate (INSPEC, 2002).
84. E. L. Wooten, K. M. Kissa, A. Y. Yan, E. J. Murphy, D. A. Lafaw, P. F. Hallemeier, D. V. Attanasio, D. J. Fritz, G. J. McBrien, and D. E. Bossi, “A review of lithium niobate modulators for fiber-optic communications systems,” IEEE J. Sel. Top. Quantum Electron. 6(1), 69–82 (2000)
85. M. Levy, R. M. Osgood, R. Liu, L. E. Cross, G. S. Cargill, A. Kumar, and H. Bakhru, “Fabrication of single–crystal lithium niobate films by crystal ion slicing,” Appl. Phys. Lett. 73, 2293–2295 (1998)
86. P. Rabiei and W. H. Steier, “Lithium niobate ridge waveguides and modulators fabricated using smart guide,”Appl. Phys. Lett. 86, 161115 (2005)
87. G. Poberaj, H. Hu, W. Sohler, and P. Günter, “Lithium niobate on insulator (LNOI) for micro-photonic devices,” Laser & Photon. Rev. 6(4), 488–503 (2012)
88. A. Guarino, G. Poberaj, D. Rezzonico, R. Degl’Innocenti, and P. Günter, “Electro- optically tunable microring resonators in lithium niobate,” Nature Photon. 1, 407- 410 (2007)
158
89. V. Stenger, J. Toney, A. Pollick, J. Busch, J. Scholl, P. Pontius, and S. Sriram, “Engineered thin film lithium niobate substrate for high gain-bandwdith electro- opic modulators, ” in CLEO 2013, San Jose, 2013.
90. R. Kou, T. Okubo, S. Kurimura, H. Nakajima, J. Ichikawa, and K. Kondou, “Quasi- phase-matched wavelength converter with adhered ridge waveguide for hybrid silicon photonics,” in CLEO: 2008, OSA Technical Digest (CD) (Optical Society of America, 2008), paper CThl4.
91. R. Rabiei, J. Ma, J. Chiles, S. Khan, and S. Fathpour, “High-Q lithium niobate microring-resonators on silicon,” in CLEO: 2013, OSA Technical Digest (online) (Optical Society of America, 2013), paper CTu1F.3.
92. Y. S. Lee, G.-D. Kim, W.-J. Kim, S.-S. Lee, W.-G. Lee, and W. H. Steier, “Hybrid Si-LiNbO3 microring electro-optically tunable resonators for active photonic devices,” Opt. Lett. 36, 1119-1121 (2011)
93. M. Wood, L. Chen, J. R. Burr, and R. M. Reano, “Optimization of electron beam patterned hydrogen silsesquioxane mask edge roughness for low-loss silicon waveguides”, J. Nanophoton. 8(1), 083098 (2014).
94. M. Wood, P. Sun, and R. M. Reano, “Compact cantilever couplers for low-loss fiber coupling to silicon photonic integrated circuits,” Opt. Express 20(1), 164-172 (2012)
95. P. Sun and R. M. Reano, “Cantilever couplers for intra-chip coupling to silicon photonic integrated circuits,” Opt. Express 17(6), 4565-4574 (2009)
96. M. S. Dresselhaus and R. Kalish, Ion Implantation in Diamond, Graphite and Related Materials (Springer, Berlin, 1992)
97. A. M. Radojevic, M. Levy, R. M. Osgood, A. Kumar, H. Bakhru, C. Tian, and C. Evans, “Large etch-selectivity enhancement in the epitaxial liftoff of single-crystal LiNbO3 films,” Appl. Phys. Lett. 74(21), 3197–3199 (1999)
98. F. Schrempel, T. Gischkat, H. Hartung, T. Höche, E. B. Kley, A. Tünnermann, and W. Wesch, “Ultrathin membranes in x-cut lithium niobate,” Opt. Lett. 34(9), 1426– 1428 (2009)
99. J. A. Rogers, M. G. Lagally, and R. G. Nuzzo, “Synthesis, assembly and applications of semiconductor nanomembranes,” Nature 477(7362), 45–53 (2011)
159
100. J. Hu, L. Li, H. Lin, P. Zhang, W. Zhou, and Z. Ma, “Flexible integrated photonics: where materials mechanics and optics meet,” Opt. Mater. Express 3(9), 1313–1331 (2013).
