Professional Documents
Culture Documents
CAMPINAS
2017
Mário César Mendes Machado de Souza
Tese apresentada ao
Instituto de Fı́sica Gleb Wataghin
da Universidade Estadual de Campinas
como parte dos requisitos exigidos para a
obtensão do tı́tulo de Doutor em ciências
Campinas
2017
Agência(s) de fomento e nº(s) de processo(s): FAPESP, 2014/04748-2, 2015/20525-6;
CNPq, 156281/2013-9
Ficha catalográfica
Universidade Estadual de Campinas
Biblioteca do Instituto de Física Gleb Wataghin
Lucimeire de Oliveira Silva da Rocha - CRB 8/9174
COMISSÃO JULGADORA:
CAMPINAS
2017
Acknowledgments
This thesis was made possible by numerous people whose support and guidance were funda-
mental to get me here.
First I thank my family. My brand new wife Solène, my parents Luiza and Augusto, my siblings
Isabel, Guto, Érika and Livia, and my nephew Lucas. Your trust has been an enormous source of
strength from the beginning. Thanks for always being there for me.
I am very glad to have had Newton Frateschi as my advisor. Thanks for teaching me a whole
lot and for helping me explore new ideas. Your constant encouragement and positive attitude
made this journey a smooth and fun one.
I want to say a special thanks to my dear friends Felipe Vallini, Luis Barea and Antônio von
Zuben “Totó”. You guys taught me a lot about physics and life and made me realize that science
and engineering are really enjoyable when you have nice people to share it with. I also want
to include Gustavo Wiederhecker to this special batch. Thanks for always being engaged in my
research and for pushing me to do my best.
All the LPD folks (Thiago, Guilherme, Felipe Santos, Gustavinho, Debora, André, Carlos Gois,
Rodrigo, Yovanny, Jorge, Laı́s, and others). Thanks for your help all around, from GDSpy scripts
to Matlab/Mathematica programs and cleanroom recipes. We all know how painful experimental
research in Brazil can be, and you guys made my life easier. Many thanks to the people from CCS
as well!
Thanks to all the UC San Diego folks, Karim, Jordan, Cheng-yi, Rajat, Joe, Andrew, Brandon
and so on. It was fun working with you all. I also appreciate all the help from the Nano3 folks,
especially Maribel Montero. Special thanks to Shaya Fainman, who welcomed me to his group and
was always open to new ideas (“go for it!”).
I want to acknowledge my dear friends Daniel, Fernando “Piá” and Anderson “Truta”. We’ve
been through a lot in the past 11 years!
Finally, I acknowledge the funding agencies FAPESP and CAPES (grants 2014/04748-2 and
2015/20525-6) for their financial support.
Resumo
Após duas décadas de progresso contı́nuo, a fotônica integrada apresenta-se como uma tecnolo-
gia indispensável, exibindo soluções para importantes demandas tecnológicas atuais como o tráfego
e processamento de sinais ópticos ultra-rápidos. Ao mesmo tempo, ela permite avanços substan-
ciais em áreas emergentes como o “laboratório-no-chip” (lab-on-a-chip). No entanto, enquanto
funcionalidades básicas necessárias para a maioria das aplicações (fontes de luz, moduladores, fil-
tros, linhas de atraso, detectores, etc.) já estão disponı́veis em uma variedade de dispositivos
e plataformas, alguns desafios ainda permanecem. Nos últimos quatro anos, estivemos interes-
sados em identificar alguns desses desafios e fornecer abordagens interessantes para enfrentá-los.
Esta tese, que engloba uma parcela importante dessas investigações, pode ser dividida em dois
tópicos. No primeiro, apresentamos microresonadores acoplados como dispositivos que permitem
um controle espectral flexı́vel e reconfigurável. Explorando as caracterı́sticas desses dispositivos,
demonstramos novas funcionalidades como o controle reconfigurável do splitting entre ressonâncias,
fornecemos novas ferramentas de modelagem como uma teoria de modos acoplados modificada e
propomos um modulador que emprega anéis acoplados, capaz de superar a limitação entre eficiên-
cia de modulação e largura de banda enfrentada por moduladores baseados em um único anel. No
segundo tópico apresentamos o desenvolvimento de um espectrômetro a transformada de Fourier
integrado em um chip, utilizando fotônica de silı́cio. Os desafios para obter esse dispositivo, como
a não-idealidade inerente à plataforma de silı́cio (dispersão e não-linearidade termo-ótica) são dis-
cutidos em detalhe, além da demonstração experimental que indica como tal dispositivo pode abrir
caminho para espectrômetros portáteis robustos e econômicos.
Abstract
After two decades of continuous progress, integrated photonics has proven its indisputable
role as an enabling technology. It addresses important technological demands of our time such as
ultrafast optical data transfer and processing while allowing substantial progress in emerging areas,
including lab-on-a-chip. Although the basic functionalities required for most applications (light
sources, modulators, filters, delay lines, detectors, etc.) are now available in a variety of designs
and platforms, significant challenges still remain and room for improvement can still be found.
During the last four years, we have been interested in identifying some of these challenges and in
providing interesting approaches to tackle them. This thesis, encompassing an important share
of such investigations, can be divided into two topics. First, we present coupled microresonators
as devices allowing for flexible and reconfigurable spectral control. Exploiting these devices, we
demonstrate novel functionalities like the reconfigurable resonance-splitting control, we provide
novel modeling tools such as a modified coupled mode theory, and we propose a coupled-ring
modulator that overcomes the trade-off between modulation efficiency and bandwidth faced by
single microrings modulators. The second topic addresses the realization of an on-chip Fourier
transform spectrometer using silicon photonics. We discuss the challenges of realizing such device
due to non-idealities inherent to the silicon platform (dispersion and thermo-optic non-linearity)
and we provide an experimental demonstration indicating how this device can pave the way for
robust and cost-effective portable spectrometers.
Preface
Before delving into technical matters, I would like to provide some background on how this
thesis came to be. I started my undergrad studies in physics at Unicamp in 2007 believing I was on
my way to become a great cosmologist or something like that. In reality, as most of my colleagues,
I had no clue what I wanted to do. Three years later, I was accepted in a double degree program
to study optical engineering at the Institut d’Optique (Supoptique) in France.
By that time I did not have a special taste to experimental work – I had not had much of that –
or optics for that matter, but at Supoptique I discovered/developed my passion for applied physics
and experimentation. In 2012 I did a six-month-long internship at Coherent Inc., a leading laser
manufacturer in the United States. At Coherent I worked on the development of high-power lasers
based on vertical-external-cavity surface-emitting-lasers (VECSELs). This was my first somehow
in-depth contact with semiconductor technology.
After my internship and having graduated from Supotique, I went back to Unicamp to finish my
degree. It was a tough time going back to a classroom to solve canonical book exercises. Willing
to get involved in experimental research, I joined the Device Research Lab (LPD) as a research
assistant in 2013. That’s when I started working with integrated photonics and microfabrication.
Later in 2013, I started my PhD in the same group. Building upon the great work done by Luis
Barea – at the end of his PhD at the time – my research would be focused on exploiting coupled
microcavities for applications in integrated photonics. A big part of this thesis consists of some of
the projects I developed on this topic.
In 2016, I went to the University of California San Diego (UCSD) as a visiting graduate re-
searcher. Working with prof. Shaya Fainman’s group, I had the chance to be involved with many
interesting projects in silicon photonics, some of which made their way to this thesis as well. The
implementation of an on-chip Fourier Transform spectrometer deserves a whole chapter, whereas
two other projects, resonators supporting multiple transversal waveguide modes and silicon mi-
crorings with ferrofluid cladding, are briefly presented in Appendix A.
Now, this thesis closes an exciting period of learning and exploration. I hope you enjoy the
reading.
List of publications
Journals
1 M. C. M. M. Souza, A. Grieco, N. C. Frateschi and Y. Fainman, “Fourier transform spec-
trometer on silicon with thermo-optic non-linearity and dispersion correction,” arXiv:1710.00061,
(2017). accepted in Nature Communications
Book chapters
9 M. C. M. M. Souza, G. F. M. Rezende, A. A. G. von Zuben, G. S. Wiederhecker, N.
C. Frateschi, and L. A. M. Barea, “Tunable Photonic Molecules for Spectral Engineering in
Dense Photonic Integration,” in Future Trends in Microelectronics: Journey into the Un-
known, S. Luryi, J. Xu, and A. Zaslavsky, eds., 1st ed. (John Wiley & Sons, Inc., 2016) pp.
337–348.
Full proceedings
10 G. F. M. Rezende, M. C. M. M. Souza, and N. C. Frateschi, “Limitations of Coupled
Mode Theory to Model Coupled Microresonators ‘Dark States’,” in 30th Symposium on Mi-
croelectronics Technology and Devices, IEEE (2015).
Conference abstracts
13 A. El Amili, M. C. M. M. Souza, F. Vallini, N. C. Frateschi, and Y. Fainman, “Silicon
Microring with Ferrofluid Cladding,” in Conference on Lasers and Electro-Optics (Optical
Society of America, 2017), p. SM1N.2.
7 Conclusion 95
Bibliography 97
A Contribution to other projects 107
A.1 Magnetically controllable microrings . . . . . . . . . . . . . . . . . . . . . . . . . . 107
A.2 Hybrid multimode resonators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
A.3 Differential sensing with coupled-resonators . . . . . . . . . . . . . . . . . . . . . . 110
Chapter 1
Introduction
Photonic Chip
Telecommunications Lab on-a-chip
Microwave Photonics
AlphaNOV
Figure 1.1: Photonic integration brings multiple optical and optoelectronic components to a singe chip. This
technology brings solutions to bottlenecks in telecommunications and optical interconnects for datacenters while
enabling novel applications such as portable spectroscopy and lab-on-a-chip.
INTRODUCTION 14
AWG are highly sensitive to fabrication variations/imperfections, and SWIFTS require out-of-chip
imaging systems and arrayed photodetectors.
The Fourier transform spectrometer (FTS) with true time-delay, ubiquitous in its bulky free-
space realization [33], represents an appealing alternative design towards broadband operation, fine
resolution and robustness. Here, we investigate the implementation of a silicon-photonics-based
on-chip Fourier Transform spectrometer (Si-FTS), its challenges and benefits.
Chapter 4 presents the basic concepts related to microring resonators and coupled-microring
designs. In Chapter 5, we discuss the modeling of coupled microring resonators based on the
transfer matrix method (TMM) and on the temporal coupled mode theory (TCMT). In this chap-
ter we also discuss a modified TCMT (m-TCMT) to deal with a coupled-ring system presenting
“quasi-dark states”. Finally, we present some applications of coupled microring resonators in Chap-
ter 6. Particularly, we discuss their use for reconfigurable resonance-splitting control, all-optical
wavelength conversion and optical modulation.
A brief description of additional projects we have worked on and which are not part of the
main body of this thesis can be found in Appendix A.
17
Chapter 2
Research groups developing integrated photonic circuits may have their experimental device
realized through in-house fabrication or through a multi-project-wafer (MPW) run, in which case
these devices are fabricated in a foundry. Both approaches have their benefits and drawbacks.
MPW runs bring the convenience of designing a complex photonic circuitry and not worrying
about the fabrication steps – provided the design is in accordance with the foundry rules. In
addition, it allows groups without access to world-class cleanroom facilities to make substantial
contributions to the field. On the other hand, it may take several months to get a device fabricated
through a MPW run, and minimum feature sizes (typically around 150 nm) still do not compete
with what can be achieved in-house using electron beam (e-beam) lithography (around 20 nm).
In addition to fast turnaround and fine resolution, in-house microfabrication allows for optimized
processes.
The devices discussed in this thesis were realized using both foundry tape-outs and in-house mi-
crofabrication. Section 2.1 discusses our involvement in tape-outs, while Section 2.2 summarizes
in-house fabrication efforts.
In addition to fabrication, this chapter presents our involvement in device characterization
in Section 2.3. We discuss the basic setup for electro-optic testing and our contribution with
semi-automation and data post-processing.
2.1. FABLESS DESIGN THROUGH FOUNDRY SERVICES 18
Figure 2.1: Schematic cross-section exemplifying the various structures that can be realized in a typical silicon
photonics foundry run (imec-ePIXfab iSiPP50G). Reproduced from www.europractice-ic.com
2.1. FABLESS DESIGN THROUGH FOUNDRY SERVICES 19
a b
10µm
10µm
Figure 2.2: Multi-project wafer (MPW) run of passive devices with IMEC. a. Schematic of the mask layout
containing multiple photonic structures. The mask dimensions are approximately 5.5 mm by 3 mm. The microring
resonators investigated in this work are highlighted by the red rectangle. b., c. Optical micrography of single and
coupled microring structures fabricated at this tape-out.
2.1. FABLESS DESIGN THROUGH FOUNDRY SERVICES 20
lateral
junction
interleaved e
junction
Shallow Si etch N++ doping
Deep Si etch P++ doping
N doping Metal via
20 µm P doping Metal
Opened window
Figure 2.3: Layout mask of the coupled-ring modulator designed for tape-out with Sandia National Labs. a.
Complete device including electrical pads (GSGSG) for differential electrical drive. b. Coupled-ring without metal
interconnect layer, showing the doping regions. c. Design of lateral and interleaved PN junctions. d. Example of
a single ring modulator, used for comparison. e. Layer library.
2.2. IN-HOUSE MICROFABRICATION 21
a b c
e f
3 µm 1 µm
100 µm 2 µm 5 µm 20 µm
a b c
Metal
SiO2 (PECVD)
Silicon
SiO2 (thermal)
HSQ
PMMA
SOI wafer + alignment marks Channel waveguides Cladding deposition
f e d
Figure 2.5: Summary of fabrication steps required to realize a silicon MZI with integrated metal microheaters.
Detailed description is provided in the text.
a
b
c
I-V +
source -
Electric
Tunable Probes
Laser
Detector
Polarization GRIN rod
...
Figure 2.6: a. Experimental apparatus showing the optical and electrical equipment including micro-lenses,
translation stages, optical microscope and electric probes. b. Zoomed-in picture of the chip mount showing a
sample together with the input-output micro-lenses and electric probes. c. Schematic of the semi-automated setup
used to characterize the transmission spectrum of microresonators with integrated microheaters. The instrument
control is done using Matlab routines via GPIB or LAN connection. I-V source: current/voltage source and meter;
TEC: temperature controller; DAQ: data acquisition module;
2.3. CHARACTERIZATION OF PHOTONIC DEVICES 24
a 0.2 0.2
f
1
Voltage (V)
c
Voltage (V)
0.1
0.1
Normalized Transmission
d
0 1
0 4 8 12 16 20 24
Normalized Transmission
Time (s)
0.5 0.5
0.2
b
0
1
Voltage (V)
0.1 e
0.5
0 0 0
1460 1480 1500 1520 1540 1560 1580 1460 1480 1500 1520 1540 1560 1580 1529.8 1530 1530.2
Wavelength (nm) Wavelength (nm) Wavelength (nm)
Figure 2.7: Post-processing steps to obtain the normalized transmission spectrum of a microring resonator. a.
As-measured temporal traces: transmitted power (blue); laser trigger (red). Using the laser trigger, the wavelength
axis is calibrated, resulting in the trace shown in b. c. Background extraction. d. Determination of the average
background level to be considered the “1” level of the normalized trace. This step is particularly important for
noisy background. e. Normalized transmission spectrum. f. The resonances are easily resolved thanks to the
high-resolution achieved with a continuous sweep of the laser source allied to high-speed data acquisition
.
output power between -9 dBm and 14 dBm. Light is coupled into and out of the chip using tapered
fibers or micro-lenses and on-chip edge-couplers based on inverted tapers [47]. The input-output
fibers are positioned using precise translation stages (Thorlabs NanoMax) driven by piezoelectric
controllers. To assure that light couples to the quasi-TE mode of the silicon waveguides, in-fiber
polarization rotators are positioned between the laser and the tapered fibers. The output fiber can
be directly connected to an InGaAs powermeter (Agilent 8163B) which delivers an analog electrical
signal proportional to the measured power to the data acquisition module (DAQ).
The samples are mounted on top of a temperature stabilized station (20 o C) and under an
optical microscope coupled either to a CCD camera for visible light or to an infra-red (IR) camera.
The CCD camera is used for imaging and alignment purposes, whereas the IR camera is used for
alignment and to image the optical mode from scattered light coming out of the resonators.
Multiple low-frequency electric probes can be used to actuate on-chip structures such as inte-
grated microheaters. They are connected to a sourcemeter (Keithley 2400) allowing for simulta-
neous precise current drive (up to 2.1 A) and voltage monitoring (up to 210 V) or vice-versa. The
micro-probes generally have 50×50 µm2 tips and introduce total parasitic series resistance of less
than 4 ohm.
All the equipment is connected to a computer and controlled via Matlab routines we imple-
mented. Besides simple and scalable, instrument control and data acquisition through Matlab
brings the advantage of allowing for data processing in the same platform. Fig.2.7 illustrates
this seamless operation, exemplifying the steps required to obtain the transmission spectrum of
2.3. CHARACTERIZATION OF PHOTONIC DEVICES 25
a microring resonator. First, the analog output from the powermeter (blue trace of Fig.2.7a) is
acquired in the data acquisition module (DAQ) along with a trigger signal delivered by the laser
source (red trace). The trigger sends a square waveform every 5 nm and allows to calibrate the
wavelength axis, resulting in the trace of Fig.2.7b. Using smoothing and interpolation techniques,
the background dependence with wavelength can be determined and filtered out (Fig.2.7c,d), re-
sulting in the normalized transmission spectrum of Fig.2.7e. In particular, slowly-varying features
such as wavelength-dependent attenuation and oscillations caused by spurious reflections are first
detected using a coarse smoothing filter (Fig.2.7c) and used for normalization of the broadband
spectrum. Subsequently, the background transmission is determined by removing points belonging
to the resonances (not shown in the figure) and its average value (bottom of red points in Fig.2.7d)
is re-normalized to 1. This last step helps to mitigate the effects of fast-varying spectral features
in the normalization of the transmission spectrum. Notice these techniques must be applied with
caution as they may filter-out spectral features that might be of interest. Fig.2.7f shows a sharply
resolved resonance. Sweeping the laser at 5 nm/s and setting the DAQ for an acquisition rate of 10
kHz, one is able to acquire a broadband window of 100 nm with high resolution of one data point
every 0.5 pm in only 20 secs. A lower sweeping speed and higher acquisition rate are possible if
very narrow features are to be detected. Also, this technique yields high wavelength repeatability,
with wavelength errors between different laser scans on the order of one picometer.
Once the spectrum is properly obtained, a variety of post-processing routines may be used to
extract useful parameters such as the quality-factor and free-spectral range (FSR). Most of the
results discussed in the next chapters were obtained using this integration between data acquisition
and processing.
2.3. CHARACTERIZATION OF PHOTONIC DEVICES 26
.
27
Chapter 3
In the introductory chapter we discussed the interest in realizing compact and robust miniatur-
ized spectrometers for portable instrumentation and we discussed the interests of realizing these
devices using silicon photonics technology. Moreover, we argued that the silicon photonics-based
Fourier transform spectrometer (Si-FTS) represents an appealing design towards broadband oper-
ation, fine resolution and robustness to fabrication variations.
In this chapter, we discuss in detail the implementation of a Si-FTS on the SOI platform
with integrated microheaters. We show how the challenges related to thermo-optic non-linearity,
thermal expansion and dispersion can be properly understood, tackled and, moreover, ultimately
result in enhanced performance.
In Section 3.1, we introduce our device and we discuss the principle of operation of an ideal
FTS in which the effects of dispersion, thermo-optic non-linearity and thermal expansion are
neglected. In Section 3.2, we show how the power spectral density (PSD) of a light source and
the interferogram measured with the real Si-FTS can still be related through a simple Fourier
transform (FT), provided the optical frequency and time delay are corrected to account for the
aforementioned non-idealities. In Section 3.3 we demonstrate the calibration of the Si-FTS using
a tunable laser source. Finally, in Section 3.4 we present the successful recovery of a broadband
spectrum — the amplified spontaneous emission (ASE) of an erbium-doped fiber amplifier (EDFA).
In Section 3.5 we discuss some interesting features of the Si-FTS design and its potential for high-
performance on-chip spectrometry.
I started working on the Si-FTS while at UCSD working with prof. Shaya Fainman and his
post-doc researcher Andrew Grieco. The subject of this chapter can be found in arXiv (1710.00061)
and, at the time of writing, is under peer-reviewing at Nature Communications.
