Design, Fabrication, and Characterization of Passive ...
Transcript of Design, Fabrication, and Characterization of Passive ...
Design, Fabrication, and Characterization
of Passive Silicon Photonic Device
by
Yang Wang
Submitted in total fulfillment of the
Requirements for the degree of
Doctor of Philosophy
Department of Electrical and Electronic Engineering
The University of Melbourne
September 2017
II
Abstract
Over the past decade, the silicon photonics technology has attracted extensive research
attention due to its great potential to leverage with the existing microelectronics facility
to produce large-scale and high-density photonic integrated circuits (LHPICs). Silicon-
based LHPICs have the potential of realizing low-cost and power-efficient tera-scale
communications for future optical networks, data-centres, supercomputers, and various
consumer electronic and photonic applications. As the fundamental building blocks of
the LHPICs, various passive optical components have been built on the silicon
photonics integration platform, and they are used to realize the basic functions of the
circuit such as light guidance, power splitting, optical reflection, and wavelength
filtering, or are used as the components to construct more complex devices such as
lasers, modulators and photodetectors. Therefore, the passive silicon photonic devices
play an important role in building high-performance and densely integrated PICs.
This thesis focuses on the design, fabrication, and characterization of passive silicon
photonic devices. Specifically, we present two novel passive silicon photonic devices
based on the popular 220 nm SOI platform, namely an ultra-broadband and low-loss
silicon adiabatic optical power splitter and an ultra-broadband and high-reflectivity
silicon circular Bragg grating mirror. The proposed adiabatic splitter can achieve a low
excess loss (<0.19 dB) and a broad operation band (500 nm) based on an adiabatic
length of only 5 µm, and more importantly, it is polarization insensitive. To the best of
our knowledge, the proposed splitter in this thesis is the shortest and the most
broadband polarization-insensitive adiabatic splitter that has been presented on the SOI
platform until now. For the proposed circular Bragg grating mirror, it solves the
challenge of achieving high reflectivity and broad reflection bandwidth simultaneously
for Bragg gratings based on the thin silicon layer SOI platform, and it has a very
compact size of only 4.49 µm × 4.54 µm, which makes it suitable for building densely
integrated devices, such as on-chip resonators, filters, and laser cavities.
Compared with existing devices, the proposed adiabatic splitter and the circular Bragg
grating mirror present significant advantages in terms of device performance and size.
Therefore, they are expected to be promising elements in LHPICs and applications
which require ultrahigh-efficiency and broadband optical power distributions and
III
reflections. The proposed silicon photonic devices are fabricated based on e-beam
lithography and deep reactive ion etching (DRIE) systems, and the development of the
fabrication process based on the 220 nm SOI wafer will also be demonstrated in this
thesis as a technical chapter.
IV
Declaration
This is to certify that
(i) the thesis comprises only my original work towards the PhD,
(ii) due acknowledgement has been made in the text to all other material used,
(iii) the thesis is less than 100,000 words in length, exclusive of tables, maps,
bibliographies and appendices.
Signature________________________
Date____________________________
V
Acknowledgement
Time flies. Now it comes to the end of my PhD research. Looking back at the past four
years, I feel lucky that I have entered a field which I have been passionately interested
in. This is definitely a field that I will be glad to be keen on throughout my entire future
career. Without doubt, the knowledge, experience, and skills that I have gained through
these four years of study will become lifetime benefits for me. This would not be
possible to happen without the help of so many people.
First of all, I would like to express my sincere gratitude to my principle supervisor,
Professor Stan Skafidas, for introducing me into this stunning world of silicon photonics,
and for his continuous support, encouragement, and advices throughout the whole stage
of my PhD research. Prof. Stan’s enthusiasm, inspiration, and rigorous attitude towards
research and work have also set an example for me.
I would also like to express my deep gratitude to my co-supervisor, Dr. Ke Wang. As a
supervisor and a friend, Dr. Ke has provided enormous help and support to me. Without
those valuable suggestions, discussions, guidance and support from him, I would not be
able to achieve those fruitful outcomes for my PhD research.
I would like to show thankfulness to Dr. Thomas Chae, for his guidance and supervision
in the early years of my PhD research. Dr. Chae has helped me knock the door of the
world of silicon photonics and build a good foundation.
I would also like to thank Prof. Thas Nirmalathas for serving as my advisory committee
member.
I would like to thank Australian Research Council (ARC) for the funding support of my
chip fabrication, and I would like to thank National ICT Australia laboratory (NICTA)
for supporting all my academic conference travels during the past four years.
I would also like to show my appreciation to the staff of Melbourne Centre for
Nanofabrication (MCN), for their training in nanofabrication facilities and the valuable
support in my chip fabrication.
Finally, I would like to show great thanks to my family, for their understanding, support,
encouragement, and love.
VI
Acronyms
APD avalanche photodetector
ASE amplified spontaneous emission
BOX buried-oxide
BPM beam propagation method
CMOS Complementary Metal Oxide Semiconductor
Cr chrome
C4F8 Octafluorocyclobutane
CVD chemical vapor deposition
DC directional coupler
DVS-BCB divinylsiloxane-benzocyclobutane
EBL e-beam lithography
EC external-cavity
EIC electronic integrated circuit
EL excess loss
Er erbium
ER extinction ratio
FDTD finite difference time domain
FEM finite element method
FMM film mode matching method
FPR free propagation region
GaSb gallium-antimonide
Ge germanium
HBG high-index-contrast Bragg grating
ICP Inductively Coupled Plasma
ICPRIE inductively coupled plasma reactive ion etching
IME Institute of Microelectronics
IMEC Interuniversity MicroElectronics Center
IL insertion loss
VII
InP indium-phosphide
IPA isopropyl alcohol
LBG low-index-contrast Bragg grating
LETI Laboratory of Electronics and Information Technologies
LPCVD low pressure chemical vapor deposition
LSF lensed single-mode fibre
LSPIC large-scale photonic integrated circuit
MMI multi-mode interference
MOS Metal Oxide Semiconductor
MZI Mach-Zehnder Interferometer
OEIC optoelectronic integrated circuit
OI optical interconnect
OSA optical spectrum analyser
PBS polarization beam splitter
PEC proximity effect correction
PECVD plasma enhanced chemical vapor deposition
PIC photonic integrated circuit
PMMA polymethyl methacrylate
PR polarization rotator
PSF Point Spread Function
Q quality factor
RF radio frequency
RIE reactive ion etching
SEM scanning electron microscope
SF6 Sulfur hexafluoride
Si silicon
SiGe silicon-germanium
SiN silicon nitride
SiO2 silicon dioxide
SOA semiconductor optical amplifier
SOI silicon-on-insulator
VIII
SRS Stimulated Raman Scattering
SRO silicon-rich-oxide
SSC spot size converter
TE transverse-electric
TIR total internal reflection
TM transverse-magnetic
VCSEL vertical-cavity surface-emitting laser
WDM wavelength division multiplexing
IX
Contents
Chapter 1 Silicon Photonics: Introduction and Review ...................... 1
1.1 Silicon Photonics Integration and Applications ................................. 1
1.2 Components ................................................................................... 5
1.2.1 Silicon Optical Modulator ......................................................................................... 6
1.2.2 Si-integrated Laser ....................................................................................................12
1.2.3 Ge-on-Si Photodetector ...........................................................................................16
1.2.4 Passive Silicon Photonic Device ...........................................................................18
1.3 Thesis Objective and Organisation ................................................. 21
Chapter 2 Passive Silicon Photonic Device Fabrication .................... 24
2.1 Single-Mode Silicon Waveguide ..................................................... 24
2.2 Fabrication and Measurement ....................................................... 30
2.2.1 Fabrication Process Based on 220 nm SOI ........................................................30
2.2.1.1 E-beam lithography .................................................................................. 31
2.2.1.2 PEC and dose test .................................................................................... 34
2.2.1.3 Silicon nano-etching ................................................................................. 35
2.2.1.4 Edge coupling ............................................................................................ 38
2.2.2 Silicon Photonic Chip Measurement Setup .......................................................42
Chapter 3 Silicon Adiabatic Optical Power Splitter ......................... 44
3.1 3-dB Silicon Optical Power Splitter ................................................. 44
3.1.1 2×2 Silicon Adiabatic Coupler ..............................................................................45
3.1.2 1×2 Silicon Adiabatic Splitter................................................................................46
3.1.3 1×2 Silicon Non-adiabatic Splitter .......................................................................47
3.2 1×2 Optical Power Splitter Based on Adiabatic Silicon Tapers ........ 48
3.2.1 Device Design ............................................................................................................48
3.2.2 Characterization .........................................................................................................55
3.2.3 Splitter with Various Power Splitting Ratio ......................................................60
3.2.4 Conclusion ...................................................................................................................61
Chapter 4 Circular Bragg Grating Mirror ...................................... 62
4.1 High-index-contrast Bragg Grating on SOI ....................................... 62
4.2 Circular Bragg Grating Based on 220 nm SOI ................................. 63
4.2.1 Device Design ............................................................................................................65
X
4.2.2 Characterization .........................................................................................................71
4.2.3 Integrated Notch Filter Based on Circular Bragg Grating Mirror ...............74
4.2.4 Solutions for Broadband and High-efficiency TM-reflection ......................76
4.2.4.1 Circular Bragg grating with thickened grating blades.......................... 77
4.2.4.2 Circular Bragg grating based on horizontal slot waveguide ............... 81
4.2.5 Extended Applications of Slot Bragg Grating ..................................................85
4.2.5.1 Polarization-selective Bragg grating mirror ........................................... 85
4.2.5.2 TM-pass/TE-block polarizer .................................................................... 89
4.2.6 Conclusion ...................................................................................................................94
Chapter 5 Conclusions and Future Work ........................................ 96
5.1 Conclusions .................................................................................. 96
5.2 Future work ................................................................................... 97
Chapter 6 References ..................................................................... 99
XI
List of Figures
Figure 1.1. The prototype of a silicon photonic integrated four-channel WDM
transceiver. ....................................................................................................................... 2
Figure 1.2. The block diagram of an on-chip OI system. ................................................. 6
Figure 1.3. Diagrams of three different structures employed in plasma-dispersion-
effect-based silicon modulators to implement the mechanisms of (a) carrier injection, (b)
carrier depletion, and (c) carrier accumulation. ............................................................. 7
Figure 1.4. The schematic diagrams of (a) an MZI modulator and (b) a microring-based
modulator [31]. ................................................................................................................ 9
Figure 1.5. The schematic of SRO embedded in a MOS structure for electrical pumping.
........................................................................................................................................ 14
Figure 1.6. Proposed wavelength-switchable laser. (a) conceptual design, (b) detailed
implementation, (c) SEM image of a broadband reflector [67]. .................................... 16
Figure 1.7. (a) Optical microscope image of the fabrication PBS. (b) SEM (scanning
electron microscope) image of the SiN waveguide (cross-section). ............................... 21
Figure 2.1. The schematic diagram of a silicon strip waveguide: (a) 3D view; (b) cross-
section view. ................................................................................................................... 25
Figure 2.2. The change of effective refractive index with waveguide width for 220-nm-
thick silicon strip waveguide with a top cladding of SiO2. The simulations were carried
out using Lumerical Mode solutions, and the refractive indices of Si and SiO2 were set
to be 3.477 and 1.444, respectively. ............................................................................... 26
Figure 2.3. (a) – (d) The mode profiles of the fundamental TE and TM mode based
different waveguide widths. ............................................................................................ 27
Figure 2.4. The silicon photonic fabrication process based on 220 nm SOI. ................ 30
Figure 2.5. (a) The schematic of the proposed splitter structure to be fabricated. The
shadowed region in the structure indicates the place where ZEP resist exists after
development. (b, c) The scanning electron microscope (SEM) images of (b) a
successfully fabricated splitter, and (c) a damaged sample caused by the resist collapse.
........................................................................................................................................ 33
Figure 2.6. (a) The SEM image of the 200-nm line and space array exposed under the
dosage of 190 µC/cm2, the picture was captured after the step of resist development.
The grooves are the exposed area where ZEP was cleared after development. (b) The
SEM image of proposed splitter. Based on the SEM measurements, the exposed gap size
was about 10 nm larger than the designed size. Therefore a 5 nm bias was applied in
the pattern layouts to get correct size. ............................................................................ 35
Figure 2.7. Examples of fabricated waveguide arrays which show (a) trapezoid profiles,
and (b) inverse trapezoid profiles with under-cut effect................................................. 37
Figure 2.8. The SEM images of (a) the fabricated silicon strip waveguide and (b) a
bending waveguide sample with a bending radius of 10 µm.......................................... 38
Figure 2.9. Fabrication process to achieve edge coupling. ........................................... 39
Figure 2.10. Examples of (a) an isotropic Bosch etch when Fomblin oil was not used,
and (b) a non-isotropic Bosch etch with vertical sidewall when Fomblin oil was used. 41
Figure 2.11. (a) The photo of the fabricated wafer after the Bosch process step. (b, c)
The optical microscope images of the fabricated chip facet. ......................................... 42
XII
Figure 2.12. (a) The experiment setup for edge-coupling. (b) The extraction of linear
waveguide propagation loss at 1550 nm in TE polarization. ......................................... 43
Figure 3.1. The schematic diagrams of (a) a conventional directional coupler and (b) a
2×2 3-dB adiabatic coupler. .......................................................................................... 45
Figure 3.2. The principle illustration of a 1×2 3-dB adiabatic coupler based on the
effective-refractive-index tapering. ................................................................................ 47
Figure 3.3. (a) The schematic diagrams of the proposed adiabatic splitter and the
intensity field distributions simulated along the device with (b) the TE and (c) the TM
polarization input at 1550 nm wavelength. The device parameters used in the
simulations are: W (400 nm), L (5 µm), G (50 nm), T (30 nm). ..................................... 49
Figure 3.4. (a) - (d) Mode profiles (1550 nm) simulated at the cross section of taper end
with W = 460 nm. (e) Effective index of the four modes along the tapers with a length of
5 µm. (f) Simulated ∆Neff at the taper end cross section with the change of W from 340
to 500 nm. (g) Change of ∆Neff along the tapers for different W and L. (h) Simulated
excess loss at 1550 nm with the change of L. ................................................................. 52
Figure 3.5. (a) The simulated transmission spectrum (output 1 and output 2 have
identical spectrum) of the proposed splitter and (b) a 90° bending waveguide with a
cross-section size of 400 nm × 220 nm and the bending radius of 15 µm. .................... 54
Figure 3.6. The change of transmission efficiency and ∆Neff with (a) T and with (b) G at
1550 nm. ......................................................................................................................... 55
Figure 3.7. (a, b) The SEM images of the fabricated adiabatic splitter. (c) The
schematic of the device measurement setup. .................................................................. 56
Figure 3.8. The optical microscope images of (a) the fabricated 1×8 splitter and (b) the
cascaded splitter. ............................................................................................................ 57
Figure 3.9. Experimentally measured spectra of the splitter for (a) the TE and (b) the
TM polarization. ............................................................................................................. 58
Figure 3.10. The extracted loss at 1550 nm for (a) the TE polarization and (b) the TM
polarization. R square is the calculated coefficient of determination. ........................... 59
Figure 3.11. The simulated power splitting ratio as a function of gap_2 when gap_1 is
fixed at 50 nm. ................................................................................................................ 60
Figure 4.1. The schematic diagrams of the proposed circular Bragg grating with two
different structures (size scaled). .................................................................................... 64
Figure 4.2. The field distribution (1550 nm) within the taper simulated with an arc
angle of (a) 270°, (b) 180°, and (c) 60°. Note that the above field distributions were
captured to illustrate the wave diffraction from the strip waveguide to the taper, and no
grating blades were set in the simulation. The reflection observed was due to the
refractive index contrast between the taper and the cladding, rather than the Bragg
reflection. The red circles in (a) and (b) indicate the branches formed in the 270° taper
and the effective-index-contrast interface formed in the 180° taper, respectively. The
dashed line in (b) shows the 60° boundary. The tapers were all covered by 2 μm SiO2. 66
Figure 4.3. The simulated reflection spectra of the proposed Bragg grating with
different values chosen for W1, Wt, and Wb. ................................................................... 68
Figure 4.4. (a) – (d): The vertical field-intensity distribution of the grating captured at
the y = 0 plane with λ set at 1.3 and 1.75 µm and W1 chosen at 100 and 181 nm. (e, f, h):
The grating spectrum simulated with different values of W1, θ, and N. (g): The change of
the reflectivity (1550 nm) and ∆λ (reflectivity >90%) with the grating blade number N.
........................................................................................................................................ 70
XIII
Figure 4.5. The schematic of the measurement setup and the SEM images of the
fabricated grating mirror. .............................................................................................. 72
Figure 4.6. (a): The tilted view of the fabricated grating mirror. (b): The measured
spectra of the fabricated grating, the reference strip waveguide, and the chip-facet
reflections. (c): The measured IL of the fabricated grating. .......................................... 73
Figure 4.7. The top-view of the fabricated notch filter with the following parameters: Lw
= 4 μm, Lc = 0 μm, rb = 5 μm ......................................................................................... 74
Figure 4.8. (a): The tilted-view of the fabricated notch filter. (b): The spectra measured
at the output port of the notch filters fabricated with different T and Lc........................ 75
Figure 4.9. (a) The 3D and (b) the vertical cross-section schematic diagrams of the
proposed circular Bragg grating mirror with thickened blades, designed for TM-
polarized light. ................................................................................................................ 78
Figure 4.10. (a) The simulated reflection spectra of the circular Bragg grating with
thickened blades (775 nm thick) and non-thickened blades (220 nm thick) with TM-
polarized light launched. (b) The vertical field intensity distributions simulated along
the gratings with 775 nm thick blades and 220 nm thick blades. The blade number was
set to 5. ............................................................................................................................ 79
Figure 4.11. The change of the grating peak reflectivity and the bandwidth with (a) the
grating arc angle θ, (b) the blade thickness, and (c) the blade number. (d) The reflection
spectra of the circular grating with thickened blades (775 nm thick) simulated for the
fundamental TE and TM modes. ..................................................................................... 80
Figure 4.12. The schematic illustration of the slot-waveguide-based circular Bragg
grating. ........................................................................................................................... 82
Figure 4.13. (a) The reflection spectra of CSG and SCG. (b, c) The vertical field
intensity distributions captured along (b) SCG and (c) CSG. ........................................ 83
Figure 4.14. (a) The change of grating peak reflectivity with slot thickness (t). (b) The
change of grating peak reflectivity and bandwidth with grating arc angle (θ). (c) The
change of grating peak reflectivity with blade number (p) for CSG and SCG. .............. 84
Figure 4.15. The schematic of a horizontal slot waveguide. .......................................... 85
Figure 4.16. (a) The 3D schematic and (b) the cross-section schematic diagrams of the
polarization-selective Bragg grating. ............................................................................. 86
Figure 4.17. The reflection spectra of (a) the TE-dedicated mirror and (b) the TM-
dedicated mirror. Dotted lines show the grating spectrum for the different polarization.
........................................................................................................................................ 87
Figure 4.18. The reflection spectra of (a) the TE-dedicated mirror and (b) the TM-
dedicated mirror. Dotted lines show the grating spectrum for the different polarization.
........................................................................................................................................ 88
Figure 4.19. The intensity field distribution simulated along the grating with different
sets of (a) slot layer thickness and (b) silicon layer thickness. ...................................... 89
Figure 4.20. (a) The 3D schematic and the 2D (b) lateral and (c) vertical illustrations of
the proposed polarizer based on horizontal slot silicon waveguide. ............................. 91
Figure 4.21. (a)-(d) The field profiles (1550 nm) captured in the horizontal and vertical
cross-sections of the device with TE- or TM-mode; and (e) and (f) The simulated
transmission spectra of the polarizer with 20 and 40 grating periods. .......................... 93
XIV
List of Publications
1. Y. Wang, S. Gao, K. Wang, H. Li, and E. Skafidas, “Ultra-broadband, compact,
and high-reflectivity circular Bragg grating mirror based on 220 nm silicon-on-
insulator platform,” Optics Express. 25(6), 6653-6663 (2017).
2. Y. Wang, S. Gao, K. Wang, and E. Skafidas, “Ultra-broadband and low-loss 3 dB
optical power splitter based on adiabatic tapered silicon waveguides,” Optics
Letters. 41(9), 2053-2056 (2016).
3. K. Wang, Y. Wang, S. Gao, A. Nirmalathas, C. Lim, K. Alameh, H. Li, and E.
Skafidas, “Silicon Integrated Optical Isolator with Dynamic Non-Reciprocity,”
IEEE Photonics Technology Letters. 29(15), 1261-1264 (2017).
4. S. Gao, Y. Wang, K. Wang, and E. Skafidas, “High contrast circular grating
reflector on silicon-on-insulator platform,” Optics Letters. 41(3), 520-523 (2016).
5. S. Gao, Y. Wang, K. Wang, and E. Skafidas, “Low-loss and Broadband 2×2
Polarization Beam Splitter Based on Silicon Nitride Platform,” IEEE Photonics
Technology Letters. 28(18), 1936-1939 (2016).
6. K. Wang, S. Gao, Y. Wang, A. Nirmalathas, C. Lim, K. Alameh, and E. Skafidas,
“Four-Wave-Mixing-Based Silicon Integrated Optical Isolator With Dynamic
Non-Reciprocity,” IEEE Photonics Technology Letters. 28(16), 1739-1742
(2016).
7. Y. Wang, S. Gao, K. Wang, and E. Skafidas, “Broadband, High-Extinction-Ratio,
and Low-Excess-Loss Polarizer Based on Horizontal Slot Silicon Bragg Grating,”
CLEO-PR/OECC/PGC 2017. Paper ID: s1127.
8. Y. Wang, S. Gao, K. Wang, and E. Skafidas, “Ultra-broadband and Low-loss
Optical Power Splitter Based on Tapered Silicon Waveguides,” IEEE Optical
Interconnects Conference 2015, April 2015.
9. Y. Wang, S. Gao, K. Wang, and E. Skafidas, “Broadband Bragg Grating Mirror
Based on Circular and Horizontal Slot Silicon Waveguides for TM0 Mode,” Asia
Communications and Photonics Conference (ACP) 2014, Nov 2014.
10. Y. Wang, C.J. Chae, S. Gao, and E. Skafidas, “Single-polarization Reflector
Based on Circular Bragg gratings and Horizontally Slotted Silicon Waveguides,”
OptoElectronics and Communications Conference (OECC) 2014, July 2014.
11. Y. Wang, C.J. Chae, S. Gao, E. Skafidas, “Circular Bragg Grating Mirror with
High Reflectance and Wide Bandwidth for TM-Polarized Waves,” IEEE Optical
Interconnects Conference 2014, May 2014.
XV
12. S. Gao, Y. Wang, K. Wang, and E. Skafidas, “High Efficient, Compact and
Broadband 2×2 Polarization Beam Splitter on Silicon Nitride,” Conference on
Lasers and Electro-Optics (CLEO) 2016, June 2016.
13. K. Wang, Y. Wang, S. Gao, A. Nirmalathas, C. Lim, K. Alameh, and E. Skafidas,
“2×2 Silicon Integrated Optical Phased Array for Beam Steering Applications,”
International Topical Meeting on Microwave Photonics 2015, October 2015.
14. S. Gao, Y. Wang, K. Wang, H. Li, and E. Skafidas, “High-Efficiency Interlayer
Coupler on Silicon Nitride,” CLEO-PR/OECC/PGC 2017. Paper ID: s2361.
15. S. Gao, Y. Wang, K. Wang, and E. Skafidas, “Polarization Insensitive Vertical
Coupler for Multi-Layer Silicon Photonic Integrated Circuits,” IEEE Optical
Interconnects Conference 2015, April 2015.
16. K. Wang, S. Gao, Y. Wang, A. Nirmalathas, C. Lim, K. Alameh, E. Skafidas,
and H. Li, “Four-Wave-Mixing Based Silicon Integrated Optical Isolator With
Dynamic Non-Reciprocity,” CLEO-PR/OECC/PGC 2017. Paper ID: s1309.
17. K. Wang, Y. Wang, S. Gao, A. Nirmalathas, C. Lim, E. Skafidas, and K. Alameh,
“Si Integrated Optical Phased Array for Efficient Beam Steering in Optical
Wireless Communications,” Globecom 2014.
18. S. Gao, Y. Wang, K. Wang, E. Skafidas, “High Index Contrast Circular Bragg
Reflector on Silicon-on-Insulator with Flat and Broadband Spectrum”, ACP 2014,
Nov 2014.
19. S. Gao, C.J. Chae, Y. Wang, and E. Skafidas, “Deeply Etched Silicon Circular
Bragg Reflector Optimized for High Reflectance, Wide Bandwidth and Reduced
Footprint,” OECC 2014, July 2014.
20. S. Gao, C.J. Chae, Y. Wang, E. Skafidas, “Fast Wavelength-Switchable Hybrid
Laser for Energy-Efficient Optical Interconnect,” IEEE Optical Interconnects
Conference 2014, May 2014.
21. K. Wang, S. Gao, Y. Wang, T. Song, T. Liang, A. Nirmalathas, C. Lim, K.
Alameh, and E. Skafidas, “Short-range infrared optical wireless communications
– Systems and integration,” Photonics Society Summer Topical Meeting Series
(SUM), 2016 IEEE, July 2016.
Chapter 1 Silicon Photonics: Introduction and Review
1
CHAPTER 1 SILICON PHOTONICS:
INTRODUCTION AND REVIEW
1.1 Silicon Photonics Integration and Applications
Silicon (Si) integrated photonics is a breakthrough technology that has been recognized
as a promising solution to build large-scale and high-density photonic integrated circuits
(LHPICs). Such LHPICs are developed to meet the requirements of low-cost, high-
volume, energy-efficient, and ultrahigh-speed data processing and communications in
future optical networks, data centres, cloud servers, supercomputers, and various
consumer applications [1-5]. This is mainly because that silicon photonics integration is
compatible with the modern complementary metal oxide semiconductor (CMOS)
fabrication technology, which enables mass production of Si-based PICs with low cost
and high yield by using existing and high-advanced CMOS facilities. In addition, the
silicon material is also an appropriate material for photonics applications.
Firstly, there is a large refractive index contrast (∆n > 2) between the silicon waveguide
core (n = 3.47) and the surrounding cladding material, typically SiO2 (n = 1.44) or air
(n = 1), which allows silicon waveguides to be built with ultra-small dimensions in the
sub-micron level. The large index contrast also enables strong optical intensity within
the silicon waveguide, which activates non-linear effects relatively easily. Based on
these non-linear interactions, important optical functions like lasing, amplification, and
wavelength conversion can be realized [6]. Secondly, silicon has a transparent
transmission window in the 1.1 - 7 µm range, which covers the traditional near-infrared
optical communication band. Therefore, silicon based optical devices has very low
absorption loss inside this band [7]. Thirdly, as a semiconductor material, the
availability of refractive index control through either thermal effect or plasma
dispersion in silicon ensures that active functions such as optical modulation can be
realized [8]. Lastly, the high-quality silicon-on-insulator (SOI) wafer which can be
produced easily nowadays provide an ideal platform to build planar PICs.