101. B. G. Yacobi, S. Martin, K. Davis, A. Hudson, and M. Hubert, “Adhesive bonding in microelectronics and photonics,” J. Appl. Phys.: Appl. Phys. Rev. 91, 6227-6262 (2002)
102. F. Niklaus, G. Stemme, J.-Q. Lu, and R. J. Gutmann, “Adhesive wafer bonding,” J. Appl. Phys. 99(3), 031101 (2006)
103. G. Roelkens, L. Liu, D. Liang, R. Jones, A. W. Fang, B. R. Koch, and J. E. Bowers, “III-V/silicon photonics for on-chip and intra-chip optical interconnects,” Laser Photonics Rev. 4(6), 751–779 (2010)
104. P. Garrou, R. Heistand, M. Dibbs, T. Mainal, C. Mohler, T. Stokich, P. Townsend, G. Adema, M. Berry, and I. Turlik,” Rapid Thermal Curing of BCB Dielectric” IEEE Trans. Compon. Hybrids Manuf. Technol. 16, 46 (1993)
105. M. Koechlin, F. Sulser, Z. Sitar, G. Poberaj, and P. Günter, “Free-standing lithium niobate microring resonators for hybrid integrated optics,” IEEE Photon. Technol. Lett. 22(4), 251–253 (2010)
106. F. Niklaus, R. Kumar, J.J. McMahon, J. Yu, J. -Q. Lu, T. S. Cale, and R. J. Gutmann, “Adhesive wafer bonding using partially cured benzocyclobutene for three-dimensional integration,” J. Electrochem. Soc. 153(4), G291−G295 (2006)
107. H. Togo, N. Shimizu, and T. Nagatsuma, “Near-field mapping system using fiber- based electro-optic probe for specific absorption rate measurement,” IEICE Trans. Electron. E90-C, 436-442 (2007)
108. R. C. J. Hsu, A. Ayazi, B. Houshmand, and B. Jalali, “All-dielectric photonic- assisted radio front-end technology,” Nature Photon. 1, 535-538 (2007)
109. V. M. N. Passaro, F. Dell'Olio, and F. De Leonardis, “Electromagnetic field photonic sensors,” Prog. Quant. Electron. 30, 45-73 (2006)
110. K. Yang, G. David, J.-G. Yook, I. Papapolymerou, L. P. B. Katehi, and J. F. Whitaker, “Electrooptic mapping and finite-element modeling of the near-field pattern of a microstrip patch antenna,” IEEE Trans. Microw. Theory Tech. 48, 288- 294 (2000)
160
111. V. S. Ilchenko, A. A. Savchenkov, A. B. Matsko, and L. Maleki, “Sub-microwatt photonic microwave receiver,” IEEE Photon. Technol. Lett. 14, 1602-1604 (2002)
112. H. Sun, A. Pyajt, J. Luo, Z. Shi, S. Hau, A. K.-Y. Jen, L. R. Dalton, and A. Chen, “All-dielectric electrooptic sensor based on a polymer microresonator coupled side- polished optical fiber,” IEEE Sensors J. 7, 515–524 (2007)
113. P. Sun and R. M. Reano, “Low-power optical bistability in a free-standing silicon ring resonator,” Opt. Lett. 35, 1124-1126 (2010).