3.1. EXPERIMENTAL DEVICE AND PRINCIPLE OF OPERATION 28
a b c
70 µm
100 µm
d e
Figure 3.1: a. Schematic of an integrated MZI with metal microheaters b. Optical micrography of the experimental
device with a total footprint of 1 mm2 . c. Dark field optical micrography of one of the MZI arms. d. SEM image
of the broadband power splitter/combiner. e. Schematic of the cross-section on top of the arms.
Our Si-FTS is based on a Mach-Zehnder Interferometer (MZI) with integrated metal micro-
heaters as shown in Fig.3.1. The device is fabricated in a 250-nm-thick SOI platform in full
compatibility with standard silicon photonics foundry processes (see Section 2.2). The external
light is butt-coupled into the chip using inverse tapers (150 nm wide at the tip) and adiabaticaly
transitions to the highly confined quasi-TE mode of the access strip waveguide (550 nm wide) be-
fore splitting in the two arms of the interferometer and subsequently recombining into the output
waveguide through broadband y-branch couplers (Fig.3.1d) [48]. The output light is coupled out
of chip directly into a photodetector. Each arm of the MZI consists of a spiral (Fig.3.1c) with total
length of 30.407 mm. The separation between adjacent waveguides in the spiral sections is 5 µm
center-to-center and the bending radius is 10 µm. Each arm is covered by independently actuated
nichrome microheaters. The propagation losses of the waveguides are estimated to be around 2
dB/cm. The device total footprint is 1 mm2 .
The power transmission of a MZI depends on the optical frequency ν and on the phase difference
between the two arms as
TMZI (ν, ∆φ) = T0 (ν) + T (ν) cos (∆φ) , (3.1)
where T0 and T are transmission amplitudes that incorporate any frequency dependence of the
system such as the limited bandwidth of the power splitter/combiner and the frequency-depended
loss of the waveguides1 . The phase difference is
2πν
∆φ(ν) = [neff,1 (ν)L1 − neff,2 (ν)L2 ] , (3.2)
c
1
see Appendix C for the dependence of T (ν) with attenuation and power coupling coefficients.
3.2. SPECTROMETER MODELING 29
where c is the speed of light, neff,i and Li are the effective index and the total length of arm i.
When light with power spectral density PSD(ν) passes through the device, the output power
will also present a constant term in addition to a ∆φ-dependent term (see Appendix C),
Now, consider the response of an idealized device. In this case T (ν) = 1, the two arms are
identical with length L, the effective indices are identical and dispersionless, neff,i (ν) ≡ neff , and
the effective index change due to temperature change ∆T depends only on a linear thermo-optic
coefficient (TOC) ∂T n, such that ∆neff = ∂T n∆T .2 The time delay between the arms of the MZI
is defined as τ = Lc ∂T n∆T and the phase difference is simply
The phase difference in the form 2π×frequency×delay establishes a direct FT relation between
I(τ ) and PSD(ν), with the conjugate variables ν and τ ,
Z +∞
I(τ ) = PSD(ν)ej2πντ dν = F [PSD(ν)] (3.6)
−∞
where F [ ] denotes the Fourier Transform. Thus, PSD(ν) can be directly obtained from the inverse
FT (IFT) of the interferogram, Z +∞
PSD(ν) = I(τ )e−j2πντ dτ = F −1 [I(τ )] (3.7)
−∞
The simplifications carried in this section depend on various waveguide parameters and therefore
are valid for the quasi-TE mode of a 250 nm-by-550 nm SOI waveguide and similar designs. They
also depend on the temperature range achieved with the heaters and the bandwidth of the incoming
light, which are considered close to our experimental conditions, namely, temperature excursion
lower than 100 K and bandwidth around 10 THz. Nonetheless, our discussion provides a general
procedure that is readily applicable to other device designs and operation conditions and allows to
identify terms that might need to be incorporated as corrections.
Compared to the discussion provided in Section 3.1, the Si-FTS with thermal tuning includes
other effects that must be taken into account. First, the strong mode dispersion of silicon waveg-
uides causes significant frequency dependence on the effective index. Second, a large temperature
excursion is required to achieve large phase unbalances and the non-linearity of the thermo-optic
response must be considered. The large temperature excursion also induces changes in the arm
length (∆L) due to thermal expansion. Finally, chip-scale variability [9, 26] and fabrication im-
perfections often introduce small differences between the two arms of the MZI, identical by design.
Such variations may affect the arm length (δL) as well as the effective index (δn(ν)). Assuming the
heater on top of arm 1 (H1 ) is actuated and including the deviations from the designed parameters
in arm 2, the effective indices and arm lengths are
The frequency-dependent effective index and the difference of effective index between the two
arms are expressed as series expansions around the central optical frequency ν0 (193.414 THz in
our case) up to third order,
1 1
neff (ν) = neff |ν0 + ∂ν n ∆ν + ∂ν 2 n ∆ν 2 + ∂ν 3 n ∆ν 3 , (3.9)
2 6
1 1
δn(ν) = δ(neff |ν0 ) + δ(∂ν n) ∆ν + δ(∂ν 2 n) ∆ν 2 + δ(∂ν 3 n) ∆ν 3 . (3.10)
2 6
As discussed in Appendix B, for temperatures rising up to 100 K above room temperature
(i.e. up to ∼400 K), it suffices to consider the first and the second order temperature-dependence
of the TOC and of the thermal expansion coefficient, respectively. These lead to a second order
and third-order temperature dependence of the effective index change ∆neff and of the arm length
change ∆L,
3.2. SPECTROMETER MODELING 31
1 2 1 3
∆neff (ν, ∆T ) = ∂T n + ∂ν,T n ∆ν + ∂ν 2 ,T n ∆ν + ∂ν 3 ,T n ∆ν ∆T + (3.11)
2 6
1 1 1
∂T 2 n + ∂ν,T 2 n ∆ν + ∂ν 2 ,T 2 n ∆ν + ∂ν 3 ,T n ∆ν ∆T 2
2 3
2 2 6
∆L(∆T ) = L α1 ∆T + α2 ∆T 2 + α3 ∆T 3 .
(3.12)
Modifications of the effective index due to thermal expansion of the waveguide in its cross-section
is negligible compared to the other effects and are not included.
The phase difference ∆φ can be expressed in an efficient way to sort and evaluate the con-
tribution of the various orders of frequency detuning ∆ν and time delay τ introduced by the
temperature difference ∆T . Substituting eqs.3.8-3.12 into eq.3.2 and manipulating the resulting
expression we write
5
" 2 3 #
X ηi i χi + ξi ∆ν χi + ζi ∆ν ζi ∆ν
∆φ(ν, τ ) = ϕ(ν) + 2π τ [ν0 + ∆ν(1 + ξi )] 1 + + + .
ηi
i=1 1
1 + ξi ν0 1 + ξi ν0 1 + ξi ν0
(3.13)
. The first term, ϕ(ν), gathers frequency-dependent terms that do not depend on τ . It is given by
L
with t0 ≡ c
and
δL δL
σ0 = δ(neff |ν0 ) + neff |ν0 and σi = δ(∂ν i n) + ∂ν i n (i = 1, 2, 3) (3.15)
L L
Assuming σ−1 = σ4 = 0, eq.3.14 can be recast in a compact form as
4
X
ϕ(ν) = t0 (σi−1 + ν0 σi ) ∆ν i . (3.16)
i=0
With no dependence in τ , ϕ(ν) does not contribute to the kernel of the transformation between
time and frequency. On the other hand, it is the only term with contribution from δ-terms. As
such, ϕ(ν) may shift and distort the interferogram, but it has no influence on the retrieved PSD
as will be shown.
For now, we focus on the summation containing the frequency×delay terms of interest. In
this form, the summation gathers the remaining terms by increasing order of τ . The terms τ i are
weighted by ηi /η1i , and the time delay itself is related to the temperature difference by τ = η1 ∆T .
3.2. SPECTROMETER MODELING 32
Assuming ∂T 0 n ≡ neff |ν0 and α0 ≡ 1, these terms can be written in a general form as
k
X 1
ηk = t0 ∂T i n αk−i . (3.18)
i=0
i!
The contribution of high order terms compared to the first order is depicted in Fig.3.2 as a function
of the time delay. These plots were obtained using the parameters summarized in Appendix B
(Table 1). The second order (blue trace) contributes significantly and therefore cannot be neglected
even for small delays. The third order term can be neglected for small delays, but it becomes
increasingly important for large delays required for high spectral resolution. It reaches 1% of the
first order contribution at 3.5 ps, and ∼ 10% at 10 ps (Fig.3.2a). Higher order terms are very small
and can be neglected even for large delays. In our experiment, we achieve a maximum time delay
around 1 ps, in which regime it suffices to keep the first and second order contributions (Fig.3.2b).
Multiplying each order in τ in eq.3.13 we have terms of increasing order in ∆ν. The first brackets
[ν0 + ∆ν(1 + ξi )] represent zeroth and first order contributions, while the second brackets assemble
a b 0.10
1.0
0.08
Term contribution
0.8
Term contribution
1
0.06
0.6 η2τ/η1
η3τ2/η1
0.4 0.04
η4τ3/η1
η5τ4/η1
0.2 0.02
0.0 0.00
0 2 4 6 8 10 0.0 0.2 0.4 0.6 0.8 1.0
Arm delay, τ (ps) Arm delay, τ (ps)
Figure 3.2: Contribution of terms is ηi τ i /η1i with respect to the first order η1 τ as a function of the time
delay in the range a. between 0 and 10 ps and b. between 0 and 1 ps – range of our experiments.
3.2. SPECTROMETER MODELING 33
higher orders terms. The parameters ξi , χi and ζi are adimensional and represent, respectively,
the first, second and third order dispersion terms in T −i :
t0 ν0 1 t0 ν02 1 t0 ν03
ξ1 = (∂ν n α1 + ∂ν,T n) χ1 = (∂ν 2 n α1 + ∂ν 2 ,T n) ζ1 = (∂ν 3 n α1 + ∂ν 3 ,T n)
η1 2 η1 6 η1
t0 ν0 1 1 t0 ν02 1 1 t0 ν03 1
ξ2 = (∂ν n α2 + ∂ν,T n α1 + ∂ 2 n) χ2 = (∂ν 2 n α2 + ∂ν 2 ,T n α1 + ∂ν 2 ,T 2 n) ζ2 = (∂ν 3 n α2 + ∂ν 3 ,T n α1 + ∂ν 3 ,T 2 n)
η2 2 ν,T 2 η2 2 6 η2 2
t0 ν0 1 1 t0 ν02 1 1 t0 ν03 1
ξ3 = (∂ν n α3 + ∂ν,T n α2 + ∂ 2 n α1 ) χ3 = (∂ν 2 n α3 + ∂ν 2 ,T n α2 + ∂ 2 2 n α1 ) ζ3 = (∂ν 3 n α3 + ∂ν 3 ,T n α2 + ∂ 3 2 n α1 )
η3 2 ν,T 2 η3 2 ν ,T 6 η3 2 ν ,T
t0 ν0 1 1 t0 ν02 1 1 t0 ν03 1
ξ4 = (∂ν,T n α3 + ∂ 2 n α2 ) χ4 = (∂ν 2 ,T n α3 + ∂ 2 2 n α2 ) ζ4 = (∂ν 3 ,T n α3 + ∂ 3 2 n α2 )
η4 2 ν,T 2 η4 2 ν ,T 6 η4 2 ν ,T
t0 ν0 1 1 t0 ν02 1 1 t0 ν03 1
ξ5 = ( ∂ν,T 2 n α3 ) χ5 = ( ∂ν 2 ,T 2 n α3 ) ζ5 = ( ∂ν 3 ,T 2 n α3 ).
η5 2 2 η5 2 6 η5 2
(3.19)
They can be presented in a compact form analogous to eq.3.18 assuming ∂ν i ,T 0 n ≡ ∂ν i n, α0 ≡ 1
and α4 = α5 ≡ 0:
k
ν0 ∂ηk t0 ν0 X 1
ξk = = ∂ν,T i n αk−i
ηk ∂ν ηk i=0 i!
k
ν02 ∂ 2 ηk 1 t0 ν02 X 1
χk = = ∂ν 2 ,T i n αk−i (3.20)
ηk ∂ν 2 2 ηk i=0 i!
k
ν03 ∂ 3 ηk 1 t0 ν03 X 1
ζk = = ∂ν 3 ,T i n αk−i .
ηk ∂ν 3 6 ηk i=0 i!
The contribution of the various terms in the second brackets is depicted in Fig.3.3 as a function
of the fractional bandwidth ∆ν/ν0 . For ν0 = 193.414 THz, a fractional bandwidth of 0.25, the
upper limit of Fig.3.3a, corresponds to a frequency range from 145 THz to 240 THz (wavelength
range from 1160 nm to 1930 nm), encompassing the full range of operation of expected Si-FTS
designs operating around this ν0 . Even in such large frequency range the contribution of high
a 0.05 b 0.005
χi+ξi ∆ν
(
χi+ζi ∆ν 2
(
ζi ∆ν 3
( (
1+ξ1 ν0 1+ξ1 ν0 1+ξ1 ν0
0.04 i=1
0.004
Term contribution
Term contribution
i=2
i=3
i=4
0.03 i=5 0.003
0.02 0.002
0.01 0.001
0.00 0.000
0.00 0.05 0.10 0.15 0.20 0.25 0.000 0.005 0.010 0.015 0.020
Fractional bandwidth, ∆ν/ν0 Fractional bandwidth, ∆ν/ν0
Figure 3.3: Contribution of each term in the second brackets of eq.3.13 as a function of the fractional
bandwidth in the range a. between 0 and 0.25 –large fractional bandwidth – and b. between 0 and 0.02
– range of our experiments.
3.2. SPECTROMETER MODELING 34
order terms is overall small, with individual terms reaching a maximum of 5% of the linear term
(corresponding to 1). From Fig.3.3a we identify the terms in order τ 2 ∆ν 2 (blue trace), τ ∆ν 3 (red-
dashed trace) and τ 3 ∆ν 3 (green-dashed trace) as those with increasingly significant contribution
for large fractional bandwidth, while the other terms are negligible. In our experiments we used a
broadband source with total bandwidth around 7 THz centered at 193.44 THz, with a maximum
fractional bandwidth around 0.02. In this case, depicted in Fig.3.3b, all the contributions are
smaller than 0.5% and considered negligible.
Based on the these results (Fig.3.2,3.3), we keep only the first and second order terms in τ
and we neglect the high orders in ∆ν altogether, leaving only the unity in the second brackets of
eq.3.13.
The phase difference then simplifies to
η2 2
∆φ(ν, τ ) = ϕ(ν) + 2π{τ [ν0 + ∆ν(1 + ξ1 )] + τ [ν0 + ∆ν(1 + ξ2 )]}. (3.21)
η12
τ + γ2 τ 2 (1 + Fξ ) [ν0 + ∆ν(1 + ξ1 )] ,
∆φ(ν, τ ) = ϕ(ν) + 2π (3.22)
with
(ξ2 − ξ1 ) ∆ν
ν0
Fξ = . (3.23)
1 + ξ1 ∆ν
ν0
a 0.10 b 0.010
0.08 0.008
Term contribution
Term contribution
0.06 0.006
0.04 0.004
∆ν > 0
∆ν < 0
0.02 0.002
0.00 0.000
0.00 0.05 0.10 0.15 0.20 0.25 0.000 0.005 0.010 0.015 0.020
Fractional bandwidth, ∆ν/ν0 Fractional bandwidth, ∆ν/ν0
Figure 3.4: Contribution of Fξ in eq.3.22 as a function of the fractional bandwidth in the range a.
between 0 and 0.25 –large fractional bandwidth – and b. between 0 and 0.02 – range of our experiments.
3.2. SPECTROMETER MODELING 35
conditions, however, the term is smaller than 1% (Fig.3.4b) and can be neglected.
Finally, the phase difference assumes the simple form
with
T ≡ τ + γ2 τ 2
(3.25)
u ≡ ν0 + ∆ν(1 + ξ1 )
as modified time delay and optical frequency. u represents the broadening of the original optical
frequency ν around ν0 by a factor (1 + ξ1 ). In our case, ξ1 is dominated by ∂ν,T n (around 90%).
T represents a correction to the original optical delay τ that includes the nonlinear contribution
γ2 τ 2 , with γ2 dominated by ∂T 2 n (around 95%). The linear coefficient η1 is dominated by ∂T n
(more than 95%).
Substituting eqs.3.24,3.25 in eq.3.4 and performing the change of variables from ν to u in the
integral, Z +∞
1
I(T ) = T (u)PSD(u)ejϕ(u) ej2πuT du
1 + ξ1 −∞
(3.26)
1 jϕ(u)
= F T (u)PSD(u)e
1 + ξ1
where F [ ] denotes the Fourier Transform. Neglecting the constant term (1 + ξ1 )−1 multiplying
the FT, the PSD is then retrieved from the absolute value of the inverse-FT of the interferogram,
normalized by the MZI transfer function T (u),
|F −1 [I(T )]|
PSD(u) = . (3.27)
T (u)
Finally, the frequency axis must be transformed back to the original optical frequency ν,
u−ν
ν= 1+ξ 0 +ν0
PSD(u) −−−−−1−−→ PSD(ν). (3.28)
3.3. SPECTROMETER CALIBRATION 36
where A and ϕν are constant amplitude and phase. Using this relation and eqs.3.25, the interfer-
a ν0
30 λ = 1600 nm, ν = 187.37 THz d
80
power (µW)
K = ∆ν(0.50 +
- 0.03) + (77.4 +
- 0.1)
Optical
15
75
K (W-1)
0
70 K = ∆ν(0.44 +
- 0.03) + (69.5 +
- 0.1)
b
10
power (µW)
65
Optical
c 1 40
optical power
γw (W-1)
Normalized
38
0
36
-1 34
0 0.1 0.2 0.3 0.4 0.5 0.6 186 188 190 192 194 196 198
Heater power (W) Optical frequency, ν (THz)
Figure 3.5: Si-FTS calibration using a tunable laser source. Measurements performed in the C-
band. Data in blue and red are related to heaters H1 and H2 , respectively. a. As-measured interferogram
at 183.37 THz (1600 nm) as a function of the heating power (W). To achieve robust non-linear fitting
we subtract the mean power (red trace) and the envelope (red trace in b) to obtain the curve in c.
The envelope in b is the absolute value of the interferogram’s Hilbert transform. c. The fitted curve
(dashed-red trace) using eq.3.30 adjusts well the experimental trace. d. Parameter K(ν) obtained from
the non-linear fit, adjusted according to eq.3.31. e. Non-linear parameters γW,1 = (36.2 ± 0.3) · 10−3 W−1
and γW,2 = (40.5 ± 0.3) · 10−3 W−1 obtained from the non-linear fit.
3.3. SPECTROMETER CALIBRATION 37
200
15.876 kΩ
160
Voltage (V)
120 15.162 kΩ
80
40
0
0 2 4 6 8 10 12
Current (mA)
Figure 3.6: Current vs voltage (IV) response of heaters H1 (blue) and H2 (red) and calculated electric resistance.
The IV curves confirm the linearity of both heaters.
ogram can be written as a function of the electric power and the original laser frequency as
with
K(ν) = κτ (1 + ξ1 ) ∆ν + κτ ν0
(3.31)
γW = κτ γ2 .
K(ν) and γW can be determined for each heater (H1 and H2 ) curve-fitting the experimental
interferograms using a cosine with non-linear argument. Following, the linear fit of K(ν) with
ν0 = 193.414 THz allows to determine κτ for each heater and ξ1 . Finally, using κτ we obtain
γ2 = γW /κτ .
The calibration results using the laser interferograms are presented in Fig.3.5a-e and the ex-
tracted parameters are summarized in Table 1. The interferogram for the laser frequency 187.37
THz (1600 nm) with heater H2 actuated and its curve-fit are shown in Fig.3.5a-c as an example of
the procedure realized for multiple frequencies and both heaters. The decrease in the mean optical
power in Fig.3.5a and the small features in the envelope of Fig.3.5b are caused by slight misalign-
ment and vibration of the input/output fibers as the heater power increases due to the thermal
expansion of the chip. In a practical packaged application, these features would not be present
as the fibers would be permanently attached to the silicon chip. Combining the interferogram
fit results for each heater, K(ν) follows a linear dependence with frequency (Fig.3.5d) while γW
has a constant value (Fig.3.5e), in agreement with the eq.3.31. The non-linear term γ2 obtained
separately for each heater has the same value within the experimental error, (101 ± 1) · 10−3 ps−1 ,
while ξ1 only slightly deviates for each heater, 0.22 ± 0.02 for H1 and 0.24 ± 0.01 for H2 , yielding
0.23 ± 0.02.