Benefited from these exceptional advantages, silicon photonics has gained extensive
research attention over the past decade, and various fundamental optical devices such as
waveguides, filters, modulators, photodetectors, and lasers, have been successfully
Chapter 1 Silicon Photonics: Introduction and Review
2
realized on the silicon photonics integration platform [6, 9]. Nevertheless, instead of
building these discrete devices, the most important feature of silicon photonics is the
integration of multiple devices onto a single chip so as to build chip-scale optical
systems with compact size and low cost [10, 11]. Figure 1.1 shows the prototype of a
silicon photonic integrated wavelength division multiplexing (WDM) system developed
by us. As illustrated in the figure, all optical components needed to construct a WDM
system, including waveguides, optical modulators, wavelength-division-multiplexer/de-
multiplexer, and photodetectors, are monolithically integrated onto a single silicon chip.
The only component which is not monolithically integrated is the laser module which is
flip-chip bonded onto the silicon substrate. Based on this prototype, the optical and the
electronic components can be integrated seamlessly to provide dynamic system-level
optical functions on chip, and these components can be produced and packaged through
a single CMOS process. Such Si-based optoelectronic integrated circuits (Si-OEICs)
have substantial applications in future low-cost and high-volume tera-scale data
transmissions.
Figure 1.1. The prototype of a silicon photonic integrated four-channel WDM
transceiver.
On-chip optical interconnect
One of the promising applications of Si-OEIC is the on-chip optical interconnect (OI).
Si-based microelectronics, under the guidance of Moore’s law, have been driving the
improvement of data processing speeds for decades. This is mainly attributed to the
Chapter 1 Silicon Photonics: Introduction and Review
3
continuous advancement of the CMOS fabrication facilities and processes, which makes
it possible to constantly downscale the size of basic electronic computing elements and
increase their integration density and complexity within electronic microprocessors [12].
However, as the electronic integrated circuits (EICs) become increasingly dense, the
metallic electrical interconnects which are responsible for the communications between
different components of an EIC or different parts of a microprocessor have to be
arranged more and more densely as well. Consequently, the resulted inherent physical
limitations of these metal wires, such as the RC circuit delay and the frequency-related
loss, signal distortion, and crosstalk noise, have severely limited the overall
performance of the EIC. In the meantime, these metal wires have contributed the
majority of the power consumption of the EICs, which is called “the interconnect
bottleneck” [13-15].
To overcome this bottleneck, a breakthrough solution has been proposed which uses
optical interconnects. By using photons instead of electrons to transmit data between
these electronic components, we can avoid those physical limitations associated with
electronics and take the exceptional advantages of high-bandwidth, low-power-
consumption, and low-interference of photonics, which so far have only been widely
used in long-haul and short-reach optical fibre communications. Optical interconnects
can be employed to realize ultrafast and energy-efficient broad-to-broad, chip-to-chip,
or even intra-chip data communications [16]. Moreover, the compatibility of Si-OEIC
with existing CMOS fabrication lines enables the possibility of producing high-
performance OIs with low cost and high yield. Due to these great potentials, more and
more enterprises have launched their long-term projects for Si-based OI chips. For
example, IBM’s SNIPER (silicon nanoscale integrated photonic and electronic
transceiver) project aims to build ultrafast on-chip WDM systems which are compatible
with their standard CMOS lines for next-generation supercomputers [17]. Intel has also
announced their plan to replace chip-to-chip electrical interconnects with optical
interconnects to achieve a higher bandwidth of 100 Gb/s to 1 Tb/s [18].
Optical fibre networks
In addition to chip-scale optical interconnections, silicon photonics can also bring
enormous benefits to long-haul fibre optic transport networks and short-reach optical
communications. With the explosive growth of multi-media users and broadband
Chapter 1 Silicon Photonics: Introduction and Review
4
Internet services, the capacity of optical networks has improved substantially over the
past decade and will continuously face significant traffic demands with a growth rate of
around 30% per year [4]. To keep this rapid growth of the network capacity sustainable,
it is not only required to expand the channel capacity of the transmission links, e.g., by
developing more advanced optical modulation formats, but is also important to improve
the physical networks to have higher capacity, lower cost and better energy efficiency.
Currently, the photonic systems used in optical networks usually comprise of discrete
optical elements such as InP lasers and photodetectors, Lithium Niobate modulators,
etc., which are bulky, expensive, and lack of energy efficiency and function scalability
[19]. To meet the demand of future green optical networks, high-density photonic
integration is unavoidable. Silicon photonic devices, benefited from their ultra-compact
footprints, dense integration, low power consumption, as well as the integration and
interaction with conventional CMOS electronics at the chip or wafer levels, are
becoming promising solutions for future optical networks.
Data centres/super computers
Another major application of silicon photonics integrated circuits is inside data centres
and supercomputers. With the explosively increasing traffic [20] generated by cloud
computing, cloud storage, and multi-media streams, the capacity demand for data
centres also increases exponentially. This will not only result in an increase in the
number of vast data centres in the future, but will also need expansions of current data
centres to hold more servers, racks, storage, and networking functions. Consequently, a
more efficient and robust solution is needed to tackle the complicated interconnection
systems inside data centres in the future [19]. Similarly, the computing performance
required for the supercomputer is estimated to grow at around 60% per year. Whilst
multi-core and parallel processing architectures have been widely deployed for this need,
the intra-system interconnection bandwidth becomes a bottleneck [2].
The vertical-cavity surface-emitting laser (VCSEL)-based multi-mode parallel
interconnections, which are the dominant deployment in current data centres and
supercomputers, is difficult to meet the demands of future interconnections for higher
capacity and longer distance (kilometre-level), due to the limited direct-modulation
bandwidth and the difficulty of integrating the WDM functionality [19]. Moreover,
based on the current deployment, expanding data centres and supercomputers imposes a
Chapter 1 Silicon Photonics: Introduction and Review
5
huge pressure on the power consumption and the cost to run the more complicated
interconnection networks and cooling systems. Therefore, considering the above issues,
the silicon photonics integration, which is able to bring ultrafast and low-cost optical
interconnections to the levels of rack-to-rack, server-to-server, or even chip-to-chip, will
be a disruptively cost- and energy-efficient solution.
Industrial milestones
Due to these promising applications, silicon photonics has been extensively studied
since the 1980s [21, 22], and significant milestones have been achieved over the past
decade towards the commercialization of Si-based OEICs. In 2004, Intel demonstrated a
standard-CMOS-compatible silicon modulator with a modulation bandwidth exceeding
1 GHz [23], which made the monolithic integration of silicon modulator with CMOS
electronics possible. In 2006, Intel and the University of California, Santa Barbara co-
announced the first electrically pumped continuous-wave hybrid silicon laser [24],
which was a milestone for realizing low-cost light sources on silicon. In 2008, Luxtera
announced the world’s first fully integrated Si-based DWDM transceiver which was
produced based on the 130 nm CMOS fabrication lines and achieved an aggregate data
rate of 40 Gb/s [25]. In 2015, IBM announced a breakthrough in silicon photonic chips,
presenting a fully silicon integrated four-channel WDM transceiver which achieved an
aggregate data rate of 100 Gb/s [26]. More and more efforts and breakthroughs have
been pushing silicon photonics to the higher level of maturity, and the real
commercialization of this technology is foreseeable in the near future.
1.2 Components
The ultimate goal of Si-OEIC is to transfer the whole optical system onto a chip. Figure
1.2 illustrates the block diagram of a basic on-chip OI system [27]. To build such an OI
chip, all the building blocks of an optical system are needed, including a laser source to
generate the light signal, an optical modulator driven by electronic circuits to encode the
signal, optical waveguides for light transmission, and a photodetector on the receiver
side to detect and convert the light into photo current. The amplifier is used to amplify
the photo current and reproduce the original electrical signals. Other optical devices
such as switches, routers, splitters, and filters can also be added to the system to provide
Chapter 1 Silicon Photonics: Introduction and Review
6
more comprehensive optical functions. In this diagram, the laser source is externally
bonded to the chip, due to the limitation of silicon material itself (indirect bandgap).
Depending on whether external electrical control signals are needed to realize the
optical function, these components can be divided into two categories, namely the active
device and the passive device. Lasers, modulators, photodetectors, and amplifiers are
active devices, while waveguides, couplers, splitters, filters, and gratings are passive
devices. Passive devices are primarily used for light transmission, splitting, switching,
coupling and reflection, or are used as the components to construct more complex
optical devices. In next section, the four main building blocks of a silicon photonic chip,
which are on-chip modulator, laser, photodetector, and passive optical device, will be
briefly reviewed.
Figure 1.2. The block diagram of an on-chip OI system.
1.2.1 Silicon Optical Modulator
Due to its centrosymmetric crystal structure, the efficiencies of the primary electrical
field effects (including the Pockels effect, the Kerr effect, and the Franz-Keldysh effect)
which are conventionally used to achieve the refractive index modulations in traditional
III-V semiconductor or Lithium Niobate based modulators are very weak in silicon at
the telecommunications wavelengths of 1.3 and 1.55 m. This left the thermo-optic
effect and the free-carrier plasma dispersion effect as the only effective means to realize
the optical modulation in silicon [28]. However, although silicon has a very high
thermo-optic coefficient of 1.86×10-4/K, the modulation speed based on thermo-optic
mechanism is too slow to meet the requirement of modern telecommunications
Chapter 1 Silicon Photonics: Introduction and Review
7
applications. Therefore, the free-carrier plasma dispersion effect, due to its much faster
modulation speed, has become the most popular approach to realize optical modulations
in silicon photonics [8]. Based on the plasma dispersion effect, a change in the density
of free carriers in silicon will cause a change in its refractive index which is eventually
transferred to a change in the phase of the optical mode. This effect can be implemented
by three different configurations: carrier injection in a p-i-n diode, carrier depletion in a
p-n diode, and carrier accumulation in a MOS capacitor, which are illustrated in Fig. 1.3.
Figure 1.3. Diagrams of three different structures employed in plasma-dispersion-effect-
based silicon modulators to implement the mechanisms of (a) carrier injection, (b)
carrier depletion, and (c) carrier accumulation.
The way to achieve carrier injection is to use a p-n junction. However, since the high
doping concentrations in the p- and n- regions will cause high optical absorption loss if
the doped regions form the parts of the waveguide where optical power is concentrated,
a p-i-n structure is more suitable. As shown in Fig. 1.3(b), the intrinsic region is where
the optical power is contained. When the device is forward biased, the carrier density
injected into the intrinsic region will increase, and this will consequently cause a change
in the refractive index of the material which will be converted to a phase change of the
propagating mode. Soref and Bennet deduced the relation between the change of the
refractive index and the change of the free carrier densities [28, 29]. At 1.55 m, it is
expressed as:
22 18 0.8[8.8 10 8.5 1 (0 ) ]e h e hn n n N N (1.1)
Where ∆ne and ∆nh are the changes in refractive index resulting from the changes in
free-electron (∆Ne) and free-hole (∆Nh) carrier concentrations respectively. The resulted
phase change of the optical mode will be:
Chapter 1 Silicon Photonics: Introduction and Review
8
0
2 nL
(1.2)
Where L is the length of the active region of the modulator. Correspondingly, there is
also an additional absorption loss for the optical power due to the change of free carrier
concentrations [8]. The forward biased p-i-n diode approach has been proven to provide
high modulation efficiency. However, due to the slow carrier generation/recombination
processes, the modulation speed is usually limited. In contrast, the reverse biased p-n
junction could cause carrier depletion which has been proven to achieve the best high
speed result. As shown in Fig. 1.3(c), the central region of the waveguide is lightly n-
type and p-type doped. When a reverse bias is applied, a depletion region forms at the p-
n junction, which expands with the bias voltage and creates a change in the refractive
index of the waveguide. The third method to achieve plasma dispersion is through
carrier accumulation. As shown in Fig. 1.3(a), an insulating oxide layer is built in the
middle of the waveguide to form a capacitor. When a voltage is applied to the structure,
carriers will accumulate at the interface of the oxide layer, resulting in a change in the
refractive index. In contrast to carrier injection, the speed of carrier accumulation is not
limited to the minority carrier lifetime, but depends on the device resistance and
capacitance [23, 30].
Based on the plasma dispersion effect, an optical phase shifter can be realized through
above three different configurations. To convert the phase change into intensity change
and realize an optical intensity modulator, two different methods are mostly used. The
first one is manipulating the relative phase difference between two propagating waves to
achieve either a constructive or destructive interference. Normally, a Mach-Zehnder
interferometer (MZI) structure is used to realize this effect.
Chapter 1 Silicon Photonics: Introduction and Review
9
Figure 1.4. The schematic diagrams of (a) an MZI modulator and (b) a microring-based
modulator [31].
MZI-based modulator
Figure 1.4(a) shows a typical 1 × 1 MZI structure. The electric field of the propagating
wave at the input port of the MZI can be represented as:
0( ) j zE z E e (1.3)
Assume the lengths of the two arms are L1 and L2. For a basic 1 × 1 MZI structure
shown in Fig. 1.4(a), when ignoring the waveguide propagation losses, the electric field
at the output port will be:
1 1 2 2
0 0
1( )
2
j L j LE E e E e
(1.4)
The output optical intensity will be:
2
0 1 1 2 2
11 cos( )
2I E E E L L (1.5)
Chapter 1 Silicon Photonics: Introduction and Review
10
From Equation. 1.5, it can be seen that the output intensity of the MZI can be
manipulated by varying the relative phase of the two arms of the interferometer, and this
is the basic principle behind an MZI modulator, where the phase change can be realized
through refractive index change.
Microring-based modulator
The second popular way to convert the phase change to intensity change is using a
microring resonator such that a small refractive index change of the resonator will cause
a change in the resonances, and consequently, there will be a significant change in the
device transmission. Figure 1.4(b) shows the first microring modulator demonstrated by
Xu et al. in 2005 [31]. In this modulator, the plasma dispersion effect was implemented
by carrier injection based on a p-i-n diode. Theoretically, for the all-pass-type microring
resonator shown in Fig. 1.4(b), the intensity transmission is given as [32]:
2 2
2
2 cos
1 2 cos ( )
pass
input
I a ra r
I ar ra
(1.6)
Where Ф = β∙2πR is the single-pass phase shift, R is the radius of the ring, a is the
single-pass amplitude transmission coefficient, and r is the self-coupling coefficient.
The ring is on resonance when the optical length of the ring is a whole number times of
the light wavelength, and the optical waves in the device will interfere constructively:
2eff
res
n R
m
1, 2, 3...m (1.7)
When the ring is on resonance (Ф = 2mπ), most of the incident power will be circulated
or preserved inside the ring, and the transmitted component will be minimum, the
intensity transmission equation becomes:
2
2
( )
(1 )
pass
input
I r a
I ra
(1.8)
In practice, the resonator is tuned such that the operating wavelength is on the slope of
the resonance peak, and by modulating the refractive index of the ring, the resonance
peak is shifted and the output intensity of the device is changed [32].
Chapter 1 Silicon Photonics: Introduction and Review
11
Compared with the MZI-based modulators which normally need a long interaction
length to realize the optical interference, the microring-based modulators could have
smaller footprints due to their looped resonant structure, and this also implies lower
insertion loss and power consumption. But on the other hand, the bandwidth of the
microring modulators is usually much narrower than MZI-based modulators, which also
indicates a higher sensitivity to fabrication errors and temperature variations [8, 33]. To
stabilize the temperature, accurate temperature controlling components and
specifications will be needed, which could instead increase the power consumption of
the device. Therefore, in practice, there will be trade-offs, and the modulator should be
chosen based on the requirements of specific applications.
Research progress
Silicon modulator was one of the earliest silicon photonic devices that researchers began
to put into study since the 1980s [34]. Initially, silicon modulators were implemented
based on the carrier injections in a p-i-n diode structure [35-38], which suffered from
relatively high optical loss and low speed, due to the high absorption of the highly
doped silicon waveguide and the long minority carrier lifetime in silicon. To overcome
these drawbacks, optimizations in modulator geometry and size had been studied, and
gigahertz modulation speeds had been presented [39-41]. In 2004, Intel demonstrated
the first monolithically integrated silicon modulator based on carrier accumulation,
which avoided the slow minority carrier lifetime in carrier-injection-type modulators
and achieved a bandwidth of >1 GHz [23]. Later on, they had improved the data rates to
10 Gb/s based on the carrier accumulation [30]. The first carrier-depletion-based
modulator was proposed by Gardes et al. in 2005, presenting a theoretical bandwidth of
50 GHz [42]. Since then, various modulators of this type based on reverse-biased p-n
diode had been presented, with data rates of >10 Gb/s [43-45]. In 2007, Intel improved
the 3 dB bandwidth of carrier-depletion-based modulator to 30 GHz with an ultrafast
data rate of up to 40 Gb/s [46]. Compared with the carrier-injection-type modulators,
the speed of carrier-depletion-type modulators is not limited by the minority carrier
lifetime due to the reverse bias, however, their smaller region for optical mode to be
interacted with the variations of the carrier density results in lower modulation
efficiency. The first microring-based modulator was experimentally demonstrated by
Xu et al. in 2005 [31], with the consideration of significantly improving the device
footprint and power consumption compared with the MZI-based modulators. The
Chapter 1 Silicon Photonics: Introduction and Review
12
modulator was based on p-i-n diode carrier injection with an initial reported data rate of
1.5 Gb/s. Later on, they had improved the performance to 16 Gb/s based on a pre-
emphasis driving signal [40]. Meanwhile, microring-based modulators based on carrier
depletions have also been studied [43, 44], and fast speed of >35 GHz was presented
[47]. In practice, the bandwidth, modulation speed, insertion loss, power consumption,
device footprint, and the CMOS compatibility etc. are some of the most important
metrics that should be taken into account when designing a silicon optical modulator [8].
But these metrics can sometimes contradict to each other, therefore trade-offs should be
carefully considered in specific applications. Meanwhile, continuous efforts are still
required to take the full advantages of the silicon integrated modulator and catalyze the
progress of Si-OEIC.
1.2.2 Si-integrated Laser
Due to the indirect bandgap structure, it is very challenging to realize lasers and
amplifiers in crystalline silicon. The misaligned momentums of conduction band and
valence band in silicon results in very poor efficiency of the radiative electron-hole
recombinations based on which the optical gain is achieved. As the gain is
insufficient to compensate for the free carrier absorption loss which is orders of
magnitude higher, the bulk silicon exhibits a net loss property under carrier
injections, and therefore it has been regarded as an incapable material for light
amplification and emission [6, 48, 49]. To overcome the hurdle and integrate light
sources onto the silicon platform, various approaches have been proposed and
studied, which are mainly based on the following several mechanisms: (1) Quantum
confinement, (2) Stimulated Raman scattering (SRS) effect, (3) Erbium (Er)-doping,
(4) Germanium (Ge)-on-silicon, (5) III-V compound hybridization.
Quantum confinement-/SRS-based laser
The basic principle of quantum confinement effect is to localize the carriers (or
confine the carriers at a small space free of defects) such that their momentums
become uncertain, and this in consequence could increase the probability of the
radiative recombination. The method to implement quantum confinement is using
Si-based nanostructures such as porous silicon, quantum wells, quantum dots, silicon
nanocrystals, etc. [50]. The SRS effect is based on the Stokes transition which is
Chapter 1 Silicon Photonics: Introduction and Review
13
induced by a pump light, and the wavelength resonant with the stokes transition can
be coherently emitted [51]. SRS was initially introduced in silica optical fibres to
realize lasers and amplifiers, and it was in 2002 when it was firstly proposed as a
way to realize lasers and amplifiers in silicon [52]. In 2007, a milestone work was
done by Intel, in which they demonstrated the first continuous-wave Raman silicon
laser [53]. Despite these progresses, the silicon lasers based on quantum
confinement and SRS effect usually suffer from a number of drawbacks which
hinder their usefulness for on-chip OIs, for example, they emit non-
telecommunications-band wavelengths and exhibit quite low emission efficiencies,
and the need of optical pumping instead of electrical pumping make them unsuitable
for on-chip integrations [54]. Therefore, in recent years, the study of silicon lasers is
mainly focused on the rest three methods.
Er-doped Si-laser
The great success of Er-doped fibre amplifiers and lasers in optical networks, as well
as their ability to emit light at wavelength of 1550 nm, have greatly motivated the
development of Er-doped silicon lasers. However, the gain of Er-doped bulk silicon
laser is quite low, both due to the low concentration of Er within bulk silicon and the
substantial non-radiative processes such as the Auger recombination and the back-
transfer of energy from the excited Er ions to silicon [55]. To overcome these issues,
the most popular solution is to dope Er into an SRO (silicon-rich-oxide) layer (a
SiO2 layer containing dispersed silicon nanocrystals) which acts as the optical gain
medium. Compared with the Er-doped silicon, the wider bandgap of silicon
nanocrystals in SRO reduces the energy back-transfer from Er ions to silicon, and
the low free carrier concentration in SiO2 could suppress the Auger recombination
[33]. Moreover, by sandwiching the Er-doped SRO layer between a metal layer and
the silicon layer (a MOS structure illustrated in Fig. 1.5), the electrical pumping can
be achieved. When a voltage is applied to the MOS structure, the Er atoms will be
excited, meanwhile, the silicon nanocrystals can also be excited, acting as the
sensitizers and transfer their energy to the nearby Er ions [6]. There are still
challenges to achieve high emission efficiency based on this method. For example,
the free carrier absorption loss is high, and the solubility of Er in SRO is still very
low [56]. To improve the efficiency, the method to increase the Er concentration in
SRO will be needed in the future. Nevertheless, good progress has been made in this
Chapter 1 Silicon Photonics: Introduction and Review
14
field, e.g., electroluminescence with a power efficiency of 10-2 and Er population
inversion of up to 20% has been demonstrated [57].
Figure 1.5. The schematic of SRO embedded in a MOS structure for electrical pumping.
Ge-on-Si laser
Same as silicon, Ge is also a material with indirect bandgap, however, the energy
difference between its direct and indirect bandgaps is only 136 meV, making it a so-
called pseudo-direct-bandgap material. In addition, Ge’s direct bandgap of 0.8 eV
corresponds to the optical telecommunications wavelength of around 1550 nm, and
Ge is compatible with the traditional silicon CMOS fabrication processes [58].
These special properties of Ge have greatly motivated researchers to pursue light
sources based on the Ge-on-Si platforms. One way to reduce the energy difference
between the direct and indirect bandgaps is introducing tensile strain in Ge layer.
According to [59], a 0.25% tensile strain reduces the energy difference to 115 meV.
To turn Ge into a complete direct-bandgap material, a 2% tensile strain will be
needed however, this will shrink the bandgap to 0.5 eV at the same time and shift
the emitted wavelength to 2500 nm. The way to reduce the energy difference while
keeping the tensile strain at a small percentage is employing n-type doping. By
filling the indirect L valleys with electrons, the energy difference can be
compensated and the emission efficiency at 1550 nm is greatly improved. The
concept of Ge-on-Si laser was first proposed in [59]. Afterwards, it has been
extensively researched and significant milestones have been achieved in both
optically and electrically pumped Ge-on-Si lasers [60, 61]. Although further
research is still needed to improve the viability of monolithic integration of Ge laser
in silicon, e.g., it is needed to deal with the conflicting impacts of the heavy n-
Chapter 1 Silicon Photonics: Introduction and Review
15
doping concentration [54], the good progress, along with the great potentials of Ge,
remain it as a promising candidate for Si-integrated on-chip light sources.
Hybrid Si-laser
Currently, the most practical method to realize light sources in silicon photonics is
integrating III-V compounds into silicon substrates so as to realize a so-called
hybrid silicon laser. The first way to do this is applying traditional hetero-epitaxial
growth of III-V layers on silicon substrates. However, the high-density threading
dislocations caused by the large difference in lattice constants and thermal
expansion coefficients of silicon and III-V materials make this approach very
difficult. One way to resolve the issue is to apply a buffer layer with an appropriate
lattice constant to alleviate the threading dislocations, such as SiGe [62] or GaSb
[63]. Another method is using quantum dots. Based on the good defect-tolerant and
threading-dislocation-filtering capabilities, various Si-based quantum-dot lasers
were presented, which have been summarized in [54]. The second method to
integrate III-V compounds to silicon substrates is the heterogeneous wafer-bonding
techniques, which is also the most appealing solution for industry use nowadays.
Based on this technique, the unpatterned III-V epitaxial layer of either a wafer or
dies can be directly molecule-bonded onto the patterned silicon substrates or through
adhesives such as divinylsiloxane-benzocyclobutane (DVS-BCB) [64], and after the
removal of III-V substrate, only the epitaxial layer exists above the passive silicon
photonic circuit, which will then be patterned to form the active gain structure. The light
generated in the active region is evanescently coupled to silicon waveguide, which
avoids the lattice-mismatch issues associated with the hetero-epitaxial growth
approach [65].
In addition to the above two methods, a more simplified way to integrate III-V
compounds is directly mounting a III-V semiconductor die onto an SOI wafer by
solder bumps. An external-cavity (EC) hybrid silicon laser is the most popular
configuration of this method. The advantages of the EC lasers include the low cost
and simplicity in fabrication process. Moreover, since the gain module and the laser
cavity are separate, they can be chosen and optimized individually to achieve better
performance of each section with great flexibilities. We proposed a wavelength-
switchable external-cavity laser structure in [66], and the concept is shown in Fig.
Chapter 1 Silicon Photonics: Introduction and Review
16
1.6. An SOA (semiconductor optical amplifier) die is flip-chip bonded onto the
silicon substrate, serving as the gain medium. Based on the cascaded active
microring resonators, we can realize a laser with compact size, switchable
wavelength, narrow linewidth and large free spectral range, which is suitable for use
in WDM networks. However, due to the different refractive indices and light
confinement properties of the gain module and the silicon waveguide, the main
challenge for this type of laser is the realization of efficient low-loss coupling
between the gain and the external cavity. The coupling can be assisted with a spot-
size converter, but accurate optical alignment will be needed, which could increase
cost and time for manufacture, thus more trade-offs remain to be considered.
Figure 1.6. Proposed wavelength-switchable laser. (a) conceptual design, (b) detailed
implementation, (c) SEM image of a broadband reflector [67].
1.2.3 Ge-on-Si Photodetector
Silicon exhibits a transparency within the optical telecommunications band of 1.3-1.55
m, making it a suitable candidate for optical waveguides. But on the other hand, this
also hints that photo-detection within this band would be impractical due to the weak
photon absorption efficiency. Ge, on the other hand, exhibits a much stronger
absorption in 1.3-1.55 m due to its smaller bandgap structure. Moreover, Ge is
compatible with the silicon CMOS fabrication processes and is much easier to be
Chapter 1 Silicon Photonics: Introduction and Review
17
integrated into silicon substrates compared with the III-V compounds which is another
viable material for photo-detection. Therefore, Ge-on-Si photodetector has become the
most popular approach for photo-detections in silicon photonics and has obtained rapid
progress over the past decade.