114. S. V. Polstyanko and J. F. Lee, “H (curl) tangential vector finite element method for modeling anisotropic optical fibers,” IEEE J. Lightwave Technol., vol. 13, no. 11, pp. 2290–2295, Nov. (1995)
115. G. Peng and J. F. Lee, “Analysis of biaxially anisotropic waveguides using tangential vector finite elements,” Microw. Opt. Tech. Lett., 9(3), 156–162, (1995)
116. F. L. Teixeira, “Time-domain finite-difference and finite-element methods for Maxwell equations in complex media,” IEEE Trans. Antennas Propagat., 56(8), 2150-2166, (2008)
117. F. L. Teixeira, "A summary review on 25 years of progress and future challenges in FDTD and FETD techniques," ACES Journal, 25(1), 1-14, (2010)
118. J. L. Volakis, A. Chatterjee, and L. C. Kempel, Finite Element Method for Electromagnetics: Antennas, Microwave Circuits, and Scattering Applications. New York: IEEE Press, (1998)
119. P. Monk, Finite Element Methods for Maxwell’s Equations. Oxford, U.K.: Oxford Univ. Press, (2003)
120. M. S. Dahlem, C. W. Holzwarth, A. Khilo, F. X. Kärtner, H. I. Smith, and E. P. Ippen, “Reconfigurable multichannel second-order silicon microring-resonator filterbanks for on-chip WDM systems,” Opt. Express 19(1), 306–316 (2011)
121. H. L. R. Lira, S. Manipatruni, and M. Lipson, “Broadband hitless silicon electro- optic switch for on-chip optical networks,” Opt. Express 17(25), 22271–22280 (2009)
122. L. Chen and R. M. Reano, “Compact electric field sensors based on indirect bonding of lithium niobate to silicon microrings,” Opt. Express 20(4), 4032-4038 (2012).
161
123. X. Wang, C.-Y. Lin, S. Chakravarty, J. Luo, A. K.-Y. Jen, and R. T. Chen, “Effective in-device r33 of 735 pm/V on electro-optic polymer infiltrated silicon photonic crystal slot waveguides,” Opt. Lett. 36(6), 882–884 (2011).
124. M. Hochberg, T. Baehr-Jones, G. Wang, J. Huang, P. Sullivan, L. Dalton, and A. Scherer, “Towards a millivolt optical modulator with nano-slot waveguides,” Opt. Express 15(13), 8401–8410 (2007).
125. M. F. Ceiler, Jr., P. A. Kohl, and S. A. Bidstrup, “Plasma-enhanced chemical vapor deposition of silicon dioxide deposited at low temperatures,” J. Electrochem. Soc., 142(6), 2067-2071 (1995).
126. E. D. Palik, ed., Handbook of Optical Constants of Solids, vol. 1 (Academic Press, 1998).
127. T. A. Ramadan, M. Levy, and R. M. Osgood, Jr., “Electro-optic modulation in crystal-ion-sliced z-cut LiNbO3 thin films,” Appl. Phys. Lett. 76(11), 1407–1409 (2000).
128. P. Dong, R. Shafiiha, S. Liao, H. Liang, N.-N. Feng, D. Feng, G. Li, X. Zheng, A. V. Krishnamoorthy, and M. Asghari, “Wavelength-tunable silicon microring modulator,” Opt. Express 18(11), 10941–10946 (2010).
129. G. Li, X. Zheng, J. Yao, H. Thacker, I. Shubin, Y. Luo, K. Raj, J. E. Cunningham, and A. V. Krishnamoorthy, “25Gb/s 1V-driving CMOS ring modulator with integrated thermal tuning,” Opt. Express 19(21), 20435–20443 (2011).
130. L. Chen, M. G. Wood, and R. M. Reano, “Compensating thermal drift of hybrid silicon and lithium niobate ring,” Opt. Lett. 40(7), 1599-1602 (2015)
131. M. F. Hossain, H. P. Chan, and M. A. Uddin, “Study of optical anisotropies in benzocyclobutene thin films for the efficient design of optical waveguide devices,” Opt. Express18(9), 8896–8905 (2010)
132. D. Dai, Y. Tang, and J. E. Bowers, “Mode conversion in tapered submicron silicon ridge optical waveguides,” Opt. Express 20(12), 13425–13439 (2012).
133. S. Kim, V. Gopalan, and A. Gruverman, “Coercive fields in ferroelectrics: A case study in lithium niobate and lithium tantalate,” Appl. Phys. Lett. 80(15), 2740-2742 (2002).