The current versus voltage (IV) curves depicted in Fig.3.6 confirm the thermo-optic origin of
the measured non-linearity and allow to understand the difference between κτ,1 and κτ,2 . The
current-voltage (I-V) plots show a fairly linear behavior for both heaters, with resistances 15,876
3.3. SPECTROMETER CALIBRATION 38
kΩ for H1 and 15,162 kΩ for H2 , and indicates that any non-linearity originating from the heaters is
small compared to the non-linearity intrinsic to the thermo-optic effect. Moreover, the difference in
electric resistance causes a difference in heater efficiency kT that explains the observed discrepancy
in the measured κτ ’s. The heater efficiency is determined such that the temperature change at the
waveguide level ∆T is related to the dissipated electric power W by ∆T = kT W , thus κτ,i = η1 kT,i .
The agreement in γ2 and ξ1 obtained independently for each heater indicates the two arms of the
MZI are fairly similar, so that η1 can be considered the same. Using η1 ≈ 1.94 · 10−2 ps/K obtained
from simulations (see Appendix B), the heater efficiencies are estimated at kT,1 ≈ 18.5 K/W and
kT,2 ≈ 20.6 K/W.
The interferometer transfer function T (ν) is obtained from the transmission spectrum of the
MZI, as depicted in Fig.3.7. Recalling eq.3.3, the MZI transmission is
where T (ν) is the envelope of oscillating term and can be obtained by adjusting the experimental
trace. It is recommended that the transmission spectrum be measured at a non-null phase difference
∆φ so that T (ν) can be decoupled from the frequency-dependence of the average transmission
power T0 (ν). In our case, the experimental transmission (black trace) represents the transmission
spectrum of the passive device when none of the heaters are actuated. The oscillations have a free-
1.0
Transmission
0.0
1.0
b
T(ν)
0.7
186 188 190 192 194 196 198
Frequency (THz)
Figure 3.7: a. Experimental (black trace) and calculated (red trace) transmission spectrum of the MZI at non-zero
optical delay (0.172 ps). The calculated transmission is obtained using eq.3.3 in order to extract the MZI transfer
function T (ν) shown in b.
3.4. BROADBAND SPECTRUM RETRIEVAL 39
spectral range (FSR) of 3.96 THz around 193.414 THz (32.7 nm around 1550 nm) originated from
slight differences between the two arms incorporated as δL and δn(ν) as discussed in the previous
section. The non-flat transmission of Fig.3.7b is attributed to the non-ideal response of the power
splitter/combiner for frequencies higher than 194 THz. In practice, the limited bandwidth of these
components will dictate the bandwidth of the Si-FTS. For the specific y-junction design used here
[48], operation over more than 25 THz (200 nm), excess loss lower than 0.3 dB and low reflection
is expected.
1.0 1 b
a
Optical power (a.u.)
PSD (a.u.)
0.0 -1
188 192 196 200 -1.0 -0.5 0 0.5 1.0
Frequency (THz) Optical delay (ps)
Figure 3.8: a. Amplified Spontaneous Emission (ASE) of a C-band EDFA used as the light source. b. Theoretical
interferogram of the ASE for an ideal (linear TOC, dispersionless, balanced) FTS.
3.4. BROADBAND SPECTRUM RETRIEVAL 40
1.0 i
i
0.5
Optical power (a. u.)
ii
0.0
ii
-0.5
-1.0
-1.0 -0.5 0 0.5 1.0
Optical delay (ps)
Figure 3.9: Experimental interferogram, shifted to 0.172 ps and distorted due to differences between the two arms
of the MZI. The optical delay axis corresponds to T . The insets show a zoom-in of the interferogram at different
optical delays superposed to a cosine (grey traces) at the ASE mean frequency νs = 193.44 THz, highlighting: i the
oscillations at νs when the envelope varies slowly; ii phase changes when the envelope varies rapidly.
tions. The zero delay (maximum envelope amplitude) is shifted to around 0.172 ps and the inter-
ferogram envelope is asymmetric. These effects result from the non-zero contribution of ϕ(ν) in
eq.3.26, which carries the contribution of the difference between arms, δL and δn(ν) (see eq.3.15).
The zero-delay shift is caused by the first-order in ν, while the envelope distortion is caused by
high-order terms.
The difference between arms that cause these observable changes in the interferogram corre-
spond to typical variations expected from the silicon photonics process, rather than strong differ-
ences due to a non-ideal fabrication process. Considering the difference in arm length δL negligible,
the zero delay is centered at
L δ(neff |ν0 ) + ν0 δ(∂ν n)
T0 ≈ . (3.32)
c 1 + ξ1
Using the known values for the other parameters, we estimate δ(neff |ν0 ) + ν0 δ(∂ν n) ≈ 2 · 10−3 . This
value is in accordance with the expected order of magnitude for effective index fluctuations due to
chip-scale variations in the silicon device layer thickness [9, 26]. The same is true for the differences
in high-order dispersion terms.
The interferogram oscillations are highlighted in the insets of Fig.3.9, where a cosine oscillation
at the ASE mean frequency (νs = 193.44 THz) is superposed (grey trace). In regions with a
smooth envelope variation the interferogram oscillates at νs (inset i), while an additional phase
is introduced as the envelope varies more abruptly (inset ii). The phase change due to envelope
variations supports the interest of using a narrow laser source with almost flat envelope to calibrate
the thermo-optic non-linearity, as the non-linear fit would be compromised by such phase change.
The PSD obtained from the experimental interferogram and the effects of thermo-optic non-
linearity (γ2 ), dispersion (ξ1 ) and MZI transfer-function (T (ν)) are presented in Fig.3.10a-c. The
3.4. BROADBAND SPECTRUM RETRIEVAL 41
ν0 ν0 ν0
1.0
a b c
0.8
0.6
PSD (a.u.)
0.4
0.2
0.0
185 190 195 200 205 210 215 188 190 192 194 196 198 200 188 190 192 194 196 198 200
Frequency (THz) Frequency (THz) Frequency (THz)
Figure 3.10: Experimental (red) and reference (black) PSD at different conditions. The red points are the ex-
perimental data obtained from the IFT and the red line is a second-order interpolation curve. a. No correction.
b. Corrected thermo-optic non-linearity, but no dispersion correction or PSD re-normalization with T (ν). c. All
effects properly accounted for. The parameters used to obtain these plots are summarized in Table 1.
spectra were calculated using the mean value of the calibration parameters of Table 1.
First, Fig.3.10a shows the PSD obtained directly from the as-measured interferogram – with
the delay axis corresponding to τ and without performing any correction. The TOC non-linearity
distorts, broadens and shifts the PSD to higher frequencies as the interferogram oscillates faster
with increasing delay change to higher frequencies.
After the optical delay axis of the interferogram is properly transformed to T , the resulting
PSD becomes very similar to the reference spectrum (Fig.3.10b). Both resolved peaks are clearly
identified and the unresolved peak is also present around 194 THz. However, since the spectrum
has not been re-scaled to the original frequency ν, it is broadened by the factor 1 + ξ1 around ν0 . In
addition, since it has not been re-normalized by T (ν), the high frequency peak appears attenuated
relatively to the low frequency peak.
Failing to re-scale the frequency axis according to eq.3.28 introduces a frequency error u − ν =
ξ1 ∆ν that degrades the absolute frequency accuracy. In this demonstration, the effect is modest
since the detuning of the peaks with respect to ν0 is small. The peak around 196 THz, for instance,
is shifted by 0.5 THz, which is roughly its full-width at half maximum (FWHM). Nonetheless, the
effect can be very significant for frequencies further from ν0 .
The PSD corrected for the thermo-optic non-linearity, dispersion and the MZI transfer function
reproduces satisfactorily well the reference spectrum (Fig.3.10c). This spectrum was obtained
performing only the aforementioned corrections, with no additional data processing such as zero-
filling or apodization of the interferogram. The experimental spectral resolution of δν = 0.38 THz
(δσ = 12.7 cm−1 , δλ = 3.05 nm) is comparable to other on-chip spectrometers aimed at broadband
operation and it is sufficient for some Raman and IR absorption spectroscopy applications [16, 25].
3.5. SI-FTS FEATURES AND PERSPECTIVES 42
in the case where the two effects are absent (κT is assumed to be identical for the two heaters). In
the presence of ξ1 and γ2 , this product is modified to
δν × Wtotal = κ−1
T [(1 + ξ1 )(1 + γ2 κT Wtotal )]
−1
and is effectively reduced, resulting in decreased power dissipation for a given resolution or, con-
versely, lower δν given a maximum power. In our case, the resolution of 0.38 THz is achieved
dissipating a total power of 5.1 W, representing a 35% decrease in the power (6.9 W) that would
be required to achieve this resolution if dispersion and thermo-optic non-linearity were absent.
43
Chapter 4
m · λm = neff · Lr (4.1)
4.1. MICRORING RESONATORS 44
a b 1
Transmission
FWHM ER
R
FSR
0
c
Footprint resonator
Norm. power
450 nm
SiO2
220 nm Si
SiO2
Si bus waveguide
1
0
λm-1 λm λm+1
Figure 4.1: a. Schematic of a microring resonator in SOI composed by a 220 × 450 nm2 channel waveguide.
The footprint is the total area occupied by the device. b. Transmission spectrum and c. Power circulating in
the microring as a function of the incident light wavelength. λm : resonance wavelength; FSR: free-spectral range;
FWHM: full-width at half-maximum, or linewidth; ER: extinction ratio.
where m is an integer representing the order of the resonance, neff is the effective index of the
transversal mode of the waveguide constituting the ring and Lr is the ring’s total length. At
λm the ring is “on resonance”, in which case the intracavity power builds up and destructively
interferes with the light coming from the bus waveguide, creating a dip in the transmission spectrum
(Fig.4.1b,c).
The various resonances of the microring are related to the different orders m and adjacent
resonances are separated by the free-spectral range (FSR) which, in wavelength, is given by
λ2
FSRλ = (4.2)
ng Lr
where ng is the group index. We present the FSR in wavelength as most of the measurements
and plots in the subsequent chapters are in wavelength. Nonetheless, it is useful to notice that in
frequency the FSR is simply the inverse of the roundtrip time of the microring τr ,
1 c
FSRν ≡ = . (4.3)
τr ng Lr
Each resonance is characterized by its depth or extinction ratio (ER), and by its linewidth or
full width at half maximum (FWHM). The ER depends on the balance between the amount of
light lost due to radiation, scattering and absorption (intrinsic losses) and the amount of light
that escapes the resonator through coupling to the bus waveguide (extrinsic losses). When the
two loss mechanisms are balanced (extrinsic = intrinsic) the ER reaches its maximum value of 1,
4.1. MICRORING RESONATORS 45
whereas for unbalanced losses the ER is somewhere between 0 and 1. In such cases, the resonator
is said to be “over-coupled” when extrinsic losses dominate or “under-coupled” when intrinsic losses
dominate. The FWHM, on the other hand, gives an indication of the total losses in the resonator,
i.e., extrinsic + intrinsic losses.
It is common to define a normalized parameter associated with the FWHM, the quality factor
(Q). In terms of measurable quantities, it is defined as
λm
Q= . (4.4)
FWHMλ
Thus, the sharper the resonance, the higher the Q. In terms of physical quantities, the Q-factor
represents the number of oscillations of the optical field before the optical power decays to 1/e of
its initial value. Defining the time for such decay as the “photon lifetime” (τp ), we have
Q = 2πντp (4.5)
where ν is the optical frequency. Following the above definition of extrinsic and intrinsic loss
mechanisms, the quality factor can be decomposed in two contributions, intrinsic Q (Qint ) and
extrinsic Q (Qext ), which relate to those mechanisms,
1 1 1
= + . (4.6)
Q Qint Qext
FSR FSRλ
F = = Q . (4.7)
FWHM λ
The finesse is an important measurable quantity that is related to the intracavity power build-up.
Defining the power enhancement factor (PE) as the ratio between the power circulating in the
microring at resonance and the incident power through the bus waveguide, the PE is proportional
to the finesse squared,
2
power in the resonator F
PE = ∝ (4.8)
power in the bus waveguide π
From the measured FSR it is possible to extract the group index using eq.(4.2), which in this
case is around 4.27 (Fig.4.2b). As the wavelength increases, ng tends to increase slightly, although
the dispersion of values around 4.25 and 4.3 makes this trend hard to visualize.
A much clearer trend is observed in the increase of ER with wavelength (Fig.4.2c). The res-
onator, fabricated with a nominal gap distance of 200 nm from the bus waveguide, is slightly
under-coupled at lower wavelengths, reaching critical coupling around 1550 nm. For longer wave-
lengths, the resonator would get over-coupled and the ER would tend to decrease as the coupling
between ring and bus increases due to a larger overlap between the modes in the two waveguides.
Finally, the total quality factor of the various resonances and their finesse are shown in Fig.4.2d.
The Q-factor ranges from 16,000 to 35,000 and the finesse from 23 to 49, however no clear trend
is discernible. At critical coupling (around 1550 nm), the intrinsic and extrinsic Q-factors are the
same and amount to twice the measured Q-factor. Thus Qint = Qext ≈ 67, 000.
A last important remark concerns the two distinct resonance types highlighted in the insets of
Fig.4.2a. While the resonance around 1553 nm presents the expected Lorentzian lineshape, the
resonance around 1542 nm presents two unresolved peaks. These two peaks are indicative of the
coupling between the counter-traveling modes in the microring. Up to now we have tacitly assumed
that the light propagating in the bus waveguide couples into the microring only in the forward
direction and that the resonant mode of the ring is therefore a “traveling wave” mode propagating
in such direction. However, a similar traveling wave mode propagating in the backwards direction
could be excited if light approaches the ring from the opposite side of the bus waveguide. These two
counter-propagating modes of the ring are degenerate, i.e., they have the same resonant wavelength.
2.3 4.35
a b
1 1 1
Group index
FSR (nm)
2.2 4.3
2.1 4.25
0.8
0 0 c 2 4.2
Normalized Transmission
0.9
0.6
0.8
0.7
0.6
0.4
0.5
d 8 50
Q−factor (x104)
6 40
Finesse
0.2 30
4
20
2
10
0
1470 1490 1510 1530 1550 1570 1470 1490 1510 1530 1550 1570
Wavelength (nm) Wavelength (nm)
Figure 4.2: a. Experimental transmission spectrum of a microring resonator and extracted parameters for each
resonance: b. FSR and group index; c. extinction ratio; d. Quality factor and finesse. The insets in a highlight the
difference between a standard resonance (1553 nm) and one where the coupling between counter-traveling modes
creates a small resonance-splitting (1541.9 nm).
4.2. COUPLED MICRO-RESONATORS 47
However, these two modes might couple through back-scattering within the microring due to
sidewall roughness or even due to back-reflections in the coupling region [58]. In this case, the
coupling between the two otherwise degenerate modes creates “supermodes” and originate the
(small) resonance-splitting observed around 1542 nm. The sidewall roughness can be interpreted
as a quasi-grating and phase matching between counter-propagating modes – and therefore strong
coupling – only occurs at specific wavelengths [58]. The concepts of mode coupling, resonance-
splitting and supermodes are explored in detail in the next section.
When multiple microring resonators are brought together to form coupled systems, the new
spectral features will depend on the number of resonators and on the way they couple to each
other. Yet, general aspects can be explained considering the optical modes of the uncoupled (bare)
resonators and the phenomenon of degeneracy breaking due to their mutual coupling [59, 60].
The work presented in this thesis focuses on investigating the type of coupled-resonators de-
picted in Fig.4.3, where one ring couples to a bus waveguide and then additional rings are coupled to
the first. The transmission traces of Fig.4.3 were obtained using the transfer matrix method (TMM)
presented in Section 5.1. As discussed in the previous section, the bus-coupled ring presents a set
of resonances separated by the FSR and individually characterized by their linewidth (FWHM)
and depth (ER) (Fig.4.3a).
When a second distinct microring is coupled to the first one (Fig.4.3b), the two rings may have
different (non-degenerate) or identical (degenerate) resonance wavelengths. In the non-degenerate
case, the resonances of the primary outer ring remain unchanged and a new resonance notch
appears due to the secondary embedded ring. When the two rings are degenerate, however, their
mutual coupling create a resonance-splitting proportional to their coupling strength.
If another embedded ring identical to the first one is introduced (Fig.4.3c), the coupling-induced
splitting will originate a duplet when the embedded and outer rings are non-degenerate. Further,
a triplet with a very sharp central notch appears if all the three rings are resonant. Notice that
although the embedded microrings are not directly coupled to each other, they do interact through
the outer microring and this interaction is strong enough to induce a substantial splitting even
when the outer ring is not resonant (non-degenerate).
When the two embedded microrings couple directly to each other (Fig.4.3d) a new set of
resonances arise. In this design, modes propagating in opposite directions are allowed to couple at
the central region where the embedded rings interact. When only the embedded rings are resonant,
a quadruplet appears in the transmission spectrum due to the coupling of four degenerate modes
4.2. COUPLED MICRO-RESONATORS 48
Non-degenerate Degenerate
a 1
FWHM ER
FSR
0
b 1
Normalized Transmission
Splitting 0
c 1
0
d 1
0
-4 -2 0 2 4 -6 -3 0 3 6
Detuning (nm) Detuning (nm)
Figure 4.3: Calculated transmission spectrum of a single microring resonator (a) and different coupled microring
designs (b–d). The detuning axis is centered at the resonance of the bare embedded rings. Non-degenerate and
degenerate refer to the condition between the bus-coupled ring and the other rings. Multiple mode-splitting appear
in the transmission spectrum depending on the number of coupled resonators and degeneracy condition.
(two counter-propagating modes of each embedded microring). When the outer and embedded
rings are degenerate, a total of six modes are coupled and a sextuplet may appear. Notice this
design is equivalent to configurations where three identical resonators are placed on the corners of
an equilateral triangle [61, 62].
the outer and embedded rings, which is anti-symmetric for the first (red) and symmetric for the
second (green). In the central resonance (blue), on the other hand, the field is localized almost
completely within the embedded rings. Such confinement towards the embedded rings reduces the
decay rate of this supermode to the bus waveguide, yielding a higher total quality factor for this
particular resonance.
The diversified spatial distribution of supermodes in coupled resonators can be harnessed as a
powerful tool in practical implementations. Selectively changing the refractive index at different
portions of the coupled system may affect supermodes differently and represents an interesting
feature to be explored in sensing applications. In addition, the supermodes do not always overlap
with each other, providing freedom to engineer mode-selective couplings to distinct bus waveguides
[64]. The spatial distribution of different supermodes will be directly exploited to achieve tunable
resonance-splitting in Section 6.1. Meanwhile, the concept of reconfigurable spectral control is
introduced next.
a b
1
Norm. Transmission
0
-3 0 3
Detuning (nm)
Figure 4.4: Spatial distribution of optical supermodes. a. Transmission spectrum and b. 2-D FDTD simulations
showing the electric field distribution in each one of the resonances of the triplet.
4.2. COUPLED MICRO-RESONATORS 50
non-degenerate at zero heating power (bottom of plots a, b and c). As the heaters are actuated,
the outer ring resonance red-shifts and approaches the identical embedded rings resonances. A
clear avoided crossing (anti-crossing) is observed in the three cases, illustrating a typical signature
of mode coupling. The experimental transmission traces at degeneracy (Fig.4.5d-f(ii)) closely
reproduce the calculated traces discussed in Fig.4.3. As the outer ring is further red-shifted, outer
and embedded rings become again non-degenerate ((Fig.4.5d-f(i)).
The control of supermodes around the anti-crossing has significant practical interest. For
instance, it can be utilized to fine-tune the extinction ratio of resonances to optimize the efficiency
of power-dependent functionalities in passive integrated structures, such as parametric frequency
comb generation [65]. In addition to ER tuning, the spectral anti-crossing can be used for local
mode dispersion compensation [66, 67]. While Fig.4.5 only presents the spectral evolution obtained
when the outer microring resonances are shifted, a variety of spectral shapes can also be obtained
by actuating the microheaters of the embedded microrings (cf. Section 6.1).
a 30 b40 c 1
(i)
Norm. transmission
Heater power (mW)
(i) 80 (i)
30
20 60
(ii)
(ii) 20 (ii)
40
10
10 20
0 0 0 0
1559 1561 1563 1577 1580 1583 1604 1608 1612
Wavelength (nm) Wavelength (nm) Wavelength (nm)
d 1
e1 f 1
Normalized transmission
0 0 0
1 1 1
(ii) (ii) (ii)
0 0 0
1559 1561 1563 1577 1580 1583 1604 1608 1612
Wavelength (nm) Wavelength (nm) Wavelength (nm)
Figure 4.5: Reconfigurable spectral control. Evolution of the transmission spectrum of coupled resonators consist-
ing of a. one, b. two uncoupled and c. two coupled embedded microrings as the outer microring resonances are
shifted using integrated microheaters. The insets show the fabricated devices (upper-left) and schematics highlight-
ing the actuated microheater (bottom-right). d.-f. Transmission traces at selected heater powers when embedded
and outer rings are (i) non-degenerate and (ii) degenerate.