The main challenge in Ge-on-Si photodetector is the growth of high-quality epitaxial Ge
layer onto silicon substrates, as the 4.2% lattice mismatch between Ge and Si leads to
severe surface roughness and threading dislocations at the interface of the two layers,
which could seriously influence the performance of the detector and the subsequent
CMOS processes [68]. The first successful approach, which was demonstrated by Luryi
et al. [69], was inserting a graded SiGe buffer layer to reduce the threading dislocations,
and this method was later improved by Fitzgerald et al. using multiple buffer layers [70,
71]. The downside of this method is that the buffer layer is very thick, which could
affect the coupling efficiency between silicon waveguide and the Ge photodetectors in a
waveguide-integrated-type Ge-on-Si photodetector in Si-OEICs [72]. The second
approach is based on a two-step epitaxial growth process [73, 74]. The first step is to
directly grow a thin layer (30-60 nm) of epitaxial Ge on silicon substrate by chemical
vapor deposition (CVD) using a low temperature of around 350C to prevent islanding,
in the second step, the temperature is increased to around 600C to grow thick high-
crystal-quality Ge absorption layer with higher growth rates. By this method, the growth
of Ge absorption layer is no longer influenced by the Ge/Si lattice mismatch therefore
could have a much smoother surface. However, the threading dislocation density of this
method is usually high and high-temperature annealing (>750C) is often needed to
effectively reduce it [68]. Various other approaches to grow Ge on silicon have also
been explored, such as by combing thin SiGe buffer layers with the two-step Ge growth
process, and good surfaceness and low threading dislocation density have been
demonstrated [75, 76].
The first type of structure of the p-i-n type Ge photodetectors demonstrated is the
normal-incidence structure, for which light is normally (or vertically) illuminated onto
the surface of the Ge detector for photocurrent generation, which is also the simplest
form and has long been used in optical communications. However, the normal-
incidence structure involves critical trade-offs in meeting the primary figures of merit of
photodetectors, which include the detector bandwidth, efficiency, and dark current noise.
Simply speaking, higher photon absorption efficiency needs a thicker Ge layer as this
Chapter 1 Silicon Photonics: Introduction and Review
18
could increase the absorption length in the intrinsic region, but on the other hand this
results in a lower bandwidth and higher dark current, as the thicker Ge also increases the
carrier transit time and the junction capacitance [33]. Fortunately, the trade-off can be
resolved with another type of structure, which is the waveguide-integrated structure. In
this structure, Ge is formed as part of the waveguide which is extended from the silicon
waveguide. Consequently, light is going through Ge detector horizontally rather than
vertically, and the Ge thickness will no longer limit the absorption efficiency and can be
made much thinner to have high bandwidth and low dark current simultaneously [72].
Moreover, the waveguide-integrated detectors have a device area which is significantly
smaller than that of the normal-incidence detector, and are more suitable for planar
CMOS processes, which greatly favours the building of large-scale Si-OEICs.
Therefore, waveguide-integrated Ge-on-Si photodetectors have become a promising
candidate in silicon photonics and have achieved good research progress, e.g., detectors
with high responsivity, wide bandwidth, and low noise have been demonstrated [77, 78].
Compared with the p-i-n type Ge photodetectors, the avalanche Ge-on-Si photodetectors
(Ge-on-Si APDs) can provide better sensitivity based on their advantage of internal
multiplication mechanism, thus have become increasingly attractive in recent years [68].
By combining the small ionization ratio (<0.1) advantage of silicon with the strong
absorption property of Ge, the performance of Ge-on-Si APDs have been demonstrated
to exceed the traditional III-V APDs in terms of higher gain-bandwidth products and
better sensitivity [79-81]. Moreover, same as the waveguide-integrated p-i-n Ge
detectors, the waveguide-integrated Ge-on-Si APD can also achieve a lower dark
current and higher bandwidth-efficiency products compared with the normal-incidence
APDs [82], which makes it a competitive candidate for future high-speed on-chip
optical systems.
1.2.4 Passive Silicon Photonic Device
As the fundamental building blocks of the complex PICs, passive components, such as
waveguides, Y-junctions, directional couplers (DCs), multi-mode interference couplers
(MMIs), Bragg gratings, etc., are used to realize various passive optical functions of the
circuit such as light guidance, power splitting, wavelength filtering, optical reflection,
polarization rotation, etc., or used as the basic constituents to build active and passive
devices with more complex functions such as lasers, modulators, and photodetectors.
Chapter 1 Silicon Photonics: Introduction and Review
19
Therefore, the performance and the size of the passive components play a pivotal role in
Si-based photonic integrations, which is also within the major interest of this thesis.
The most fundamental passive function within a silicon photonic circuit is the guidance
of light. The two most popular waveguide structures are rib waveguide and strip
waveguide. Rib waveguides benefit from smaller propagation loss, lower fabrication
requirement, and easier coupling with optical fibres due to their relatively large cross-
section size (multi-micron). But strip waveguides, despite their critical feature size,
more challenging fibre-coupling, and higher requirement on waveguide smoothness, are
the most competitive candidate to meet the large-scale and high-density photonic
integration requirements for Si-based PICs. This is because strip waveguides have ultra-
small dimensions of sub-micron level, and they can be used to realize tight waveguide
bends with radius of only several microns. The major challenge with strip waveguide is
the high scattering loss resulting from the strong interaction of the confined mode with
the imperfectly fabricated waveguide sidewalls, which is also the main constituent of
the waveguide propagation loss. Particularly, the transverse-electric (TE) polarized light
suffer more from the sidewall roughness than the transverse-magnetic (TM) polarized
light, due to their stronger optical field concentration at waveguide sidewalls. Without
further optimization on waveguide roughness, propagation loss of around 3 dB/cm at
1550 nm was typically reported for dry-etched strip waveguides [83-85]. By wet-
etching or applying high-temperature thermal oxidations, the sidewall roughness can be
effectively reduced, and propagation loss of <1 dB/cm has been presented [86-88].
The second challenge with strip waveguide is obtaining low-loss coupling with external
fibre, due to the significant mode-size mismatch. Two common methods are normally
used, which are the vertical coupling through surface grating couplers and the edge
coupling through spot size converters (SSCs). In comparison, grating couplers usually
have a relatively large coupling loss of several dB and a narrow 3 dB bandwidth of only
tens of nanometres [89-91], and the non-fully-etched structure requires more complex
fabrication process. In contrast, edge couplers can achieve a low sub-dB coupling loss
and a much wider 3 dB bandwidth of hundreds of nanometres [92-94], which are
suitable for wideband operations, and normally the fabrication of them only needs one
step of full etch down to the substrate. But on the other hand, the alignment tolerance of
the edge coupling is much lower than that of the grating coupling, and sub-micron
alignment accuracy is needed to avoid substantial coupling losses [95]. In contrast,
Chapter 1 Silicon Photonics: Introduction and Review
20
grating couplers have much more relaxed alignment tolerances. An alignment error of
±2 µm in grating coupler leads to an additional coupling loss of <1 dB [90, 96], which
makes it a more commercial approach due to the reduced packaging cost [97].
Another major issue in silicon photonic circuit is the polarization handling. The planar
structures of silicon waveguides and devices show significant birefringence, which
makes them strongly polarization sensitive, therefore a polarization-diversity scheme is
usually applied in the design of Si-PICs. To realize an on-chip polarization-diverse
system, an integrated polarization manipulation device is normally needed, such as a
polarizer, a polarization rotator (PR), or a polarization beam splitter (PBS). A polarizer
only transmits light with the desired polarization while blocks the other, and its
principle is mainly based on the leakage or cut-off of light with unwanted polarization.
The leakage-type polarizer usually needs a sufficiently long device length to ensure a
high leakage loss (1 mm in [98]), while the cut-off-type polarizer, such as the
plasmonic-type polarizers [99, 100], can be made very short (several microns) due to
the strong mode attenuation, but it comes with the introduction of new materials such as
metal into the silicon substrate, which increases the fabrication complexity and
introduces high insertion loss. The principle of PR is based on mode hybridization,
where a TE/TM mode is gradually rotated or converted to a TM/TE mode, normally
through a specific asymmetric or multi-level structures [101], and the fabrication of
such structures can be very complex. In comparison, PBS is the simplest and most
fabrication-friendly polarization handling device. It works by separating the input light
with hybrid polarizations into two outputs with single polarization, and can be realized
through various planar structures, such as MZIs, MMIs, or DCs [102-104]. Based on the
large birefringence of silicon waveguides, silicon PBS can have high extinction ratio
(ER) and compact device size. But on the other hand, the large birefringence of silicon
platform also indicates a high sensitivity to fabrication errors [105], which could
seriously influence the ER of the PBS.
As an alternative, we have proposed an MZI-based PBS in silicon nitride (SiN) platform
[106] (shown in Fig. 1.7). The availability for growth with either low-pressure chemical
vapor deposition (LPCVD) or plasma-enhanced chemical vapor deposition (PECVD),
as well as the compatibility with silicon CMOS fabrications make SiN another attractive
platform. In our PBS, the low refractive index of SiN provides a higher fabrication
tolerance and a lower waveguide scattering loss compared with Si-based PBSs, and
Chapter 1 Silicon Photonics: Introduction and Review
21
based on an optimized phase control section, a high ER of >20 dB has been achieved
over a wide 80 nm wavelength range (1530 nm – 1610 nm) with a minimum insertion
loss of only 0.13 dB, and the whole device length is only 113 m, which is significantly
shorter than the previously demonstrated works such as the SiN-based PBS in [107].
Figure 1.7. (a) Optical microscope image of the fabrication PBS. (b) SEM (scanning
electron microscope) image of the SiN waveguide (cross-section).
1.3 Thesis Objective and Organisation
Chapter 1 has introduced the promising applications of silicon photonics and reviewed
the major components of the silicon photonic system, which include silicon integrated
laser, modulator, photodetector, and passive optical device. Amongst these key silicon
photonic building blocks, the passive device will be the focus of this thesis. Specifically,
in this thesis, two main types of passive optical devices, namely the optical power
splitter and the Bragg grating mirror, will be studied and investigated based on the
silicon photonics integration platform. A power splitter is widely used for light power
distributions or the construction of more complex optical devices, such as optical
switches, modulators, multiplexers, etc. An integrated Bragg grating mirror is used for
on-chip reflections which normally have important applications such as lasers,
resonators, or wavelength filters, etc. Therefore, these two devices are the two most
fundamental building blocks of the optical systems, and play an important role in
realizing highly integrated and high-performance Si-PICs.
The key contributions of this thesis will include the design of a silicon integrated ultra-
broadband and low-loss adiabatic optical power splitter and a silicon integrated ultra-
broadband and high-reflectivity circular Bragg grating mirror. In addition, the
fabrication process of these two devices will be developed based on the 220 nm SOI
platform, and will also be presented in this thesis.
Chapter 1 Silicon Photonics: Introduction and Review
22
The specific thesis organisation will be as following:
Chapter 2 has two major sections. In the first section, the principle of single-mode
silicon strip waveguide is presented. In the second section, the fabrication process of the
single-mode strip waveguide and the passive silicon photonic devices is developed and
demonstrated based on the 220 nm SOI platform. The developed process presented in
this chapter is the fundamental step to realize the low-loss silicon waveguide as well as
the devices proposed in Chapter 3 and Chapter 4 of this thesis. In this chapter, the
silicon photonic chip measurement setup is also demonstrated.
In Chapter 3, a 3-dB optical power splitter based on adiabatic silicon tapers is proposed
and experimentally demonstrated. The proposed splitter breaks the bottleneck of the
conventional silicon integrated 3-dB optical power splitters, and for the first time, it
achieves the properties of ultrabroad band, low excess loss, polarization insensitivity,
and compact device size simultaneously. The exceptional advantages of the splitter
make it a promising element for use in building large-scale and high-density Si-PICs.
In Chapter 4, a novel circular Bragg grating mirror is proposed and experimentally
demonstrated based on the 220 nm SOI platform. The proposed circular Bragg grating
mirror solves the difficulty of achieving high reflectivity and broad reflection bandwidth
simultaneously for conventional Bragg gratings based on the thin-silicon SOI wafer,
and it has a very compact device size of only several microns. As one of the promising
applications of the proposed circular Bragg grating mirror, a compact, high extinction-
ratio, and low-loss integrated notch filter is also demonstrated in this chapter.
In Chapter 4, in addition to the circular Bragg grating presented on the 220 nm SOI
platform, which is designed for TE-polarized light, two different solutions to realize
broadband and high-efficiency TM-reflections are also proposed and investigated based
on FDTD simulations. These solutions are based on non-220 nm SOI platform. The first
solution is a circular Bragg grating with thickened grating blades, and the second
solution is a circular Bragg grating based on the horizontal slot silicon waveguide. Both
solutions are proved to be effective for TM-reflection. Finally, in this chapter, two
extended applications of the proposed slot grating, including a polarization-selective
Bragg grating mirror, and a broadband high-extinction-ratio TM-pass/TE-block
polarizer is proposed and investigated based on simulations.
The conclusions and future work of this thesis are presented in Chapter 5.
Chapter 1 Silicon Photonics: Introduction and Review
23
The references are shown in Chapter 6.
Chapter 2 Passive Silicon Photonic Device Fabrication
24
CHAPTER 2 PASSIVE SILICON
PHOTONIC DEVICE FABRICATION
2.1 Single-Mode Silicon Waveguide
In this chapter, the principle of single-mode silicon strip waveguide and its fabrication
process will be discussed. Specifically, in this section, the major properties of silicon
strip waveguide, which include the single-mode condition, typical size, the fundamental
modes, and the source of the waveguide propagation loss will be discussed. In the next
section, the developed fabrication process of single-mode silicon strip waveguide and
the proposed passive devices in this thesis will be presented in detail. In addition, the
experimental measurement setup will also be introduced at the end of this chapter.
Optical waveguide is used for light guidance and routing in PICs. It is also used as the
fundamental building block to construct various integrated optical devices with more
complex functions such as directional couplers (DCs), MZIs, polarization rotators,
micro-ring resonators, etc. As introduced in Chapter 1, the two most commonly used
waveguide structures in silicon photonics are rib waveguide and strip waveguide. Both
types of waveguide have merits and demerits, depending on specific applications. But in
terms of realizing the true advantage of silicon photonics which is large-scale and high-
density photonic integration, strip waveguide is recognized as the only competitive
candidate due to its ultra-small dimension of hundreds of nanometres. Moreover, based
on the tight optical confinement, the bending strip waveguide can be made with very
small radius of only a few microns, while still keeping a low bending loss. This is
especially advantageous in building high-density PICs. In contrast, the bending rib-
waveguide normally needs a radius of at least hundreds of microns to achieve a low
bending loss. Benefited from the large refractive index of silicon and the tight optical
confinement of strip waveguide, silicon photonic devices can normally be made to be
very compact and have good performance. In this thesis, the strip waveguide is used in
the design of silicon photonic devices and chips.
Chapter 2 Passive Silicon Photonic Device Fabrication
25
Figure 2.1. The schematic diagram of a silicon strip waveguide: (a) 3D view; (b) cross-
section view.
Figure 2.1 shows the schematic of a typical SOI-based strip waveguide with a cross-
section size denoted as W × T. The strip waveguide made of silicon has a very high
refractive index of 3.47 and sits above a so-called BOX (Buried-Oxide) layer which
has a much lower refractive index of 1.44. Based on such a large refractive index
difference, the waveguide could provide a very strong optical confinement and can be
made very compact to satisfy the “total internal reflection” (TIR) condition. Normally,
the BOX layer is made thick enough (1 - 3 m) to avoid evanescent coupling to the
silicon substrate, and in practice the waveguide can be covered by either SiO2 or air as
the upper cladding.
Theoretically, depending on its cross-section size (W × T), the waveguide shown in Fig.
2.1 can support single or multiple “propagation modes”, with each mode having a
different effective refractive index (neff). From a physical point of view, these
propagation modes can be regarded as different energy transmission forms within the
waveguide. From a mathematical perspective, these modes correspond to the eigenmode
solutions of the Maxwell equations. In practice, to avoid intermodal dispersion, the
waveguide size is determined such that it will only support one mode, which is the
“fundamental mode”, and optical devices are normally designed specifically based on
this fundamental mode. Therefore, retaining a single-mode transmission scheme is
important in Si-PICs.
In practice, the most commonly used thickness for single-mode silicon strip waveguides
is about 200 - 250 nm. The aspect ratio of the waveguide width to thickness is normally
Chapter 2 Passive Silicon Photonic Device Fabrication
26
designed to be about 2:1, which is based on a number of considerations such as the
single mode transmission condition, waveguide propagation loss, waveguide bending
loss, and the ease of fabrication, etc. The exact waveguide thickness is usually chosen
based on specific applications. Particularly, a thickness of 220 nm has been taken as a
standardized size by some major silicon photonics fabrication foundries in the world
such as IME, IMEC, and LETI [108] and therefore has been widely adopted for use in
many academic and industrial research groups around the world. In this thesis, the
silicon photonic device and the chip design is based on the 220 nm SOI platform.
Theoretically, it is complicated to obtain the exact analytical solution of a non-planar
waveguide such as a silicon strip waveguide. Instead, a variety of numerical methods
have been proposed for the analysis of guided modes, such as the finite element method
(FEM), film mode matching method (FMM), beam propagation method (BPM), and
finite difference time domain method (FDTD), etc. [109] In this thesis, the FDTD-based
Maxwell solver Lumerical FDTD solutions and MODE solutions are utilized for the
analysis of optical modes and the device design.
Figure 2.2. The change of effective refractive index with waveguide width for 220-nm-
thick silicon strip waveguide with a top cladding of SiO2. The simulations were carried
out using Lumerical Mode solutions, and the refractive indices of Si and SiO2 were set
to be 3.477 and 1.444, respectively.
Chapter 2 Passive Silicon Photonic Device Fabrication
27
Figure 2.3. (a) – (d) The mode profiles of the fundamental TE and TM mode based
different waveguide widths.
Figure 2.2 shows the simulated effective refractive indices of the fundamental mode and
the first higher-order mode at 1550 nm with the change of waveguide width (W) for a
220-nm thick silicon strip waveguide. When both the upper and lower claddings are set
to SiO2, the waveguide can be called as a “symmetric waveguide”, as it has the same
refractive index for the top and bottom claddings. In this case, the single mode
condition is satisfied when W is ≤ 460 nm, as indicated in the figure. Figures. 2.3(a) and
2.3(b) show the simulated mode profiles of the fundamental TE and TM mode for a
460-nm wide silicon strip waveguide. As shown in the figure, while the electric field
intensity of the fundamental TE mode is mainly confined within the waveguide core,
the intensity of the fundamental TM mode is mainly concentrated at the top and bottom
interfaces of the waveguide with a large cladding penetration. As a result, neff-TE is larger
than neff-TM as shown in Fig. 2.2. The mode concentration at the waveguide boundaries is
caused by the discontinuity of the normal component of the electric field at the
waveguide boundaries which can be expressed by the boundary condition as following:
1 1 2 2n nE E (2.1)
Chapter 2 Passive Silicon Photonic Device Fabrication
28
Therefore,
2 211 2
2 2
2 12
n
n
EI
I E
(2.2)
Where ε1 and ε2 are the permittivity of different mediums across the boundary, En is the
normal component of the electric field which is also the only electric component of the
fundamental TE mode, and I is the optical intensity. In practice, as the sidewall
roughness of the fabricated waveguide is usually worse than the roughness of its top and
bottom interfaces, the fundamental TE mode usually has higher propagation loss than
the fundamental TM mode due to its stronger interaction with the sidewalls. Compared
to an “asymmetric waveguide” which has different refractive indices for the top and
bottom claddings, a special property of the symmetric waveguide is that the
fundamental modes will never be cut off regardless of what W is chosen, which means
that mathematically, the solution of the eigenmode equation always exists for the
fundamental TE and TM mode. On the contrary, for an asymmetric waveguide, once the
waveguide dimension is below a certain value, even the fundamental modes could be
cut off. But on the other hand, with a narrow waveguide width, e.g., 200 nm, as shown
in Figs. 2.3(c) and 2.3(d), the mode confinement will be very weak, and in practice,
such a weak mode confinement could be easily lost when light is going through tight
waveguide bends or transmitting in a non-ideal index environment where the refractive
indices of the top and bottom claddings are not exactly the same [108]. Furthermore, the
waveguide propagation loss would normally be higher for a narrower width, due to the
stronger mode interaction with the waveguide sidewalls. Therefore, in practice, to retain
both the single-mode transmission and low propagation loss, 460 × 220 nm has become
the most popularly used size for silicon strip waveguides.
The propagation loss of a passive silicon waveguide is mainly composed of three
components, namely the intrinsic material absorption loss (inter-band absorption), the
waveguide bending loss, and the mode scattering loss which is caused by the waveguide
surface roughness. As introduced in Chapter 1, silicon has a transparent window which
covers the major 1.3 - 1.55 μm telecommunications band within which it has negligible
absorption loss, therefore this part of loss is not a major contribution of the total
propagation loss. To understand the cause of the waveguide bending loss, it is more
intuitive to consider it in a ray optics model. Bending a straight waveguide would lead
Chapter 2 Passive Silicon Photonic Device Fabrication
29
to a decrease of the light incident angle, and once the incident angle is decreased below
the critical angle, the TIR would be broken and a part of light would be leaked into the
claddings, which results in a part of radiation loss. The bending loss is mainly
determined by the lateral mode confinement ability of the waveguide as the bending
occurs in the lateral plane. Due to the high index contrast and the small dimension of the
silicon strip waveguide, the bending loss is very low even with a small bending radius,
e.g., a 220-nm-thick 90° silicon strip waveguide bend with a bending radius of 2 μm
only exhibited a bending loss of around 0.1 dB [83]. Normally, the bending loss of a
silicon bending strip waveguide would be negligible if the bending radius is ≥ 5 μm.
The propagation loss of a silicon strip waveguide mainly comes from the waveguide
surface scattering loss. The high-density mode confinement within the small waveguide
core indicates a strong interaction of the guided mode with the waveguide surfaces,
therefore the propagation loss is very sensitive to the waveguide surface roughness. In
addition, as the top and bottom surfaces of the waveguide are relatively smooth based
on the crystal growth of the silicon layer, the scattering loss is mainly determined by the
roughness of the waveguide sidewalls which is dependent on the applied fabrication
process. Therefore, achieving smooth sidewall roughness is one of the important goals
of silicon photonic fabrication.
Chapter 2 Passive Silicon Photonic Device Fabrication
30
2.2 Fabrication and Measurement
2.2.1 Fabrication Process Based on 220 nm SOI
Figure 2.4. The silicon photonic fabrication process based on 220 nm SOI.
The fabrication process of silicon strip waveguide is illustrated in Fig. 2.4. The
fabrication started with SOITEC’s UNIBOND 200mm (8') SOI wafer with a 220-nm-
thick silicon layer and a 2-μm-thick BOX layer. The 8' SOI wafer was cleaved into
square pieces with a dimension of 7cm × 7cm such that it could be processed by the
employed fabrication facilities which support a wafer size of up to 4 inch. The photonic
structures were defined through E-beam lithography (EBL), with ZEP520A used as the
e-beam resist. After development of the positive-tone resist, Chrome (Cr) was deposited
by E-beam evaporation to serve as the etching mask for the silicon device layer, and the
silicon device layer was then etched through deep reactive-ion etching (Oxford Plasma
Lab100). After Cr removal, SiO2 was deposited on chip by plasma-enhanced chemical
vapor deposition (PECVD) to serve as the upper cladding layer.
Chapter 2 Passive Silicon Photonic Device Fabrication
31
2.2.1.1 E-beam lithography
EBL was implemented based on the Vistec EBPG5000 Plus system. During the e-beam
writing, all the silicon structures were set into a single writing layer which includes strip
waveguides, spot size converters (used for coupling between strip waveguide and fibre),
proposed adiabatic-taper-based optical power splitters (Chapter 3), and proposed
circular Bragg grating mirrors (Chapter 4). As a rule of thumb, the beam spot size for e-
beam exposure is set to be around one quadrant of the pattern critical dimension, and the
beam step size (BSS) is selected to enable a 50-60% overlap between two consecutive
beam shots so that a consistent exposure can be achieved with the Gaussian beam. The
illustration of the beam spot size and the beam step size is shown in the following figure,
and they are selected based on the equations below.
Beam Spot Size = 0.25 Pattern Critical Dimension
Beam Step Size = 0.5-0.6 Beam Spot Size
Since the proposed adiabatic splitter has a very small critical dimension of ≦30 nm, as
shown in Fig. 2.5, a sufficiently small beam spot size of 4 nm was chosen for e-beam
writing based on a beam current of 0.1 nA at 100 kV. In addition, a beam step size of 2
nm was selected to enable a 50% overlap between two consecutive beam shots.
ZEP520A was chosen as the positive-tone e-beam resist for EBL. Compared to another
popularly used positive-tone e-beam resist PMMA (polymethyl methacrylate), ZEP can
provide a comparably good resolution (minimum 20 nm based on our tests), a higher
sensitivity, and a stronger dry-etch resistance, therefore it has also been widely used in
EBL to realize nano-scale structures. In our case, the higher sensitivity of ZEP means
that it can be exposed with a much lower dose compared to PMMA, which could
significantly save the writing time and the cost for us.
Chapter 2 Passive Silicon Photonic Device Fabrication
32
The thickness of the resist had a significant impact to our fabrication. Without dilution,
the thinnest ZEP that we can spin-coat is about 350 nm with a spin speed of 5000 rpm.
This produced a significant challenge when fabricating the proposed adiabatic splitters.
The proposed splitter consists of three silicon tapers, as shown in Fig. 2.5(a), and the
gap between two adjacent tapers is as small as 50 nm. As a result, after ZEP resist
development [99% amyl acetate for 1 min followed by isopropyl alcohol (IPA)], the
unexposed resist strips located within the gaps [denoted by the shadowed regions in Fig.
2.5(a)] would have a high aspect ratio of 1:7 for its width to height. Such a challenging
aspect ratio could easily cause collapse of the resist strips, especially when the sample
was being blow-dried by nitrogen after IPA rinsing. The collapse of the resist strips
would lead to incomplete Cr mask, as they will cover parts of the exposed area, and the
Cr deposited onto these areas will be removed during the lift-off step. Consequently,
after deep RIE etching, some of the samples showed damaged blocks which were etched
down to the substrate. Fig. 2.5(c) shows one of the unsuccessfully fabricated splitters
caused by the collapse of the resist strips.