162
134. M. Luennemann, U. Hartwig, G. Panotopoulos, and K.Buse, “Electrooptic Properties of Lithium Niobate Crystals for Extremly High External Electric Fields,” Appl. Phys. B 76, 403 (2003).
135. S. S. Djordjevic, L. W. Luo, S. Ibrahim, N. K. Fontaine, C.B. Poitras, B. Guan, L. Zhou, K. Okamoto, Z. Ding, M. Lipson, and S. J. B. Yoo, “Fully reconfigurable silicon photonic lattice filters with four cascaded unit cells,” IEEE Photon. Tech. Lett., 23(1), 42-44 (2011).
136. L. Chen, M. G. Wood, and R. M. Reano, “12.5 pm/V hybrid silicon and lithium niobate optical microring resonator with integrated electrodes, Opt. Express 21, 27003-27010 (2013)
137. J. T. Robinson, K. Preston, O. Painter, and M. Lipson, “First-principle derivation of gain in high-index-contrast waveguides,” Opt. Express 16, 16659–16669 (2008)
138. R. Ding, T. Baehr-Jones, Y. Liu, R. Bojko, J. Witzens, S. Huang, J. Luo, S. Benight, P. Sullivan, J.- M. Fedeli, M. Fournier, L. Dalton, A. Jen, and M. Hochberg, “Demonstration of a low VπL modulator with GHz bandwidth based on electro-optic polymer-clad silicon slot waveguides,” Opt. Express 18, 15618–15623 (2010)
139. J. L. Nightingale, R. A. Becker, R. C. Willis, and J. S. Vrhel, “Characterization of frequency dispersion in Ti-diffused lithium niobate optical devices,” Appl. Phys. Lett. 51, 716-718 (1987)
140. R. L. Jungerman and C. A. Flory, “Low-frequency acoustic anomalies in lithium niobate Mach-Zehnder interferometers,” Appl. Phys. Lett. 53, 1477–1479 (1988)
141. W. K. Burns, M. M. Howerton, R. P. Moeller, R. Krähenbühl, R. W. McElhanon, and A. S. Greenblatt, “Low drive voltage, broad-band LiNbO3 modulators with and without etched ridges,” J. Lightwave Technol., 17, 2551–2555, (1999)
142. C. D. Watson, M. Poirier, J. M. Heaton, M. Lewis, and M. Boudreau, “Acoustooptic resonance in deep-etched GaAs-AlGaAs Electrooptic modulators,” J. Lightwave Technol. 22(6), 1598-1603 (2004)
143. M. Dubois, “Thin film bulk acoustic wave resonators: a technology overview,” in MEMSWAVE 03, 1-4 (2003)
144. J. P. Berenger, “A perfectly matched layer for the absorption of electromagnetic waves,” J. Comput. Phys., 114(2), 185–200, (1994)
163
145. Z. S. Sacks, D. M. Kingsland, R. Lee, and J. F. Lee, “A perfectly matched anisotropic absorber for use as an absorbing boundary condition,” IEEE Trans. Antennas Propag., 43, 1460–1463, (1995)
146. F. L. Teixeira and W. C. Chew, “Complex space approach to perfectly matched layers: A review and some new developments,” Int. J. Numer. Model., 13, 441–455, (2000)
147. M. Yoshida, S. Okamoto, T. Omiya, K. Kasai, and M. Nakazawa, “256 QAM digital coherent optical transmission using Raman amplifiers,” IEICE Trans. Commun., E94-B, 417–424, (2011)
148. Z. Ren, P. J. Heard, J. M. Marshall, P. A. Thomas, and S. Yu, “Etching characteristics of LiNbO3 in reactive ion etching and inductively coupled plasma,” J. Appl. Phys. 103, 034109 (2008)
149. K. Banaszek, A. B. U'Ren, and I. A. Walmsley, "Generation of correlated photons in controlled spatial modes by down conversion in nonlinear waveguides," Opt. Lett. 26, 1367-1369 (2001)
150. L. Chen, J. Chen, J. Nagy, and R. M. Reano, “Highly linear ring modulator from hybrid silicon and lithium niobate,” Opt. Express 23(9), 2015 (to appear)
151. A. Karim and J. Devenport, “Noise figure reduction in externally modulated analog fiber- optic links,” IEEE Photon. Technol. Lett. 19, 312–314 (2007)
152. M. Song, L. Zhang, R. G. Beausoleil, and A. E. Willner, “Nonlinear distortion in a silicon microring-based electro-optic modulator for analog optical links,” IEEE J. Sel. Top. Quantum Electron. 16(1), 185–191 (2010)
153. A. Yariv, “Universal relations for coupling of optical power between microresonators and dielectric waveguides,” Electron. Lett. 36(4), 321-322 (2000).