4.2. COUPLED MICRO-RESONATORS 51
.
53
Chapter 5
Resonant structures are ubiquitous in nanophotonics [34–36, 39, 40, 60, 71, 72]. Their wide
use is facilitated by the availability of powerful and simple mathematical tools such as the transfer
matrix method (TMM) [73] and temporal coupled-mode theory (TCMT) [74–76].
In TMM, presented in Section 5.1, the response of a resonant system is calculated directly
from the combined interference of light propagating through multiple optical paths and therefore
it is suitable to describe systems in which these optical paths are well known, such as optical
resonators composed of waveguides or using free-space optics. The TMM description, however,
may become cumbersome for some coupled-resonator structures such as those allowing coupling
between counter-propagating modes as described in Section 4.2 [71, 77, 78]. To tackle these cases
while remaining in the scope of TMM, we describe a method based on Mason’s gain rule for linear
systems. This simple approach was introduced in our group by Guilherme Rezende and has proved
very useful.
The TCMT, discussed in Section 5.2, is based on a lumped resonator model and the “su-
permodes” of a resonant system are calculated from the perturbative coupling of the lumped
resonators. As a perturbative model, TCMT is valid when describing low-loss and weakly-coupled
systems excited around their resonant frequency (small detuning). In addition to conventional
resonators in waveguides, its simple formulation appeals to a variety of less conventional electro-
magnetic resonant structures [79–82]. Furthermore, TCMT is based on time differential equations
and provides a very simple tool to evaluate dynamic responses of resonators [35, 36].
Relying on TCMT to model coupled resonators might however be misleading in some circum-
stances due to the lumped-element nature of the model. In Section 5.3 we tackle an important
limitation of TCMT related to the prediction of dark states in a three-ring system. We iden-
tify the limitation in the TCMT model to be related to the mechanism of excitation/decay of
the supermodes and we propose a correction that effectively reconciles the model with expected
results.
5.1. TRANSFER MATRIX METHOD 54
where α is the attenuation coefficient. These two can be combined in a propagation term
P = ejφ A. (5.3)
The beamsplitter relations depend on the number of waveguides coupled in the same region of
space. Here, we are only interested in the simple directional coupler formed by two waveguides
type of splitter [73]. In this case, calling E and S the optical field propagating through the two
waveguides, their incoming and outgoing amplitudes are related through a 2 × 2 coupling matrix
composed by the effective field transmission and coupling coefficients t and k,
! ! !
Eout t jk Ein
= (5.4)
Sout jk t Sin
We consider lossless splitters respecting t2 + k 2 = 1. With this definition, t and k are positive real
quantities.
The transmission of ring resonators is obtained from the accumulated propagation and coupling
contributions. The propagation terms and beamsplitters considered in the modeling of systems
of ring resonators is exemplified in Fig.5.1, along with a verification of the accuracy of the TMM
model compared to experimental transmission traces.
Single ring
sout 2 t0 − P0 2
T = =
1 − t0 P0 .
(5.5)
sin
5.1. TRANSFER MATRIX METHOD 55
a b c
A0,φ0 A0,φ0
A 0,φ0
1
Normalized transmission
0.8
0.6 A0 0.965
A0 0.963 k0 0.47
0.4 k0 0.34 A1,2 0.994
A0 0.955 A1,2 0.997 k1,2 0.445
0.2
k0 0.35 k1,2 0.20 k3 0.3
0
1546.5 1547 1547.5 1548 1548.5 1549 1536 1538 1540 1542 1544 1546 1605 1606 1607 1608 1609 1610 1611 1612
Wavelength (nm) Wavelength (nm) Wavelength (nm)
Figure 5.1: Modeling of microresonators via TMM. The upper schematics show the modeling parameters while the
bottom plots show the experimental transmission (black traces) and TMM curves (dashed red lines) for different
resonator designs: a. single microring resonator; b. three-ring system with two internal rings coupled to the external
one; c. three-ring system with direct coupling between the internal cavities. a and b can be easily modeled using
the standard TMM procedure, whereas the direct coupling between the inner cavities in c, allowing the excitation of
counter-propagation modes, makes the calculation somewhat cumbersome. In this case, the Mason’s Rule provides
a good technique to obtain the spectral response of the coupled cavity. In b and c, the inner rings are identical.
From this expression, the various properties of the resonance discussed in the previous chapter can
be derived, including the extinction ratio
t0 − A0 2
ER = 1 −
, (5.6)
1 − t0 A0
For low coupling and low losses (k0 1, A0 ≈ 1), the power enhancement can be written
simply as a function of the finesse and k,
2
F
PE0 ≈ k . (5.10)
π
5.1. TRANSFER MATRIX METHOD 56
Now consider a new ring resonator with propagation term P1 is coupled to the first one, and
that the field transmission and coupling coefficients between the two rings are t1 and k1 . From the
first ring’s perspective, the second ring contribution can be summarized in a complex phase term
t1 − P1
χ1 = . (5.11)
1 − t1 P1
sout 2 t0 − P0 χ1 2
T =
=
(5.12)
sin 1 − t0 P0 χ1
It is easy to generalize the above result for the case when N rings are coupled to one ring,
but not coupled among themselves or to the bus waveguide. In this case, assuming that Pi is the
propagation term of ring i and that the coupling to the first ring is represented by (ti , ki ), the
power transmission coefficient of the coupled system is
2
sout 2 t0 − P0 N
Q
i=1 χi
T =
= (5.13)
sin 1 − t0 P0 N
Q
i=1 χi
with
ti − Pi
χi = (5.14)
1 − ti Pi
The results of the above expression applied for the case of two identical inner rings are depicted
in Fig.5.1b. It accurately reproduces the transmission features of this type of coupled-resonator
system including the resonance-splitting and high-Q supermode.
For some configurations of coupled resonators, the TMM approach becomes cumbersome due to
a large number of possible optical paths for light propagation. A very convenient way of obtaining
the spectral response of such devices using the TMM formalism utilizes the matrix form of Mason’s
rule [83]. The method consists of building a square matrix A in which each row and correspondent
column corresponds to one port of a beamsplitter. All beamsplitters must be taken into account.
The transfer function between the m-th and n-th ports (which depend on the specific way A was
build) is given by the element Bmn of the matrix B = (I − A)−1 . In particular, the transmission of
the coupled system is obtained this way from the element connecting the column associated with
sout and the row associated with sin 1 .
1
An informative discussion on the use of Mason’s rule in this context including examples with coupled rings
can be found (in Portuguese) in the master thesis of Guilherme Rezende entitled “Projeto, caracterização e análise
5.1. TRANSFER MATRIX METHOD 57
The three-ring design with coupled inner rings provides a good example to put the Mason’s rule
in use (Fig.5.1c) As discussed in the previous chapter, this system couples ring modes propagating
in both directions. Using the procedure described above the transmission is easily obtained,
t0 + P0 (t0 P0 ζ1 + (1 + t2 ) ζ2 ) 2
ζ3 0
T = (5.15)
P
1 + t0 ζ (t0 P0 ζ1 + 2 ζ2 )
0
3
The “not-so-simple” nature of these expressions are indicative of how useful it is to have such
tool!
Since the Mason’s rule provides the transfer function between any m-th and n-th ports, this
tool is very useful for obtaining the power circulating in any given point of the resonant system
and therefore the power enhancement factor. For instance, the power enhancement factors in each
ring of the three-ring system of Fig.5.1b at the central high-Q supermode are
2
A 0 χ1 χ 2
P E0 = jk0 (5.17)
1 − t0 A0 χ2 χ2
2 2
1/4
A 1 A 0
P E1 = jk0 jk1 (5.18)
1 − t1 A1 1 − t0 A0 χ1 χ2
2 2
3/4
A 2 A0 χ 1
P E2 = jk0 jk2 (5.19)
1 − t2 A2 1 − t0 A0 χ1 χ2
where Pi should be replaced by Ai in the expressions for χi . These expressions will be used in
Section 5.3 when evaluating our modified-TCMT model.
de microressonadores óticos acoplados em plataforma SOI” (Unicamp) http://repositorio.unicamp.br/jspui/
handle/REPOSIP/276922.
5.2. TEMPORAL COUPLED MODE THEORY (TCMT) 58
s out
1 s out
3
a1
s in1 a2 s in3
s in2 aN s inM
s out
2 s out
M
Figure 5.2: Representation of a general resonant system with N resonant modes (ai ) and M input/output ports
(sin out
i /si ).
In a regime of small losses and weak coupling, coupled resonators and waveguides can be
described using TCMT as a system incorporating multiple lumped resonances and input/output
ports, schematically illustrated in Fig. 5.2. The general TCMT equations describing the coupling
of N resonators connected to M ports are [75]
d~a
= (jΩ − Γ) · ~a + K T · ~sin (5.20)
dt
~sout = C · ~sin + K · ~a. (5.21)
is the mode vector comprising the N mode amplitudes ai of the bare system, i.e., the system
composed by individual resonators not coupled to each other. The M ports comprising the in-
put/output power amplitudes are represented by
~sin = (sin in in
1 (t) s2 (t) · · · sM (t))
T
(5.23)
~sout = (sout out out T
1 (t) s2 (t) · · · sM (t)) .
These amplitudes are normalized so that |ai (t)|2 represents the total energy stored in mode i at
time t, while |sin 2 out 2
j (t)| and |sj (t)| represent the total power flowing in and out the system through
port j.
Matrix ΩN ×N comprises the resonance frequencies ωi of the bare modes (ai ) as its diagonal
elements and the coupling between modes κij , calculated by their overlap integral [76], as its
off-diagonal elements.
5.2. TEMPORAL COUPLED MODE THEORY (TCMT) 59
Matrix ΓN ×N represents the decay of the mode amplitudes. It can be written as Γ = Γport +Γloss ,
where Γport describes the decay due to coupling to the ports and Γloss accounts for intrinsic losses.
The resonant modes can be excited by the incoming light and decay into the output ports
through the coupling matrix KM ×N , while the direct power flow through ports (without the influ-
ence of resonances) is represented by the unitary and symmetric scattering matrix CM ×M .
Energy conservation and time-reversal symmetry considerations require that matrices C, K
and Γport satisfy the following relations [75]:
CK ∗ = −K (5.24)
† port
K K=2Γ (5.25)
It is useful to define traveling wave amplitudes Ai (t) normalized such that |Ai (t)|2 represents
the total power flowing through any cross section of resonator i. These power amplitudes are
related to the mode amplitude ai (t) through [76]
vgi
|Ai (t)|2 = |ai (t)|2 (5.26)
Li
where Li is the length of resonator i and vgi is the group velocity of the associated transverse
mode.
Equations (5.20, 5.21) represent the dynamics of the coupled system written in the basis of
the modes of the uncoupled system {ai }, or bare modes. Once matrix M = jΩ − Γ is known, the
supermodes bi of the coupled system and their associated resonance frequencies ωib can be found as
the eigenvectors and eigenvalues of M, respectively. Furthermore, the transformation S (similarity
matrix) that diagonalizes M is composed by the column eigenvectors of M, S = (b1 · · · bN ). Using
S, the CMT equations can be rewritten in the basis of the supermodes {bi }:
d~b
= Mλ · ~b + Kb T · ~sin (5.27)
dt
~sout = C · ~sin + Db · ~b (5.28)
The relation between matrices and vectors on the two sets of equations, (5.20, 5.21) and (5.27,
5.28), are as follows:
~b = S −1 · ~a (5.29a)
Mλ = S −1 · M · S (5.29b)
Kb = K · (S −1 )T (5.29c)
Db = K · S (5.29d)
5.2. TEMPORAL COUPLED MODE THEORY (TCMT) 60
For resonant systems composed of microring resonators the loss and coupling coefficients Γloss ,
K and κij can be very conveniently rewritten in terms of their equivalent TMM parameters for
lossess and coupling, α and k, defined in the previous section. First, Γloss can be written in terms
of the intrinsic loss rate of each resonator γi as
int
Γloss = diag(γ1int γ2int · · · γN ), (5.30)
while each γi can in turn be written in terms of the distributed loss coefficients αi
vgi αi
γiint = . (5.31)
2
The coupling vector can be written as K = [jµqi ] where j is the imaginary unit and µqi is a real
number representing the coupling of resonator i to port q in TCMT. Now, µqi can be expressed as
s
vgi
µqi = kqi (5.32)
Li
where kqi is the field coupling coefficient between port q and resonator i defined in TMM.
Finally, the coupling between resonators i and j, κij , can also be written in terms of the coupling
coefficients kij as s
vgi vgj
κij = kij . (5.33)
Li Lj
~a = a0 , Ω = ω0 .
Moreover, we consider the optical mode on the bus waveguide propagating in the forward
direction only, thus the system contains one input/output port and, from eqs.5.24,5.25
µ2
C = 1, K = jµ and Γport = ≡ γ0ext
2
. The total loss rate is γ0 = γ0ext + γ0int and the loaded quality factor is Q = ω0 /(2γ0 ).
5.2. TEMPORAL COUPLED MODE THEORY (TCMT) 61
a b
1.0 1.0
Normalized transmission
Normalized transmission
0.8 0.8
0.6 2 γ0 0.6
γ0
γ0
a0, ω0 a0, ω0
0.4 0.4 γ1
a1,ω1 κ
TMM TMM
0.2 0.2
s in µ s out TCMT µ 2κ TCMT
s in s out
0.0 0.0
- 0.2 - 0.1 0.0 0.1 0.2 - 0.6 - 0.4 - 0.2 0.0 0.2 0.4 0.6
Detuning (rad) Detuning (rad)
Figure 5.3: TCMT model of a. a single ring resonator and b. a dual-ring system. The parameters for the model are
represented in the inset schematics. The transmission spectra coincide perfectly with that obtained using the TMM
model. The parameters used in a and b are (k0 , A0 ) = (0.15, 0.98) and (k0 , A0 , k1 , A1 ) = (0.23, 0.98, 0.15, 0.98),
respectively, and the radii of the bus-coupled ring and internal ring are 20 µm and 5 µm.
where a0 and a1 represent the bare modes of each individual ring. Notice we only consider the
modes propagating in one direction as coupling between counter-propagating modes is neglected.
The system matrix is composed by the bare resonance frequencies of the individual rings and by
the coupling between them: !
ω0 κ
Ω=
κ ω1
Finally, we have again one input/output port yielding, along with eqs.5.24,5.25,
!
µ2
0
C = 1, K = (jµ 0) , Γport = 2
.
0 0
Notice this exact same model of a dual-mode system could be used to describe a single microring
with coupling between counter-propagating modes.
As will be discussed in the next section, the TCMT model of the three-ring system of Fig.5.1b
deserves more attention due to the presence of a quasi-dark state.
5.3. MODIFIED-TCMT (M-TCMT) 62
The experimental realization of the three-ring coupled resonator and its spectral response are
presented in Fig.5.4a-d. The device consists of two identical microring resonators coupled to a third
dissimilar microring that is coupled to a bus waveguide (Fig.5.4a). The transmission spectrum in
Fig.5.4b shows a triplet resulting from the coupling-induced mode-splitting when the three rings
are degenerate, while the light distribution in each of the three resonances is illustrated in the
infra-red (IR) micrographs of Fig.5.4c. The lateral resonances of the triplet have similar Q-factor
(14,000) and extinction ratio, whereas the central resonance constitutes a so-called quasi-dark state
with significantly higher Q-factor (66,000) as the light is mostly localized in the embedded rings,
effectively reducing the extrinsic (coupling) losses. A typical anti-crossing diagram (Fig.5.4d) is
obtained from the transmission spectrum when the detuning between outer and embedded rings
is controlled by means of an integrated microheater (H1 in the inset micrograph of Fig.5.4d). The
experimental anti-crossing is subjected to an overall red-shift of the resonances due to thermal
crosstalk, but the evolution of the supermodes remain clear.
The spectral features obtained experimentally are well reproduced by the TMM model (Fig.5.4e-
h) already discussed in Section 5.1. The transmission spectrum (Fig.5.4f) shows a similar triplet
5.3. MODIFIED-TCMT (M-TCMT) 63
a b c d 40
Norm. Transmission
20
(ii)
(i) (iii)
0 0
Normalized Transmission
e f g h 180
Norm. Transmission
60
ω1 - ω0 (2π GHz)
1
Normalized Power
2
P1 ,φ1 | A1
2
| A2
k3 P3 ,φ3 P2 ,φ2 k2 2
30 | A3 0
s in k1 s out 0 0 -180
i j k l 180
Norm. Transmission
ω1 - ω0 (2π GHz)
γ1 1
Normalized Power
2
γ3 a1, ω1 | A1
2
γ2 | A2
(i)
κ a3,ω0 a2 ,ω0 κ 30 | A3 2
0 (i) x (ii) x (iii) x
(ii)
(iii)
0
s in µ1 s out 0 0 -180
1578 1579 1580 1581 1582 1578 1579 1580 1581 1582 1578 1579 1580 1581 1582
Wavelength (nm) Wavelength (nm) Wavelength (nm)
Figure 5.4: Experimental data a-d, TMM model e-h and TCMT model i-l of a three-ring resonator system. a.
Fabricated device and b. transmission spectrum showing a triplet with high-Q quasi-dark state in the center when
the three rings are degenerate. c IR-micrograph of the scattered light at each resonance. d. Anti-crossing obtained
when the outer ring is detuned using a microheater (H1 in the inset micrograph). An overall red-shift is present
due to thermal crosstalk affecting the embedded rings. e. TMM parameters: sin /sout are the input/output fileds;
k1 and k2 are coupling coefficients and φi and Pi are the accumulated phase and attenuation in each microring,
respectively. f. TMM triplet similar to the experimental observation and g. intracavity power spectrum with high
power enhancement for the central resonance, in which case light is confined to the embedded rings. |A1 |2 , |A2 |2
and |A3 |2 represent the power circulating in the outer ring, first and second embedded rings, respectively. The blue
and green curves closely overlap. h. TMM anti-crossing showing the evolution of the supermode resonances in the
absence of thermal crosstalk. i. TCMT model (parameters described in the text). In contrast to the experimental
data and TMM, no central resonance is observed in the j. transmission spectrum and k. intracavity power spectrum.
l. TCMT anti-crossing obtained from the transmission spectrum and from the eigenvalues of Ω (dashed-blue lines).
The central mode is predicted by the eigenvalues but not excited, constituting a dark state in the TCMT model.
Inset: spatial distribution of each supermode at degeneracy, representing the eigenvectors of Ω.
with resonance-splitting dictated by the coupling between embedded and outer ring resonators
(k2 ) and with a high-Q central resonance. The intracavity power for each ring (Fig.5.4g) — nor-
malized to sin — provides a quantitative assessment of the power distributions observed in the
IR-micrographs: for both lateral resonances light circulates in all the three rings, while for the cen-
tral resonance light is localized within the embedded rings resulting in a small effective coupling
to the bus waveguide and high power enhancement. Finally, the TMM anti-crossing diagram of
Fig.5.4h provides the expected evolution of the supermodes in the absence of thermal crosstalk.
Unlike TMM, however, the TCMT model disagrees with the experimental observations as it
predicts a dark state for the central supermode (Fig.5.4i-l). No central resonance notch appears
in the transmission spectrum (Fig.5.4j) and no light circulates in the cavity (Fig.5.4k) since the
supermode cannot be excited by the incoming light. These results are calculated using the general
formulation of the orthogonal TCMT [75] presented in the previous section,
5.3. MODIFIED-TCMT (M-TCMT) 64
d~a
= (jΩ − Γ) · ~a + K T · ~sin (5.34)
dt
~sout = C · ~sin + K · ~a (5.35)
with the following parameters for the three-ring model (Fig.5.4i) [88, 89]: the mode amplitude of
the individual resonators are grouped in the mode vector ~a = (a1 a2 a3 )T and their bare resonance
frequencies (ω1 for the external ring and ω0 for the identical embedded rings) and mutual coupling
(κ) constitute the system matrix
ω1 κ κ
Ω = κ ω0 0 ; (5.36)
κ 0 ω0
the single bus waveguide is described by the incoming and outgoing power amplitudes sin and sout
and requires C = 1; the coupling between the bus waveguide and the resonant system, occurring
only through the outer ring, is represented by the coupling vector
K= jµ1 0 0 (5.37)
where µ1 represents the coupling of the first resonator to the bus waveguide; the decay matrix
Γ = Γloss + Γport completes the model, accounting for the intrinsic losses in each resonator,
Γloss = diag(γ1int γ2int γ3int ) and for the extrinsic loss term Γport = diag(µ21 /2 0 0). The
transmission spectrum is calculated as |sout /sin |2 while the intracavity power spectra are calcu-
lated using eq.(5.26).