Chapter 2 Passive Silicon Photonic Device Fabrication
33
Figure 2.5. (a) The schematic of the proposed splitter structure to be fabricated. The
shadowed region in the structure indicates the place where ZEP resist exists after
development. (b, c) The scanning electron microscope (SEM) images of (b) a
successfully fabricated splitter, and (c) a damaged sample caused by the resist collapse.
To solve the problem, the aspect ratio of the resist strips needs to be reduced, and this
was realized using diluted ZEP (ZEP: Anisol 1:2) as the e-beam resist. The diluted ZEP
was spin-coated based on a two-step spin cycle. Firstly, the spin coater was ramped to
100 rpm with a 50 rpm/s acceleration and then held for 10 seconds. Secondly, the spin
coater was ramped to 1000 rpm with a 500 rpm/s acceleration and held for 42 seconds.
Finally, this would give a much thinner resist coating of about 100 nm. The film was
then pre-baked on a hotplate with a temperature of 180°C for 3 mins before the e-beam
exposure. Based on the diluted ZEP, the aspect ratio of the resist strips was reduced
significantly to 1:2, which successfully resolved the resist-collapse challenge.
Chapter 2 Passive Silicon Photonic Device Fabrication
34
2.2.1.2 PEC and dose test
To realize accurate size of the small features in our fabrication, proximity effect
correction (PEC) was applied to EBL, and it could help to avoid the potential problems
of pattern distortion and overexposure caused by the electron backscattering effect. The
proximity effect is normally characterized by a Point Spread Function (PSF) which
specifies the deposited energy as a function of distance from the incident beam, and it is
strongly dependent on the chosen acceleration voltage (100 kV) of the e-beam exposure
and the stack materials (ZEP on SOI in our case) and their thicknesses. In our
fabrication, the PSF was calculated based on a Monte Carlo simulation software
TRACER.
The dose test was implemented by exposing 200-nm-wide line and space arrays, and the
base dose was determined as the smallest dose that could expose the patterns with no
resist residues left. Based on this method, the base dose was tested to be around 180 to
200 µC/cm2, and therefore 190 µC/cm2 was chosen as the correct base dose in our
fabrication. With PEC applied, the width of the exposed 200-nm-wide line and space
array was measured to be about 10 to 20 nm wider than the designed width, as shown in
Fig. 2.6(a). Meanwhile, the exposed gap size of the proposed splitter was measured to
be about 10 nm larger than the designed value, as shown in Fig. 2.6(b). therefore, to
achieve correct device size, in our fabrication, a 5-nm size bias was applied in the
pattern layouts (10 nm smaller in total). It should be noted that, in reference [110], the
base dose for ZEP520A is also tested to be 190 µC/cm2, which is consistent with our
test results. But the ZEP520A is not diluted in [110], which suggests that the dilution of
ZEP doesn’t have a substantial impact on the base dose.
Chapter 2 Passive Silicon Photonic Device Fabrication
35
Figure 2.6. (a) The SEM image of the 200-nm line and space array exposed under the
dosage of 190 µC/cm2, the picture was captured after the step of resist development.
The grooves are the exposed area where ZEP was cleared after development. (b) The
SEM image of proposed splitter. Based on the SEM measurements, the exposed gap
size was about 10 nm larger than the designed size. Therefore a 5 nm bias was applied
in the pattern layouts to get correct size.
2.2.1.3 Silicon nano-etching
In our fabrication, the silicon device layer was etched through the pseudo Bosch silicon
etch recipe with a mixed SF6/C4F8 chemistry, and the etch was implemented based on
the Oxford Plasma Lab100 inductively coupled plasma reactive ion etching (ICPRIE)
system. Based on the pseudo Bosch principle, the silicon layer is etched by reactive ions
with the etched sidewalls protected by a generated polymer chain layer during the etch
process. As a result, an anisotropic etch with a vertical etching profile could be realized.
Specifically, the silicon layer is etched with fluorine ions and radicals which are
generated through injection and ionization of the SF6, in the meanwhile, a CF2 polymer
chain layer is generated with the injection and ionization of C4F8, and the generated
polymer is deposited on both the horizontal surface of the silicon layer and the sidewalls.
Under a radio frequency (RF) electric field, the silicon etching rate in the horizontal
surface could be faster than the polymer deposition rate, and as a result, the horizontal
surface of the silicon layer is continuously etched down with the sidewalls protected by
the deposited polymer. Compared to the conventional Bosch process for which the
Chapter 2 Passive Silicon Photonic Device Fabrication
36
etching and passivation are implemented cyclically, the continuous etching and
passivation in this pseudo Bosch process result in a much slower etching rate, on the
other hand, the inherent rippling effect on the sidewalls for the conventional Bosch
process is avoided. Based on the slow etching rate and the smooth sidewalls, the pseudo
Bosch recipe is suitable for realization of sub-micron structures.
Cr was selected as the material of the etching mask due to its strong etch resistance and
the ease of removal, and a thickness of 30 nm was applied through E-beam evaporation.
The selection of the etching mask material does have an influence on the silicon etch
result, but is without the scope of this thesis. Based on the ICPRIE system, the pseudo
Bosch etch result is mainly influenced by several factors, including the injected gas
ratio, the RF Forward power, the ICP power, and the chamber pressure. Each of these
parameters could affect the silicon etching result, and the final etching rate and quality
is a result of the combined effect of the selected parameters.
Without describing the working principle of the ICPRIE system in details (which can be
found in previous literatures), the general etching test results that we obtained are
briefly summarized here. In our tests, the chamber pressure of ICPRIE was fixed at 10
mTor. The square wafer piece was placed on a 4 inch carrier wafer with Fomblin oil
used as the adhesive. Firstly, it was found that inappropriate ratio of SF6/C4F8 could
result in non-vertical etching profiles. If the proportion of SF6 was too low, the
fabricated waveguide would show trapezoid profile, as shown in Fig. 2.7(a). The
trapezoid profile was mainly due to the insufficient fluorine ions which led to the
accumulation of polymer chains at the bottom of the waveguide. But if the proportion of
SF6 was higher than C4F8, the fabricated waveguide would show an interesting inverse
trapezoid profile with an undercut effect, as shown in Fig. 2.7(b). This could be
attributed to the fact that the polymer removal rate is faster that its re-deposition rate,
which resulted in the isotropic-like effect. Based on our tests, an injected gas ratio of
around 1:2 for SF6/C4F8 was proved to be able to provide nearly 90° sidewalls for the
etched waveguide.
Chapter 2 Passive Silicon Photonic Device Fabrication
37
Figure 2.7. Examples of fabricated waveguide arrays which show (a) trapezoid profiles,
and (b) inverse trapezoid profiles with under-cut effect.
The RF Forward power and the ICP power could affect the silicon etching rate.
Generally, these two parameters could influence the mechanical etching rate of the
chemical reactive ions and the number of ions that could be generated, respectively. The
silicon etching rate was found to be increasing with the Forward power while
decreasing with the ICP power. This is because the momentum of the fluorine ions
directed to the silicon substrate becomes higher with a larger Forward power, which
could increase the mechanical etching rate of the reactive ions. However, with a larger
ICP power, although the total number of the generated reactive ions could increase, the
amount of polymer chains increases more than fluorine ions due to the larger proportion
of C4F8, and this instead resulted in an overall slower etching rate. With a Forward
power of 15 W and an ICP power of 1200 W, the silicon etching rate was tested to be
around 160 nm/min. In our fabricated devices, a slight over-etch was carried out to
ensure that the silicon layer could be fully etched, thanks to the strong etch resistance of
the Cr mask. Figure. 2.8 shows the final results of the fabricated waveguide based on
the optimized etching recipe. As shown in the figure, relatively vertical and smooth
sidewalls were achieved for the etched silicon strip waveguides.
Chapter 2 Passive Silicon Photonic Device Fabrication
38
Figure 2.8. The SEM images of (a) the fabricated silicon strip waveguide and (b) a
bending waveguide sample with a bending radius of 10 µm.
2.2.1.4 Edge coupling
The edge coupling method was used for coupling between the fabricated silicon
photonic chip and the lensed single-mode fibre (LSF). As introduced in Chapter 1, when
compared with the grating coupler, the edge coupler has much wider operation
bandwidth, which is specially advantageous in our applications where both the proposed
adiabatic splitter and the circular Bragg grating mirror have the property of ultrabroad
bandwidth. Linear inverse tapers with a length of 200 µm and a tip width of 180 nm was
used as the spot size converter (SSC) to reduce the coupling loss caused by the mode
size mismatch between the strip waveguide and the LSF [111]. The LSF used has a spot
diameter of 2.5 µm. Figure. 2.9 shows the fabrication process to achieve the edge
coupling. In the process, the oxide cladding (both the upper cladding deposited by
PECVD and the buried oxide) was etched down to the silicon substrate by ICPRIE to
achieve a vertical and smooth coupling facet. A small gap of 4 µm was left between the
SSC tip and the etched oxide facet such that only one material (SiO2) would be etched
throughout the RIE process. This approach could avoid multi-step etching caused by the
stack of different materials (Si and SiO2) and protect the SSC tip to be distorted during
the facet etching process, therefore, it could help reduce the coupling loss and simplify
the process. In the meanwhile, the extra coupling loss caused by the introduction of the
small oxide gap proved to be negligible [92].
Chapter 2 Passive Silicon Photonic Device Fabrication
39
Figure 2.9. Fabrication process to achieve edge coupling.
Before implementing the SiO2 etch, SU-8 photoresist was spin-coated onto the wafer to
serve as the etching mask for both the SiO2 etch and the deep silicon substrate
etch/Bosch etch. The deep Bosch etch was implemented following the SiO2 etch to
allow fibre attach. The SU-8 model used was SU-8 2005. Before coating the SU-8, a
thin film of OmniCoat was spin-coated onto the SiO2 substrate to improve the SU-8
adhesion, otherwise, based on our tests, there would be a risk for SU-8 to peel off from
the SiO2 substrate during the development step. SU-8 2005 was spin-coated based on a
two-step spin cycle. In the first cycle, the spin coater was ramped to 500 rpm with a 100
rpm/s acceleration, in the second cycle, the spin coater was ramped to 1000 rpm with a
300 rpm/s acceleration and held for 30 seconds, and this would finally lead to a film
thickness of about 7 - 8 µm. After spin-coating, soft-bake was implemented to densify
the coated film. Firstly, the wafer was pre-baked on a hotplate under a temperature of
65°C for 1 min, and then the hotplate temperature was ramped to 95°C and the wafer
was soft-baked for 2 mins. The slower the temperature ramping, the better the coating
fidelity and the resist adhesion. After cooling down to the room temperature, the resist
was patterned with UV exposure (SUSS MA-6 Mask Aligner). Following the exposure,
a two-step post-expose-bake was applied. In the first step, the wafer was baked under
65°C for 1 min, and in the second step, the wafer was baked under 95°C for another 1
min. After cooling down, the wafer was developed under SU-8 developer for 1 min
followed by IPA rinsing and dried with nitrogen. Before being sent to the RIE etching,
Chapter 2 Passive Silicon Photonic Device Fabrication
40
the SU-8 resist was checked with an optical microscope to ensure that there was no
resist cracking or peel-off, especially at the film corners.
The oxide cladding was etched through ICPRIE with a C4F8/O2 chemistry, and the etch
rate was about 200 nm/min. The etch selectivity for SU-8 resist was about 1:1, therefore,
after the oxide etch, the thickness of the SU-8 left was about 3 - 4 µm. Following the
oxide etch, the standard Bosch process was carried out to achieve a deep substrate etch
so that the chip facet can be accessed by the LSF. The etching ratio of silicon to SU-8
resist was about 20:1 for the Bosch process, therefore, 3 - 4 µm SU-8 would be able to
etch 60 - 80 µm silicon substrate. For the sake of fibre alignment, the etch depth of the
substrate should be at least larger than half of the fibre diameter so that the bottom of
the fibre would not be contacted by the substrate. As the diameter of a stripped SMF 28
fibre is 125 µm, a 60 - 80 µm etch depth would be risky. Therefore, when the SU-8
resist was etched up, the Bosch process was continuously implemented with the SiO2
serving as the etching mask. The SiO2 has a much higher etch selectivity, and the
etching ratio of silicon to SiO2 is about 100:1. Therefore, a 0.5-µm oxide etch was
enough to offer a total substrate etch of more than 100 µm deep.
It should be noted that, during the Bosch process, if the entire wafer is not used, such as
in our case where square wafer pieces were used, it is important to provide proper
thermal conductivity for the wafer piece. In our case, Fomblin oil was used as the
adhesive as well as the thermal conductor between the wafer piece and the carrier wafer.
This is because the Bosch process relies on the combination of etching and passivation
steps to protect the sidewalls while etching downwards, and the passivation step can
only happen when there is proper cooling of the wafer substrate. Therefore, if there is
insufficient thermal conductivity provided between the wafer piece and the carrier wafer,
there will be improper passivation which will finally lead to the isotropic etching result.
Figure. 2.10(a) shows an example of the isotropic etching result which was obtained
when Fomblin oil was not used during the Bosch process. Figure. 2.10(b) shows a non-
isotropic etching result. As shown in the figure, the isotropic etch caused by the
improper thermal conductivity resulted in the non-vertical sidewall and the under-cut
effect.
Chapter 2 Passive Silicon Photonic Device Fabrication
41
Figure 2.10. Examples of (a) an isotropic Bosch etch when Fomblin oil was not used,
and (b) a non-isotropic Bosch etch with vertical sidewall when Fomblin oil was used.
Figure. 2.11(a) shows the final fabricated wafer after the Bosch process. In each
fabrication round, six chips were fabricated, and the chip size was designed to be 3mm
× 15mm. Figures. 2.11(b) and 2.11(c) show the optical microscope images of the
fabricated chip facet. As shown in the figure, a deep-etched smooth facet was achieved,
and the SSC tip was well protected within the oxide cladding. After going through a
Piranha cleaning, the wafer was cleaved to get chips which were ready for testing. The
<100> wafer could be cleaved manually with a scriber or through a dicing saw. If the
wafer is cleaved manually, the cleaved edge of the wafer should be kept sufficiently far
from the chip facet to ensure that the facet would not be damaged during the cleaving.
Chapter 2 Passive Silicon Photonic Device Fabrication
42
Figure 2.11. (a) The photo of the fabricated wafer after the Bosch process step. (b, c)
The optical microscope images of the fabricated chip facet.
2.2.2 Silicon Photonic Chip Measurement Setup
Figure 2.12(a) shows the experiment setup used to implement the edge coupling. Two
Thorlab’s 3-axis NanoMax Flexure Stages were used to realize sub-micron-precision
fibre alignment and coupling. The lensed fibres (input and output) were mounted onto
the stages through a fibre rotator, and the chip was placed on a mounting plate which
was fixed between the two stages. The propagation loss of the fabricated silicon strip
waveguide was measured based on the cut-back method. Waveguides with different
Chapter 2 Passive Silicon Photonic Device Fabrication
43
lengths were fabricated and measured. As shown in Fig. 2.12 (b), at 1550 nm, the
fabricated strip waveguide exhibited a propagation loss of around 2.425 dB/cm.
Figure 2.12. (a) The experiment setup for edge-coupling. (b) The extraction of linear
waveguide propagation loss at 1550 nm in TE polarization.
Chapter 3 Silicon Adiabatic Optical Power Splitter
44
CHAPTER 3 SILICON ADIABATIC
OPTICAL POWER SPLITTER
3.1 3-dB Silicon Optical Power Splitter
A 3-dB optical power splitter is an important optical device that has been widely used
for light power distributions and the construction of more complex optical devices, such
as optical switches, modulators, multiplexers, and optical phased arrays, etc. [112-115].
Normally, the most important figures of merit that are utilized to characterize a 3-dB
integrated optical power splitter include the excess loss (EL), the operation bandwidth,
the polarization sensitivity, and the device footprint. Based on the silicon photonics
integration platform, 3-dB power splitters are mainly realized through Y-branches,
directional couplers (DCs), and multi-mode interference (MMI) couplers [116-121].
However, these approaches are conventionally limited by a number of drawbacks which
makes it challenging for the splitters to satisfy all the above metrics simultaneously.
Specifically, a conventional Y-branch would suffer from relatively high excess loss
when the branching angle is not sufficiently small. The high excess loss is mainly
caused by the mode scattering from the non-adiabatic branch corner. To achieve an
adiabatic power splitting for a conventional Y-branch, a sufficiently small branching
angle and a long device length are needed. Unfortunately, the small branching angle
would lead to a high fabrication requirement and a high sensitivity to fabrication errors,
and the long device length would scarify the device’s competitiveness in building large-
scale and high-density PICs. On the other hand, the DC-based splitters can achieve a
low excess loss based on a relatively short device length, however, the 3-dB coupling
ratio of a conventional directional coupler is highly sensitive to wavelength and
coupling length, as a result, the operation bandwidth of the DC-based splitters is
relatively limited, and the splitters are also sensitive to the fabrication variations. To
avoid these drawbacks, the MMI-based splitters are popularly used, owing to their
relatively wide bandwidth and high fabrication tolerance. Nevertheless, MMI coupler is
still a wavelength-dependent device, as the effective width of an MMI coupler for self-
imaging is inherently sensitive to wavelength, and the low-loss bandwidth of an MMI
coupler is usually less than 100 nm.
Chapter 3 Silicon Adiabatic Optical Power Splitter
45
3.1.1 2×2 Silicon Adiabatic Coupler
Recently, to overcome the wavelength dependence issue associated with the traditional
DC and MMI-coupler based splitters, 2×2 adiabatic couplers with wide bandwidth and
high fabrication tolerance have been proposed [122-126]. Figure. 3.1 shows the
schematic diagrams of a conventional DC and a typical 2×2 3-dB adiabatic coupler. In
contrast to the conventional DC for which the fundamental odd and even modes are
both excited which then interfere with each other along the mode coupling region and
result in the power oscillation (coupling ratio) between the two coupled waveguides, in
the adiabatic coupler, only one mode (the fundamental even/odd mode) is excited and
propagates all the way along the coupler. During the mode evolution, the excited mode
is adiabatically (slowly) transferred to the even/odd mode of the coupled waveguides.
Consequently, the 3-dB power splitting ratio achieved at the end of the adiabatic region
is independent of the coupler length and wavelength, given the adiabatic condition is
satisfied. Moreover, as there is no power oscillation between the two coupled
waveguides, the 3-dB splitting ratio is less sensitive to fabrication variations. Normally,
the bandwidth of an adiabatic coupler is much wider than that of a conventional DC or
MMI coupler, which is up to hundreds of nanometres.
Figure 3.1. The schematic diagrams of (a) a conventional directional coupler and (b) a
2×2 3-dB adiabatic coupler.
Chapter 3 Silicon Adiabatic Optical Power Splitter
46
Despite the advantages of the wide bandwidth and the high fabrication tolerance, the
existing Si-integrated adiabatic couplers have two major drawbacks, which are the
relatively large device size and the polarization-sensitive property. Firstly, to satisfy the
adiabatic-transition condition, a sufficiently long mode evolution region is needed for
an adiabatic coupler. The coupler length that can satisfy the adiabatic evolution is
usually larger than 100 µm [122-126]. Secondly, the existing silicon adiabatic couplers
are normally polarization dependent. Due to the different coupling strength for the TE
and TM polarizations, the mode evolution length required for the adiabatic transition is
different for the two polarizations. Consequently, the adiabatic coupler designed for
TE/TM is not adiabatic for TM/TE. Although polarization dependence can be rendered
by using a longer coupler length, it will further scarify the device footprint. One way to
achieve a short polarization-independent adiabatic coupler length is balancing the
coupling strength of the two polarizations as demonstrated in [125]. Nevertheless, a
considerably long coupling length of 300 µm is still needed for the proposed
polarization-independent adiabatic coupler. The large device size and the high
polarization sensitivity could limit the practical applications of the existing silicon
adiabatic coupler, and to the best of our knowledge, an ultra-compact (sub-10 µm),
broadband (hundreds of nanometres), low-loss (<0.2 dB), and polarization-independent
Si-integrated adiabatic coupler has not been demonstrated until now.
3.1.2 1×2 Silicon Adiabatic Splitter
In addition to the 2×2 adiabatic couplers which are based on the tapering design of the
coupled waveguides, various 1×2 power splitters based on the tapered waveguides have
also been presented, and these splitters can be generally divided into two categories,
namely the adiabatic-taper and non-adiabatic-taper based splitters. Normally, for an
adiabatic-taper-based 1×2 power splitter, the adiabatic mode division is realized through
an adiabatic variation of the effective refractive index of the input and the output
waveguides, which can be illustrated in Fig. 3.2. In principle, the input waveguide is
supposed to have a gradually decreasing effective refractive index along the light
propagation direction, and the two output waveguides which are placed on two sides of
the input waveguide are supposed to have a gradually increasing effective refractive
index along the light propagation direction. Based on this adiabatic-index-tapering
design, the power of the fundamental mode of the input waveguide is gradually shifted
Chapter 3 Silicon Adiabatic Optical Power Splitter
47
towards the two output waveguides with low loss and is independent of wavelength. For
the previously presented adiabatic-taper-based splitters, the refractive index variation of
the output waveguides was usually realized through tapered waveguides, as illustrated
in Fig. 3.2. But for the input waveguide, the index variation was usually realized
through a complex structure which is hard to fabricate. For instance, in [127], the
linearly decreasing refractive index variation of the input waveguide (n2 as illustrated in
Fig. 3.2) was realized through etching the rib waveguide for a variable rib height, which
made the fabrication of the 3D structure considerably complex. Moreover, these
splitters were not based on the silicon integration platform, which normally required
considerably long tapers (hundreds of microns) to realize the adiabatic transition. To the
best of our knowledge, a compact and easy-to-fabricate adiabatic-taper-based 1×2
power splitter based on the silicon photonics integration platform has not been
demonstrated yet.
Figure 3.2. The principle illustration of a 1×2 3-dB adiabatic coupler based on the
effective-refractive-index tapering.
3.1.3 1×2 Silicon Non-adiabatic Splitter
In addition to the adiabatic-taper-based 1×2 splitters, non-adiabatic-taper-based 1×2
splitters have also been demonstrated. The antenna-coupled Y-branch splitters [128-130]
are a major representative of the non-adiabatic-taper-based splitters. Instead of being
based on an adiabatic power transition from the input waveguide to the output
waveguides, this type of splitters work in a non-adiabatic manner where the input light
is radiated from a truncated input waveguide and is re-launched into the two parallel
output waveguides through two tapers. The idea of the antenna-coupled Y-branch was
originally introduced in [130]. Generally, this type of splitters could be made much
shorter than the adiabatic splitters, but they suffered from a relatively high radiation loss.
To achieve a low radiation loss in the transition region, the refractive index near the
Chapter 3 Silicon Adiabatic Optical Power Splitter
48
branching point of the splitter could be designed to be smaller than the refractive index
of the claddings, which on the other hand could increase the fabrication complexity.
In conclusion, in terms of achieving a low excess loss and a wide operation bandwidth,
the adiabatic-type splitters are inherently advantageous compared with the non-adiabatic
splitters. However, both the existing 2×2 adiabatic couplers and the 1×2 adiabatic
splitters suffer from either a large device size or a complex structure which is difficult to
fabricate, and those splitters were polarization dependent. In this chapter, to overcome
these issues and realize an ultra-compact, broadband, low-loss, and polarization-
independent optical power splitter, we propose a novel 1×2 adiabatic-taper-based
splitter based on the 220 nm SOI platform.
3.2 1×2 Optical Power Splitter Based on Adiabatic Silicon Tapers
3.2.1 Device Design
The proposed new adiabatic optical splitter structure is shown in Fig. 3.3(a). As shown
in the figure, the proposed structure mainly consists of three regions, namely input
waveguide, adiabatic tapers (one input taper and two output tapers), and output bending
waveguides. A single-mode input strip waveguide with a dimension of W × 220 nm is
gradually tapered down (tip width T) along a length L, and the taper is symmetrically
inserted into two output tapers which have the same length (L), end width (W) and tip
width (T) as the input taper. The ends of the output tapers are connected with two
bending waveguides. The gap between two adjacent tapers is denoted as G.
Chapter 3 Silicon Adiabatic Optical Power Splitter
49
Figure 3.3. (a) The schematic diagrams of the proposed adiabatic splitter and the
intensity field distributions simulated along the device with (b) the TE and (c) the TM
polarization input at 1550 nm wavelength. The device parameters used in the
simulations are: W (400 nm), L (5 µm), G (50 nm), T (30 nm).
Different from a conventional 2×2 3-dB adiabatic coupler, for which the adiabatic
coupling occurs between two tapered waveguides, in the proposed splitter, light is
adiabatically coupled from one input waveguide to two symmetric output waveguides
through three tapers and, as long as the adiabatic mode evolution is satisfied for both TE
and TM polarizations, the 3-dB power splitting ratio can be achieved for both
polarizations at the end of the tapered region. The signal after passing through the tapers
Chapter 3 Silicon Adiabatic Optical Power Splitter
50
is then split into the fundamental modes through the bending waveguides. Moreover,
different from a conventional 1×2 3-dB adiabatic splitter as shown in Fig. 3.2, where the
adiabatic tapers were only used for the output waveguides of the splitter, and the
adiabatic index variation of the input waveguide was realized following a specific
function which makes the device hard to fabricate, in the proposed splitter, both the
input and output waveguides are realized through adiabatic tapers, and the three
identical tapers are placed in a complementary manner with tiny gaps to realize the
adiabatic mode transition. Based on the 220 nm SOI wafer, the proposed splitter only
requires a single step of etching for device patterning, which indicates a much simpler
device fabrication process. Figures. 3.3(b) and 3.3(c) show the simulated mode
evolution process for the TE and TM polarizations, respectively (in different color
scales). It can be seen that when the TE mode is launched, due to the electric-field
discontinuities at the tapered region, the light field is first enhanced and concentrated
within the tiny gaps and, along a certain propagation distance, light can be seen as
lossless eigenmodes propagating in a double-slot waveguide [131]. On the other hand,
for the TM mode, as the electric field is mainly concentrated at the top and bottom
interfaces of the waveguide, light field is not squeezed into the gaps throughout the
tapers. Consequently, in practice, the mode transition in TE polarization is more
sensitive to the sidewall roughness within the tapered region, which has to be taken into
account in device optimizations.
To achieve low loss and wide bandwidth for both TE and TM polarizations, the
adiabatic mode evolution needs to be maintained for both polarizations so that only the
fundamental mode is excited and propagates along the tapers. An eigenmode solver was
used for the mode analysis in the tapered region. The input and output waveguides
width was first selected at 460 nm, and L was selected at 5 µm. As shown in Fig. 3.4(e),
it is found that for either TE or TM polarization, two modes are supported along the
tapers (denoted as the fundamental and first-order modes). Furthermore, the effective
refractive index difference between these fundamental and first-order modes becomes
smaller in the taper region along the light propagation direction. The effective indices at
the taper end of the two modes in TE polarization are very close (∆Neff-TE = 0.05) while
the difference of the TM polarization is comparatively larger (∆Neff-TM = 0.2). The
corresponding mode profiles at the taper end cross section are shown in Figs. 3.4(a) –
3.4(d). Therefore, in practice, when the TE-polarized signal is utilized, given the
Chapter 3 Silicon Adiabatic Optical Power Splitter
51
fabrication imperfection, the adiabatic transition condition is difficult to maintain and
the higher-order mode can be excited, which can lead to mode interferences and cause
extra losses [132]. Therefore, the device needs to be optimized to suppress the higher-
order mode excitation, while still keeping the device compact.