154. H. Tazawa and W. H. Steier, “Linearity of ring resonator-based electrooptic polymer modulator,” Electron. Lett. 41(23), 1297-1298 (2005)
155. L. Chen, Q. Xu, M. G. Wood, and R. M. Reano, "Hybrid silicon and lithium niobate electro-optical ring modulator," Optica 1, 112-118 (2014).
156. P. Rabiei, W. H. Steier, C. Zhang, and L. R. Dalton, “Polymer Micro-Ring Filters and Modulators,” IEEE J. Lightw. Technol. 20(11), 1968-1975 (2002).
164
157. A. Vorobiev, J. Berge, S. Gevorgian, M. Löffler, and E. Olsson, “Effect of interface roughness on acoustic loss in tunable thin film bulk acoustic wave resonators,” J. Appl. Phys. 110, 024116 (2011).
165
APPENDIX A: Fabrication Processes for Hybrid Silicon and LiNbO3
Microring Modulator
This appendix describes the detailed fabrication processes of the hybrid silicon and
LiNbO3 microring modulator presented in Chapter 5. The fabrication was mostly conducted at the Nanotech West Laboratory of the Ohio State University. Ion implantation was conducted by an industrial vendor.
A.1 Silicon Waveguide and Slab
1. Starting with 4” silicon-on-insulator (SOI) wafer (<100> orientation) with 340 nm
silicon layer and 1-µm -thick buried oxide layer. Thin silicon thickness to 250 nm
through wet oxidation process at 900 °C for 82 min.
2. Wet etch the SOI wafer in 7:1 buffered hydrofluoric acid (BHF) for 4 min 30 sec;
Rinse: DI H2O for 10 minutes; Blow-dry the substrate with nitrogen gun.
3. Dehydrate: 200 °C for 10 minutes on a hotplate; Wait for substrate to cool.
4. Spin coat Dow Corning XR-1541 (4% hydrogen silsesquioxane (HSQ) in Methyl
Isobutyl Ketone (MIBK) with a spin coater using a three-step procedure: 750 rpm,
350 rpm/s, 2 sec; 1850 rpm, 500 rpm/s, 3 sec; 3000 rpm, 2000 rpm/s, 45 sec.
5. Bake: 50 °C for 40 minutes in OVN1
166
6. E-beam lithography in a Leica EBPG-5000 to pattern silicon waveguides at 100 kV,
beam current = 3 nA, resolution = 5 nm, aperture = 200 µm, beam step size = 5 nm,
2 dose 5000 µC/cm . Silicon markers are also patterned for layer-to-layer alignment.
7. Develop the sample in 25% Tetra-Methyl Ammonium Hydroxide (TMAH) for 1
minute at 25 °C. Rinse in DI water for 1 min and blow-dry with nitrogen gun.
8. Rapid thermal annealing (RTA) at 1000 °C for 1 minutes with O2 ambient.
9. Etch silicon waveguide using inductively coupled plasma reactive ion etch (ICP-RIE)
in a PlasmaTherm SLR-770 etcher using the recipe: Cl2 = 49 sccm, O2 = 2 sccm,
pressure = 6 mTorr, RF power = 205 W, ICP power = 390 W, backside He cooling 5
sccm, DC Bias ~ 470 V, etching time 55 sec. Etching depth is 205 nm.
10. Spin-coat and bake HSQ as in step 4 and 5.
11. E-beam lithography to pattern the slab at 100 kV, beam current = 15 nA, resolution =
2 5 nm, aperture = 200 µm, beam step size = 10 nm, dose = 4000 µC/cm . Develop the
HSQ as in step 7.