The prediction of a dark state by TCMT can be understood considering the interaction between
the supermodes of the coupled system and the bus waveguide. First, we calculate the eigenvalues
and normalized eigenvectors of Ω, which give the supermodes’ resonance frequencies ωib ’s and mode
amplitudes bi ’s. The ωib ’s are depicted in Fig.5.4l (blue traces) and they follow closely the spectral
evolution expected from the experimental results and TMM. At degeneracy, the eigenfrequencies
and eigenmodes are √ T
−1
ω1b = ω0 + 2 κ , b1 = √ 1 1
2 2 2 T
b −1 √1
ω2 = ω0 , b2 = 0 √ 2 2
(5.38)
√ T
ω3b = ω0 − 2 κ , b3 = √12 12 12 .
These expressions show that b2 , the supermode corresponding to the (quasi-)dark state, is com-
pletely confined to the embedded rings while b1 and b3 have components in the outer ring as
illustrated in the inset of Fig.5.4l. Since b2 vanishes in the outer ring it cannot be excited by the
incoming light which only couples to a1 (see eq.(5.37)). The effective zero drive for supermode b2
5.3. MODIFIED-TCMT (M-TCMT) 65
b µ1
K = j√ −1 0 1 (5.39)
2
FDTD simulations
FDTD simulations of two distinct three-ring designs allow to understand the limitations of
the TCMT model and how it can be modified to properly describe the quasi-dark state. The
transmission spectrum and the mode profiles of each supermode are presented in Fig.5.5a,c for
a design similar to the one described in Fig.5.4 with two embedded rings coupled to the outer
ring at different positions, while Fig.5.5b,d shows these results for a design where both rings are
coupled to the outer ring at the same point. The FDTD simulation for the first design is consistent
with the previously discussed experimental and TMM results as it also predicts the excitation of
the quasi-dark state. In the second design, on the other hand, the supermode is not excited. It
constitutes, therefore, an effective dark state in agreement with the TCMT prediction. In a lumped
element model such as TCMT, however, these two designs are equivalent: two identical resonators
weakly coupled to a third one which in turn is coupled to a bus waveguide, with system matrix
and coupling vector given by eq.(5.36,5.37) and supermodes given by eq.(5.38).
1
(i) (ii) (iii)
Normalized Transmission
(ii)
a c
(i) (iii)
0
1
(ii) (i) (ii) (iii)
b d
(i) (iii)
0
1440 1450 1460 1470
Wavelength (nm)
Figure 5.5: 2D-FDTD simulations. a., b.: Transmission spectrum; c., d.: steady-state electric field amplitude of
the supermodes of the three-ring device at degeneracy in two different configurations. a., c.: When the embedded
rings coupled to the outer ring at different positions a weak field circulates in the outer ring allowing the excitation
of the quasi-dark state (ii). b., d.: When the embedded rings couple to the outer ring at the same position the
destructive interference in the outer ring prevents the excitation of supermode (ii), originating a dark state.
5.3. MODIFIED-TCMT (M-TCMT) 66
The fundamental difference between the two designs, not accounted in the TCMT model, lies
on the effect of the embedded rings in the roundtrip phase of the outer ring. In the absence of
embedded rings the outer ring is resonant at ω0 and its accumulated roundtrip phase is a multiple
of 2π. When the two embedded rings couple to the outer ring at different positions, each of them
introduces a zero or π phase-shift depending on its coupling regime (undercoupled or overcoupled,
respectively [91]) and the accumulated roundtrip phase in the outer ring remains a multiple of
2π. Even though no resonant light circulates in the outer ring as indicated by b2 in eq.(5.38) an
amount of non-resonant light is able to propagate over the outer ring to feed the embedded rings
as illustrated in Fig.5.5c(ii). In other words, the outer ring acts as a waveguide in this situation,
allowing the communication between the supermode b2 confined in the embedded rings and the bus
waveguide. The contribution of this non-resonant light is not considered in the TCMT model, which
only accounts for the resonant mode amplitudes. On the other hand, when the coupling between
rings occurs in the exact same point a π phase shift is introduced in the accumulated roundtrip
phase of the outer ring, resulting in destructive interference and preventing the excitation of b2 , as
shown in Fig.5.5d(ii). Notice that we used a racetrack as the outer resonator to assure the coupling
between outer and embedded rings is the same in both designs. Detailed information regarding
the parameters used in the FDTD simulations can be found in Appendix D.
The TCMT model can nonetheless be altered to deliver a description of the three-ring design
that allows for the excitation of the quasi-dark state b2 . This is accomplished with a modified
coupling vector which will be derived in the next section.
where vg is the group velocity in the outer ring (we will assume the same group velocity for all
5.3. MODIFIED-TCMT (M-TCMT) 67
It is expected that supermodes b1 and b3 be equally coupled to the bus waveguide (|K1b |2 = |K3b |2 )
as they have the same mode profile, thus requiring µ2 and µ3 to satisfy µ3 = −µ2 . This allows to
simplify K b to
b 1
K =j √ −µ1 −2 µ2 µ1 . (5.42)
2
This expression gives the same coupling to b1 and b3 as eq.(5.39) which was already in agreement
with the expected results. As for mode b2 , it can now be excited by a non-null µ2 .
The term µ2 represents the indirect coupling between bus waveguide and embedded rings and
its dependence with k1 and k2 is determined using power conservation [75, 76]. Consider a lossless
system (Γloss = 0) with mode b2 excited to energy |b02 |2 at t = 0. With no incoming light (sin = 0)
the energy in the resonator decays and the power flowing through the output port is |sout (t)|2 =
2 µ22 |b2 (t)|2 . The same scenario can be described using a power-normalized amplitude B2 (t) that
couples to the outgoing wave sout (t) through a power coupling coefficient defined as kb so that
|sout (t)|2 = kb2 |B2 (t)|2 . The equivalence between the two pictures requires
On the other hand, the relation between circulating power and stored energy given by eq.(5.26)
requires
vg
|B2 (t)|2 = |b2 (t)|2 (5.44)
2 L2
where L2 is the length of each identical embedded ring so that 2 L2 is the effective length of
supermode b2 . Eqs.(5.43, 5.44) allow to write µ2 in terms of kb :
r
kb vg
µ2 = (5.45)
2 L2
Finally, the power coupling coefficient kb is given by k12k2 , as follows. The supermode confined
to the embedded rings, with power |B2 |2 , is fed by a certain amount of power circulating in the
outer ring |Aπ1 |2 by means of k2 , so that
At frequency ω2b , for which this correction is derived, |Aπ1 |2 is the circulating power in a microring
5.3. MODIFIED-TCMT (M-TCMT) 68
out-of-resonance and can be estimated using TMM along with the TCMT assumptions of low loss
(A1 → 1) and weak coupling (k1 1):
Combining eq.(5.46) and eq.(5.47) and remembering that kb was defined as the coupling coefficient
between B2 and the bus waveguide, we have kb = k12k2 .
The modification of the TCMT model is therefore complete, consisting of a new coupling vector
µ µ
K = j(µ1 µ2 − µ2 ) and additional elements Γport
ij = i 2 j to the decay matrix, where
r r
vg k1 k2 vg
µ1 = k1 , µ2 = . (5.48)
L1 4 L2
due to the fact that µ3 = −µ2 in our model. On the other hand, the power imbalance between |A2 |2
and |A3 |2 in TMM (inset of Fig.5.6c) reflects the fact that light arrives at the second embedded
ring modified by the resonance of the first one, as can be seen in the expression
|A3 |2
= A1 |χ1 |2 (5.50)
|A2 |2
where A1 is the roundtrip attenuation factor of the outer ring and χ2 is the complex transmission
of the first embedded ring (see Section 5.1).
5.3. MODIFIED-TCMT (M-TCMT) 69
a e
Transmission
2
Normalized Experimental 1
TMM
TMM
m-TCMT
0
m-TCMT
0
b 60 60 f 1 k1 = 0.4
Norm. Power
m-TCMT
| A1 2 0.3
30 0 2 0.5 0.2
- 0.1 0 0.1 | A2
2
| A3
0.1
0 0
c 60 60 g
Norm. Power
TMM 30 k1 = 0.4
2 0.3
| A1
30 0
- 0.1 0 0.1 | A2 2 20 0.2
2
| A3
10 0.1
0
d 60 60
h
(|A 2 2 + |A 3 2 )/ 2
Norm. Power
k1 = 0.4
120
Norm. Power
0.3
TMM
30 m-TCMT 0.2
0
- 0.1 0 0.1 60
0.1
0 0
1578 1579 1580 1581 1582 0 0.1 0.2 0.3 0.4 0.5
Wavelength (nm) k2
Figure 5.6: Comparison between modified-TCMT (m-TCMT) and TMM. a. Transmission spectrum and b., c.
intracavity power spectrum calculated with the parameters used in Fig.5.4. Insets: detail of the central peaks. d.
Average intracavity power for the quasi-dark state calculated with TMM and m-TCMT. e. Resonance splitting
for the triplet and f. extinction ration, g. linewidth and h. average intracavity power for the quasi-dark state
calculated for various coupling coefficients. The m-TCMT calculations agree with TMM for a wide range of coupling
strengths, while in the standard TCMT curves (f -h) would vanish.
The new terms in the m-TCMT equations depend on the coupling coefficients k1 and k2 , there-
fore the model must be validated over a wide range of these parameters. The m-TCMT and TMM
models yield very close predictions for various values of coupling coefficients as demonstrated
in Fig.5.6e-h, which shows results for the resonance splitting (Fig.5.6e), the extinction ratio and
linewidth of the quasi-dark state resonance (Fig.5.6f,g) and its average intracavity power (Fig.5.6h).
Particularly, the agreement between TMM and m-TCMT for the average intracavity power indi-
cates that the power imbalance captured in the TMM model does not significantly affect the total
power in the supermode. The two models show slight discrepancies only for combinations of large
coupling strengths, when the weak-coupling assumption of TCMT starts to fail. Notice that, ex-
cept for the resonance splitting, the calculated quantities would vanish in the standard TCMT
model. A similar comparison for the lateral resonances of the triplet (supermodes b1 and b3 ) is
unnecessary as they were already well described by the standard TCMT and their steady-state
values are not affected by µ2 , as predicted by eq.(5.42).
5.3. MODIFIED-TCMT (M-TCMT) 70
In addition to the steady-state response, the m-TCMT model provides a good description
of the transient behavior of the coupled system as confirmed by a comparison with 2D-FDTD
simulations (Fig.5.7) — see Appendix D for simulation parameters. The transient evolution of the
√
lateral resonances (Fig.5.7b) presents fast oscillations determined by the resonance splitting (2 2κ)
which are also reproduced by m-TCMT (Fig.5.7c) as a result of the additional non-diagonal terms
in eq.(5.49). For the quasi-dark state, the FDTD transient presents a power imbalance between
embedded rings similar to that observed in the TMM calculations (Fig.5.7d). Once again, it is the
average power in the supermode that corresponds to the m-TCMT solution (Fig.5.7e). This good
description of the transient response of the optical supermodes makes m-TCMT a suitable model
to describe dynamic perturbations on the coupled-ring system such as optical modulation through
refractive index perturbation.
b 25 d 20 FDTD
Normalized Power
Normalized Power
FDTD
20 (ii)
(i), (iii) 15
2
15 | A1
a 2 10 | A2 2
1 10 | A1
2
| A2 2 | A3
Transmission
5 5 (|A 2 2 + |A 3 2 )/ 2
Normalized
2
| A3
(ii) 0 0
m-TCMT
c 25 e 20 m-TCMT
Normalized Power
2D-FDTD
Normalized Power
m-TCMT
(i) (iii) (ii)
0 20 (i), (iii)
15
1445 1450 1455 1460 1465 15
Wavelength (nm) 2 10 | A1 2
10 | A1
2
| A2 2 | A2
5 5 2
| A3 2 | A3
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Time (ps) Time (ps)
Figure 5.7: Comparison between m-TCMT and 2D-FDTD simulations. a. Transmission spectrum and b.-e.
transient intracavity power evolution. The FDTD results are presented in b for the lateral resonances and in d for
the quasi-dark state, while the corresponding m-TCMT solutions are presented in c and e. The m-TCMT model
reproduces the transient evolution for the lateral modes including the fast oscillations presented in the FDTD
simulation. For the quasi-dark state it describes the average power circulating in the resonator.
5.3.4 Conclusion
We demonstrated an important limitation of the TCMT model: the incorrect prediction of
dark-states in coupled resonators instead of the actual high-Q quasi-dark states. Through the
analysis of a three-ring resonator system, we showed this inaccurate prediction occurs due to the
inability of TCMT to account for non-resonant light circulating in the system. The existence
of such non-resonant light can be properly accounted for by introducing extra drive terms in a
modified-TCMT (m-TCMT) model.
5.3. MODIFIED-TCMT (M-TCMT) 71
The TCMT formalism is applied to many fields and the limitation demonstrated here might
be present in distinct resonant structures. This discussion may be helpful to prevent the misrepre-
sentation of quasi-dark states and will provide insight on how the standard TCMT model can be
modified to allow accurate results.
This modified TCMT model is used in Section 6.3 to provide a small signal modulation response
of the three-ring design under refractive index modulation.
5.3. MODIFIED-TCMT (M-TCMT) 72
.
73
Chapter 6
In this chapter we utilize coupled microrings and the idea of coupling-dictated resonance-
splitting and supermodes in practical applications.
In Section 6.1 we demonstrate a mechanism for reconfigurable control of mode-splitting with
extremely reduced resonance shifts. Exploiting the controlled excitation of counter-propagating
modes, we show that resonances can be tuned from single-notches to large mode-splitting (80 GHz)
with a minimum of resonance-shift limited by thermal crosstalk. This work has been published in
Optics Letters [71].
Section 6.2 presents the experimental demonstration of multi-channel all-optical wavelength
multicasting. Here, the coupled resonators are utilized to achieve dense wavelength conversion and
low power operation simultaneously. This work has been published in Optics Express [60].
In Section 6.3 we propose a coupled ring modulator capable of delivering better modulation
efficiency×bandwidth than a single microring modulator with no significant increase in footprint.
The experimental design of such coupled-ring modulators was presented in Section 2.1 when dis-
cussing the tape-out with Sandia National Labs. As mentioned in that section, we were unable to
test the devices due to unexpected delays.
6.1. RECONFIGURABLE SPLITTING CONTROL 74
a b
splitting resonance
shift
λ0 λ λ0 λ
Figure 6.1: a.Ideally, one should be able to control the splitting while maintaining the central wavelength λ0
unchanged. b. In practice, however, most of the splitting control mechanisms come associated with very large
resonance shifts.
6.1. RECONFIGURABLE SPLITTING CONTROL 75
a cw b
ccw H1
a2cw a3ccw
a1cw a1ccw
a2ccw
a3cw H23
H2 H3
10 µm
sin
Figure 6.2: Coupled microring device. a. The direct coupling between two embedded microrings enables the
excitation of CW- and CCW-traveling modes acw,ccw2 and acw,ccw
3 even when the incident light sin travels in one
cw ccw
direction only. The coupling between a1 and a1 is strong only if the detuning between outer and embedded
rings is small. b. Optical micrography of the fabricated device with Ni-Cr microheaters. H1 , H2 and H3 tune each
microring separately, while H23 allows to tune both embedded rings simultaneously.
6.1. RECONFIGURABLE SPLITTING CONTROL 76
a 1 Transmission b 1 Transmission
0 0
Wavelength Wavelength
Transmission Transmission
1 1
Heater power
Heater power
0 0
Wavelength Wavelength
Transmission Transmission
1 1
0 0
Wavelength Wavelength
Figure 6.3: Splitting control approaches using microheaters. The transmission traces show the resonances of the
coupled cavity for different detuning conditions and the resonances in which mode-splitting is controlled (blue and
green traces). The microring device bellow each resonance group depicts the spatial distribution of the associated
supermodes and the actuated microheaters (orange rectangles). a. Embedded and outer ring supermodes are
initially uncoupled (bottom transmission). The embedded ring resonances are simultaneously red-shifted towards
the outer ring resonance by means of heater H23 , increasing the CW-CCW mode-coupling in the outer ring. b.
Starting with the three microrings degenerate (bottom transmission) a single embedded ring is detuned using heater
H3 , reducing the CW-CCW-induced mode-coupling for all supermodes.
controllable CW-CCW mode-splitting is created as mode accw 1 is increasingly excited through the
embedded rings, as shown in the blue-highlighted trace in Fig.6.3a. Alternatively, starting with the
three coupled rings degenerate (Fig.6.3b, bottom transmission), all the six CW and CCW modes
(acw,ccw
1,2,3 ) are coupled but only four resonance notches appear due to an accidental degeneracy [60].
Using heater H3 to red-shift only R3 effectively reduces the CW-CCW coupling for all supermodes.
Although the actuated ring supermodes strongly red-shift, the position of the remaining CW-CCW
supermodes of cavities R1,2 (blue and green traces) are practically unaffected.
The proposed splitting control approaches feature important advantages towards widely tun-
able splitting with no resonance-shift. The controlled CW-CCW mode coupling is mediated by the
coupling between microrings, therefore the maximum CW-CCW induced splitting can be tailored
through cavity design (gap separation and coupling length). On the other hand the controlled
mode-splitting can always be reduced to a single-notch resonance, since the uncoupled CW and
CCW modes are degenerate. In both schemes described in Fig.6.3 the resonance-shifts are ef-
fectively mitigated as we minimize the overlap between the spatial distribution of the controlled
6.1. RECONFIGURABLE SPLITTING CONTROL 77
supermodes (blue and green traces) and the cavity sections in which the refractive index is changed
(around the orange heaters).
a 25 1
b 1 e 35 1 f 1
30
Normalized transmission
Normalized transmission
20 0 17.7 mW 0 35 mW
1 1
Heater power (mW)
25
Heater power (mW)
15
20
0 13 mW 0 18 mW
1 1
10 15
0 8 mW 10 0 9 mW
5 1 1
5
0 0 0 mW 0 0 0 mW
0 0
1547.5 1548.5 1549.5 1550.5 1551.5 1547.5 1549.5 1551.5 1566 1567.5 1569 1570.5 1572 1566 1568 1570 1572
Wavelength (nm) Wavelength (nm) Wavelength (nm) Wavelength (nm)
c 0.5 d g h
Phase shift (rad) 1.8 Phase shift (rad)
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.9 1.8
80
0.2
Resonance Splitting (GHz)
28 1.35
Phase shift (rad)
60
21
-0.1 0.9
40
14
-0.4 0.45 20
7
-0.7 0 0.0 0
1547.5 1548.5 1549.5 1550.5 1551.5 0 4.5 9 13.5 18 1566 1567.5 1569 1570.5 1572 0 17.5 35
Wavelength (nm) Heater Power (mW) Wavelength (nm) Heater Power (mW)
Figure 6.4: Spectral evolution of the coupled-cavity resonances using heaters H23 (a-d) and H3 (e-h). Measured
(a) and calculated (c) optical transmission as H23 power is increased. The blue-dashed line follows the central
wavelength of the controlled resonance. b. Transmission traces corresponding to specified heater powers. The
blue color highlights the controlled resonance. c. Calculated transmission as the phase of the embedded rings
are simultaneously varied. d. Experimental (circles) and calculated (line) resonance-splitting as a function of the
power dissipated in H23 and equivalent phase shift between embedded and outer rings (upper horizontal scale),
respectively. Measured (e) and calculated (g) optical transmission as H3 power is increased. The blue- and green-
dashed lines follow the central wavelength of the controlled resonances. f. Transmission traces corresponding to
specified heater powers. The blue and green color highlights the controlled resonances. h. Measured (circles) and
calculated (lines) resonance-splitting for the set of split resonances, with green and blue indicating the leftmost and
central resonances respectively.
6.1. RECONFIGURABLE SPLITTING CONTROL 78
transmission spectra (blue-dashed line) due to the mode anti-crossing dispersion and a small ther-
mal crosstalk. This heating configuration shifts the embedded microring resonances at a rate of
9.5 GHz/mW while a spurious shift of the outer ring resonance occurs at a rate of 1.5 GHz/mW
due to the thermal crosstalk. To accurately measure the spurious shift we used a spectral region
(not shown in Fig.6.4) where the embedded and outer rings resonances are uncoupled throughout
the full range of heating power.