Theoretically, it is easier to maintain the adiabatic transition with a longer taper length
(L), a smaller gap size (G) and a sharper tip width (T). However, considering fabrication
limitations, G and T are chosen at 50 and 30 nm, respectively, based on the typical E-
beam lithography (EBL) resolution. The impact of waveguide width (W) and taper
length (L) is then investigated and optimized based on 3D-FDTD simulations. Figure
3.4(f) shows the simulated ∆Neff at the taper end cross section when W is varied over a
range from 340 to 500 nm. It can be seen that ∆Neff decreases with W for both TE and
TM polarizations, while overall it remains relatively large for the TM polarization.
Since it is easier to maintain the adiabatic transition with a larger ∆Neff, W is chosen at
400 nm here by taking the propagation and bending losses into consideration
simultaneously. In addition to the larger ∆Neff, another advantage of choosing W = 400
nm is that the change of ∆Neff along the tapered region is also smoother, as shown in Fig.
3.4(g).
Chapter 3 Silicon Adiabatic Optical Power Splitter
52
Figure 3.4. (a) - (d) Mode profiles (1550 nm) simulated at the cross section of taper end
with W = 460 nm. (e) Effective index of the four modes along the tapers with a length of
5 µm. (f) Simulated ∆Neff at the taper end cross section with the change of W from 340
to 500 nm. (g) Change of ∆Neff along the tapers for different W and L. (h) Simulated
excess loss at 1550 nm with the change of L.
With W fixed at 400 nm, the device performance in terms of EL at 1550 nm is then
investigated by changing the taper length (L) from 1 to 7 µm, and the results are shown
in Fig. 3.4(h). Note that in the simulation, the impact of sidewall roughness was not
included in EL. The excess loss here is defined by the following equation:
Chapter 3 Silicon Adiabatic Optical Power Splitter
53
Excess Loss (dB) = 1 210*lgin
P P
P
(3.1)
Where P1 and P2 are the output power of the splitter, and Pin is the input power. In
general, longer tapers are preferred to maintain the adiabatic transition. It is clear that
for L ˃ 3 µm, the EL for the TE polarization becomes negligible, indicating that no
higher-order mode is excited, while for the TM polarization, due to the larger ∆Neff, the
EL remains low, even for L < 3 µm. Considering compact size, as well as the adiabatic
requirements for both polarizations, a conservative value of 5 µm is chosen for L. As
shown in Fig. 3.4(g), when L = 5 µm, ∆Neff also changes more smoothly than the case
when L = 3 µm. Compared with the previous polarization-independent 3-dB adiabatic
couplers [125], the weak mode confinement at the input taper, as well as the small gap
size between tapers, leads to much stronger coupling strength for both TE and TM
polarizations, which leads to a significantly shorter (by two orders of magnitude) taper
length which can satisfy the adiabatic evolution for both polarizations simultaneously.
Based on the optimized parameters, the simulated transmission spectra of the splitter are
shown in Fig. 3.5(a). In the simulation, 90° bending waveguides were used as the output
branches of the splitter, and the bending radius was set to a relatively large value of 15
µm to avoid substantial bending losses. As can be seen, the splitter achieves an
extremely high and flat transmission efficiency of nearly 50% for each output port over
the whole simulated wavelength range. For the TM polarization, the spectrum gradually
drops for wavelengths > 1550 nm, and this is due to the gradually increasing bending
loss within the bending waveguide. To confirm this, we simulated the spectrum of the
bending waveguide used in the splitter, and the results are shown in Fig. 3.5(b) below.
As can be seen, for the TM polarization, the spectrum of the bending waveguide
matches with the spectrum of the splitter, which indicates that the loss within the
adiabatic tapered region is negligible. To improve the efficiency for wavelengths > 1550
nm, wider waveguide or larger bending radius can be utilised. However, as explained
previously, 400 nm waveguide width is the optimum choice for the proposed device,
and a larger bending radius also leads to less compact device. In addition, although the
efficiency gradually drops after 1550 nm, the wavelength range with high efficiency
(above 49%) is still very wide for the TM polarization (1200 – 1625 nm).
Chapter 3 Silicon Adiabatic Optical Power Splitter
54
Figure 3.5. (a) The simulated transmission spectrum (output 1 and output 2 have
identical spectrum) of the proposed splitter and (b) a 90° bending waveguide with a
cross-section size of 400 nm × 220 nm and the bending radius of 15 µm.
The splitter’s tolerances to the variations of the non-zero taper tip width (T) and the
small gap size (G) were also investigated. Firstly, T was changed from 0 to 80 nm,
while other parameters were kept at the optimized values. As shown in Fig. 3.6(a), the
transmission efficiency degradation is negligible (<0.3%) for both polarizations. The
impact on ∆Neff is not substantial either, with ∆Neff-TE dropping from 0.12 to 0.07 and
∆Neff-TM dropping slightly from 0.25 to 0.23. Therefore, the splitter performance has
relatively high tolerance to the taper tip width variation. Then G was increased from 50
nm to a relatively large value of 130 nm, while other parameters remained unchanged.
As shown in Fig. 3.6(b), while the transmission ratio of the TE mode slightly decreases
by around 0.6%, the transmission ratio of the TM mode remains stable. However, ∆Neff
reduces sharply for both polarizations, with ∆Neff-TM dropping from 0.25 to 0.12 and
∆Neff-TE dropping from 0.1 to only 0.02. As ∆Neff-TM remains high (>0.12), even with
large gaps, it is still difficult to excite the higher-order TM modes. For the TE
Chapter 3 Silicon Adiabatic Optical Power Splitter
55
polarization, as ∆Neff rapidly drops to <0.05 when G is larger than 70 nm, the adiabatic
evolution is difficult to maintain in practice. Therefore, the proposed splitter has higher
fabrication tolerance in TM polarization, while relatively precise fabrication is needed
for the TE polarization.
Figure 3.6. The change of transmission efficiency and ∆Neff with (a) T and with (b) G at
1550 nm.
3.2.2 Characterization
The proposed splitter was fabricated based on the fabrication process presented in Fig.
2.3, and the SEM images of the fabricated splitter are shown in Figs. 3.7(a) and 3.7(b).
Chapter 3 Silicon Adiabatic Optical Power Splitter
56
Based on the SEM measurements, the fabrication variation on T and G was about ±5 nm.
The device was measured based on the measurement setup illustrated in Fig. 3.7(c). An
ASE source (THORLABS ASE730) with an operation wavelength of 1530 to 1610 nm
was used as the wideband light source, and the output passed through an in-line
polarizer and a polarization controller (PC) to generate the single-polarized light. The
edge-coupling approach illustrated in Fig. 2.11 was utilized for the coupling between
the fabricated silicon chip and the LSF. The output spectrum of the splitter was
monitored by an optical spectrum analyzer (OSA). Because the operation wavelength of
the polarizer used was limited to 1600 nm, only the spectrum within the range of 1530
to 1600 nm was extracted for device performance characterizations.
Figure 3.7. (a, b) The SEM images of the fabricated adiabatic splitter. (c) The schematic
of the device measurement setup.
To characterize the splitter performance more reliably, three stages of 1×2 splitters were
cascaded to form a 1×8 splitter, as illustrated in Fig. 3.8(a). The average EL of a single
element was then estimated based on the following equation [133]:
Chapter 3 Silicon Adiabatic Optical Power Splitter
57
0( ) ( ) / 3 3wgEL dB P P (3.2)
Where P0 is the average output power of all output ports and Pwg is the measured output
power of the reference strip waveguide. The factors 1/3 and 3 correspond to the number
of cascaded stages and the 50% splitting ratio, respectively.
Figure 3.8. The optical microscope images of (a) the fabricated 1×8 splitter and (b) the
cascaded splitter.
The device measurement results are shown in Fig. 3.9.
Chapter 3 Silicon Adiabatic Optical Power Splitter
58
Figure 3.9. Experimentally measured spectra of the splitter for (a) the TE and (b) the
TM polarization.
Figures. 3.9(a) and 3.8(b) show the measured excess loss of the splitter for TE and TM
polarizations, respectively. The insertion loss of three random ports [as indicated in Fig.
3.8(a)] were also extracted to show the uniformity of the 1×8 splitter. It can be seen that
a flat spectrum and low loss profile can be achieved for both polarizations over the
entire measured wavelength range. Due to the higher sensitivity to sidewall roughness,
Chapter 3 Silicon Adiabatic Optical Power Splitter
59
the performance of TE-polarized light has larger fluctuations across the operation
bandwidth. For better performance characterizations, regression lines were extracted
based on the robust locally weighted regression method [134]. It is clear that the excess
loss is lower than 0.19 dB over the whole measured wavelength range for the TE
polarization and better than 0.14 dB for the TM polarization. Compared with simulation
results, the measurement results agree well for the TM polarization, while they are
slightly worse for the TE polarization. This is mainly due to the extra scattering losses
caused by sidewall roughness of the tiny gaps between tapers where the TE field is
mainly concentrated. This can be further improved by optimizing the fabrication
processes (e.g., oxidation) to have smoother sidewalls.
More accurate results on the EL at 1550 nm were extracted by linearly fitting the
measured output power after passing through different numbers of splitters, as shown in
Fig. 3.8(b), and the results are shown in Fig. 3.10. At 1550 nm, an ultra-low EL of 0.11
and 0.08 dB can be achieved for the TE and the TM polarizations, respectively, and the
results also agree well with the measured wideband spectrum shown in Fig. 3.9.
Figure 3.10. The extracted loss at 1550 nm for (a) the TE polarization and (b) the TM
polarization. R square is the calculated coefficient of determination.
The power uniformity of the splitter was also estimated, and it was defined as the
difference between the maximum output power and the minimum output power across
all output ports of the splitter. The power uniformity increases with the number of
cascade stages. For the cascaded 1×8 splitter, as shown by the insertion loss spectra in
Figs. 3.9(a) and 3.9(b), uniform outputs are achieved over a wide wavelength range. At
Chapter 3 Silicon Adiabatic Optical Power Splitter
60
1550 nm, the power uniformity across all 8 output ports was measured to be only 0.47
dB and 0.17 dB for the TE and TM polarizations, respectively. For a single 1×2 splitter,
the best measured uniformity was only 0.07 and 0.02 dB, respectively.
3.2.3 Splitter with Various Power Splitting Ratio
For the proposed optical splitter, it is possible to realize various power splitting ratios.
For the TE polarization, due to the strong confinement of light within the tiny gaps, the
coupling ratio is very sensitive to the gap size variation. Additional FDTD simulations
have been carried out. As shown in Fig. 3.11 below, when one gap (gap 1) was fixed at
50 nm while the other gap (gap 2) was varied from 50 nm to 130 nm, a splitting ratio of
about 0.833 : 0.159 can be realized. On the other hand, for the TM polarization, as the
light field is mainly distributed at the top and bottom interfaces of the tapers, the power
splitting ratio is more sensitive to the shape of the entire tapered region rather than small
variations of the gap size. FDTD simulation results of the TM polarization are shown in
Fig. 3.11 as well. It can be seen that with small gap size variation, the power splitting
ratio can be altered. However, the variation is much smaller than the scenario with TE
polarized signals. All the results presented here are only based on simulations which
have not been confirmed by experiments (additional chips need to be fabricated).
Figure 3.11. The simulated power splitting ratio as a function of gap_2 when gap_1 is
fixed at 50 nm.
Chapter 3 Silicon Adiabatic Optical Power Splitter
61
3.2.4 Conclusion
In conclusion, in this chapter, an ultra-broadband, compact, low-loss, and polarization
insensitive adiabatic 3-dB optical power splitter has been proposed and experimentally
demonstrated. The proposed splitter has been shown to be only 5 µm in length, and the
EL is <0.19 dB for the TE polarization and <0.14 dB for the TM polarization over the
entire measured band of 1530 to 1600 nm. The relatively small taper tip width and the
gap size would make the splitter hard to realize through today’s standard CMOS process.
However, with the rapid development of the fabrication technology, a much smaller
feature size can be realized, such as using the emerging extreme-UV lithography, and
the proposed device is expected to have a strong potential in applications that need
ultra-high-efficiency optical power distributions and broadband operations. For the
proposed splitter, it is possible to realize various splitting ratios by using asymmetric
gap sizes between tapers, which will be investigated in future research.
Chapter 4 Circular Bragg Grating Mirror
62
CHAPTER 4 CIRCULAR BRAGG
GRATING MIRROR
4.1 High-index-contrast Bragg Grating on SOI
Bragg gratings are important optical components which are widely used as the optical
mirrors to reflect light within a specific band of wavelengths. The optical reflection of a
Bragg grating is realized based on the principle of Bragg condition which is satisfied
through a periodic variation of the effective refractive index of the waveguide. In
integrated photonics and silicon photonics, Bragg gratings are usually formed in
waveguides with a periodic structural corrugation to realize the periodic effective index
modulation. Generally, based on the index modulation strength, the waveguide Bragg
gratings can be categorized into two groups, namely the low-index-contrast Bragg
gratings (LBGs) and the high-index-contrast Bragg gratings (HBGs).
The LBGs normally have a weak grating index contrast, and the weak index contrast is
usually formed through shallowly- (in vertical directional) or partially-etched (in
horizontal direction) waveguide corrugations. Due to the weak grating strength, the
LBGs normally require a large number of grating periods to achieve high reflectivity
(hundreds of periods are usually needed), and the reflection bandwidth of an LBG is
usually very narrow (e.g., within one nanometer) [135, 136]. On the other hand, the
HBGs are usually formed through deeply- or fully-etched waveguide corrugations to
realize a strong grating index contrast. Attributed to the strong grating strength, the
HBGs can have a much broader reflection bandwidth of up to hundreds of nanometers
based on only a few grating periods, therefore, they have been widely utilized as
integrated broadband mirrors to provide broad and flat reflection spectra in a wide
variety of applications, such as filters, resonators and semiconductor lasers [137-142].
Typically, based on the silicon integration platform, the HBGs are usually built using a
relatively thick silicon layer (normally >1 μm thickness) with deeply- or fully-etched
grating trenches to achieve the high index contrast [141, 142]. However, this approach
becomes problematic when implemented on the popular silicon photonics integration
platform based on SOI wafers with a thin silicon layer of 220 nm, where waveguides
typically with a dimension of 460 × 220 nm for single-mode transmission are utilized.
Chapter 4 Circular Bragg Grating Mirror
63
In this 220 nm SOI platform, the grating index contrast becomes very limited due to the
thin silicon device layer, even with deeply- or fully-etched grating trenches. Moreover,
if a conventional deeply- or fully-etched Bragg grating structure is implemented based
on the sub-wavelength waveguide in this 220 nm SOI platform, the grating would suffer
from high mode diffraction loss due to the small size of the waveguide. As a result, it is
challenging to realize both broad bandwidth and high reflectivity simultaneously for a
Bragg grating based on this popular thin SOI platform, and to the best of our knowledge,
no previous work has attempted to tackle this issue.
4.2 Circular Bragg Grating Based on 220 nm SOI
To overcome this difficulty, in this chapter, we propose a novel circular Bragg grating
mirror based on the 220 nm SOI platform which could achieve the properties of high
reflectivity, broad bandwidth and compact device size simultaneously. In contrast to the
conventional HBGs which have rectangular-shape grating teeth/trenches, in the
proposed circular Bragg grating mirror, the grating blades (teeth) are bent around the
strip waveguide end to form a circular shape such that they could match with the
diffracted circular waves from the strip waveguide end. Consequently, the circular
waves diffracted from the strip waveguide end are recollected back to the waveguide,
which avoids the high diffraction loss of the conventional HBGs based on the thin-Si-
layer SOI platform. Moreover, although the grating is still made on the thin silicon
device layer, the beam divergence in the lateral direction leads to a high effective index
for the circular grating blades, and a high index contrast can be achieved for the grating.
Circular Bragg gratings with two different structures have been proposed, as shown in
Fig. 4.1. The initially proposed circular Bragg grating structure is shown in Fig. 4.1(a),
and its detailed device optimization process is presented in [143]. Based on the
optimized parameters, the proposed grating mirror shown in Fig. 4.1(a) could achieve a
reflectivity of >90% over a wide bandwidth of 385 nm, based on a compact footprint of
only 4.03 µm × 4.32 µm. To the best of our knowledge, it is the first broadband and
high-reflectivity Bragg grating mirror demonstrated on the 220 nm SOI platform.
Unfortunately, the initially proposed circular Bragg grating mirror has very stringent
fabrication requirements, which limit its practical applications. For instance, to match
with the circular wave diffraction from the strip waveguide end, the circular grating
Chapter 4 Circular Bragg Grating Mirror
64
blades were designed to be centered at the end of the waveguide with inner-radiuses (r)
selected as:
( 1)p tr w p (4.1)
Where wt is the grating trench width (170 nm), Λ is the grating period (360 nm), and p is
the blade number. Consequently, circular blades with very small radiuses are needed,
which is difficult to realize based on standard fabrication lines. Particularly, from the
equation above, the first blade has a critical radius (r1) of only 170 nm, which
significantly increases the fabrication challenge, and in practice, extremely high-
resolution lithography (a sub-10-nm EBL was used in [143]) is required to realize a
smooth circle with such a tiny radius. Furthermore, a small gap is needed between the
waveguide end and the first blade in this design, and the gap needs to be ≤30 nm to
ensure satisfying grating performance, which further increases fabrication challenges.
Figure 4.1. The schematic diagrams of the proposed circular Bragg grating with two
different structures (size scaled).
Chapter 4 Circular Bragg Grating Mirror
65
To resolve these fabrication issues, we propose another circular Bragg grating mirror
which is shown in Fig. 4.1(b). The new design has significantly relaxed fabrication
requirements and a better grating performance in terms of a much broader high-
reflectivity (>95%) grating bandwidth. The minimum radius of the grating blade in this
design can be increased by more than tenfold to >2 µm, while overall the circular
grating still has a very compact size of only 4.49 μm × 4.54 μm. In addition, the
minimum device feature size is increased substantially to 100 nm, which allows much
larger fabrication tolerance and greatly relieves fabrication challenges. The proposed
circular grating is studied via numerical simulations, and results show an ultrabroad
operation band of 500 nm with a reflectivity of >90%, and a broad 397-nm band with a
high reflectivity of >95% covering the entire E- and U-bands. In this chapter, the
detailed optimization process and the experimental characterizations of the proposed
new circular Bragg grating shown in Fig. 4.1(b) will be presented.
4.2.1 Device Design
As shown in Fig. 4.1(b), the proposed new circular Bragg grating mainly consists of
three parts: a single-mode strip waveguide (460 × 220 nm), a pie-shaped taper which is
spread from the end of the waveguide, and several circular blades following the taper
with the same arc angle to serve as the grating teeth. The arc angle of the taper and the
blades is denoted as θ, and the length of the taper is denoted as R. To maximize the
grating bandwidth, the grating period is selected using the first order reflection together
with a 50% duty cycle (defined as the ratio of the blade width to the grating period),
which is capable of providing the maximized grating strength [144, 145]. The Bragg
condition can be expressed as follows:
0
2t t b bnW n W
(4.2)
Where Wt and Wb are the width of the grating trenches and the blades, respectively, nt
and nb are the effective refractive index for the grating trenches and the blades,
respectively, and λ0 is the Bragg wavelength. In the proposed new circular grating
design, although the grating blades are still centered at the waveguide end similar to the
design shown in Fig. 4.1(a), the insertion of the taper significantly expands their
radiuses, now the minimum radius of the circular structure is increased to >R (the center
Chapter 4 Circular Bragg Grating Mirror
66
of the taper is slightly overlapped with the waveguide), which greatly relieves
fabrication challenges associated with the tiny radius in the previous design. In addition,
as the waveguide is gradually spread to the circular shape which is directly aligned with
the following grating blades, a small gap (wg ≤ 30 nm) between the waveguide end and
the first grating blade is no longer required, and it further reduces fabrication challenges.
Figure 4.2. The field distribution (1550 nm) within the taper simulated with an arc angle
of (a) 270°, (b) 180°, and (c) 60°. Note that the above field distributions were captured
to illustrate the wave diffraction from the strip waveguide to the taper, and no grating
blades were set in the simulation. The reflection observed was due to the refractive
index contrast between the taper and the cladding, rather than the Bragg reflection. The
red circles in (a) and (b) indicate the branches formed in the 270° taper and the
effective-index-contrast interface formed in the 180° taper, respectively. The dashed
line in (b) shows the 60° boundary. The tapers were all covered by 2 μm SiO2.
In the new device design, the pie-shaped taper is intentionally used to create a circular
wave diffraction to match with the subsequent circular grating blades. In this case, the
taper functions like a free propagation region (FPR), and the wave diffraction through
the strip waveguide to the taper is similar to a Huygens-Fresnel diffraction, according to
which the transversely confined light transmitting to an aperture (a) which is small
enough relative to its wavelength (a ≤ ) will be diffracted to circular waves [146]. The
taper angle is a critical parameter which could influence the shape of the diffracted
wavefront. Figure. 4.2 shows the simulated diffraction field distribution within the taper
when it has different arc angles. In principle, to perform like an FPR, θ should be
chosen as ≥180° so that the diffracted wavefront would not be obstructed along the y-
Chapter 4 Circular Bragg Grating Mirror
67
direction. Figure. 4.2(a) shows the simulated field distribution when the taper has an arc
angle of 270°. In this case, although circular waves can be formed within the taper, the
reflected mode transmitting back towards the waveguide will be split by the sharp V-
shaped branching corners formed by the non-vertical taper boundary and the waveguide,
and results in the additional propagation routes inside the taper, as shown in the figure.
These unwanted fields will experience multiple resonance routes inside the grating,
which could seriously interfere with the intended Bragg reflection and destruct the
Bragg spectrum. In the case of θ = 180, as shown in Fig. 4.2(b), the sharp branches are
avoided, and perfect circular waves can be observed in the taper. However, the effective
refractive index contrast at the interface of the waveguide and the taper is relatively
large (2.37 and 2.85 respectively, calculated by Lumerical Mode Solutions), which
could result a considerable amount of power to be reflected again at this interface, and
the disturbing resonant effect can be generated. To avoid this problem, we set θ to a
smaller value of 60°, and the simulated light field distribution within the taper is shown
in Fig. 4.2(c). The taper here can still cover the dominant angular beam divergence
observed in Fig. 4.2(b), while in the meantime, the index contrast at the interface is
reduced (2.37 and 2.47 for the waveguide and near the taper end, respectively). Such
design also enables a much smaller device size.
Another important design parameter in the proposed circular Bragg grating is the taper
length R, which needs to be ≫λ to create circular wavefronts [146, 147] (λ here is the
wavelength inside the silicon taper). Due to the sub-wavelength size of the silicon strip
waveguide and the high material refractive index, circular waves can be obtained even
using a short taper, and here a relatively small value of R = 2 μm was selected to
achieve a compact device size. It should be noted that, theoretically, the longer the wave
evolution length, the closer for the diffracted waves to become circular [146]. Therefore,
a longer taper provides a better match between the diffracted wavefront and the circular
blades, at the cost of a larger device size. Compared to the design in Fig. 4.1(a), the
circular grating proposed in this paper enables a much longer wave evolution path
within the taper. On the contrary, for the design in Fig. 4.1(a), a relatively large gap
between the waveguide and blades can lead to high diffraction loss.
Chapter 4 Circular Bragg Grating Mirror
68
Figure 4.3. The simulated reflection spectra of the proposed Bragg grating with
different values chosen for W1, Wt, and Wb.
The performance of proposed circular grating was investigated based on the three-
dimensional finite difference time domain (3D-FDTD) simulations. In the simulation,
𝑊𝑡 and 𝑊𝑏 were set to 181 nm based on the following parameters: λ0 = 1.55 μm, nt =
1.44, nb = 2.85. Here, nt was selected as the refractive index of SiO2, and nb was
approximated as the effective index of a 220 nm thick slab silicon waveguide as it is
difficult to calculate the accurate effective index of a circular-shaped blade. The number
of grating blades was set to 6. The simulated reflection spectrum is shown by the red
dotted line in Fig. 4.3. As can be seen, the grating has an ultrabroad spectrum ranging
approximately from 1260 to 1760 nm benefited from its strong index contrast. However,
the reflection has a gradually declining tendency towards the shorter wavelengths,
resulting in a non-flat top for the spectrum. This is mainly caused by the extra mode
diffraction loss in the vertical direction inside grating trenches, due to the strong vertical
confinement within the thin silicon layer. This effect is shown in Figs. 4.4(a) and 4.4(b),
where the vertical diffraction is clearly stronger for smaller wavelength, due to its
tighter confinement within the taper (equivalently, a larger neff). Therefore, the vertical
diffraction loss needs to be reduced to achieve a flat-top reflection spectrum.
Theoretically, the diffraction loss in a fully-etched waveguide grating can be reduced
with narrower grating trenches [148, 149]. The yellow dashed line in Fig. 4.3 shows the
reflection spectrum when Wt was reduced to 100 nm. The trench width was not further
reduced considering the fabrication difficulty (correspondingly, Wb was increased to
221 nm based on Equation (1) to maintain the Bragg condition). As can be seen, the
Chapter 4 Circular Bragg Grating Mirror
69
back-reflection variation becomes smaller within the operation bandwidth and a flat-top
spectrum is realized. However, such an approach of changing the duty cycle also results
in weaker grating strength, which leads to a much narrower reflection bandwidth.
In order to achieve both flat spectrum with high reflectivity and large device bandwidth
simultaneously, instead of changing the grating duty cycle, here we propose a new
solution of only reducing the width of the first trench, which is denoted as W1 in Fig.
4.1(b). The field distributions with different W1 values are simulated and results are
shown in Fig. 4.4. As shown in Fig. 4.4(a), due to the strong index contrast of the
circular grating, light is reflected within only a few grating periods, and the mode
intensity decreases sharply along the propagation direction due to the strong reflection
at each blade. Consequently, since the mode intensity is strongest within the taper, the
diffraction occurring at the first trench is much stronger than others. Therefore, the
diffraction loss can be suppressed considerably as long as W1 is reduced. Figure. 4.4(c)
shows the field distribution when W1 is reduced to 100 nm for the wavelength of 1.3 m.