12. Etch silicon slab with ICP-RIE using the recipe in step 9. Etching time is 15s.
13. Bake on a hotplate at 350 oC for 5 min
14. Wet etch in 10:1 HF for 20s to remove the HSQ mask for the slab patterning. The
first HSQ mask over the waveguide is not removed.
15. Spin coat adhesion promoter (Dow Chemical AP3000) and diluted BCB (Dow
Chemical CYCLOTENE 3022-35) mixed with mesitylene (Dow Chemical T1100) to
form a mask for wet-etching the first HSQ mask. AP3000, 4000 rpm, 30 sec; BCB,
3000 rpm, 30 sec. Bake at 115 oC for 1 min. The BCB thickness is 180 nm.
167
16. Etch back BCB with ICP-RIE to expose the HSQ mask over silicon waveguide using
the recipe: O2 = 40 sccm, CF4 = 5 sccm, pressure = 30 mTorr, DC voltage = 50V, ICP
power = 250 W, etching time = 101s. The etching depth is 41 nm.
17. Bake in the oven filled with nitrogen ambient at 275 oC to cure BCB.
18. Wet etch in 7:1 BHF for 30s to remove the HSQ mask over the waveguide.
19. Remove BCB using Piranha solution (75 ml H2SO4 and 25 mL H2O2) for 1 min.
Rinse in DI water for 1 min.
20. Ash in O2 plasma for 3 min to remove organic residues that might still exist on the
slab.
A.2 Contact and Bottom Electrode
1. Deposit 20 nm plasma-enhanced chemical vapor deposition (PECVD) silicon dioxide
to protect the sample from contamination using the recipe: pressure = 900 mTorr,
SiH4 = 100 sccm, N2O = 300 sccm, temperature = 250 °C, deposition time 55 sec.
+ 13 2 2. Blanket implant the sample using BF2 ions at 45 keV with a dose of 1.8×10 /cm .
The ion implantation is conducted by Leonard Kroko, Inc. in CA.
3. RTA at 1250 °C for 1min to activate and diffuse the ions in Ar ambient. a. Dispense P(MMA-MAA 8.5%) EL11 copolymer using a syringe, and a 0.2 µm
PTFE/nylon filter. Spin coat at 2000 rpm for 90 sec.
4. Bake on a hotplate at 200 °C for 5 min.
168
5. E-beam lithography to pattern the ion implantation windows for contacts at 50 kV,
beam current = 75 nA, resolution=10 nm and beam step size = 50 nm, aperture = 200
µm, dose = 300 µC/cm2.
6. Develop in 3:1 IPA: MIBK developer for 3 minutes. Ash in O2 plasma for 30 sec.
+ 15 2 7. Implant BF2 ions at 45 keV with a dose of 3×10 /cm .
8. Remove the PMMA mask using N-Methyl-2-pyrrolidone (NMP) solution on a 150 oC
hotplate for 15 min. Ash in O2 plasma for 10 min.
9. RTA at 1000 °C for 12 sec for dopant activation and repairing damaged silicon.
10. Wet etch in 10:1 HF for 20 sec to etch the 20 nm PECVD silicon dioxide.
11. Spin coat P(MMA-MAA 8.5%) EL11 copolymer at 2000 rpm for 90s. Bake at 200 oC
for 5 min. Spin coat PMMA (950K PMMA A4) at 1500 rpm for 90s. Bake at 200 oC
for 5 min.
12. E-beam lithography to pattern the nickel silicidation windows at 50 kV, beam current
= 75 nA, resolution=10 nm and beam step size = 50 nm, aperture = 200 µm, dose =
2 800 µC/cm . Develop the resist as in step 6. Ash in O2 plasma for 3 min to remove
residue.
13. Wet etch in 10:1 BHF for 25 sec and rinse in DI water for1 min; Blow dry.
14. Clean the silicon in the contact window with ICP-RIE using the recipe: Ar = 50
sccm, RF1 power = 100W, ICP power = 150 W, pressure = 10 mTorr, etching time =
1min.