A qualitatively distinct tunable mode-splitting spectrum is observed when the three microrings
are initially degenerate and only one embedded ring (R3 ) is actively detuned (Fig.6.4e,f). At
low heating powers, the six supermodes exhibit a quasi-fourfold mode-splitting, highlighted in
the bottom transmission trace in Fig.6.4f. As R3 is detuned, the coupling strength between CW
and CCW-modes gradually vanishes along with the associated mode-splitting for all supermodes.
The two resonances of interest are shown in Fig.6.4f in green and blue traces. We show the
resonance-splitting dependence on the heating power for these resonances in Fig.6.4h. Although
this dependence is clearly nonlinear this approach yields larger tunable splitting when compared
to the previous one, with a maximum splitting of 80 GHz observed for the central (blue) resonance
when the heater is off. On the other hand, this approach is also more severely affected by thermal
crosstalk as indicated by the dashed lines in Fig.6.4e. Heater H3 shifts R3 at 8.7 GHz/mW, while
thermal crosstalk impinges resonance-shifts of 1.56 GHz/mW in R2 and 4.6 GHz/mW in R1 .
resonance-splitting over resonance-shift. The experimental results yield splitting/shift ratios of 0.46
using heater H23 and 0.42 and 0.54 for the blue and green resonances using heater H3 (Fig.6.4f).
Ignoring the resonance shifts caused by thermal crosstalk, this values increase to 0.83, 2.67 and
2.53 for the three cases respectively. Although clearly detrimental to this proof-of-principle demon-
stration, the effects of thermal crosstalk can be eliminated employing distinct dispersion control
mechanisms such as carrier effects in silicon (PN junctions instead of heaters) or electro-optic ef-
fects in other platforms, which exhibit extremely reduced crosstalk. Moreover these mechanisms
would enable low-power, ultrafast mode-splitting modulation.
6.1.5 Conclusion
In conclusion, using the coupled microring device allowing for coupling between CW and CCW
modes we demonstrated an efficient mechanism for achieving reconfigurable mode-splitting con-
trol with reduced resonance shifts. In our experimental demonstration, we achieved continuous
splitting tuning from single notch resonances to 80 GHz. The limitation of large resonance shifts
associated with the splitting tuning could be overcome by controlling the coupling between counter-
propagating modes in one part of the coupled-resonator system while changing the optical phase
in another part of the system. The application of such mechanism using carrier effects instead
of thermal tuning would mitigate undesired spurious shifts due to crosstalk and find interesting
applications in reconfigurable optical filtering and microwave photonics.
6.2. ALL-OPTICAL MULTICASTING 80
a b
λ1
Multicasting λ1 λ2 λ3 λ4
Network
λ2 λ4
λ3 1530 nm 1630 nm
~50 GHz
Figure 6.5: a. Wavelength multicasting consists of replicating the information transmitted in one wavelength
channel (λ1 ) into multiple other wavelengths (λ2,3,4 ). b. In a flexible DWDM network, this conversion should be
achievable either between closely-spaced channels (λ1 to λ2 ) or across the DWDM grid (λ1 to λ3 , λ4 ).
6.2. ALL-OPTICAL MULTICASTING 81
Control
PG
VOA
PC EDFA BPF PC
Mod
Signal
Figure 6.6: Experimental setup for four-channel wavelength multicasting. PC: polarization controller; PG: pattern
generator; Mod: optical modulator; EDFA: erbium-doped fiber amplifier; VOA: variable optical attenuator; BPF:
band-pass optical filter.
ried by the control signal can be simultaneously converted to multiple probe signals with different
wavelengths. Leveraging resonance-splitting (40-60 GHz) and high-Q resonances (30,000), these
devices achieve wavelength conversion between closely-spaced channels at very low input power
and tight footprint.
The experimental setup is depicted in Fig.6.6. Both control and probe light are generated
by continuous wave (cw) tunable lasers. The control light is modulated with non-return-to-zero
(NRZ) 27 − 1 pseudo-random bit sequence at 622 Mbit/s. Control and probe channels are coupled
through a 90:10 coupler and launched into the waveguide. Accounting for insertion losses around
7 dB, the control power launched into the waveguide is 0 dBm (1 mW) and the probe power is
-12.5 dBm in each channel. The output of the microring resonator is filtered to separate control
and converted signals, which are then amplified and individually analyzed.
The transmission spectrum of the coupled ring device in the wavelength range used for this
experiment is show in Fig.6.7. The control light is fixed at the short-wavelength edge of a sharp
resonance located at 1547.9 nm (λC ) and the probe lasers are tuned to the quadruplet resonances
around 1617 nm.
The control waveform and converted signals are shown in Fig.6.8. When the probe wavelengths
are located on resonance, the probe transmission increases as the high control signal (logic 1)
generates free carriers and blue-shifts all the resonances, resulting in a non-inverted wavelength
conversion (Fig.6.8a). Accordingly, when the probe wavelength is located at the shorter-edge of
the resonance, the probe transmission decreases as the resonances are blue-shifted, resulting in
inverted wavelength conversion (Fig.6.8b). The time constant of this device is ≈ 670 ps, which is
dominated by the free carrier lifetime of intrinsic silicon (450 ps) [102]. This limits the maximum
modulation rate to less than 1 Gbit/s, which is obviously of no use for high-speed optical networks.
However, it has been experimentally demonstrated that much shorter free carrier lifetimes and thus
6.2. ALL-OPTICAL MULTICASTING 82
1
Norm. transmission
0
1550 1560 1570 1580 1590 1600 1610 1620
Wavelength (nm)
Figure 6.7: Broadband transmission spectrum of the coupled-ring device and detail of the resonances used for the
conversion of a signal close to 1550 nm (λC ) to four adjacent channels separated by 50 GHz centered around 1617
nm (λ1−4 ).
higher modulation speeds can be achieved by employing reverse-biased p-i-n junctions [103], which
will extract the carriers as soon as they are generated, or by introducing traps to decrease the
free-carrier lifetime, as through ion implantation [104, 105].
The control power required to achieve all-optical modulation based on resonance shift caused by
FCD is proportional to V Q−3/2 [101], where V is the volume of the mode and Q is the quality factor.
An SOI single microring with 50-GHz channel spacing (FSR) would require a total length around
1.415 mm. Such length is more than 10 times larger than the total length of the two embedded
rings supporting the supermodes associated with the quadruplet resonances. As a result, assuming
the single ring would have similar Q (30,000) this device would require 10 times more power to
achieve a similar modulation.
C
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Time (ns) Time (ns)
Figure 6.8: Four-channel wavelength multicasting. Control and converted waveforms for a. non- inverted and b.
inverted wavelength conversion at 622 Mbit/s. The control power is 0 dBm and the probe power is −12 dBm in
each channel. Control and converted signal wavelengths are, respectively, λC = 1547.9 nm, λ1 = 1616.40 nm, λ2 =
1616.77 nm, λ3 = 1617.31 nm, and λ4 = 1617.65 nm.
6.3. COUPLED-RING MODULATOR 83
6.3.1 Motivation
Microring modulator interest and trade-off
Efficient optical modulators are essential for the deployment of integrated photonics in optical
interconnects for data-centers and high-performance computing. So far, Mach-Zehnder interfer-
ometer (MZI)-based modulators are the standard choice in commercial systems as it brings the
advantage of broadband operation. However, these devices are relatively large (arm length in the
millimeter range) and power hungry compared to microring resonators, which can operate in the
fJ/bit range with low peak-to-peak voltages (Vpp ) [106]. The need for low-voltage, compact and
energy efficient modulators is particularly strong in approaches that integrate CMOS electronics
with photonics in the same process [107], in which scenario microring resonators become the design
of choice.
Although highly efficient microring resonators have been demonstrated, these devices present a
fundamental trade-off between the modulation efficiency and energy consumption in one hand and
the modulation bandwidth on the other hand. This trade-off, dictated by the resonance linewidth,
is illustrated in Fig.6.9 where two resonators with different quality factors are driven by the same
index change ∆n (same voltage Vpp ). The resonator with lower quality factor (blue) presents larger
3 dB bandwidth at the expense of lower optical modulation amplitude (OMA). The energy-per-
bit of a lumped microring modulator with total capacitance Cd transmitting a non-return-to-zero
(NRZ) pseudo-random-bit-sequence (PRBS) is [108]
Cd Vpp2
Ebit = . (6.1)
4
a b c
OMA
Figure 6.9: Trade-off between modulation efficiency and bandwidth of microring resonators. a. Two critically-
coupled resonators with distinct quality-factors. When modulated by the same index change, these resonators
deliver distinct b. 3 dB bandwidth and c. optical modulation amplitude (OMA).
6.3. COUPLED-RING MODULATOR 84
Thus, to achieve the same OMA as the high-Q (red) modulator the low-Q modulator would require
a larger voltage swing and the energy consumption would be greatly increased.
The 3-dB bandwidth that is dictated by the resonance linewidth and illustrated in Fig.6.9b
is called the “optical” bandwidth (fopt ) to differentiate it from the cutoff frequency of the elec-
tric circuit driving the modulator (fRC ), dictated by its RC constant. The total bandwidth of
the electro-optic modulator (f3dB ) is smaller than either fopt or fRC and depends on these two
contributions as
1 1 1
2
= 2 + 2 . (6.2)
f3dB fRC fopt
Different approaches have been proposed to overcome this efficiency/bandwidth trade-off. For
instance, modulation of the coupling coefficient instead of the propagation phase have been demon-
strated to work well at high frequencies [109], but it is less efficient at low-frequencies. In addition,
the coupling section in this approach is effectively a long MZI which prevents further scaling down
of the device size. Another approach utilizes a microring with FSR of a few tens of gigahertz
so that the modulation side-bands would couple into adjacent resonances to achieve an extended
optical bandwidth [110]. To achieve such small FSR, however, the device length must be very large
(millimeter range). In both approaches the requirement for a large device not only brings with it
the cost of footprint, but it also has a detrimental effect on the bandwidth. In large devices, the
capacitance of the PN junction will bring the RC constant up and the electric bandwidth down.
Ultimately, it hinders the use of the extended optical bandwidth obtained with such designs.
Double-ring modulators have been investigated as a device that could deliver extended band-
width [35, 111–113]. While promising for radio-frequency (RF) analog modulation [113], their
use for digital baseband modulation is less significant since the extended bandwidth does come
associated with a reduction in modulation efficiency [35].
Here, we discuss how to go beyond the bandwidth×efficiency limit of single microring resonators
by exploiting the supermodes of a three-ring device (Fig.6.10d-f). Exploiting the triplet with GHz-
range resonance spacing dictated by the coupling strength between rings, this design utilizes the
concept of coupling to adjacent modes while remaining compatible with compact ring resonators.
Our discussion is based on a small-signal analysis using TCMT and focuses on the optical part of
the modulation, while assuming the electric properties of single and coupled-ring modulators to be
equivalent.
6.3. COUPLED-RING MODULATOR 85
a b c
γ1int
1.0
a1, ω1 Laser
Δn
0.8
Transmission
0.6
0.4
Modulation
0.2 ON
OFF
s in µ1 s out
0.0
1549.8 1549.9 1550.0 1550.1 1550.2
Wavelength (nm)
d e f Laser
γ3int γ1int
a1, ω1 1.0
κ a3,ω3 a2 ,ω2 κ
Transmission
0.6
+ Δn 0.4
Modulation
0.2 ON
OFF
0.0
s in µ1 s out
1549 1550
Wavelength (nm)
1551
Figure 6.10: Optical modulator designs compared on this section. a.,d. TCMT model for single microring and
coupled microrings. b.,e. Illustration of the refractive index modulation in both designs. The inner rings are
modulated in a differential manner. c.,f. Steady state response of the two designs when the index change is turned
on (red traces) and off (black traces). In the single ring, the laser is positioned at the resonance slope, while in the
coupled ring it is tuned to the central resonance.
In this section, we present the small-signal analysis of the two designs depicted in Fig.6.10. Both
designs are subjected to refractive index modulation ∆n. In the coupled-ring, the modulation is
applied to the inner rings in a differential manner, i.e., positive change in one ring and negative
change in the other. Also, while in the single ring the laser detuning with respect to the resonance is
a free operational parameter, in the coupled-ring design the laser is tuned to the central resonance
of the triplet.
The analysis is based on the TCMT equations already presented in Section 5.2,
d~a
= (jΩ − Γ) · ~a + K T · ~sin (6.3)
dt
~sout = C · ~sin + K · ~a. (6.4)
The validity of TCMT relies on the assumption that the mode amplitudes vary in a time-
scale much longer than the optical period (| dadtm | |ωm am |). This assumption is compatible with
modulation rates ranging from tens to a few hundred GHz (101 − 102 GHz) as they are at least
three orders of magnitude slower than the optical field oscillations, on the order of hundreds of
THz (≈ 2 · 105 GHz at 1550 nm).
6.3. COUPLED-RING MODULATOR 86
The TCMT model of the single ring was presented in Section 5.2, while for the coupled ring
we use the modified model presented in Section 5.3. The single ring is described by
µ21
~a = a1 , Ω = ω1 , C = 1, K = jµ1 , Γ = γ1 = γ1ext + γ1int , γ ext =
2
The index modulation is incorporated in the model directly as a frequency modulation, δω cos (Ωt),
where δω is the amplitude of the frequency modulation and Ω is the modulation angular frequency.
We are interested in silicon optical modulators based on carrier depletion which also suffer loss
modulation, introduced in the model as δγ cos (Ωt). For typical silicon PN junctions with waveg-
uide doping concentrations around 1017 -1018 cm−3 , the effective index modulation efficiency is
around −5 · 10−5 V−1 while the loss modulation efficiency is around 1 dB·cm−1 V−1 . Combining
these two numbers we can assume δγ = β · δω with β ≈ 0.057.
The effects of frequency and loss modulation then induce a change δa(t) in the optical mode
amplitude. Thus, the modulation response is obtained incorporating these terms in the model,
ω1 → ω1 + δω cos (Ωt)
γ1int → γ1int + δγ cos (Ωt) (6.5)
a1 → a1 + δa(t)
keeping only terms up to 1st order in δ and solving for the output power,
where PDC is the continuous output power term and PΩ is the modulated output power amplitude.
In the following, we will be interested in comparing PΩ for the different modulator designs.
For the coupled ring design, the TCMT parameters are
a1 ω1 κ κ
~a = a2 , Ω = κ ω2 0 , C = 1, K = j(µ1 µ2 − µ2 ),
a3 κ 0 ω3
µ21 −µ1 µ2
µ1 µ2
2 2 2
µ22 −µ22
Γ = Γloss + Γport , Γloss = diag(γ1int γ2int γ3int ), Γport = µ1 µ2
.
2 2 2
−µ1 µ2 −µ22 µ23
2 2 2
6.3. COUPLED-RING MODULATOR 87
The modulation of the inner rings and changes in optical mode amplitudes are incorporated by
int int
ω1 → ω0 γ1 → γ1 a1 → a1 + δa1 (t)
ω2 → ω0 + δω cos (Ωt) , γ2int → γ2int + δγ cos (Ωt) , a2 → a2 + δa2 (t) (6.7)
ω3 → ω0 − δω cos (Ωt)
γ int → γ int − δγ cos (Ωt)
a3 → a3 + δa3 (t).
3 3
Notice the different sign for the frequency and loss modulation in the two inner rings. Again, we
assume δγ = β · δω.
Single ring
The modulated optical amplitude PΩ is proportional to the applied frequency swing δω. How-
ever, it can be put in a form that normalizes δω by the intrinsic decay rate γ1int ,
δω
PΩ = PNΩ (ΩN , ∆N , η, β). (6.8)
γ1int
The resulting modulation amplitude PNΩ , in units of δω/γ1int , is a function of the already defined β
and of the normalized modulation frequency (ΩN ), normalized laser detuning (∆N ) and criticality
factor (η), defined as
Ω ∆ω γ ext
ΩN = int , ∆N = int , η = 1int . (6.9)
γ1 γ1 γ1
The expression for PNΩ is
Using PNΩ in terms of the normalized parameters defined above is very convenient as it allows to
carry a general assessment of the single ring modulation response as a function of ∆N and η for
any given γ1int . Conversely, given γ1int which is ultimately limited by the platform and fabrication
parameters, one can decide on the other parameters to achieve a specific performance target. This
procedure will become clear when we compare single and coupled ring modulators using typical
design parameters.
6.3. COUPLED-RING MODULATOR 88
The single ring trade-off between modulation amplitude and bandwidth is illustrated in Fig.6.11.
As defined in Fig.6.11a, PN0 is the low-frequency modulation amplitude and PNpeak is the peak
modulation amplitude. The normalized bandwidth Ω3dB N represents the 3-dB bandwidth of the
receiver, i.e., the frequency at which (PNΩ )2 drops by half. Notice this definition is more stringent
than assuming that PNΩ should drop by half. The normalized gain-bandwidth product (GBPN ) is
defined as the low-frequency modulation amplitude times the 3-dB bandwidth:
As shown in Fig.6.11c, PN0 is maximized for an under-coupled ring (green trace) at ∆N ≈ 0.9.
However, at this condition Ω3dB
N is small. At larger detuning Ω3dB
N increased at the expense of PN0 .
0.5 20
Peak 0.3
PN
10
0.2
0 5
PN 0.1
0
PN 0.0 0
2.0
Normalized GBP, GBPN
d
ΩN3dB 1.5
Normalized modulation freq., ΩN
1.0
1.0 0.5
0.8 b
0.0
Tranmsission
10
0.6 e
8
η = γ1int
ext
PNpeak/PN0 (dB)
0.4 γ1
0.5 6
1
0.2 2
4
0.0 2
6 4 2 0 2 4 6
Normalized detuning, ∆N 0
0 1 2 3 4 5
Normalized detuning, ∆N
Figure 6.11: Single ring modulator. a. Illustration of the modulation response versus modulation frequency
defining PN0 , PNpeak and Ω3dB 0 3dB
N . b. Transmission spectrum, c. PN (left axis) and ΩN (right axis), d. GBPN and e.
peak 0
Ratio PN /PN versus normalized laser detuning ∆N at three coupling conditions: under-coupling (green), critical
coupling (red) and over-coupling (blue). The blue dot in d shows the condition chosen for comparison with the
coupled ring modulator.
6.3. COUPLED-RING MODULATOR 89
Also, differently from the modulation amplitude, the bandwidth is maximized for over-coupled
rings (blue trace). This trade-off is manifested in Fig.6.11d as a maximum in GBPN for each
coupling condition.
The increase in bandwidth with ∆N is a result of the “peaking” effect [36] in which the apperance
of the peak around the laser detuning frequency extends the bandwidth at the expense of optical
modulation efficiency. As a result, this effect is also accompanied by an increase in the peak-to-
baseline ratio (Fig.6.11e). For η larger than 2 the GBPN would continue to increase due to peaking
enhancement of Ω3dBN , but the low-frequency modulation amplitude would be further compromised.
For such reason, we take η = 2 and ∆N = 3.3 (blue point in Fig.6.11d) as an optimal set of
parameters as GBPN is maximized while maintaining a descent modulation amplitude – roughly
half of the maximum achievable.
Coupled rings
In analogy with the single ring case, the modulated optical amplitude of the coupled-ring design
can be written in terms of parameters normalized by the intrinsic decay rate of the modulated rings
(inner rings):
δω
PΩ = PNΩ (ΩN , ηi , ηe , ηL , κN ). (6.11)
γ2int
The normalized parameters upon which PNΩ depends are
√ 2
γ int γ ext
Ω 2κ k2 L1
ΩN = int , ηi = 1int , ηe = 1int , κN = , ηL = (6.12)
γ2 γ2 γ2 γ2int 4 L2
Notice that ηi and ηe are the normalized intrinsic and extrinsic losses of the outer ring. κN is the
normalized splitting between the central and lateral resonances. The parameter ηL is related to
the ratio between outer and inner rings length (L1 and L2 , respectively) but it also depends on the
coupling strength between outer and inner rings k2 , where k22 represents the power coupling as in
the TMM model. For the coupled-ring modulator, PNΩ does not depend on β and is it given by
2
p ηe −ζ0 (1+ζ2 ) 2ζ0 +jΩ N ζ1 −Ω N
PNΩ = 4κN ηe ζ2
ηe +ζ0 (1+ζ2 ) [ηe +ζ0 (1+ζ2 )]+jΩN [ζ1 +2ηe +ζ2 (1+ηi )]−Ω2N [2+ηi +ηe +ζ2 ]−jΩ3N
with
ζ0 =ηi + κ2N
ζ1 =1 + 2ηi + κ2N
ζ2 =2ηe ηL .