It is clear that the diffraction at the first trench has been significantly weakened, and the
diffraction at other trenches still remains weak. In addition, as the duty cycle of most
periods is unchanged, the reduction of W1 imposes a negligible impact on the grating
strength. Figure. 4.4(e) shows the circular grating reflection spectrum for different
values of W1 ranging from 181 to 100 nm. It can be seen that the grating reflectivity is
gradually improved with decreasing W1. It is also clear from the figure that the
reflectivity improvement gradually becomes insignificant for longer wavelengths. This
is mainly because the mode diffraction becomes weaker, as shown by Figs. 4.4(b) and
4.4(d). It is clear from these figures that the reduction of W1 from 181 to 100 nm has a
negligible impact on the diffraction at the wavelength of 1.75 µm.
With W1 chosen at 100 nm, the simulated circular grating reflection spectrum is shown
by the blue solid line in Fig. 4.3, which presents a flat reflectivity of >90% over an
ultra-broad band of 500 nm (1263 - 1763 nm) with a high / of 32.3%, and a high
reflectivity of >95% over a broadband of 397 nm (1340 - 1737 nm) with a / of 25.6%
which covers the entire E- to U- bands. At 1550 nm, the reflectivity is 97.6%.
Compared to the previous design which has a high-efficiency (>95%) bandwidth of
only 171 nm (1410 - 1581 nm), the high-efficiency band of the proposed new circular
grating has been improved significantly. In addition, as shown by Fig. 4.4(e), even for
W1 = 120 nm, the grating still shows a comparably good performance, which has a 498-
Chapter 4 Circular Bragg Grating Mirror
70
nm band with a reflectivity of >90% and a 298-nm band with a high reflectivity
of >95%, indicating an even larger fabrication tolerance.
Figure 4.4. (a) – (d): The vertical field-intensity distribution of the grating captured at
the y = 0 plane with λ set at 1.3 and 1.75 µm and W1 chosen at 100 and 181 nm. (e, f, h):
The grating spectrum simulated with different values of W1, θ, and N. (g): The change of
the reflectivity (1550 nm) and ∆λ (reflectivity >90%) with the grating blade number N.
Chapter 4 Circular Bragg Grating Mirror
71
The impact of grating angle θ on the device performance was investigated as well, and
simulation results are shown in Fig. 4.4(f). When θ is relatively small, increasing θ
expands the beam divergence, which results in a larger 𝑛eff and more-circular wavefront.
Therefore, the wavefront matches better with the following circular blades, and both
reflectivity and bandwidth can be improved, as shown by the simulations results of θ =
20o and 60o. When θ continues to increase, no notable improvement on the grating
performance can be observed, as both neff and the wavefront shape remain relatively
stable (the main lobe is already fully captured). Therefore, as shown in Fig. 4.4(f), with
θ further increased to 180°, the spectrum is only slightly wider than the spectrum with θ
= 60°. On the other hand, for θ = 180°, a periodic fluctuation in the spectrum can be
seen, which is mainly due to the mode resonance inside the taper, as explained
previously. This problem can be further confirmed by calculating the corresponding
resonance cavity length (from the resonance peak spacing), and it is about 2.56 μm
[150]. The calculated cavity length is slightly larger than the taper length, and this is due
to the mode penetration into the grating, as can be illustrated in Fig. 4.4. When θ is
further increased to 270° (the case shown in Fig. 1(c)), the Bragg reflection is severely
destructed due to the strong disturbing fields inside the taper, and the grating reflection
becomes unusable. Therefore, θ = 60° was chosen as the optimal angle in our design to
achieve both a wide and flat reflection spectrum and a compact device size.
The impact of the number of grating blades (N) on the reflectivity and
(reflectivity >90%) was also investigated, and simulation results are shown in Fig.
4.4(g). It is clear that both and reflectivity improve sharply with a larger N until N
reaches 6. For N > 6, the reflectivity and the bandwidth remain almost unchanged,
whilst the spectrum shows sharper edges, as shown in Fig. 4.4(h). To summarize, based
on the device optimization, the main design parameters of the proposed circular grating
are chosen to be: θ = 60°, R = 2 µm, 𝑊1 = 100 nm, and N = 6 to ensure that a high-
efficiency and broadband spectrum with a flat top can be achieved based on a very
compact device footprint.
4.2.2 Characterization
The fabrication and measurement of the proposed circular Bragg grating was realized
based on the fabrication process shown in Fig. 2.4 and Fig. 2.9. The inset of Fig. 4.5
shows the SEM image of the fabricated device in top-view and the schematic of the
Chapter 4 Circular Bragg Grating Mirror
72
measurement setup. Figure. 4.6(a) shows the tilted view of the fabricated grating. It can
be seen from Fig. 4.6(a) that vertical sidewalls were achieved for the fabricated grating
blades. To measure the fabricated grating mirror, as illustrated in Fig. 4.5, a broadband
light source (THORLABS ASE730) with an operation wavelength of 1530 to 1610 nm
followed by an in-line polarizer and a polarization controller (PC) was used to provide
the TE-polarized light. The input light passed through a circulator (Port 1→2) before
being injected to the chip. After being back-reflected by the circular grating, the signal
was routed by the circulator (Port 2→3) to an optical spectrum analyzer (OSA) for
spectrum characterizations (shown by Setup 1). To measure the insertion loss (IL) of the
grating, a reference strip waveguide was measured with the setup indicated by the
dotted connection (Setup 2) in Fig. 4.5, and the length of the reference waveguide was
chosen as twice the length of the waveguide connecting the circular grating to the spot
size converter, considering the round-trip propagation of the light reflected by the
circular grating. It should be noted that, during the input light injection in Setup 1, there
were chip-facet reflections which could also be collected by Port 3 of the circulator.
This facet-reflected power was measured at Port 3 of the circulator using Setup 2, and
the measured facet reflection is shown in Fig. 4.6(b). Although this reflection was
relatively weak, it was measured for more reliable performance characterization of the
circular grating. Furthermore, the IL of the circulator for Port 2→3 was also measured
and compensated to the device characterization results.
Figure 4.5. The schematic of the measurement setup and the SEM images of the
fabricated grating mirror.
Chapter 4 Circular Bragg Grating Mirror
73
Figure 4.6. (a): The tilted view of the fabricated grating mirror. (b): The measured
spectra of the fabricated grating, the reference strip waveguide, and the chip-facet
reflections. (c): The measured IL of the fabricated grating.
Chapter 4 Circular Bragg Grating Mirror
74
The device measurement results are shown in Fig. 4.6(b). It can be seen that the
measured reflection spectrum of the grating is similar with the transmission spectrum of
the reference waveguide, indicating the high reflection efficiency of the device. The
measured IL of the grating is shown in Fig. 4.6(c), which presents fluctuations that are
mainly caused by the unoptimized device sidewall roughness, and at some points it is
even above 0 dB. For reliable characterization, a regression line was calculated based on
the robust locally weighted regression method [134]. Based on the regression line
results, the grating exhibits a low IL of -0.32 to -0.1 dB (corresponding to a reflectivity
of ~ 93% to 98%), which generally matches well with the simulation results.
4.2.3 Integrated Notch Filter Based on Circular Bragg Grating Mirror
Based on the proposed circular grating mirror, we also designed and fabricated a notch
filter [151]. The filter consisted of a resonant cavity formed by two circular grating
mirrors connected by a strip waveguide. A bending waveguide coupler with a bending
radius of rb and a coupling length of Lc was used as the input/output. The SEM images
of the fabricated filter in top-view and tilted-view are shown in Fig. 4.7 and Fig. 4.8(a),
respectively. This filter is equivalent to a 2-port version of a micro-ring based add-drop
filter, with its input and output ports corresponding to the drop and pass ports of the
micro-ring filter, respectively, and the total cavity length corresponding to half of the
round trip length of a micro-ring filter. The input light was coupled into the resonant
cavity through the coupler, and after the cavity the resonant wavelengths were reflected
back to the input port (drop port) of the coupler, whilst other wavelengths passed
through to the output port (pass port) of the coupler. The total cavity length (T) equals
Lw + 2R.
Figure 4.7. The top-view of the fabricated notch filter with the following parameters: Lw
= 4 μm, Lc = 0 μm, rb = 5 μm
Chapter 4 Circular Bragg Grating Mirror
75
Figure 4.8. (a): The tilted-view of the fabricated notch filter. (b): The spectra measured
at the output port of the notch filters fabricated with different T and Lc.
Circular grating based filters with different cavity and coupling lengths were fabricated.
The gap between the coupler and the cavity was fixed at 200 nm. The IL of the
fabricated filters are all within 0.5 dB, as shown by the measured output spectra in Fig.
4.8(b), which indicates a low loss of the cavity and thus verifies the high reflectivity of
the proposed grating mirror. With a cavity length of 44 μm (corresponding to a micro-
Chapter 4 Circular Bragg Grating Mirror
76
ring filter with a radius of ∼14 μm and a footprint of ~ 616 µm2), the filter shows a free-
spectral range (FSR) of about 6.34 nm at around 1550 nm. It can also be seen from the
measurement results that for a given cavity length, the notch depth of the filter increases
with the coupling strength while the quality factor (Q) decreases with the coupling
strength. With a coupling length of 2 μm and 4 μm (corresponding to a power coupling
ratio of about 7.6% and 15% at 1550 nm), the notch depth of the filter was measured to
be >5 and 10 dB, respectively, and the maximum Q was estimated to be about 14312
and 4683. In practice, the notch depth and Q can be tailored to desired values based on
specific applications. When T was reduced to a much smaller value of 8 μm
(corresponding to a micro-ring radius of ∼2.5 μm), a larger FSR of 30 nm was achieved,
as shown in the figure. However, as the coupling ratio of this filter was not optimized at
the time of fabrication (which is only about 1.4% with Lc set at 0 μm), the notch depth
was very limited to <3 dB. In practice, the notch depth can be significantly increased by
increasing the coupling strength, e.g., based on simulation results, a deep notch of >20
dB can be achieved for the short-cavity notch filter under critical coupling condition.
The proposed notch filter has low IL regardless of the cavity length, mainly attributed to
the low loss of the proposed circular grating mirror, and this becomes specially
advantageous in applications where large FSR and low loss are required. On the other
hand, the IL of a micro-ring filter based on this thin-Si-layer SOI could increase
substantially for <3 μm radius, due to the bending loss.
4.2.4 Solutions for Broadband and High-efficiency TM-reflection
In last section, a broadband circular Bragg grating mirror has been proposed based on
the standard 220 nm SOI platform. While the proposed grating mirror could achieve
high reflectivity over a broad bandwidth for TE-polarized light, it doesn’t work
efficiently for the TM-polarized light. Based on the mode profiles shown in Figs. 2.3(a)
and 2.3(b), unlike the fundamental TE mode whose power is mostly confined within the
silicon waveguide, the fundamental TM mode has a large cladding-penetration, and thus
effective refractive index becomes very low. As the TM mode is penetrated into the
claddings mainly in the vertical direction, even based on the proposed circular grating
structure, the index contrast for the grating is still relatively small, which is <0.5 for the
fundamental TM mode (1550 nm). The weak index contrast makes it hard to achieve
wide reflection bandwidth and high reflectivity, even with a large number of grating
Chapter 4 Circular Bragg Grating Mirror
77
periods. In addition, the large cladding-penetration of the TM mode also leads to higher
mode diffraction losses at the grating trenches. As a result, neither broad bandwidth nor
high reflectivity could be easily realized for the TM-polarized light based on the circular
Bragg grating mirror proposed on the 220 nm SOI platform. The polarization
dependence could limit the practical applications of the grating mirror in a polarization-
diverse integrated photonics environment. To overcome this difficulty, in sections
4.2.4.1 and 4.2.4.2, two potential solutions will be proposed based on the SOI platform
to realize broadband and high-efficiency TM-reflections.
4.2.4.1 Circular Bragg grating with thickened grating blades
From a physical point of view, as the confinement factor of the fundamental TM mode
is less than 0.3 and not all, but part of the waves that hit the grating blades gets reflected,
resulting reflectivity is obviously very low. Therefore, the first solution to increase the
grating reflectivity for the fundamental TM mode is increasing the thickness of the
grating blades so that more waves in the grating structure will be pushed for the
constructive interferences. In the meanwhile, as the effective index of the grating blades
increases for larger thickness, the grating index contrast becomes higher which could
increase the reflection bandwidth. The schematic illustration of the proposed circular
Bragg grating with thickened grating blades is shown in Fig. 4.9. In this TM-reflection
grating mirror, the blade thickness is designed to be >220 nm.
Chapter 4 Circular Bragg Grating Mirror
78
Figure 4.9. (a) The 3D and (b) the vertical cross-section schematic diagrams of the
proposed circular Bragg grating mirror with thickened blades, designed for TM-
polarized light.
3D-FDTD simulations were run to investigate properties of the proposed TM-designed
grating mirror. The grating parameters used in the simulations were: r = 5 µm, θ = 36°,
W1 = 370 nm, W2 = 286 nm, T = 775 nm, p = 5. It should be noted that, the widths of the
grating blades and the trenches here are designed based on second-order Bragg
condition to achieve grating parameters which are all compatible with the standard UV
lithography, and this is different from the previous TE-reflection grating which was
designed based on first-order Bragg condition with the consideration of achieving
maximized grating strength. The equations used for Bragg condition is expressed as:
0 0
1 2
1 2
3,
4 4eff eff
W Wn n
(4.3)
Where neff1 and neff2 are the effective refractive indices for the grating blades and the
trenches, respectively. Two different structures were compared: circular grating with
non-thickened blades (h = T = 220 nm), and circular grating with thickened blades (h =
220 nm, T = 775 nm). Spectral responses are shown in Fig. 4.10(a), which clearly
demonstrates that the grating with thickened blades is far more advantageous than the
Chapter 4 Circular Bragg Grating Mirror
79
other in terms of a higher reflectivity and a broader bandwidth. The vertical field
intensity distributions for the two structures were captured in Fig. 4.10(b), which
suggests the effectiveness of the raised blade height in reducing vertical diffractions.
Figure 4.10. (a) The simulated reflection spectra of the circular Bragg grating with
thickened blades (775 nm thick) and non-thickened blades (220 nm thick) with TM-
polarized light launched. (b) The vertical field intensity distributions simulated along
the gratings with 775 nm thick blades and 220 nm thick blades. The blade number was
set to 5.
The impact of the grating arc angle θ, the blade thickness, and the blade number on the
grating performance were investigated as shown in Fig. 4.11. As shown in Fig. 4.11(a),
both the grating peak reflectivity and the bandwidth (reflectivity >0.8) increase with θ,
but become nearly unaffected when θ exceeds 36°. Figure. 4.11(b) shows the change of
the grating peak reflectivity with the blade thickness. As shown in the figure, the
optimum blade thickness required to achieve the highest peak reflectivity is around 775
nm. For either thinner or thicker blades, the mode size mismatch between the blades and
the taper results in degraded grating performances, and this is mainly because the
fundamental TM mode is concentrated at the upper and lower interfaces of the
waveguide, which results in a relatively larger mode dimension than the fundamental
TE mode, which can be illustrated from Figs. 2.3(a) and 2.3(b). The grating peak
reflectivity also increases with the blade number, as shown in Fig. 4.11(c), and it
becomes stable (∼0.97) when the blade number exceeds 6. In contrast, based on 220 nm
thick blades, the peak reflectivity is <0.5 even with more than 6 blades.
Chapter 4 Circular Bragg Grating Mirror
80
When the fundamental TE mode was launched into the grating designed for TM modes,
it is interesting that the TM-designed grating is also good for TE modes, as shown in
Fig. 4.11(d). This is because the effective indices of the grating blade become very close
for the fundamental TE and TM modes, based on the larger thickness, where neff-TE ≈ 3.3,
neff-TM ≈ 3.2 at 1550 nm, and therefore the Bragg condition is satisfied for both
polarizations simultaneously. Nevertheless, the grating spectra for the two polarizations
are still different, and carefully observing them in Fig. 4.11(d) suggests two possibilities
in the future. The first one is that broadband grating mirrors can be designed for both
polarizations with proper engineering. The other is that the design can be elaborated for
only one polarization, TM or TE, while strongly discriminating the undesired one.
Figure 4.11. The change of the grating peak reflectivity and the bandwidth with (a) the
grating arc angle θ, (b) the blade thickness, and (c) the blade number. (d) The reflection
spectra of the circular grating with thickened blades (775 nm thick) simulated for the
fundamental TE and TM modes.
Chapter 4 Circular Bragg Grating Mirror
81
4.2.4.2 Circular Bragg grating based on horizontal slot waveguide
In section 4.2.4.1, it has been shown that circular grating with thickened grating blades
is effective in reflecting the TM-polarized light. However, the proposed structure shown
in Fig. 4.9 has one weakness, which is the relatively complex fabrication requirement.
This is because the waveguide and the grating blades are not based on the same silicon
substrate, as shown in Fig. 4.9(b), which makes the practical implementation of the
proposed structure very complicated. In this section, another solution for broadband and
high-efficiency TM-reflection is proposed, which is called a circular slot Bragg grating
mirror. The slot Bragg grating has a double-layer-silicon structure and is designed based
on the horizontal slot silicon waveguides. Compared to the structure proposed in last
section, the slot grating has an easy-to-fabricate structure.
As shown in Fig. 4.12, the proposed grating based on SOI substrate has three layers,
including two layers of silicon with the same thickness T and an embedded SiO2 slot
layer with the thickness t. The device mainly consists of four parts: an input strip
waveguide, an inverse taper of length lt used for mode transition from the fundamental
TM mode of strip waveguide to the slot mode of slot waveguide, a horizontal slot
waveguide of length ls, and the circular slot Bragg grating. The circular slot Bragg
grating region consists of a slot taper, and several circular slot blades. The arc angle and
radius of the taper is denoted as 𝜃 and r, the widths of grating blade (W1) and trench (W2)
are determined based on Equation 4.3.
Different from the structure in Fig. 4.9 which resolves the vertical diffraction loss issue
of the fundamental TM mode by using thicker grating blades [principle illustrated in Fig.
4.10(b)], the slot grating mirror proposed here in Fig. 4.12 eliminates the vertical
diffraction loss based on the conversion of the fundamental TM mode of strip
waveguide to the slot mode of slot waveguide. In a horizontal slot waveguide [152], due
to the electric-field discontinuity at the upper and lower waveguide boundaries, the
mode intensity of the fundamental TM mode is enhanced and concentrated within the
small slot region, and the TM mode could propagate through the slot waveguide
losslessly. Although TM mode is concentrated within the oxide slot, the effective index
of the slot waveguide can still be larger than the effective index of the strip waveguide,
if the slot thickness is properly chosen. Moreover, compared to the strip-waveguide-
based grating, the slot-waveguide-based grating has thicker grating blades [(2T + t) as
Chapter 4 Circular Bragg Grating Mirror
82
shown in Fig. 4.12], which leads to weaker vertical mode diffractions. As a result,
gratings based on the slot waveguide could be advantageous than gratings based on the
strip waveguide in terms of a larger index contrast and lower diffraction loss, given the
TM-polarized light is launched. Moreover, compared to the structure in Fig. 4.9, the slot
grating structure shown in Fig. 4.12 is relatively easy to fabricate. The fabrication can
be implemented based on a horizontal slot waveguide which is easy to realize [153], and
various similar structures based on horizontal slot waveguides have been demonstrated
before [154-156].
Figure 4.12. The schematic illustration of the slot-waveguide-based circular Bragg
grating.
The grating performance was investigated based on 3D-FDTD simulations. The lengths
of mode transition taper, slot waveguide, and circular grating taper were set to 5 µm (lt),
1 µm (ls) and 5 µm (r), respectively, and the taper tip width was set to 180 nm. Blade
and trench widths of 515 nm and 286 nm were selected for a Bragg central wavelength
of 1550 nm, and the grating angle 𝜃 was initially set at 20°. Based on the chosen
parameters, the proposed device is compact and is compatible with standard CMOS
fabrication processes. The performances of proposed circular slot grating (CSG) and the
single-layer circular grating (SCG) [Fig. 4.1 (b)] were both simulated. In both structures,
six grating blades were used for short device length. The thickness of the two silicon
layers of the CSG was set at 240 nm, and the SiO2 slot thickness was set to 150 nm. In
addition, devices were covered by SiO2 upper cladding in the simulation. The reflection
spectra of the two gratings are shown in Fig. 4.13(a).
Chapter 4 Circular Bragg Grating Mirror
83
As shown in Fig. 4.13(a), when compared to the SCG, the proposed CSG has achieved a
much higher reflectivity and a broader bandwidth. Vertical field intensity distributions
captured in Figs. 4.13(b) and 4.13(c) suggest the effectiveness of the CSG in reducing
vertical diffractions. Compared to the proposed grating in last section [Fig. 4.10(a)], the
relatively smaller effective index (2.36) of the slot waveguide results in a slightly
smaller reflectivity and bandwidth, nevertheless, when compared to the SCG (neff =
1.76), the CSG is still significantly advantageous in terms of a much higher peak
reflectivity (>0.9) and a broader bandwidth (>100 nm for reflectivity >0.8). It should be
noted that, in the proposed structure, the reason to choose the silicon layers at 240 nm
instead of 220 nm is because the optimal reflectivity of the grating will degrade based
on the 220 nm silicon thickness.
Figure 4.13. (a) The reflection spectra of CSG and SCG. (b, c) The vertical field
intensity distributions captured along (b) SCG and (c) CSG.
The impact of slot thickness was studied and the results are shown in Fig. 4.14(a). Here
the thickness of silicon layers (T) was fixed at 240 nm and the thickness of SiO2 slot (t)
Chapter 4 Circular Bragg Grating Mirror
84
was changed from 30 to 270 nm. It can be seen from the results that, Rpeak initially
increases with t, but after t = 150 nm, Rpeak starts to decrease. The reason is that for very
thin slot layer, there will be strong wave diffractions occurring through the slot to
grating trenches due to the tight slot mode confinement, which can be verified by the
field profile in the inset of Fig. 4.14(a) where t was set to a small value of 30 nm. On the
other hand, when the slot thickness increases beyond 150 nm, the slot mode condition
[152] is no longer satisfied, which also results in high propagation loss. It should also be
noted that extra excess loss is generated during the mode transition process, and longer
tapers (lt) can be used to reduce this loss. The impact of grating arc angle and blade
number were also investigated, as shown in Figs. 4.14(b) and 4.14(c). From Fig. 4.14(b),
it is clear that Rpeak and bandwidth increase with 𝜃 until around 30°, then both
parameters keep nearly constant. Furthermore, from Fig. 4.14(c), Rpeak increases with
the period number for both CSG and SCG. While in CSG, light is almost completely
reflected within 8 periods, the SCG shows low reflectivity even using a large number of
blades.
Figure 4.14. (a) The change of grating peak reflectivity with slot thickness (t). (b) The
change of grating peak reflectivity and bandwidth with grating arc angle (θ). (c) The
change of grating peak reflectivity with blade number (p) for CSG and SCG.
Chapter 4 Circular Bragg Grating Mirror
85
4.2.5 Extended Applications of Slot Bragg Grating
In section 4.2.4.2, a circular Bragg grating mirror was proposed based on the horizontal
slot silicon waveguide to realize broadband high-efficiency TM-reflections. Figure. 4.15
shows the schematic of a horizontal slot silicon waveguide. While the fundamental TM
mode of a slot waveguide is concentrated within the oxide slot layer due to the electric
field discontinuity at the upper and lower boundaries of the silicon layers, the
fundamental TE mode is mainly confined within the silicon layers, as illustrated in the
figure. Consequently, based on the differentiated mode profiles of the two polarizations
(a substantial effective index contrast exists between two different polarizations), Bragg
gratings can be designed to be highly polarization-selective. A polarization-selective
mirror should be highly reflective over a broad bandwidth and be operated only for TE
or TM with the other mode severely suppressed. Such a mirror could be used to reduce
the polarization degeneracy in applications such as tunable lasers, or to maintain the
polarization stability in a polarization-sensitive photonic integrated circuit. In next
section, slot-waveguide-based Bragg gratings are optimized to achieve high reflectivity,
wide bandwidth, as well as a strong polarization selectivity simultaneously.
Figure 4.15. The schematic of a horizontal slot waveguide.
4.2.5.1 Polarization-selective Bragg grating mirror
The schematic illustration of a polarization-selective Bragg grating mirror is shown in
Fig. 4.16. The structure was investigated based on 3D-FDTD simulations based on the
following parameters: θ = 24°, r = 5 µm, T = 240 nm, t = 150 nm, p = 6. Based on the
chosen parameters, the effective refractive indices for the fundamental TE and TM
Chapter 4 Circular Bragg Grating Mirror
86
modes at 1550 nm are about 2.84 and 2.41, respectively. Based on Equation 4.3,
different sets of grating blade/trench widths (W1/W2) were chosen to satisfy the Bragg
condition for the two polarizations. For TE polarization: W1 = 370 nm, W2 = 286 nm.
For TM polarization: W1 = 510 nm, W2 = 286 nm.
Figure 4.16. (a) The 3D schematic and (b) the cross-section schematic diagrams of the
polarization-selective Bragg grating.
Simulated spectral responses of the grating are shown in Figs. 4.17(a) and 4.17(b). From
the figures, we can see that the mirrors have good performances in terms of reflectivity
and bandwidth. The grating peak reflectivity can reach 0.956 and 0.914, respectively,
for TE and TM polarizations. For TE, high reflectivity (R > 0.8) can be obtained over a
broad bandwidth (BW) of 159 nm from 1490 to 1649 nm. For TM, a BW of 101nm from
1497 to 1598nm is estimated, covering the full C and L bands. When a TM mode is
launched into the TE-designed mirror, it would suffer very high loss as indicated by the
dotted line in Fig. 4.17(a). Extinction ratio (γ = RTE/RTM) was estimated to be 21.4 dB at
1550 nm, indicating an excellent polarization selectivity. Likewise, γ for TM-intended
design was calculated to be 5.05 dB, which could be made larger with more slot silicon
blades. Further optimisations to enlarge γ leave for future works.
Chapter 4 Circular Bragg Grating Mirror
87
Figure 4.17. The reflection spectra of (a) the TE-dedicated mirror and (b) the TM-
dedicated mirror. Dotted lines show the grating spectrum for the different polarization.
The influences of blade angle θ, blade number p, silicon layer thickness T and slot layer
thickness t, on the grating performance were investigated as shown in Fig. 4.18. As
shown in Fig. 4.18(a), Rpeak increases with θ, but becomes generally unaffected after 12º,
although there are maximums showing up at 24º. Fig. 4.18(b) suggests that 6 blades are
enough to reach the maximum Rpeak for both polarizations. To get Rpeak of 0.8, only 2
and 4 blades are needed for TE and TM polarizations, respectively, hinting a very small
device footprint. The values of T and t have been proved to have substantial impact on
grating mirror performance. Figure. 4.18(c) shows Rpeak as a function of t when T was
fixed at 240 nm. It is found that Rpeak for TM modes increases initially with t, then
decreases after reaching its maximum at 150 nm. This is because for too small slots,
there will be strong wave diffractions from the slot layer when light propagates through
the grating, due to the strong slot mode confinement. The same problem happens in the
structure shown in Fig. 4.12. For too thick slots, the condition for slot mode
transmission is not satisfied [152, 154], which leads to relatively high propagation loss.