15. Evaporate 25 nm nickel using a Denton 502A evaporator.
16. Remove PMMA as in step 8. Ash in O2 plasma for 5 min.
169
17. RTA at 550 °C for 1 min in nitrogen ambient to form nickel silicide.
18. Wet etch in Piranha solution to remove unreacted nickel. Etch in 4:1 H2SO4: H2O2
solution for 10 min and rinse in DI water for 1 min.
19. Spin coat and bake PMMA and MMA copolymer as in step 11.
20. E-beam lithography to pattern the bottom electrode at 50 kV, beam current = 75 nA,
resolution=10 nm and beam step size = 50 nm, aperture = 200 µm, dose = 850
2 µC/cm . Develop as in step 6. Ash in O2 plasma for 90 sec.
21. ICP-RIE Ar milling for 30s using the recipe: Ar = 50 sccm, RF1 power = 100W, ICP
power = 150 W, pressure = 10 mTorr, etching time = 30 sec
22. Evaporate 100 nm aluminum using a Denton 502A evaporator.
23. Lift-off the aluminum electrode as in step 8.
A.3 LiNbO3 Thin Film
1. Clean implanted z-cut LiNbO3 sample in acetone with ultrasonic for 10 min, in IPA
with ultrasonic for 5 min, and in DI water for 10 min. Blow dry.
o 2. RTA the LiNbO3 sample at 300 C for 8 sec in 5% H2 and 95% N2 ambient.
3. Flip the LiNbO3 sample on an unpolished silicon substrate. Heat the samples on a
hotplate from room temperature to 300o C in 5 min, and hold at 300 oC for 5 min
before cooling down to room temperature. LiNbO3 thin films are blistered and
exfoliated from the LiNbO3 substrate and transferred to the silicon substrate.
o 4. RTA the LiNbO3 thin films at 1000 C for 30 sec in Ar ambient.
170
A.4 Bonding Process
1. Spin coat the SOI device sample with AP 3000 and BCB: T1100. Spin coat AP3000
at 4000 rpm for 30 sec. Spin coat BCB: T1100 at 4200 rpm for 30 sec.
2. Bake on a hotplate at 70 oC for 5 min.
3. Partially cure the BCB thin film in an oven filled with nitrogen. Ramp from 30 oC to
150o C in 20 min, and hold for 1 hour. Ramp to 190 oC in 10 min, and hold for 15
min. Ramp down to room temperature. The BCB thickness is around 350 nm.
4. Etch back BCB with ICP-RIE using the recipe: CF4 = 5 sccm, O2 = 40 sccm, pressure
= 30 mTorr, DC bias = 50 V, ICP power = 250W. Pause 1 min after every 30 sec
etch. Etch time rate is 46 nm/min. Adjust the etching time to minimize the BCB
thickness on top of the waveguide.
5. Pick up, transfer, and bond LiNbO3 thin films to SOI waveguides using a micro-
vacuum tip on a probe station.
6. Anneal the hybrid silicon and LiNbO3 devices in an oven filled with nitrogen at 250
oC for 1 hour to cure BCB.
A.5 Via and Top Electrode
1. Etch BCB not covered by LN with ICP-RIE using the recipe:
a. CF4 = 5 sccm, O2 = 40 sccm, pressure = 30 mTorr, DC bias = 50 V, ICP power =
250W, etching time = 5 min
b. O2 = 20 sccm, pressure = 10mT, RF power = 100W, ICP power = 200W, etching
time = 1min 30 sec
171
2. Deposit 1000 nm PECVD SiO2 as the optical cladding layer for the waveguides.
3. Spin coat P(MMA-MAA 8.5%) EL11 copolymer at 2000 rpm for 90s. Bake at 200 oC
for 5 min. Spin coat PMMA (950K PMMA A4) at 1500 rpm for 90s. Bake at 200 oC
for 5 min.
4. E-beam lithography to pattern the contact via at 50 kV, beam current = 70 nA,
resolution=10 nm and beam step size = 50 nm, aperture = 200 µm, dose = 800
µC/cm2.