The transmission spectrum of the coupled ring and the modulated optical amplitude as a
function of the modulation frequency are illustrated in Fig.6.12. The baseline (PN0 ) and peak PNpeak
6.3. COUPLED-RING MODULATOR 90
Laser
a 1.0 b
Modulation response
0.8 PN
Peak
Transmission
0.6 PN
0
Valley
0.4 PN
0
PN
0.2 N
0.0
-4 -2 0 2 4 ΩN3dB
Normalized modulation freq., ΩN
Normalized detuning, ∆N
Figure 6.12: a. Transmission spectrum and b. normalized modulated amplitude for the coupled ring modulator.
In addition to PN0 and PNpeak , we define PNvalley as the modulated amplitude at the valley of between the baseline
and the peak.
modulation amplitudes as well as the 3-dB bandwidth (Ω3dB N ) are defined in the same way as for
the single ring as shown in Fig.6.12b. An additional amplitude PNvalley is defined, associated to the
valley between the baseline and the peak.
While the design of the single ring modulator involves only two normalized parameters (∆N
and η), the design of coupled ring involves four (κN , ηi , ηe and ηL ). As long as PNvalley does not fall
beyond 3 dB, Ω3dB N is directly proportional to the normalized splitting κN , Ω3dB
N ≈ 1.3κN . Thus,
κN must be the first parameter to be fixed given the bandwidth requirements.
The outer ring normalized loss parameter ηi does not influence PN0 significantly, provided κN
is large enough to cause resolved resonance splitting – which is the regime we are interested at.
This result is illustrated in Fig.6.13a for different κN while fixing ηL (0.02) and ηe (5) 1 . A smaller
ηi , on the other hand, increases the peak amplitude as shown in Fig.6.13b. These results indicate
that ηi should be minimized and that an undoped waveguide should be used for the outer ring 2 .
a b 10
0.4
N
0
Low-freq. amplitude, PN
10
15
5
20
Ppeak/P0 (dB)
0.3
0
0.2
5
0.1 10
0 2 4 6 8 10 0 2 4 6 8 10
i i
Figure 6.13: a. Low-frequency (baseline) modulation amplitude(PN0 ) and b. peak-to-baseline amplitude ratio
(PNpeak /PN0 ) as a function of the normalized intrinsic loss of the outer ring ηi for various normalized splitting κN .
ηL and ηe are fixed at 0.02 and 5, respectively.
1
While the trend would be the same for different values of ηL and ηe , the specific values used here are very close
to the parameters used in the comparison with the single ring, discussed ahead. The values used for κN and ηi in
Fig.6.14 are also those used in the comparison
2
Notice that for baseband digital modulation a large peak-to-baseline ratio is acceptable since in a real modulator
the electric bandwidth would attenuate the modulation amplitude at high frequencies.
6.3. COUPLED-RING MODULATOR 91
Notice the results are consistent with what should be expected considering the power distribu-
tion of supermodes discussed in sections 4.2,5.3: at low modulation frequency, the optical power is
distributed within the inner rings and the losses or the outer ring are unimportant; at the modula-
tion peak (mod. frequency equals the splitting frequency), however, the supermode populates both
inner and outer rings and the higher the losses in the latter the lower the modulation efficiency.
The dependence of the modulation response with the remaining parameters ηe and ηL is ana-
lyzed in Fig.6.14 for fixed ηi (0.1) and κN (12.5). As previously mentioned, it is necessary to assure
that the valley-to-peak ratio is greater than -3 dB (region above the dashed red line in the plots)
so that the bandwidth is dictated by the splitting. ηL and ηe must therefore be chosen to maximize
GBPN while satisfying the the above condition (red star in the plots). The red star in the plots
indicate the parameters used in the numerical comparison between coupled-ring and single ring
modulators presented next.
Before proceeding to compare the two designs with a realistic numerical example, it is worth
noticing that a trade-off is still present for the coupled-ring. For different κN , the plots in
Fig.6.14c,d would change but there would still exist a clear region in which the modulation am-
plitude and GBPN would be maximized. This is a result of the resonant nature of the modulator
and the general rule that the modulation efficiency increases with the finesse. What we will show
next is that this trade-off is less stringent for the coupled ring modulator.
As a final remark, notice that we could have chosen the operation point (red star in Fig.6.14d)
at higher ηe and slightly lower ηL to achieve maximum GBP. However, for higher ηe the coupling
between outer ring and bus waveguide becomes too strong, in which case our modeling based on
TCMT might provide less accurate results.
-0.75 12
9
0.15 -1.50 0.15
6
-2.25
3
-3
-3.75
-6
0.05 -4.50 0.05
-9
-5.25
-12
0.00 0.00
0 5 10 15 20 0 5 10 15 20
e e
c Low-freq. amplitude, PN
0
d Normalized GBP, GBPN
0.20 0.20
0.35 4.0
3.5
0.15 0.30 0.15
3.0
0.25
2.5
0.10 0.20 0.10
L
2.0
0.15
1.5
0.05 0.5
0.00 0.00
0 5 10 15 20 0 5 10 15 20
e e
Figure 6.14: Coupled-ring modulator: dependence with ηL and ηe . a. Valley-to-baseline amplitude ratio
(PNvalley /PN0 ). This value must be higher than - 3 dB (20Log10 (PNvalley /PN0 ) > 1/2) indicated by the dashed-red line.
b. Peak-to-baseline ratio (PNpeak /PN0 ). c. Low-frequency modulation amplitude. d. Normalized gain-bandwidth
product. The red star indicates the set of parameters used in next section when comparing coupled-ring and single
ring modulators. In all the plots ηi and κN are fixed at 0.1 and 12.5 respectively.
6.3. COUPLED-RING MODULATOR 93
it may also be interpreted as a fourfold reduction in energy consumption for the same modulation
amplitude.
We conclude this section briefly discussing how the coupled-ring modulation efficiency can scale
with reduced intrinsic losses.
Our previous discussion of the single ring modulator in terms of normalized parameters shows
that the device’s modulation efficiency cannot scale-up if the propagation losses are reduced. Sup-
pose that, as in the previous example, we want to achieve 50-GHz bandwidth with optimized GBPN
for η = 2, but now the propagation losses are 3 dB/cm instead of 25 dB/cm. Since η is fixed and
the bandwidth is dictated by the resonance linewidth we have no choice but to increase γ1int back to
the value corresponding to 25 dB/cm by introducing a drop-port. Consequently, nothing changes
a 1.0 b 2.0 c 5
0.8 0
1.5
0.6 -5
1.0
0.4 -10
0.5
0.2 -15
Figure 6.15: Performance comparison between single ring and coupled-ring modulators designed for a 3-dB band-
width of 50 GHz (parameters given in the text). a. Transmission spectrum with solid dots indicating the position
of the laser. b. Modulation amplitude in linear scale and c. Normalized modulation amplitude in dB. The coupled
ring modulation efficiency is roughly twice that of the single ring modulator.
0.30
Coupled-ring modulation efficiency
5
Single ring modulation efficiency
Coupled-ring energy-per-bit
0.25
Single ring energy-per-bit
4
0.20
0.15 3
0.10
2
0.05
1
0.00
5 10 15 20 25
Propagation losses (dB/cm)
Figure 6.16: Modulation efficiency (blue) and energy-per-bit (red) of the coupled ring modulator with respect to
the single ring as the propagation losses of the two modulators decrease for a fixed 3-dB bandwidth of 50 GHz.
6.3. COUPLED-RING MODULATOR 94
Chapter 7
Conclusion
This thesis summarized our main contributions to the field of integrated silicon photonics.
It presented our involvement in several projects from device conceptualization through micro-
fabrication and testing. Further, it demonstrated how our work contributed to enabling novel
functionalities in different applications, from on-chip spectroscopy to optical signal processing.
In the first part, we demonstrated the realization of a Fourier Transform spectrometer with true
time delay based on silicon photonics technology. We showed how issues related to thermo-optic
non-linearity, thermal expansion and dispersion can be properly understood and incorporated in a
simple manner. We derived a FT relation between PSD and interferogram with modified optical
frequency and arm delay accounting for these effects and we demonstrated a calibration procedure
using a tunable laser source. Further, we demonstrated the retrieval of a 7-THz-wide light source
around 193.4 THz with spectral resolution of 0.38 THz (12.7 cm−1 , 3.05 nm) using a 1 mm2
device with total electric power dissipation around 2.5 W per heater. The Si-FTS shows intrinsic
resilience to fabrication variations that allows scalability of its resolution and power consumption
performance, enabling robust and versatile portable spectrometers.
In the second part, we demonstrated how coupled resonators can be used to achieve closely-
spaced resonances (40-60 GHz) obtained through mode-splitting while maintaining high loaded
quality-factors (30,000-60,000) in compact footprints. We utilized these devices to achieve low-
power four-channel wavelength multicasting based on the TPA-induced FCD and to demonstrate
an efficient mechanism for achieving reconfigurable mode-splitting control (from single notch to
80 GHz) with reduced resonance shifts. In addition, we proposed a coupled-ring modulator that
overcomes the trade-off between modulation efficiency and bandwidth faced by single microrings
modulators.
Our investigations on coupled microring systems also uncovered an important limitation of
the temporal coupled mode theory (TCMT): the incorrect prediction of dark-states in coupled
resonators instead of the actual high-Q quasi-dark states. Through the analysis of a three-ring
resonator system, we showed this inaccurate prediction occurs due to the inability of TCMT to
96
account for non-resonant light circulating in the system. The existence of such non-resonant light
was properly accounted for by introducing extra drive terms in a modified-TCMT (m-TCMT)
model.
We hope this thesis will become a useful reference for future research on the various topics
discussed here. Significant improvements are still to be incorporated to our proof-of-principle
demonstrations, in one hand. On the other hand, these photonic devices may be proven useful in
so far unforeseen applications.
97
Bibliography
[1] Chan, J. et al. Laser cooling of a nanomechanical oscillator into its quantum ground state.
Nature 478, 89–92 (2011). 14
[3] Sibson, P. et al. Integrated Silicon Photonics for High-Speed Quantum Key Distribution.
Optica 4, 172–177 (2016). 14
[4] Paesani, S. et al. Experimental Bayesian Quantum Phase Estimation on a Silicon Photonic
Chip. Physical Review Letters 118, 100503 (2017). 14
[5] Koblmüller, G. et al. Roadmap on silicon photonics. Journal of Optics 18, 073003 (2016).
14
[6] Chrostowski, L. & Hochberg, M. Silicon Photonics Design (Cambridge University Press,
Cambridge, 2015). 14, 42
[7] Hochberg, M. et al. Silicon photonics: The next fabless semiconductor industry. IEEE
Solid-State Circuits Magazine 5, 48–58 (2013). 14
[8] Lim, A. E. J. et al. Review of Silicon Photonics Foundry Efforts. IEEE Journal on Selected
Topics in Quantum Electronics 20 (2014). 14
[9] Zortman, W. a., Trotter, D. C. & Watts, M. R. Silicon photonics manufacturing. Optics
Express 18, 23598–23607 (2010). 14, 30, 40
[10] Hochberg, M. & Baehr-Jones, T. Towards fabless silicon photonics. Nature Photonics 4,
492–494 (2010).
[11] Subramanian, A. Z. et al. Silicon and silicon nitride photonic circuits for spectroscopic
sensing on-a-chip. Photonics Research 3, 47–59 (2015). 14
[12] Li, J.-y., Lu, D.-f. & Qi, Z.-m. Miniature Fourier transform spectrometer based on wavelength
dependence of half-wave voltage of a LiNbO3 waveguide interferometer. Optics Letters 39,
3923 (2014). 14
[14] Coutant, O., De Mengin, M. & Le Coarer, E. Fabry-Perot optical fiber strainmeter with an
embeddable, low-power interrogation system. Optica 2, 400–404 (2015). 14
[15] Sabry, Y. M., Khalil, D. & Bourouina, T. Monolithic silicon-micromachined free-space optical
interferometers on chip. Laser & Photonics Reviews 9, 1–24 (2015).
[18] Kyotoku, B. B. C., Chen, L. & Lipson, M. Sub-nm resolution cavity enhanced micro-
spectrometer. Optics Express 18, 102–107 (2010). 14
[19] Florjaczyk, M. et al. Development of a slab waveguide spatial heterodyne spectrometer for
remote sensing. In Proceedings of SPIE, vol. 7594, 75940R–75940R–9 (2010). 14
[20] Ma, X., Li, M. & He, J. J. CMOS-compatible integrated spectrometer based on echelle
diffraction grating and MSM photodetector array. IEEE Photonics Journal 5, 6600807
(2013). 14
[21] Redding, B., Liew, S. F., Sarma, R. & Cao, H. Compact spectrometer based on a disordered
photonic chip. Nature Photonics 7, 746–751 (2013).
[22] Redding, B., Liew, S. F., Bromberg, Y., Sarma, R. & Cao, H. Evanescently coupled multi-
mode spiral spectrometer. Optica 3, 956–962 (2016).
[25] Nie, X., Ryckeboer, E., Roelkens, G. & Baets, R. CMOS-compatible broadband co-
propagative stationary Fourier transform spectrometer integrated on a silicon nitride pho-
tonics platform. Optics Express 25, A409–A418 (2017). 14, 41
[26] Rahim, A. et al. Expanding the Silicon Photonics Portfolio With Silicon Nitride Photonic
Integrated Circuits. Journal of Lightwave Technology 35, 639–649 (2017). 14, 30, 40
[27] Malik, A. et al. Ge-on-Si and Ge-on-SOI thermo-optic phase shifters for the mid-infrared.
Optics Express 22, 28479–28488 (2014). 14
BIBLIOGRAPHY 99
[29] Zhou, Z., Yin, B. & Michel, J. On-chip light sources for silicon photonics. Light: Science &
Applications 4, e358 (2015). 14
[30] Crosnier, G. et al. Hybrid indium phosphide-on-silicon nanolaser diode. Nature Photonics
11, 297–300 (2017).
[31] Volet, N. et al. Semiconductor optical amplifiers at 2.0-µm wavelength on silicon. Laser and
Photonics Reviews 11 (2017). 14
[33] Griffiths, P. R. The Early Days of Commercial FT-IR Spectrometry: A Personal Perspective.
Applied Spectroscopy 71, 329–340 (2017). 15
[34] Xu, Q. Silicon dual-ring modulator. Optics Express 17, 20783–20793 (2009). 15, 53
[35] Yu, H. et al. Silicon dual-ring modulator driven by differential signal. Optics Letters 39,
6379–6382 (2014). 53, 84
[36] Müller, J. et al. Optical peaking enhancement in high-speed ring modulators. Scientific
Reports 4, 6310 (2014). 15, 53, 89
[37] Mesaritakis, C., Papataxiarhis, V. & Syvridis, D. Micro ring resonators as building blocks
for an all-optical high-speed reservoir-computing bit-pattern-recognition system. J. Opt. Soc.
Am. B 30, 3048–3055 (2013). 15
[39] Huang, C., Fan, J. & Zhu, L. Dynamic nonlinear thermal optical effects in coupled ring
resonators. AIP Advances 2 (2012). 15, 53
[40] Mancinelli, M., Borghi, M., Ramiro-Manzano, F., Fedeli, J. M. & Pavesi, L. Chaotic dynam-
ics in coupled resonator sequences. Optics express 22, 14505–16 (2014). 15, 53
[41] Santos, F. G. S. et al. Hybrid confinement of optical and mechanical modes in a bullseye
optomechanical resonator. Optics Express 25, 508–528 (2017). 19
BIBLIOGRAPHY 100
[42] Benevides, R., Santos, F. G. S., Luiz, G. O., Wiederhecker, G. S. & Alegre, T. P. M.
Ultrahigh-Q optomechanical crystal cavities fabricated in a CMOS foundry. Scientific Re-
ports 7, 2491 (2017).
[43] Luiz, G. O. et al. Efficient anchor loss suppression in coupled near-field optomechanical
resonators. arXiv:1701.03321 (2017). 19
[44] Barea, L. A. M., Vallini, F., Jarschel, P. F. & Frateschi, N. C. Silicon technology compati-
ble photonic molecules for compact optical signal processing. Applied Physics Letters 103,
201102 (2013). 19, 74
[45] Yu, H. et al. Performance tradeoff between lateral and interdigitated doping patterns for high
speed carrier-depletion based silicon modulators. Optics Express 20, 12926–12938 (2012).
20
[46] Zortman, W. a., Lentine, A. L., Trotter, D. C. & Watts, M. R. Low-voltage differentially-
signaled modulators 19, 26017–26026 (2011). 20
[47] Almeida, V. R., Panepucci, R. R. & Lipson, M. Nanotaper for compact mode conversion.
Optics letters 28, 1302–1304 (2003). 24
[48] Zhang, Y. et al. A compact and low loss Y-junction for submicron silicon waveguide. Optics
Express 21, 1310–6 (2013). 28, 39, 42
[49] Yun, H. et al. Broadband 2×2 adiabatic 3 dB coupler using silicon-on-insulator sub-
wavelength grating waveguides. Optics Letters 41, 3041–3044 (2016). 42
[50] Dong, P. et al. Low loss shallow-ridge silicon waveguides. Optics Express 18, 14474–14479
(2010). 42
[51] Yurtsever, G., Weiss, N., Kalkman, J., van Leeuwen, T. G. & Baets, R. Ultra-compact
silicon photonic integrated interferometer for swept-source optical coherence tomography.
Optics Letters 39, 5228–31 (2014).
[52] Teng, C. C. et al. Fiber-pigtailed silicon photonic sensors for methane leak detection. In
Conference on Lasers and Electro-Optics, AM3B.2 (Optical Society of America, 2017). 42
[53] Sedky, S., Witvrouw, A., Bender, H. & Baert, K. Experimental Determination of the Maxi-
mum Post-Process Annealing Temperature for Standard CMOS Wafers. IEEE Transactions
on Electron Devices 48, 377–385 (2001). 42
[54] Patel, D. et al. High-speed compact silicon photonic Michelson interferometric modulator.
Optics Express 22, 26788–26802 (2014). 42
BIBLIOGRAPHY 101
[55] Dong, P. et al. Low power and compact reconfigurable multiplexing devices based on silicon
microring resonators. Opt. Express 18, 9852–9858 (2010). 42
[56] Bogaerts, W. et al. Silicon microring resonators. Laser Photonics Reviews 6, 47–73 (2012).
43
[57] Feng, S. et al. Silicon photonics: from a microresonator perspective. Laser & Photonics
Reviews 6, 145–177 (2012). 43
[58] Borselli, M., Johnson, T. & Painter, O. Beyond the Rayleigh scattering limit in high-Q
silicon microdisks: theory and experiment. Optics Express 13, 1515–1530 (2005). 47
[59] Barea, L., Vallini, F., De Rezende, G. & Frateschi, N. C. Spectral engineering with CMOS
compatible SOI photonic molecules. IEEE Photonics Journal 5, 2202717 (2013). 47
[60] Souza, M. C. M. M. et al. Embedded coupled microrings with high-finesse and close-spaced
resonances for optical signal processing. Optics Express 22, 10430–10438 (2014). 47, 53, 73,
75, 76
[61] Schmid, S. I., Xia, K. & Evers, J. Pathway interference in a loop array of three coupled
microresonators. Physical Review A 84 (2011). 48
[62] Franchimon, E. F., Hiremath, K. R., Stoffer, R. & Hammer, M. Interaction of whispering
gallery modes in integrated optical microring or microdisk circuits: hybrid coupled mode
theory model. Journal of the Optical Society of America B 30, 1048–1057 (2013). 48
[63] Schmidt, C. et al. Near-field mapping of optical eigenstates in coupled disk microresonators.
Physical Review A 85, 033827 (2012). 48
[65] Miller, S. A. et al. Tunable frequency combs based on dual microring resonators. Optics
Express 23, 21527 (2015). 50
[66] Gentry, C. M., Zeng, X. & Popovic, M. A. Tunable coupled-mode dispersion compensation
and its application to on-chip resonant four-wave mixing. Optics Letters 39, 5689 (2014).