The impact of slot layer thickness can be illustrated by the vertical field intensity
distributions of the grating which were simulated with different t values, as shown in
Fig. 4.19(a). For TE modes, in case of t >150 nm, diffraction loss caused by mode
mismatch between waveguide and blades increases with t, thus peak reflection goes
gradually down. Figure. 4.18(d) shows variation of Rpeak as a function of T while the slot
thickness t is fixed at 150 nm. We see there is also an optimum value for TM modes.
The reason is that, for thinner silicon layers, the power leakage to claddings is larger
Chapter 4 Circular Bragg Grating Mirror
88
while, for thicker silicon layers, slot mode confinement becomes weak which causes
relatively high propagation loss. The impact of silicon layer thickness is illustrated in
Fig. 4.19(b). In contrast, the influence of T for TE modes is so negligible that Rpeak
almost remains unchanged. This is because TE mode power is mostly distributed inside
silicon layers no matter how thick they are, once the slot thickness is chosen optimally
for low diffraction loss. In addition, the high index contrast of the TE modes always
helps achieve a high reflectance.
Figure 4.18. The reflection spectra of (a) the TE-dedicated mirror and (b) the TM-
dedicated mirror. Dotted lines show the grating spectrum for the different polarization.
Chapter 4 Circular Bragg Grating Mirror
89
Figure 4.19. The intensity field distribution simulated along the grating with different
sets of (a) slot layer thickness and (b) silicon layer thickness.
4.2.5.2 TM-pass/TE-block polarizer
In last section, a polarization-selective Bragg grating mirror was proposed based on the
horizontal slot silicon waveguide, and the polarization selectivity was realized by taking
advantage of the large effective index difference between the TE and TM polarizations.
The polarization-selective mirror could reflect light with a single polarization (TE or
Chapter 4 Circular Bragg Grating Mirror
90
TM) over a broad bandwidth while strongly discriminating the other. Particularly, for
the TE-reflection mirror, attributed to the large effective index of the fundamental TE
mode within the double-layer slot silicon waveguide, very high reflectivity and broad
reflection bandwidth can be achieved. In the meanwhile, the TM-polarized light
propagates through the slot layer of the grating and gets negligible reflection with most
of its power diffused into the claddings, contributing to a high polarization extinction
ratio for the grating. This inspires us to propose a solution to collect the non-reflected
TM modes which transmit through a TE-reflection slot Bragg grating, such that a
broadband and high extinction ratio TE-block/TM-pass polarizer could be realized.
Due to the large birefringence of the silicon photonics integration platform, photonic
integrated devices and circuits are polarization sensitive. Various types of polarization
manipulation devices have been demonstrated, amongst which the polarizer is an
important component which transmits light with only the selected polarization (TE or
TM) while blocks the other. However, in previous studies [157-159], the polarizer
usually suffers from a number of drawbacks, including the low extinction ratio (ER)
[159], high excess loss (EL) [158], and narrow operation bandwidth [157]. In this
section, we propose a TE-block/TM-pass polarizer based on the horizontal slot silicon
Bragg grating which overcomes these issues. Based on 3D-FDTD simulation results, the
polarizer could achieve an ultra-high ER of ˃ 30 dB over an ultra-broad wavelength
range of 264 nm (1412 – 1676 nm), and a low EL of < 0.3 dB. In addition, the proposed
polarizer is very compact, having a length of only 10.64 µm for the entire device,
making it suitable for large-scale photonic integrations.
Chapter 4 Circular Bragg Grating Mirror
91
Figure 4.20. (a) The 3D schematic and the 2D (b) lateral and (c) vertical illustrations of
the proposed polarizer based on horizontal slot silicon waveguide.
Different from the TE-reflection mirror demonstrated in last section, the TE-block/TM-
pass polarizer is not only required to be highly reflective for TE-polarized light, but also
needs to transmit the TM-polarized light with low loss. If circular Bragg grating is
employed to realize the polarizer, although high reflectivity can be achieved for the TE
modes, the TM modes would be hard to recollect, as the mechanism to recollect the
diffracted circular waves back to the straight waveguide is relatively complicated to
realize. Therefore, in the proposed polarizer, the grating blades were designed to be
Chapter 4 Circular Bragg Grating Mirror
92
straight rather than circular, as shown in Fig. 4.20, such that there would be no circular
diffractions occurring inside the grating, and the TM modes could transmit through the
slot grating with low loss. Two tapered mode converters are used in the proposed device
which connect the input/output waveguides with the Bragg grating. The input taper is
used to guide the input light into the grating mirror, and the output taper is used to guide
the transmitted TM modes into the output waveguide. To avoid high mode diffraction
losses for the TM modes, the grating is not fully etched in the lateral direction, which
leaves a nano-bridge connecting across the grating blades, as shown in the figure. The
use of nano-bridge ensures that the TM modes could be well confined within the slot
layer throughout the grating, which consequently avoids the high diffraction loss at
grating trenches compared with the case where there is no bridge.
In order to block the TE polarization while transmitting the TM polarization with low
loss, the grating period is selected so that the Bragg condition is only satisfied for the
TE polarization: Neff1-TEW1 + Neff2-TEW2 = λ0/2, and the grating period is smaller than
half of the TM polarization operating wavelength: Neff1-TMW1 + Neff2-TMW2 < λ0/2, where
Neff1 and Neff2 are the effective refractive index in the blade and the bridge regions,
respectively, and W1 and W2 are the lengths of the grating blades and the nano-bridges,
respectively. Consequently, the device works as a highly reflective mirror for the TE
polarization while as a slot sub-wavelength grating (SWG) waveguide which supports
the lossless Bloch mode [4] for the TM polarization. It should be noted that, W1 and W2
were selected based on first-order Bragg condition to realize a wide bandwidth and high
extinction ratio for the proposed polarizer.
In the grating section, high index contrast is of great importance in realizing a wideband
and high reflectance device. To enlarge the index contrast between grating blades and
nano-bridges for the TE modes, the blade width (Wt) is selected at a relatively large
value of 1 µm while the bridge width (Wb) is selected at a much smaller value of 100
nm, which enables a large index contrast of ˃1.4 for the TE mode (1550 nm). In
addition, the choosing of 100 nm bridge width ensures the cutoff of TE mode at the
bridge section, which leads to extra scattering losses for the TE mode and contributes to
a higher extinction ratio.
Chapter 4 Circular Bragg Grating Mirror
93
Figure 4.21. (a)-(d) The field profiles (1550 nm) captured in the horizontal and vertical
cross-sections of the device with TE- or TM-mode; and (e) and (f) The simulated
transmission spectra of the polarizer with 20 and 40 grating periods.
The grating duty cycle could influence the bandwidth and insertion loss of the proposed
polarizer. To realize broad bandwidth and low insertion loss simultaneously for the
Chapter 4 Circular Bragg Grating Mirror
94
proposed polarizaer, the following optimized parameters were used: W1 = 0.202 µm, W2
= 0.13 µm, based on a central Bragg wavelength of 1550 nm. In addition, the device
was covered by 2 µm thick SiO2 as the upper cladding. Figures. 4.21(a) – 4.21(d) show
the field propagation profiles (1550 nm) in both horizontal and vertical cross-sections
with TE- or TM-mode, respectively. It is clear that the TE mode is almost fully blocked
through the strong reflections of the grating. On the other hand, the TM mode
propagates through the slot layer with negligible loss. The transmission spectra are
shown in Figs. 4.21(e) and 4.21(f). It can be seen that with 20 grating periods (with a
total device length of only 10.64 µm), ultra-high ER of ˃ 30 dB can be achieved over a
broad band of 264 nm (1412 -1676 nm), and the EL for the TM mode is < 0.3 dB from
1434 to 1690 nm. With 40 periods (with a total device length of 17.28 µm), the ER is
further improved to ˃ 40 dB over a broader range of 276 nm (1409 -1685 nm), and there
is no substantial increase in the EL over the entire band.
4.2.6 Conclusion
In conclusion, in this chapter, a compact circular Bragg grating mirror with relaxed
fabrication requirements and a broader high-reflectivity (>95%) bandwidth has been
proposed based on the 220 nm SOI platform. Compared to the previously presented
circular grating in [143], the minimum radius of the circular grating proposed in this
chapter is significantly enlarged from 170 nm to >2 µm without sacrificing the overall
device footprint, and the minimum device feature size is increased from 30 to 100 nm,
which also allows a large fabrication tolerance. Simulation results have shown that the
proposed grating can achieve a high reflectivity of >95% over an ultrabroad band of 397
nm which is more than twice as wide as the high-efficiency band achieved in [143],
while experimental measurement results have shown that the fabricated circular grating
can achieve a high reflectivity of 93% to 98% within the measured band of 1530 to
1610 nm, which matches with the simulation results. Based on the proposed grating
mirror, a compact notch filter with large notch depth, high Q, and low transmission loss
has also been designed and demonstrated, and the proposed notch filter could be utilized
in various optical applications such as optical monitoring, sideband suppression, optical
modulation, etc., and sensing applications such as thermal and biochemical sensors [32,
160, 161]. With the advantages of broad bandwidth, high reflectivity and compact size,
the proposed circular Bragg grating is expected to be a promising element for wide use
Chapter 4 Circular Bragg Grating Mirror
95
in large-scale photonic integrated circuits and applications that require ultra-broadband
and high-efficiency on-chip optical reflections. In this chapter, in addition to the
presented circular Bragg grating mirror designed for TE polarization, two potential
solutions to realize high-efficiency TM-reflections have also been proposed and studied,
which include a circular Bragg grating with thickened blades and a slot circular Bragg
grating based on horizontal slot waveguide. These solutions have been proved to be
effective in realizing wideband and high reflectivity for TM-polarized light based on the
SOI platform. Finally, in this chapter, two extended applications of proposed slot Bragg
grating, including a polarization-selective mirror and a TM-pass/TE-block polarizer has
also been demonstrated and investigated through simulations.
Chapter 5 Conclusions and Future Work
96
CHAPTER 5 CONCLUSIONS AND
FUTURE WORK
5.1 Conclusions
In this thesis, the design, fabrication, and experimental characterizations of two novel
silicon photonic devices, namely a broadband, low-loss, and polarization-insensitive
adiabatic optical power splitter and a broadband high-reflectivity circular Bragg grating
mirror, have been presented. To the best of our knowledge, compared existing same
devices, both proposed devices in this thesis have achieved the best device performance
and the smallest device size. These promising properties make them highly competitive
in building large-scale high-density photonic integrated circuits.
In Chapter 2, we have developed the fabrication process for silicon photonic device
based on 220 nm SOI wafer and assembled the device measurement setup. Low
propagation loss of 2.4 dB/cm has been achieved for the fabricated single-mode silicon
strip waveguide.
The 3-dB adiabatic optical power splitter has been presented in Chapter 3. FDTD
simulation results have shown that the proposed splitter can achieve a transmission
efficiency of nearly 50% for the two output ports over an ultrabroad band from 1200 to
1700 nm, which is literally wavelength insensitive. Experimental measurement results
have shown that the fabricated splitter exhibits a low excess loss of <0.19 dB and <0.14
dB for the TE and TM polarizations, respectively, over the entire measured band from
1530 to 1600 nm, while having an adiabatic taper length of only 5 µm. The splitter has
also exhibited a good power uniformity. The measured uniformity of the cascaded 1×8
splitter has been only 0.47 and 0.17 dB for the TE and TM polarizations, respectively.
In Chapter 4, a broadband and high-reflectivity circular Bragg grating mirror has been
demonstrated. FDTD simulation results have shown that the proposed grating mirror
can achieve a reflectivity of >90% over an ultrabroad bandwidth of 500 nm (1263 -
1763 nm), and a high reflectivity of >95% over a broad bandwidth of 397 nm (1340 -
1737 nm), which covers the entire E- to U-bands. Experimental measurement results
have exhibited a high reflectivity of 93% - 98% within the measured band of 1530 to
1610 nm, which agrees well with simulations. The proposed Bragg mirror has a very
Chapter 5 Conclusions and Future Work
97
compact size of only 4.49 µm × 4.54 µm, making it suitable to build densely integrated
devices such as filters, resonators, and on-chip laser cavities. As one of its applications,
a compact notch filter with high rejection ratio (>10 dB) and low transmission loss
(<0.5 dB) has also been fabricated and presented in this chapter. In this chapter, in
addition to the grating mirror designed for TE-polarized light, solutions for high-
efficiency TM-reflections have been proposed and investigated. Moreover, two
extended applications of the proposed slot grating structure in this chapter have been
presented through simulations.
5.2 Future work
The outstanding properties of the proposed devices in this thesis can enable various
promising applications in the future. One of the promising applications of the proposed
adiabatic splitter in Chapter 3 is the silicon integrated optical phased array proposed in
[162]. The proposed phased array can be used to realize beam steering function in high-
speed short-range infrared optical wireless communication systems. Silicon integrated
power splitter is one of the key components of the proposed optical phased array, and
the splitter is required to have a compact size, low insertion loss and broad bandwidth.
In addition, it needs to be polarization independent. The proposed adiabatic splitter in
this thesis meets all these requirements and will be utilized in the proposed optical
phased array in our future work. In addition, as mentioned in section 3.2.2, it is possible
to realize various power splitting ratios for the proposed splitter by using asymmetric
gaps between tapers. Free choice of splitting ratio is an important feature which can be
used to realize dynamic control and efficiency management of optical power in photonic
systems [163], and is also needed in various applications such as asymmetric MZI [164]
and signal monitoring [165]. Therefore, realizing various splitting ratio is another
improvement of the proposed adiabatic splitter in the future.
The proposed broadband circular Bragg grating mirror in Chapter 4 can be used to
construct integrated filters, resonators, and lasers. In addition to the integrated notch
filter presented in Chapter 4, another promising application of the proposed circular
Bragg mirror is the fast wavelength-switchable hybrid silicon laser demonstrated in [66].
Compared with a loop mirror which is also commonly used to build external laser
cavities [166, 167], the proposed circular Bragg grating mirror has the advantages of flat
reflectivity, high polarization extinction ratio, and smaller footprint. These properties
Chapter 5 Conclusions and Future Work
98
make it capable of building lasers with stable lasing output, stronger polarization
stability, and higher level of integration. While in [66], only the initial concept of the
hybrid laser is presented, the experimental demonstrations will be left for our future
work. In addition, in Chapter 4, several new multi-layer Bragg grating structures with
novel applications have also been proposed. Normally, silicon photonic structures with
multiple silicon layers can be realized based on the deposition of amorphous silicon,
which has been extensively researched in previous literatures [153, 154, 156]. Due to
the facility limitation, the proposed multi-layer devices in this thesis are only estimated
through simulations, and the fabrication and experimental characterization will also
leave for the future work.
Chapter 6 References
99
CHAPTER 6 REFERENCES
[1] Y. A. Vlasov, "Silicon CMOS-Integrated Nano-Photonics for Computer and Data
Communications Beyond 100G," IEEE Communications Magazine, vol. 50(2), pp. s67-
s72, Feb 2012.
[2] P. Pepeljugoski et al., "Low Power and High Density Optical Interconnects for Future
Supercomputers," presented at the Optical Fiber Communication Conference, San
Diego, California, United States, Mar, 2010.
[3] C. Sun et al., "Single-chip microprocessor that communicates directly using light,"
Nature, vol. 528, pp. 534-538, Dec 2015.
[4] K. Yamada et al., "High-performance silicon photonics technology for
telecommunications applications," Science and Technology of Advanced Materials, vol.
15(2), Apr 2014.
[5] P. Dong, Y.-K. Chen, G.-H. Duan, and D. T. Neilson, "Silicon photonic devices and
integrated circuits," Nanophotonics, vol. 3, no. 4-5, pp. 215-228, 2014.
[6] B. Jalali and S. Fathpour, "Silicon Photonics," JOURNAL OF LIGHTWAVE
TECHNOLOGY, vol. 24(12), pp. 4600-4615, Dec 2006.
[7] B. Jalali, V. Raghunathan, R. Shori, S. Fathpour, D. Dimitropoulos, and O. Stafsudd,
"Prospects for Silicon Mid-IR Raman Lasers," IEEE JOURNAL OF SELECTED
TOPICS IN QUANTUM ELECTRONICS, vol. 12(6), pp. 1618-1627, Nov/Dec 2006.
[8] G. T. Reed, G. Mashanovich, F. Y. Gardes, and D. J. Thomson, "Silicon optical
modulators," Nature Photonics, vol. 4, pp. 518-526, Jul 2010.
[9] H. Subbaraman et al., "Recent advances in silicon-based passive and active optical
interconnects," Optics Express, vol. 23(3), pp. 2487-2510, Feb 2015.
[10] L. TSYBESKOV, D. J. LOCKWOOD, and M. ICHIKAWA, "Silicon Photonics:
CMOS Going Optical," Proceedings of the IEEE, vol. 97(7), pp. 1161-1165, Jul 2009.
[11] R. Soref, "The Past, Present, and Future of Silicon Photonics," IEEE JOURNAL OF
SELECTED TOPICS IN QUANTUM ELECTRONICS, vol. 12(6), pp. 1678-1687,
Nov/Dec 2006.
[12] L. Pavesi and D. J. Lockwood, "Integrated Photonics," in Silicon Photonics: Springer-
Verlag Berlin Heidelberg, 2004.
[13] L. Pavesi and D. J. Lockwood, "Monolithic Silicon Microphotonics," Springer-Verlag
Berlin Heidelberg, 2004.
[14] Z. Zhou, Z. Tu, T. Li, and X. Wang, "Silicon Photonics for Advanced Optical
Interconnections," JOURNAL OF LIGHTWAVE TECHNOLOGY, vol. 33(4), pp. 928-
933, Feb 2015.
[15] D. A. B. MILLER, "Rationale and Challenges for Optical Interconnects to Electronic
Chips," Proceedings of the IEEE, vol. 88, no. 6, pp. 728-749, Jun 2000.
Chapter 6 References
100
[16] D. A. B. Miller, "DeviceRequirementsforOptical Interconnects to Silicon Chips,"
Proceedings of the IEEE, vol. 97(7), pp. 1166-1185, Jul 2009.
[17] L. D. Paulson, "IBM Project Proposes Using Light to Make Chips Faster," Computer,
vol. 44, no. 2, pp. 14-17, Feb 2011.
[18] D. McGrath. (2009) Intel developing optical chip-to-chip interconnects. Electronic
Engineering Times.
[19] L. Pavesi and D. J. Lockwood, "Merits and Potential Impact of Silicon Photonics," in
Silicon Photonics III: Systems and Applications: Springer-Verlag Berlin Heidelberg,
2016.
[20] Available: https://www.cisco.com/c/dam/en/us/solutions/collateral/service-
provider/global-cloud-index-gci/white-paper-c11-738085.pdf
[21] R. A. Soref and J. P. Lorenzo, "Single-crystal silicon: a new material for 1.3 and 1.6 μm
integrated-optical components," Electronics Letters, vol. 21, no. 21, pp. 953-954, 1985.
[22] R. A. Soref and J. P. Lorenzo, "All-silicon active and passive guided-wave components
for λ = 1.3 and 1.6 µm," IEEE Journal of Quantum Electronics, vol. 22, no. 6, pp. 873-
879, Jun 1986.
[23] A. Liu et al., "A high-speed silicon optical modulator based on a metal–oxide–
semiconductor capacitor," Nature, vol. 427, pp. 615-618, Feb 2004.
[24] B. R. Koch, A. W. Fang, O. Cohen, and J. E. Bowers, "Mode-locked silicon evanescent
lasers," Optics Express, vol. 15(18), pp. 11225-11233, Sep 2007.
[25] A. Narasimha et al., "A Fully Integrated 4 × 10-Gb/s DWDM Optoelectronic
Transceiver Implemented in a Standard 0.13 μm CMOS SOI Technology," IEEE
JOURNAL OF SOLID-STATE CIRCUITS, vol. 42(12), pp. 2736-2744, Dec 2007.
[26] S. Assefa et al., "A 90nm CMOS integrated Nano-Photonics technology for 25Gbps
WDM optical communications applications," presented at the Electron Devices Meeting
(IEDM), 2012 IEEE International, San Francisco, CA, 2012.
[27] M. Haurylau et al., "On-Chip Optical Interconnect Roadmap: Challenges and Critical
Directions," IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS,
vol. 12(6), pp. 1699-1705, Nov/Dec 2006.
[28] G. T. Reed, "Optical Modulators in Silicon Photonic Circuits," in SILICON
PHOTONICS: THE STATE OF THE ART: John Wiley & Sons Ltd, 2008.
[29] R. Soref and B. Bennett, "Electrooptical effects in silicon," IEEE Journal of Quantum
Electronics, vol. QE-23(1), pp. 123-129, Jan 1987.
[30] L. Liao, D. Samara-Rubio, M. Morse, A. Liu, and D. Hodge, "High speed silicon Mach-
Zehnder modulator " Optics Express, vol. 13(8), pp. 3129-3135, Apr 2005.
[31] Q. Xu, B. Schmidt, S. Pradhan, and M. Lipson, "Micrometre-scale silicon electro-optic
modulator," Nature, vol. 435, pp. 325-327, May 2005.
[32] W. Bogaerts et al., "Silicon microring resonators," LASER & PHOTONICS REVIEWS,
vol. 6(1), pp. 47-73, 2012.
Chapter 6 References
101
[33] Z. Fang and C. Z. Zhao, "Recent Progress in Silicon Photonics: A Review," ISRN
Optics, vol. 2012, 2012.
[34] R. A. Soref and B. R. Bennett, "Kramers-Kronig Analysis Of Electro-Optical Switching
In Silicon," Proc. SPIE 0704, Integrated Optical Circuit Engineering IV, vol. 704, pp.
32-37, Mar 1987.
[35] C. K. Tang and G. T. Reed, "Highly efficient optical phase modulator in SOI
waveguides," Electronics Letters, vol. 31(6), pp. 451-452, Mar 1995.
[36] J. P. Lorenzo and R. A. Soref, "1.3 μm electro‐optic silicon switch," Applied Physics
Letters, vol. 51, pp. 6-8, 1987.
[37] A. Cutolo, M. Iodice, P. Spirito, and L. Zeni, "Silicon electro-optic modulator based on
a three terminal device integrated in a low-loss single-mode SOI waveguide," Journal
of Lightwave Technology, vol. 15, no. 3, pp. 505-518, Mar 1997.
[38] C. A. Barrios, V. R. Almeida, R. Panepucci, and M. Lipson, "Electrooptic modulation
of silicon-on-insulator submicrometer-size waveguide devices," Journal of Lightwave
Technology, vol. 21, no. 10, pp. 2332-2339, Oct 2003.
[39] C. E. Png, S. P. Chan, S. T. Lim, and G. T. Reed, "Optical phase modulators for MHz
and GHz modulation in silicon-on-insulator (SOI)," Journal of Lightwave Technology,
vol. 22, no. 6, pp. 1573-1582, Jun 2004.
[40] S. Manipatruni, Q. Xu, B. Schmidt, J. Shakya, and M. Lipson, "High Speed Carrier
Injection 18 Gb/s Silicon Micro-ring Electro-optic Modulator," presented at the Lasers
and Electro-Optics Society, 2007. LEOS 2007. The 20th Annual Meeting of the IEEE,
Lake Buena Vista, FL, Oct, 2007.
[41] W. M. J. Green, M. J. Rooks, L. Sekaric, and Y. A. Vlasov, "Ultra-compact, low RF
power, 10 Gb/s silicon Mach-Zehnder modulator," Optics Express, vol. 15, no. 25, pp.
17106-17113, 2007.
[42] F. Y. Gardes, G. T. Reed, N. G. Emerson, and C. E. Png, "A sub-micron depletion-type
photonic modulator in Silicon On Insulator," Optics Express, vol. 13, no. 22, pp. 8845-
8854, 2005.
[43] F. Y. Gardes et al., "High-speed modulation of a compact silicon ring resonator based
on a reverse-biased pn diode," Optics Express, vol. 17, no. 24, pp. 21986-21991, 2009.
[44] J.-B. You, M. Park, J.-W. Park, and G. Kim, "12.5 Gbps optical modulation of silicon
racetrack resonator based on carrier-depletion in asymmetric p-n diode," Optics Express,
vol. 16, no. 22, pp. 18340-18344, 2008.
[45] N.-N. Feng et al., "High speed carrier-depletion modulators with 1.4V-cm VπL
integrated on 0.25μm silicon-on-insulator waveguides," Optics Express, vol. 18, no. 8,
pp. 7994-7999, 2010.
[46] L. Liao et al., "40 Gbit/s silicon optical modulator for highspeed applications,"
Electronics Letters, vol. 43(22), Oct 2007.
[47] D. M. Gill et al., "Internal Bandwidth Equalization in a CMOS-Compatible Si-Ring
Modulator," IEEE Photonics Technology Letters, vol. 21, no. 4, pp. 200-202, Dec 2008.
Chapter 6 References
102
[48] D. Liang and J. E. Bowers, "Recent progress in lasers on silicon," Nature Photonics, vol.
4, pp. 511-517, Jul 2010.
[49] G. T. Reed, "Silicon Lasers," in SILICON PHOTONICS: THE STATE OF THE ART:
John Wiley & Sons Ltd, 2008.
[50] L. Pavesi and D. J. Lockwood, "Silicon Fundamentals for Photonics Applications," in
Silicon Photonics: Springer-Verlag Berlin Heidelberg, 2004.
[51] G. T. Reed and A. P. Knights, "Prospects for Silicon Light-emitting Devices," in Silicon
Photonics: An Introduction: John Wiley & Sons Ltd, 2004.
[52] R. Claps, D. Dimitropoulos, Y. Han, and B. Jalali, "Observation of Raman emission in
silicon waveguides at 1.54 μm," Optics Express, vol. 10, no. 22, pp. 1305-1313, 2002.
[53] H. Rong et al., "A continuous-wave Raman silicon laser," Nature, vol. 433, pp. 725-728,
Feb 2005.
[54] Z. Zhou, B. Yin, and J. Michel, "On-chip light sources for silicon photonics," Light:
Science & Applications vol. 4. e358, 2015.
[55] A. J. Kenyon, "Erbium in silicon," SEMICONDUCTOR SCIENCE AND
TECHNOLOGY, vol. 20, no. 12, pp. R65-R84, 2005.
[56] M. J. DEEN and P. K. BASU, "Light Emitters in Si," in Silicon Photonics:
Fundamentals and Devices: John Wiley & Sons Ltd, 2012.
[57] O. Jambois, F. Gourbilleau, A. J. Kenyon, J. Montserrat, R. Rizk, and B. Garrido,
"Towards population inversion of electrically pumped Er ions sensitized by Si
nanoclusters," Optics Express, vol. 18, no. 3, pp. 2230-2235, 2010.