5. Develop in 3:1 IPA: MIBK developer for 3 minutes. Ash in O2 plasma for 3 min.
6. Etch PECVD SiO2 in the via with ICP-RIE using the recipe:
a. CHF3 = 30 sccm, pressure = 35mT, DC bias = 400 V, ICP power = 300W, etch 3
min and pause for 1 min.
b. Repeat step a for a total active etching time of 30 min
7. Remove the PMMA mask using N-Methyl-2-pyrrolidone (NMP) solution on a 150 oC
hotplate for 15 min. Ash in O2 plasma for 5 min.
8. Spin coat the SOI device sample with AP 3000 and BCB: T1100 to form a slope over
the LiNbO3 thin film edges. Spin coat AP3000 at 4000 rpm for 30 sec. Spin coat
BCB: T1100 at 4000 rpm for 30 sec. Bake on a hotplate at 115 oC for 1 min. The
BCB thickness is 170 nm.
9. Etch back BCB with ICP-RIE as in step 1a.
10. Anneal the sample in an oven filled with nitrogen at 230 oC for 4 hours.
11. Spin coat and bake MMA and PMMA as in step 3.
172
12. E-beam lithography to pattern the top electrode at 50 kV, beam current = 70 nA,
resolution = 10 nm and beam step size = 50 nm, aperture = 200 µm, dose = 850
µC/cm2. Develop as in step 5.
13. Remove oxide on Al bottom electrode with ICP-RIE Ar milling using the recipe: Ar
= 50 sccm, pressure = 15 mT, RF power = 100W, ICP power = 150 W, etching time =
45 sec.
14. Evaporate 300 nm aluminum using a Denton 502A evaporator.
15. Lift-off the aluminum electrode as in step 7.
A.6 Cantilever Coupler
1. Spin coat the sample with AP 3000 and BCB: T1100 to protect the Al electrode. Spin
coat AP3000 at 4000 rpm for 30 sec. Spin coat BCB: T1100 at 4000 rpm for 30 sec.
Bake on a hotplate at 115 oC for 1 min. The BCB thickness is 170 nm.
2. Anneal the sample in an oven filled with nitrogen at 230 oC for 4 hours.
3. Deposit 150 nm of PECVD SiO2.
4. Evaporate 200nm chromium mask for cantilever patterning
5. Deposit 300 nm PECVD SiO2.
6. Spin coat P(MMA-MAA 8.5%) EL11 copolymer using spin coater (COT3) at 2000
rpm for 90s and bake on a hotplate at 200 oC for 5 min.
7. E-beam lithography to pattern the cantilever couplers at 50 kV, beam current = 10
nA, resolution=10 nm and beam step size = 50 nm, aperture = 200 µm, dose = 400
2 µC/cm . Develop in 3:1 IPA: MIBK developer for 3 minutes. Ash 60 sec.
173
8. Etch the PECVD SiO2 mask with ICP-RIE using the recipe: CHF3 = 30 sccm,
pressure = 35mT, DC bias = 400 V, ICP power = 300W, etch 3 min 30 sec and pause
1 min, repeat for a total active etching time of 10 min 30 sec.
9. Remove the PMMA mask using N-Methyl-2-pyrrolidone (NMP) solution on a 150 oC
hotplate for 15 min. Ash in O2 plasma for 5 min.
10. Etch chromium mask with ICP-RIE using the recipe:
a. Ar = 50 sccm, pressure = 10 mT, RF power = 100W, ICP power =150W, etching
time = 2 min.
b. Cl2 = 20 sccm, O2 = 5 sccm, pressure = 5 mT, RF power = 150W, ICP power =
500W, etching time = 4 min.
11. Form the cantilever couplers using SF6 plasma with a LAM 490 etcher
a. SF6 =100 sccm, He = 30 sccm, pressure = 375 mTorr, electrode gap=1.35 cm, RF
power = 500 W. Etch 3 min and pause for 3 min.
b. Repeat step a for a total active etching time of 36 min.
12. Remove the chromium mask. Wet etch in CR-7s etcher for 15 min.
174