50, 74
[67] Lu, X., Rogers, S., Jiang, W. C. & Lin, Q. Selective engineering of cavity resonance for
frequency matching in optical parametric processes. Applied Physics Letters 105, 151104
(2014). 50, 74
BIBLIOGRAPHY 102
[68] Marpaung, D. et al. Integrated microwave photonics. Laser & Photonics Reviews 7, 506–538
(2013). 51
[69] Dumeige, Y. & Féron, P. Coupled optical microresonators for microwave all-optical genera-
tion and processing. Optics Letters 40, 3237 (2015). 51
[70] Dong, J. et al. Compact notch microwave photonic filters using on-chip integrated microring
resonators. IEEE Photonics Journal 5, 0–7 (2013). 51
[71] Souza, M. C. M. M. et al. Spectral engineering with coupled microcavities: active control of
resonant mode-splitting. Optics Letters 40, 3332–3335 (2015). 53, 73
[72] Zhang, M., Shah, S., Cardenas, J. & Lipson, M. Synchronization and Phase Noise Reduction
in Micromechanical Oscillator Arrays Coupled through Light. Physical Review Letters 115,
163902 (2015). 53
[73] Yariv, A. Universal relations for coupling of optical power between microresonators and
dielectric waveguides. Electronics Letters 36, 321–322 (2000). 53, 54
[74] Haus, H. A. & Huang, W. Coupled-mode theory. Proceedings of the IEEE 79, 1505–1518
(1991). 53
[75] Suh, W., Wang, Z. & Fan, S. Temporal coupled-mode theory and the presence of non-
orthogonal modes in lossless multimode cavities. IEEE Journal of Quantum Electronics 40,
1511–1518 (2004). 58, 59, 63, 67
[76] Little, B. E., Chu, S. T., Haus, H. A., Foresi, J. & Laine, J.-P. Microring resonator channel
dropping filters. Journal of Lightwave Technology 15, 998–1005 (1997). 53, 58, 59, 66, 67
[77] Wu, J. et al. Compact on-chip 1 × 2 wavelength selective switch based on silicon microring
resonator with nested pairs of subrings. Photonics Research 3, 9–14 (2015). 53
[78] Haldar, R., Das, S. & Varshney, S. K. Theory and Design of Off-Axis Microring Resonators
for High-Density On-Chip Photonic Applications. Journal of Lightwave Technology 31, 3976–
3985 (2013). 53
[79] Verslegers, L., Yu, Z., Ruan, Z., Catrysse, P. B. & Fan, S. From Electromagnetically Induced
Transparency to Superscattering with a Single Structure : A Coupled-Mode Theory for
Doubly Resonant Structures. Physical Review Letters 108, 083902 (2012). 53
[80] Karalis, A. & Joannopoulos, J. D. Temporal coupled-mode theory model for resonant near-
field thermophotovoltaics. Applied Physics Letters 107, 141108 (2015).
BIBLIOGRAPHY 103
[81] Jia, Y. et al. Theory of Half-Space Light Absorption Enhancement for Leaky Mode Resonant
Nanowires. Nano letters 15, 5513–5518 (2015).
[82] Zhen, B. et al. Spawning rings of exceptional points out of Dirac cones. Nature 525, 354–358
(2015). 53
[83] Mason, S. Power Gain in Feedback Amplifier. Transactions of the IRE Professional Group
on Circuit Theory CT-1, 20–25 (1954). 56
[84] Chak, P., Poon, J. K. S. & Yariv, A. Optical bright and dark states in side-coupled resonator
structures. Optics Letters 32, 1785 (2007). 62
[85] Benisty, H. Dark modes, slow modes, and coupling in multimode systems. J. Opt. Soc. Am.
B 26, 718–724 (2009). 62
[86] Scheuer, J., Sukhorukov, A. A. & Kivshar, Y. S. All-optical switching of dark states in
nonlinear coupled microring resonators. Optics Letters 35, 3712 (2010). 62
[87] Gentry, C. M. & Popovic, M. A. Dark state lasers. Optics Letters 39, 4136 (2014). 62
[88] Sandhu, S. & Fan, S. Lossless intensity modulation in integrated photonics. Optics Express
20, 4280 (2012). 62, 64
[89] Hamam, R. E., Karalis, A., Joannopoulos, J. & Soljacic, M. Efficient weakly-radiative
wireless energy transfer: An EIT-like approach. Annals of Physics 324, 1783–1795 (2009).
62, 64
[90] Souza, M. C. M. M., Rezende, G. F. M., Barea, L. A. M., Wiederhecker, G. S. & Frateschi,
N. C. Modeling quasi-dark states with Temporal Coupled-Mode Theory. Optics Express 24,
18960–18972 (2016). 62
[91] Wu, J. et al. Nested Configuration of Silicon Microring Resonator With Multiple Coupling
Regimes. IEEE Photonics Technology Letters 25, 580–583 (2013). 66
[92] Willner, A. A. E., Khaleghi, S., Chitgarha, M. M. R. & Yilmaz, O. O. F. All-optical signal
processing. Journal of Lightwave Technology 32, 660–680 (2014). 74
[95] Chang, Q. et al. A tunable broadband photonic RF phase shifter based on a silicon microring
resonator. IEEE Photonics Technology Letters 21, 60–62 (2009). 74
[97] Sun, X. et al. Investigation of Coupling Tuning in Self-Coupled Optical Waveguide Res-
onators. IEEE Photonics Technology Letters 25, 936–939 (2013). 74
[98] Poon, J. K. S. J., Scheuer, J. & Yariv, A. Wavelength-selective reflector based on a circular
array of coupled microring resonators. IEEE Photonics Technology Letters 16, 1331–1333
(2004). 75
[100] Lin, R., Zhong, W.-D. D., Bose, S. K. & Zukerman, M. Light-tree configuration for multicast
traffic grooming in WDM mesh networks. Photonic Network Communications 20, 151–164
(2010). 80
[101] Xu, Q., Almeida, V. R. & Lipson, M. Micrometer-scale all-optical wavelength converter on
silicon. Optics letters 30, 2733–2735 (2005). 80, 82
[102] Almeida, V. R., Barrios, C. A., Panepucci, R. R. & Lipson, M. All-optical control of light
on a silicon chip. Nature 431, 1081–1084 (2004). 81
[103] Preble, S. F. F., Xu, Q., Schmidt, B. S. S. & Lipson, M. Ultrafast all-optical modulation on
a silicon chip. Optics letters 30, 2891–2893 (2005). 82
[104] Wright, N. M. et al. Free carrier lifetime modification for silicon waveguide based devices.
Optics Express 16, 19779–19784 (2008). 82
[106] Timurdogan, E. et al. An ultralow power athermal silicon modulator. Nature Communica-
tions 5, 1–11 (2014). 83
[107] Sun, C. et al. Single-chip microprocessor that communicates directly using light. Nature
528, 534–538 (2015). 83
[108] Li, G. et al. Ring resonator modulators in silicon for interchip photonic links. IEEE Journal
on Selected Topics in Quantum Electronics 19 (2013). 83
BIBLIOGRAPHY 105
[109] Sacher, W. D. et al. Coupling modulation of microrings at rates beyond the linewidth limit.
Optics Express 21, 9722–9733 (2013). 84
[110] Tzuang, L. D., Soltani, M., Lee, Y. H. D. & Lipson, M. High RF carrier frequency modulation
in silicon resonators by coupling adjacent free-spectral-range modes. Optics letters 39, 1799–
1802 (2014). 84
[111] Li, Y. et al. Coupled-ring-resonator-based silicon modulator for enhanced performance. Op-
tics Express 16, 13342–13348 (2008). 84
[112] Wade, M. T., Zeng, X. & Popovic, M. A. Wavelength conversion in modulated coupled-
resonator systems and their design via an equivalent linear filter representation. Optics
Letters 40, 107 (2015).
[113] Ehrlichman, Y., Dostart, N. & Popovic, M. A. Dual-Cavity Optically and Electrically Res-
onant Modulators for Efficient Narrowband RF/Microwave Photonics. In 2016 IEEE Inter-
national Topical Meeting on Microwave Photonics (MWP), TuMP24 (IEEE, 2016). 84
[114] Pantouvaki, M. et al. Comparison of silicon ring modulators with interdigitated and lateral
p-n junctions. IEEE Journal on Selected Topics in Quantum Electronics 19 (2013). 91
[115] Weigel, P. O. et al. Lightwave Circuits in Lithium Niobate through Hybrid Waveguides with
Silicon Photonics. Scientific Reports 6, 22301 (2016). 94
[116] El Amili, A., Souza, M. C. M. M., Vallini, F., Frateschi, N. C. & Fainman, Y. Magnetically
controllable silicon microring with ferrofluid cladding. Optics Letters 41, 5576–5579 (2016).
108
[117] Davis, J. A., Grieco, A., Souza, M. C. M. M., Frateschi, N. C. & Fainman, Y. Hybrid
multimode resonators based on grating-assisted counter-directional couplers. Optics Express
25, 16484–16490 (2017). 109
[118] Frey, B. J., Leviton, D. B. & Madison, T. J. Temperature dependent refractive index of silicon
and germanium. In Proceedings of SPIE - The International Society for Optical Engineering,
vol. 6273 II (2006). 111, 112
[120] Della Corte, F. G., Esposito Montefusco, M., Moretti, L., Rendina, I. & Cocorullo, G.
Temperature dependence analysis of the thermo-optic effect in silicon by single and double
oscillator models. Journal of Applied Physics 88, 7115–7119 (2000). 112
BIBLIOGRAPHY 106
[121] Li, H. H. Refractive index of silicon and germanium and its wavelength and temperature
derivatives. Journal of Physical and Chemical Reference Data 9, 561–658 (1980). 112
[122] Okada, Y. & Tokumaru, Y. Precise determination of lattice parameter and thermal expansion
coefficient of silicon between 300 and 1500 K. Journal of Applied Physics 56, 314–320 (1984).
112, 113
[123] Yariv, A. Optical electronics (Rinehart & Winston, New York, 1985), 3 edn. 114
[124] Popovic, M., Manolatou, C. & Watts, M. Coupling-induced resonance frequency shifts in
coupled dielectric multi-cavity filters. Optics express 14, 1208–1222 (2006). 116
107
Appendix A
(d) 1.0
(a) (c)
transmission
Normalized
0.8
50 µm 0 Oe
0.6 28 Oe
55 Oe
83 Oe
I-V +
0.4
110 Oe
source - 1556.4 1557.0 1557.6 1558.2
Wavelength (nm)
Helmholtz coils (e) 0
(b)
Tunable Laser PM fiber
∆ λ res (pm)
-100
B=0 B≠0 B
MF
Detector
-200
(f) 6000
loaded-Q
DAQ 5600
Figure A.1: (a) Microscope and SEM micrographs of the partially unclad silicon microring resonator. (b) Illustra-
tion of the MF on top of the microring for zero and non-zero external magnetic fields. The formation of nanoparticle
columns effectively reduces the local concentration of the MF cladding. (c) Experimental setup. (d) Transmission
spectrum (e) resonance shift and (f) loaded quality factor for magnetic fields between 0 and 110 Oe.
My contribution to this work included full fabrication and characterization. The resulting
publication in Optics Letters [116] includes M. Souza and A. El Amili as co-first authors.
A.2. HYBRID MULTIMODE RESONATORS 109
Mirror
heaters
Cavity
heater
Figure A.2: (a) Schematics of a hybrid multimode resonator. Bragg mirrors route light to the through port (on
resonance) or drop port (off resonance) at different transversal mode orders. Scanning electron microscope image
of the cavity and surrounding mirrors. (b) Measured spectral response for the 1st order quasi-TE over the phase
matched region for the through, drop, and add ports. (Inset) associated image of the 1st and 2nd order modes at
the output of the through port (on resonance) and drop port (off resonance), respectively. (c) Long-cavity (100
µm) hybrid resonator with integrated microheaters and (d) transmission spectra at different heating powers. The
heaters actuated in this case are over the mirror sections.
A.3. DIFFERENTIAL SENSING WITH COUPLED-RESONATORS 110
(a) (b)
Figure A.3: (a) Optical and SEM micrographs of the coupled resonator with the inner silicon ring exposed. (b)
Transmission spectra of the coupled resonator exposed to various concentrations of EG:H2 O solutions. The sharp
resonance of the inner disk is clearly shifted while the resonances of the unexposed outer ring remain stationary
and act as a reference.
111
Appendix B
In this section we present the properties of the silicon-on-insulator (SOI) waveguides used in
our realization of the chip-scale silicon photonics Fourier Transform spectrometer (Si-FTS). Our
design exploits the quasi-TE mode of a 250 nm-by 550-nm SOI waveguide. The effective index,
its dispersion and its thermo-optic derivatives are summarized in Table 1. Here, we have used
a contracted notation for partial derivatives, ∂n∂xeff ≡ ∂x n. These coefficients were obtained from
finite difference element (FDE) simulations in the frequency range from 180 THz to 210 THz and
for temperatures between 300 K and 400 K.
The simulations were performed using the FDE solver Lumerical MODE with modified bulk
refractive index models for both the silicon core and the silica cladding.
The models include a Sellmeier dependence for both silicon[118] and silica[119]. The refractive
index models presented in these references also include temperature dependence in the range from
20 K to 300 K, whereas we are interested in the temperature range from 300 K to 400 K. For the
x10-4
a 1.90 b 0.20
dTn + dT2n ∆T + dT3n ∆T2
Term contribution
(K-1)
1.80 0.10
∆nSi
∆T
T0 = 300 K
1.75 dTn = 1.8 x10-4 K-1 0.05
dT2n = 2.4 x10-7 K-2
dT3n = - 2.1 x10-10 K-3 (dT3n ∆T2)/dTn
1.70 0.00
280 300 320 340 360 0 20 40 60 80 100
T (K) ΔT (K)
Figure B.1: Thermo-optic coefficient (TOC) of silicon. a. Around 300 K, the TOC has a second
order dependence with temperature with coefficients shown in the figure. b. Contribution of first and
second order terms with respect to zeroth order. For a temperature change of 100 K, the first order
contributes around 14%. The second order contribution is around 1% and can be neglected. The model
is based on [120].
silica cladding, the thermo-optic coefficient (TOC) is practically constant around 300 K (3 · 10−5
K−1 ) and we use the thermo-optic model provided in [119]. Silicon’s TOC, on the other hand,
varies significantly with temperature and the model provided in [118] does not deliver consistent
results in the temperature range of interest. For this reason, we consider an index model for silicon
that combines the dispersion from [118] with the TOC obtained from investigations comprising
temperatures from 300 K to 400 K[120]. We introduce the thermo-optic behavior as independent
∂2n
additional terms in the refractive index model since the crossed-dependence ∂ν∂TSi is negligible in
the frequency and temperature ranges of interest [121].
The thermo-optic behavior of silicon around 300 K is presented in Fig.B.1. In this model, the
TOC has a second-order dependence with temperature (Fig.B.1a),
∆nSi
= dT n + dT 2 n∆T + dT 3 n∆T 2
∆T
where ∆T = T − T0 with T0 = 300K. The contribution of the temperature dependent terms with
respect to the zeroth order in depicted in Fig.B.1b. It shows that the first order has a non-negligible
contribution reaching close to 14% of the zeroth order for a temperature change of 100 K, while
the second order reaches a maximum of 1% in the same temperature range and can be neglected
in practice.
In addition to the thermo-optic effect, thermal expansion changes the total length of the inter-
ferometer’s arm and must be accounted for. The thermal expansion coefficient of silicon around
300 K is presented in Fig.B.2a and presents a strong dependence with temperature[122]. The
113
x10-6
a 3.0 b 0.35
α1 + α2 ∆T + α3 ∆T + α4 ∆T
2 3
2.9 0.30
Term contribution
2.8 0.25 (α2∆T)/α1
0.20
αSi (K-1)
2.7
T0 = 300 K 0.15
2.6 α1 = 2.5 x10-6 K-1
(α3∆T2)/α1
α2 = 8.5 x10-9 K-2 0.10
2.5 α3 = - 2.3 x10-11 K-3
α4 = 4.5 x10-14 K-4
0.05 (α4∆T3)/α1
2.4 0.00
280 300 320 340 360 0 20 40 60 80 100
T (K) ΔT (K)
Figure B.2: Thermal expansion coefficient of silicon. a. Around 300 K it has a strong dependence
with temperature, modeled using a third order polynomial with the coefficients shown in the figure. b.
Contribution of first, second and third order terms with respect to zeroth order. For a temperature change
of 100 K, the first and second order contribute around 33% and 10%. The third order contribution is
around 2% and is neglected. The model is based on [122].
αSi = α1 + α2 ∆T + α3 ∆T 2 + α4 ∆T 3 .
The contribution of the temperature-dependent terms with respect to the zeroth order in depicted
in Fig.B.2b. The first and second order terms contribute to 33% and 10% of the zeroth order
value for temperature changes of 100 K. The third order reaches a maximum of 2% in the same
temperature range. This last contribution is neglected since the uncertainty on the non-linear
terms obtained in the calibration process is higher than 2%. The coefficients used in our model
(zeroth, first and second order) are also summarized in Table 1.
114
Appendix C
Neglecting proportionality constants, the power spectral density can be defined as [123]
The MZI is composed of input and output couplers with frequency-dependent field coupling
coefficients (t1 (ν), k1 (ν)) and (t2 (ν), k2 (ν)), respectively, and two arms with attenuation factors
R1 (ν) and R2 (ν). Thus, after splitting into arms 1 and 2 and recombining at the output waveguide,
the electric field that propagated through each arm is
Z +∞
E1 (t) = T1 (ν)S(ν)ej2πνt ejφ1 (ν) dν (C.5)
−∞
Z +∞
E2 (t) = T2 (ν)S(ν)ej2πνt ejφ2 (ν) dν (C.6)
−∞
where T1 (ν) = t1 (ν)t2 (ν)R1 (ν) and T2 (ν) = k1 (ν)k2 (ν)R2 (ν) are real transmission factors and
115
2πν
φi (ν) = βi (ν)Li = neff,i (ν)Li (C.7)
c
where βi , Li and neff,i are the propagation constant, length and effective index of arm i and c is
the speed of light in vacuum. The real nature of the output fields requires that
Ti (ν) = Ti (−ν)
(C.8)
φi (ν) = −φi (−ν) −→ neff,i (ν) = neff,i (−ν).
The output power is then obtained time averaging the squared output electric field through
many field oscillations,
2
Iout ∝ Eout (t)
∝ [E1 (t) + E2 (t)]2
The first two terms in the last line of eq.C.9 contribute to the mean output power, while the last
term contains the interference term of interest. In the following we will focus on manipulating such
term, which after some manipulation can be written as
Z +∞
I = 2 hE1 (t)E2 (t)i ∝ T (ν)PSD(ν)ej∆φ(ν) dν (C.10)
−∞
2πν
∆φ = [neff,1 (ν)L1 − neff,2 (ν)L2 ] . (C.11)
c
Dropping the proportionality sign and taking I equal to the right hand side (RHS) of eq.C.10 gives
the relation between the oscillatory output power from the MZI – the interferogram – and the PSD
in the general case.
116
Appendix D
The 2D-FDTD simulations presented in the main text were performed using the software RSoft
FullWAVE. In our analysis we used the TE polarization, defined such that the electric field is
normal to the simulation plane. The devices were scaled down in order to reduce computation
time. The bus-coupled resonator consisted of a racetrack with 3-µm bending radius and 2-µm
straight sections while the embedded rings had a 1-µm bending radius. The gap between outer
ring and bus waveguide and between coupled resonators were 250 nm and 200 nm, respectively, and
all waveguides were 200-nm wide. The index of refraction of the surrounding background was 1.45
while different indices were used for the waveguide sections as follows: 3.48 for the bus waveguide
and the racetrack; 3.497 for the embedded rings in the first design (coupling to the racetrack at
different points); 3.502 for the embedded rings in the second design (coupling to the racetrack at
the same point). The slight difference in refractive index between embedded rings and racetrack
assures the degeneracy of all the three resonators. Notice that the index used for the embedded
rings is different in each design in order to compensate for the phenomenon of coupling induced
resonance frequency shift (CIFS) [124]. Losses in the system were accounted by introducing a
non-null imaginary part to the refractive index of all waveguides nimg = 10−4 .
The transmission spectra were calculated using a pulsed simulation. A broadband pulse was
launched into the bus waveguide and the outgoing light in the through port was recorded until
the pulse has completely exited the simulation domain. The transmission spectra were obtained
from the Fast Fourier Transform (FFT) of the time monitor in the through port. Once the spectral
response was calculated, the transient response of the devices was obtained launching a continuous-
wave excitation at wavelengths corresponding to each resonance of interest.
The FDTD results are compared to modified-TCMT calculations in Fig.5.7 of the main text.
The parameters used in the modified-TCMT calculations are: ng = 3.06, k1 = 0.285, k2 = 0.29,
P1 = 0.982 and P2 = 0.996.