[58] J. Liu et al., "Ge-on-Si optoelectronics," Thin Solid Films, vol. 520, pp. 3354–3360,
2012.
[59] J. Liu et al., "Tensile-strained, n-type Ge as a gain medium for monolithic laser
integration on Si," Optics Express, vol. 15, no. 18, pp. 11272-11277, 2007.
[60] X. Sun, J. Liu, L. C. Kimerling, and J. Michel, "Direct gap photoluminescence of nn-
type tensile-strained Ge-on-Si," Applied Physics Letters, vol. 95, no. 1, 2009, Art. no.
011911.
[61] J. Liu, X. Sun, R. Camacho-Aguilera, L. C. Kimerling, and J. Michel, "Ge-on-Si laser
operating at room temperature," Optics Letters, vol. 35, no. 5, pp. 679-681, 2010.
[62] M. E. Groenert et al., "Monolithic integration of room-temperature cw GaAs/AlGaAs
lasers on Si substrates via relaxed graded GeSi buffer layers," Journal of Applied
Physics, vol. 93, no. 1, pp. 362-367, 2003.
[63] L. Cerutti, J. B. Rodriguez, and E. Tournie, "GaSb-Based Laser, Monolithically Grown
on Silicon Substrate, Emitting at 1.55 μm at Room Temperature," IEEE Photonics
Technology Letters, vol. 22, no. 8, pp. 553-555, 2010.
[64] G. Roelkens, D. V. Thourhout, R. Baets, R. Nötzel, and M. Smit, "Laser emission and
photodetection in an InP/InGaAsP layer integrated on and coupled to a Silicon-on-
Insulator waveguide circuit," Optics Express, vol. 14, no. 18, pp. 8154-8159, 2006.
Chapter 6 References
103
[65] G.-H. Duan et al., "Hybrid III–V on Silicon Lasers for Photonic Integrated Circuits on
Silicon," IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS,
vol. 20, no. 4, 2014, Art. no. 6100213.
[66] S. Gao, C.-J. Chae, Y. Wang, and E. Skafidas, "Fast Wavelength-Switchable Hybrid
Laser for Energy-Efficient Optical Interconnect," presented at the 2014 Optical
Interconnects Conference, San Diego, CA, May 2014, 2014.
[67] C.-J. Chae, E. Skafidas, and D.-Y. Choi, "Compact bragg grating reflectors in silicon
waveguides and their application to resonator filters," presented at the Optical Fiber
Communications Conference and Exhibition (OFC), 2014, San Francisco, CA, March
2014, 2014.
[68] J. Michel, J. Liu, and L. C. Kimerling, "High-performance Ge-on-Si photodetectors,"
Nature Photonics, vol. 4, pp. 527-534, Aug 2010.
[69] S. Luryi, A. Kastalsky, and J. C. Bean, "New infrared detector on a silicon chip," IEEE
Transactions on Electron Devices vol. 31, no. 9, pp. 1135-1139, Sep 1984.
[70] E. A. Fitzgerald, "Dislocations in strained-layer epitaxy: theory, experiment, and
applications," Materials Science Reports, vol. 7, no. 3, pp. 87-142, Nov 1991.
[71] S. B. Samavedam, M. T. Currie, T. A. Langdo, and E. A. Fitzgerald, "High-quality
germanium photodiodes integrated on silicon substrates using optimized relaxed graded
buffers," Applied Physics Letters, vol. 73, pp. 2125-2127, 1998.
[72] J. Wang and S. Lee, "Ge-Photodetectors for Si-Based Optoelectronic Integration,"
Sensors, vol. 11, pp. 696-718, Jan 2011.
[73] H.-C. Luan et al., "High-quality Ge epilayers on Si with low threading-dislocation
densities," Applied Physics Letters, vol. 75, no. 19, pp. 2909-2911, 1999.
[74] J. M. Baribeau, T. E. Jackman, D. C. Houghton, P. Maigné, and M. W. Denhoff,
"Growth and characterization of Si1−x Ge x and Ge epilayers on (100) Si," Journal of
Applied Physics, vol. 63, pp. 5738-5746, 1988.
[75] Z. Huang, J. Oh, and J. C. Campbell, "Back-side-illuminated high-speed Ge
photodetector fabricated on Si substrate using thin SiGe buff er layers," Applied Physics
Letters, vol. 85, pp. 3286-3288, 2004.
[76] J. Nakatsuru, H. Date, S. Mashiro, and M. Ikemoto, "Growth of high quality Ge
epitaxial layer on Si(100) substrate using ultra thin Si0.5Ge0.5 buff er," MRS
Proceedings, vol. 891, pp. 315-320, 2006.
[77] L. Vivien et al., "42 GHz p.i.n germanium photodetector integrated in a silicon-on-
insulator waveguide," Optics Express, vol. 17, no. 8, pp. 6252-6257, 2009.
[78] D. Feng et al., "High-speed Ge photodetector monolithically integrated with large
cross-section silicon-on-insulator waveguide," Applied Physics Letters, vol. 95, 2009,
Art. no. 261105.
[79] Y. Kang et al., "Monolithic germanium/silicon avalanche photodiodes with 340 GHz
gain–bandwidth product," Nature Photonics, vol. 3, pp. 59-63, 2009.
Chapter 6 References
104
[80] W. S. Zaoui et al., "Frequency response and bandwidth enhancement in Ge/Si
avalanche photodiodes with over 840GHz gain-bandwidth-product," Optics Express,
vol. 17, no. 15, pp. 12641-12649, 2009.
[81] Y. Kang et al., "Ge/Si Waveguide Avalanche Photodiodes on SOI Substrates for High
Speed Communication," ECS Transactions, vol. 33, no. 6, pp. 757-764, 2010.
[82] Y. Kang et al., "Monolithic Ge/Si avalanche photodiodes," presented at the 2009 6th
IEEE International Conference on Group IV Photonics, San Francisco, CA, Sep, 2009.
[83] Y. A. Vlasov and S. J. McNab, "Losses in single-mode silicon-on-insulator strip
waveguides and bends," Optics Express, vol. 12, no. 8, pp. 1622-1631, 2004.
[84] T. Tsuchizawa et al., "Microphotonics devices based on silicon microfabrication
technology," IEEE Journal of Selected Topics in Quantum Electronics, vol. 11, no. 1,
pp. 232-240, 2005.
[85] P. Dumon et al., "Low-loss SOI photonic wires and ring resonators fabricated with deep
UV lithography," IEEE Photonics Technology Letters, vol. 16, no. 5, pp. 1328-1330,
2004.
[86] K. Debnath et al., "Low-Loss Silicon Waveguides and Grating Couplers Fabricated
Using Anisotropic Wet Etching Technique," frontiers in Materials, vol. 3, Feb 2016,
Art. no. 10.
[87] F. Y. Gardes et al., "Sub-micron optical waveguides for silicon photonics formed via
the local oxidation of silicon (LOCOS)," SPIE Proceedings, Jan 2008.
[88] K. K. Lee, D. R. Lim, L. C. Kimerling, J. Shin, and F. Cerrina, "Fabrication of ultralow-
loss Si/SiO2 waveguides by roughness reduction," Optics Letters, vol. 26, no. 23, pp.
1888-1890, 2001.
[89] G. Z. Masanovic et al., "A high efficiency input/output coupler for small silicon
photonic devices," Optics Express, vol. 13, no. 19, pp. 7374-7379, 2005.
[90] D. Vermeulen et al., "High-efficiency fiber-to-chip grating couplers realized using an
advanced CMOS-compatible Silicon-On-Insulator platform," Optics Express, vol. 18,
no. 17, pp. 18278-18283.
[91] X. Chen, C. Li, and H. K. Tsang, "Fabrication-Tolerant Waveguide Chirped Grating
Coupler for Coupling to a Perfectly Vertical Optical Fiber," IEEE Photonics
Technology Letters, vol. 20, no. 23, pp. 1914-1916, Aug 2008.
[92] J. Cardenas, C. B. Poitras, K. Luke, L.-W. Luo, P. A. Morton, and M. Lipson, "High
Coupling Efficiency Etched Facet Tapers in Silicon Waveguides," IEEE Photonics
Technology Letters, vol. 26, no. 23, pp. 2380-2382, Sep 2014.
[93] T. Shoji, T. Tsuchizawa, T. Watanabe, K. Yamada, and H. Morita, "Low loss mode size
converter from 0.3 μm square Si wire waveguides to singlemode fibres," Electronics
Letters, vol. 38, no. 25, pp. 1669-1670, Dec 2002.
[94] C. Kopp et al., "Silicon Photonic Circuits: On-CMOS Integration, Fiber Optical
Coupling, and Packaging," IEEE Journal of Selected Topics in Quantum Electronics vol.
17, no. 3, pp. 498-509, Oct 2010.
Chapter 6 References
105
[95] S. J. McNab, N. Moll, and Y. A. Vlasov, "Ultra-low loss photonic integrated circuit
with membrane-type photonic crystal waveguides," Optics Express, vol. 11, no. 22, pp.
2927-2939, 2003.
[96] Z. Yu et al., "High efficiency and broad bandwidth grating coupler between
nanophotonic waveguide and fibre," Chinese Physics B, vol. 19, no. 1, 2010.
[97] L. Pavesi and D. J. Lockwood, "Packaging of Silicon Photonic Devices," in Silicon
Photonics III: Systems and Applications: Springer-Verlag Berlin Heidelberg, 2016.
[98] D. Dai, Z. Wang, N. Julian, and J. E. Bowers, "Compact broadband polarizer based on
shallowly-etched silicon-on-insulator ridge optical waveguides," Optics Express, vol. 18,
no. 26, pp. 27404-27415, 2010.
[99] Y. Huang, S. Zhu, H. Zhang, T.-Y. Liow, and G.-Q. Lo, "CMOS compatible horizontal
nanoplasmonic slot waveguides TE-pass polarizer on silicon-oninsulator platform,"
Optics Express, vol. 21, no. 10, pp. 12790-12796, 2013.
[100] M. Z. Alam, J. S. Aitchison, and M. Mojahedi, "Compact and silicon-on-insulator-
compatible hybrid plasmonic TE-pass polarizer," Optics Letters, vol. 37, no. 1, pp. 55-
57, 2012.
[101] D. Dai, J. Bauters, and J. E. Bowers, "Passive technologies for future large-scale
photonic integrated circuits on silicon: polarization handling, light non-reciprocity and
loss reduction," Light: Science & Applications, 2012.
[102] T. K. Liang and H. K. Tsang, "Integrated polarization beam splitter in high index
contrast silicon-on-insulator waveguides," IEEE Photonics Technology Letters, vol. 17,
no. 2, pp. 393-395, 2005.
[103] B.-K. Yang, S.-Y. Shin, and D. Zhang, "Ultrashort Polarization Splitter Using Two-
Mode Interference in Silicon Photonic Wires," IEEE Photonics Technology Letters, vol.
21, no. 7, pp. 432-434, Feb 2009.
[104] Y. Yue, L. Zhang, J.-Y. Yang, R. G. Beausoleil, and A. E. Willner, "Silicon-on-
insulator polarization splitter using two horizontally slotted waveguides," Optics Letters,
vol. 35, no. 9, pp. 1364-1366, 2010.
[105] D. Dai, Z. Wang, J. Peters, and J. E. Bowers, "Compact Polarization Beam Splitter
Using an Asymmetrical Mach–Zehnder Interferometer Based on Silicon-on-Insulator
Waveguides," IEEE Photonics Technology Letters, vol. 24, no. 8, pp. 673-675, Jan
2012.
[106] S. Gao, Y. Wang, K. Wang, and E. Skafidas, "Low-Loss and Broadband 2 × 2
Polarization Beam Splitter Based on Silicon Nitride Platform," IEEE Photonics
Technology Letters, vol. 28, no. 18, pp. 1936-1939, Jun 2016.
[107] J. Feng and R. Akimoto, "Silicon nitride polarizing beam splitter with potential
application for intersubband-transition-based all-optical gate device," Japanese Journal
of Applied Physics, vol. 54, no. 4S, 2015.
[108] D.-X. Xu et al., "Silicon Photonic Integration Platform—Have We Found the Sweet
Spot?," IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, vol.
20, no. 4, pp. 189-205, 2014.
Chapter 6 References
106
[109] D. J. Lockwood and L. Pavesi, "Silicon Photonic Wire Waveguides: Fundamentals and
Applications," in Silicon Photonics II: Components and Integration: Springer-Verlag
Berlin Heidelberg, 2011.
[110] N. Devlin and D. Brown. (2010). Dose Determination Using ZEP520A Resist as a
Model. Available: http://nanolithography.gatech.edu/training/Determining_dose_r3.pdf
[111] J. Cardenas;, C. B. Poitras;, K. Luke;, L.-W. Luo;, P. A. Morton;, and M.
Lipson;, "High Coupling Efficiency Etched Facet Tapers in Silicon Waveguides,"
IEEE PHOTONICS TECHNOLOGY LETTERS, vol. 26, no. 23, pp. 2380-2382, 2014.
[112] D. Kwong, A. Hosseini, Y. Zhang, and R. T. Chen, "1 × 12 Unequally spaced
waveguide array for actively tuned optical phased array on a silicon nanomembrane,"
APPLIED PHYSICS LETTERS, vol. 99, 2011, Art. no. 051104.
[113] S. Tomofuji, S. Matsuo, T. Kakitsuka, and K.-i. Kitayama, "Dynamic switching
characteristics of InGaAsP/InP multimode interference optical waveguide switch,"
Optics Express, vol. 17, no. 26, pp. 23380-23388, 2009.
[114] T. Li et al., "Low-voltage, high speed, compact silicon modulator for BPSK modulation
" Optics Express, vol. 21, no. 20, pp. 23410-23415, 2013.
[115] S. Chen, Y. Shi, S. He, and D. Dai, "Compact monolithically-integrated hybrid
(de)multiplexer based on silicon-on-insulator nanowires for PDM-WDM systems,"
Optics Express, vol. 23, no. 10, pp. 12840-12849, 2015.
[116] H. Yamada, T. Chu, S. Ishida, and Y. Arakawa, "Optical directional coupler based on
Si-wire waveguides," IEEE Photonics Technology Letters, vol. 17, no. 3, pp. 585-587,
2005.
[117] Y.-j. Quan, P.-d. Han, Q.-j. Ran, F.-p. Zeng, L.-p. Gao, and C.-h. Zhao, "A photonic
wire-based directional coupler based on SOI," Optics Communications, vol. 281, no. 11,
pp. 3105-3110, 2008.
[118] Z. Sheng et al., "A Compact and Low-Loss MMI Coupler Fabricated With CMOS
Technology," IEEE Photonics Journal, vol. 4, no. 6, pp. 2272-2277, 2012.
[119] H. Zhou, J. Song, C. Li, H. Zhang, and P. G. Lo, "A Library of Ultra-Compact
Multimode Interference Optical Couplers on SOI," IEEE Photonics Technology Letters,
vol. 25, no. 12, pp. 1149-1152, 2013.
[120] X. Tang, J. Liao, H. Li, L. Zhang, R. Lu, and Y. Liu, "A novel scheme for 1×N optical
power splitter," Optics Express, vol. 18, no. 21, pp. 21697-21704, 2010.
[121] K. K. Chung, H. P. Chan, and P. L. Chu, "A 1 × 4 polarization and wavelength
independent optical power splitter based on a novel wide-angle low-loss Y-junction,"
Optics Communications, vol. 267, no. 2, pp. 367-372, 2006.
[122] J. Xing et al., "Silicon-on-insulator-based adiabatic splitter with simultaneous tapering
of velocity and coupling," Optics Letters, vol. 38, no. 13, pp. 2221-2223, 2013.
[123] L. Cao, A. Elshaari, A. Aboketaf, and S. Preble, "Adiabatic couplers in SOI
waveguides," presented at the Conference on Lasers and Electro-Optics 2010, San Jose,
California United States, 16–21 May 2010, 2010.
Chapter 6 References
107
[124] H. Yun, Z. Lu, Y. Wang, W. Shi, L. Chrostowski, and N. A. Jaeger, "2x2 Broadband
Adiabatic 3-dB Couplers on SOI Strip Waveguides for TE and TM modes," presented
at the CLEO: 2015, San Jose, California United States, 10–15 May 2015, 2015.
[125] J. Xing, Z. Li, Y. Yu, and J. Yu, "Design of polarization-independent adiabatic splitters
fabricated on silicon-on-insulator substrates," Optics Express, vol. 21, no. 22, pp.
26729-26734, 2013.
[126] K. Solehmainen, M. Kapulainen, and M. Harjanne, "Adiabatic and Multimode
Interference Couplers on Silicon-on-Insulator," IEEE Photonics Technology Letters, vol.
18, no. 21, pp. 2287 - 2289, 2006.
[127] M. H. Hu, J. Z. Huang, R. Scarmozzino, M. Levy, and R. M. Osgood, "A Low-Loss and
Compact Waveguide Y-Branch Using Refractive-Index Tapering," IEEE PHOTONICS
TECHNOLOGY LETTERS, vol. 9, no. 2, pp. 203-205, 1997.
[128] M. Rangaraj, M. Minakata, and S. Kawakami, "Low loss integrated optical Y-branch,"
Journal of Lightwave Technology, vol. 7, no. 5, pp. 753 - 758, 1989.
[129] A. Sakai, T. Fukazawa, and T. Baba, "Low loss ultra-small branches in a silicon
photonic wire waveguide," IEICE TRANS. ELECTRON., vol. E85-C, no. 4, pp. 1033-
1038, 2002.
[130] O. Hanaizumi, M. Miyagi, and S. Kawakami, "Wide Y-Junctions with Low Losses in
Three-Dimensional Dielectric Optical Waveguides," IEEE Journal of Quantum
Electronics vol. QE-21, pp. 168-173, 1985.
[131] S.-H. Yang, M. L. Cooper, P. R. Bandaru, and S. Mookherjea, "Giant birefringence in
multi-slotted silicon nanophotonic waveguides," Optics Express, vol. 16, no. 11, pp.
8306-8316, 2008.
[132] Z. Xiao et al., "Ultra-compact low loss polarization insensitive silicon waveguide
splitter," Optics Express, vol. 21, no. 14, pp. 16331-16336, 2013.
[133] X. Li, H. Xu, X. Xiao, Z. Li, J. Yu, and Y. Yu, "Compact and low-loss silicon power
splitter based on inverse tapers," Optics Letters, vol. 38, no. 20, pp. 4220-4223, 2013.
[134] W. S. Cleveland, "Robust Locally Weighted Regression and Smoothing Scatterplots,"
Journal of the American Statistical Association, vol. 74, no. 368, pp. 829-836, 1979.
[135] T. E. Murphy, J. T. Hastings, and H. I. Smith, "Fabrication and characterization of
narrow-band Bragg-reflection filters in silicon-on-insulator ridge waveguides," Journal
of Lightwave Technology, vol. 19, no. 12, pp. 1938-1942, 2001.
[136] X. Wang, W. Shi, H. Yun, S. Grist, N. A. F. Jaeger, and L. Chrostowski, "Narrow-band
waveguide Bragg gratings on SOI wafers with CMOS-compatible fabrication process,"
Optics Express, vol. 20, no. 14, pp. 15547-15558, 2012.
[137] K.-C. Shin et al., "Low threshold current density operation of GaInAsP-InP laser with
multiple reflector microcavities," IEEE Photonics Technology Letters, vol. 7, no. 10, pp.
1119-1121, 1995.
[138] T. Mukaihara, N. Yamanaka, M. N. Iwai, S. Arakawa, T. Ishikawa, and A. Kasukawa,
"Integrated GaInAsP Laser Diodes with Monitoring Photodiodes Through
Chapter 6 References
108
Semiconductor/Air Bragg Reflector (SABAR)," IEEE Journal of Selected Topics in
Quantum Electronics, vol. 5, no. 3, pp. 469-475, 1999.
[139] C. F. R. Mateus, M. C. Y. Huang, Y. Deng, A. R. Neureuther, and C. J. Chang-Hasnain,
"Ultrabroadband Mirror Using Low-Index Cladded Subwavelength Grating," IEEE
Photonics Technology Letters, vol. 16, no. 2, pp. 518-520, 2004.
[140] T. C. Kleckner et al., "Design, Fabrication, and Characterization of Deep-Etched
Waveguide Gratings," Journal of Lightwave Technology, vol. 23, no. 11, pp. 3832-3842,
2005.
[141] C. A. Barrios, V. R. Almeida, R. R. Panepucci, B. S. Schmidt, and M. Lipson,
"Compact Silicon Tunable Fabry-Perot Resonator With Low Power Consumption,"
IEEE Photonics Technology Letters, vol. 16, no. 2, pp. 506-508, 2004.
[142] M. W. Pruessner, T. H. Stievater, and W. S. Rabinovich, "Integrated waveguide Fabry-
Perot microcavities with silicon/air Bragg mirrors," Optics Letters, vol. 32, no. 5, pp.
533-535, 2007.
[143] S. Gao, Y. Wang, K. Wang, and E. Skafidas, "High contrast circular grating reflector on
silicon-on-insulator platform," Optics Letters, vol. 41, no. 3, pp. 520-523, 2016.
[144] T. E. Murphy, "Design, Fabrication and Measurement of Integrated Bragg Grating
Optical Filters," Ph.D., Department of Electrical Engineering and Computer Science,
MIT, 2001.
[145] K. Ghoumid, I. Elhechmi, S. Mekaoui, C. Pieralli, and T. Gharbi, "Analysis of optical
filtering in waveguides with a high index modulation using the extended coupled mode
theory by hybridization of a matrix method," Optics Communications, vol. 289, pp. 85-
91, 2013.
[146] E. Hecht, "Diffraction," in Optics5th ed.: Pearson, 2017.
[147] M. Born and E. Wolf, "Elements of the theory of diffraction," in Principles of Optics7th
ed.: Cambridge University Press, 1999.
[148] P. J. Bock et al., "Subwavelength grating periodic structures in silicon-on-insulator: a
new type of microphotonic waveguide," Optics Express, vol. 18, no. 19, pp. 20251-
20262, 2010.
[149] R. Halir et al., "Waveguide sub-wavelength structures: a review of principles and
applications," LASER & PHOTONICS REVIEWS, vol. 9, no. 1, pp. 25-49, 2015.
[150] G. T. Reed, "Passive Silicon Photonic Devices," in Silicon Photonics: The State of The
Art: John Wiley & Sons, 2008.
[151] C. J. Chae, D. Y. Choi, E. Skafidas, and Y. T. Lee, "Compact Waveguide Resonator
Filter for Wavelength-Selective Reflection and Rejection in Silicon Waveguides,"
presented at the Opto-Electronics and Communications Conference 2015, 2015.
[152] P. Sanchis, J. Blasco, A. Martinez, and J. Marti, "Design of Silicon-Based Slot
Waveguide Configurations for Optimum Nonlinear Performance," Journal of Lightwave
Technology, vol. 25, no. 5, pp. 1298-1305, 2007.
Chapter 6 References
109
[153] R. Sun et al., "Horizontal single and multiple slot waveguides: optical transmission at λ
= 1550 nm " Optics Express, vol. 15, no. 26, pp. 17967-17972, 2007.
[154] H. Zhang et al., "Polarization splitter using horizontal slot waveguide," Optics Express,
vol. 21, no. 3, pp. 3363-3369, 2012.
[155] J. Zhang, H. Zhang, and S. Chen, "Mode converter between channel waveguide and slot
waveguide," presented at the Photonics Global Conference (PGC), 2012, 2012.
[156] H. Zhang et al., "Efficient and broadband polarization rotator using horizontal slot
waveguide for silicon photonics," Applied Physics Letters, vol. 101, no. 021105, 2012.
[157] X. Guan, P. Chen, S. Chen, P. Xu, Y. Shi, and D. Dai, "Low-loss ultracompact
transverse-magnetic-pass polarizer with a silicon subwavelength grating waveguide,"
Optics Letters, vol. 39, no. 15, pp. 4514-4517, 2014.
[158] L. Sánchez, S. Lechago, and P. Sanchis, "Ultra-compact TE and TM pass polarizers
based on vanadium dioxide on silicon," Optics Letters, vol. 40, no. 7, pp. 1452-1455,
2015.
[159] P. Ma, P. Strasser, P. Kaspar, and H. Jackel, "Compact and Integrated TM-Pass
Photonic Crystal Waveguide Polarizer in InGaAsP–InP," IEEE Photonics Technology
Letters, vol. 22, no. 24, pp. 1808-1810, 2010.
[160] H. Sun, A. Chen, and L. R. Dalton, "A reflective microring notch filter and sensor,"
Optics Express, vol. 17, no. 13, pp. 10731-10737, 2009.
[161] M. S. Rasras et al., "Demonstration of a Tunable Microwave-Photonic Notch Filter
Using Low-Loss Silicon Ring Resonators," JOURNAL OF LIGHTWAVE
TECHNOLOGY, vol. 27, no. 12, pp. 2105-2110, 2009.
[162] K. Wang et al., "Short-range infrared optical wireless communications — Systems and
integration," presented at the Photonics Society Summer Topical Meeting Series (SUM),
2016 IEEE, 2016.
[163] S. Tao, B. Yang, H. Xia, H. Wang, and G.-Q. Lo, "An Optical Power Splitter With
Variable Power Splitting Ratio," IEEE Photonics Technology Letters, vol. 23, no. 14,
pp. 1004-1006, 2011.
[164] M. Uenuma and T. Motooka, "Temperature-independent silicon waveguide optical
filter," Optics Letters, vol. 34, no. 5, pp. 599-601, 2009.
[165] U. Koren et al., "A 1.3-μm wavelength laser with an integrated output power monitor
using a directional coupler optical power tap," IEEE Photonics Technology Letters, vol.
8, no. 3, pp. 364-366, 1996.
[166] K. Nemoto, T. Kita, and H. Yamada, "Narrow spectral linewidth wavelength tunable
laser with Si photonic-wire waveguide ring resonators," presented at the Group IV
Photonics (GFP), 2012 IEEE 9th International Conference on, 2012.
[167] T. Matsumoto et al., "Narrow spectral linewidth full band tunable laser based on
waveguide ring resonators with low power consumption," presented at the Optical Fiber
Communication (OFC), collocated National Fiber Optic Engineers Conference, 2010
Conference on (OFC/NFOEC), 21-25 March 2010, 2010.
Minerva Access is the Institutional Repository of The University of Melbourne
Author/s:
Wang, Yang
Title:
Design, fabrication, and characterization of passive silicon photonic device
Date:
2017
Persistent Link:
http://hdl.handle.net/11343/191856
File Description:
Design, fabrication, and characterization of passive silicon photonic device
Terms and Conditions:
Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the
copyright owner. The work may not be altered without permission from the copyright owner.
Readers may only download, print and save electronic copies of whole works for their own
personal non-commercial use. Any use that exceeds these limits requires permission from
the copyright owner. Attribution is essential when quoting or paraphrasing from these works.