Enhancement of Light Matter Interaction of Thin Film Materials in Optoelectronic Devices: Plasmonic Antennas, Electro-Optic Modulators, and Solar Cells

by Mohammadhossein Tahersima

B.S. in Electronics System Engineering, April 2012, Kyoto Institute of Technology

A Dissertation submitted to

The Faculty of The School of Engineering and Applied Science of the George Washington University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

August 31, 2018

Dissertation directed by

Volker J. Sorger Associate Professor of Electrical and Computer Engineering The School of Engineering and Applied Science of The George Washington

University certifies that Mohammadhossein Tahersima has passed the Final

Examination for the degree of Doctor of Philosophy as of April 30, 2018. This is the final and approved form of the dissertation.

Enhancement of Light Matter Interaction of Thin Film Materials in Optoelectronic Devices: Plasmonic Antennas, Electro-Optic Modulators, and Solar Cells

Mohammadhossein Tahersima

Dissertation Research Committee:

Volker J. Sorger, Associate Professor of Electrical and Computer Engineering, Dissertation Director

Can Korman, Professor of Electrical and Computer Engineering, Committee Member

Mona Zaghloul, Professor of Electrical and Computer Engineering, Committee Member

Shahrokh Ahmadi, Professor of Electrical and Computer Engineering, Committee Member

Anastas Popratiloff, Director, Center for Microscopy and Image Analysis, Committee Member

iii

Dedicated to my parents.

iv Acknowledgments

First, I would like to thank my advisor, Prof. Volker Sorger, for agreeing to supervise this interdisciplinary project. I am also indebted to Profs. Can Korman, Anastas

Popratiloff, Shahrokh Ahmadi, Mona Zaghloul, Ergun Simsek, Cesare Soci, Ludwig

Bartels, Evan Reed, Bala Pesala, and Masahiro Yoshimoto for the many discussions about research, academia, and life in general. I am also grateful to Drs. Danang Birowosuto, Ke

Liu, Yigal Lilach, and Christine Brantner, for providing me with precise feedback during my research.

I would like to thank my lab mates Zhizhen Ma, Konstantinos Oikonomou,

Sikandar Khan, Ameen Elikkottil, Shuai Sun, Rubab Amin, for having a great time working with them. Thanks also go out to my friends: Matteus, Blaire, Jimmy, Ehsan, and Kimberly.

Special thanks go to my loving family, Vanessa, Ali, Hanif, Mina, Maryam,

Mahmoud and Robab. You define who I am.

This work was supported by the National Science Foundation (NSF) Designing

Materials to Revolutionize and Engineer our Future (DMREF), East Asia and Pacific

Summer Institutes for U.S. Graduate Students (EAPSI), and George Washington

University (GWU) fellowships. Furthermore, the Nanofabrication and Imaging Center of

GWU and the Center for Nanoscale and Science and Technology (CNST) of National

Institute of Standards and Technology (NIST) is hereby acknowledged for technical support.

v Abstract of Thesis

Enhancement of Light Matter Interaction of Thin Film Materials in Optoelectronic Devices: Plasmonic Antennas, Electro-Optic Modulators, and Solar Cells

The most often cited challenge in the field of nanoscale optoelectronics is the

weak light matter interaction that has traditionally led to bulky optoelectronic

components in scales comparable to the wavelength of light (~500 nm). Recently ultra-

thin film (0.5-20 nm) materials have demonstrated to have unique potential for

applications in planar optoelectronic and integrated photonics. However, the small

optical path across such ultra-thin film materials is the major limiting factor in their

optoelectronic performance. In this dissertation, I discuss my PhD research activities in

enhancement of light matter interaction of ultra-thin film materials in optical resonant

cavities for photo-emission, photo-absorption, and electro-optic modulation application

by localizing optical energy in Plasmonic, Fabry-Perot, and Micro-Ring cavities.

Transition metal dichalcogenides (TMDs) are stable and naturally occurring

semiconductors of two-dimensional (2D) materials that offer well-defined tunable direct

band gaps when thinned down to a nanometer. To increase the visible light emission

from direct bandgap of TMD monolayers for application in LEDs, nanoscale plasmonic

antennae offer a substantial increase of the electric field strength over very short

distances, comparable to the native thickness of the TMD. Here I report on the emission

enhancement generated in TMD films by several nanoantenna geometries compared to

their intrinsic emission.

Next, to increase the photo-absorption of TMD thin films further, to compete with

thick classical materials, I propose and investigate a novel stack of 2D material

vi heterostructure forming a core-shell light-concentrating optical cavity. This structure is motivated by deploying the mechanical flexibility of 2D materials to enable a multilayer solar cell without the necessity to contact each of the layers separately. We further investigate and demonstrate a spectral filtering metasurface for selective guiding of solar spectrum for smart power windows.

Finally, Indium tin oxide, that is already an industrial transparent conducting oxide material, shows strong electro-optic tunability in their thin films (~10 nm). I study its application in a novel micro ring reservoir coupling as a wavelength scale CMOS compatible phase modulator on silicon photonic platform.

In conclusion, novel nano-photonic components have been proposed and demonstrated to outperform traditional optoelectronics by taking advantage of the unique properties of atomically thin film materials and optical cavities. These finding are important for fast growing application of photonics in lighting, telecommunication, and optical energy conversion.

vii Table of Contents

Dedication ……………………………………………………………………………… iv

Acknowledgement ……………………………………………………………………… v

Abstract ………………………………………………………………………………… vi

List of Figures …………………………………………………………..……………… x

List of Tables ………………………………………………………...………….….… xiii

Abbreviations ………………………………………...……………………….……… xiv

Chapter 1 Introduction ...... 1

1.1 Objective...... 1

1.2 Problem Statement ...... 3

1.3 Summary of Main Contributions ...... 7

1.4 Organization of the Dissertation ...... 7

Chapter 2 Optoelectronic Properties of Nanoscale Materials ...... 9

2.1 2D Transition Metal Dichalcogenides...... 9

2.2 Plasmonic Materials ...... 15

2.3 Indium Tin Oxide...... 19

Chapter 3 Fabry-Perot Cavity in Spiral Solar Cells ...... 25

3.1 Introduction ...... 26

3.2 Solar Spectrum Absorption of 2D Materials ...... 27

3.3 Van der Waal Heterostructures and Spiral Solar Cell ...... 31

3.4 Numerical Methods ...... 36

3.5 Results and discussion ...... 38

3.6 Conclusion and Outlook ...... 47

viii Chapter 4 Light Emission Enhancement of 2D Materials in Optical Antennae ... 51

4.1 Introduction ...... 52

4.2 Mie Scattering ...... 59

4.3 Field Enhancement ...... 64

4.4 Quality Factor and Purcell Effect ...... 71

4.5 Analytical and Numerical Method ...... 75

4.6 Device Fabrication ...... 81

4.7 Measurement ...... 85

4.8 Experimental Results and Discussion ...... 87

Chapter 5 Diffraction Grating for Light Path Engineering ...... 93

5.1 Introduction ...... 93

5.2 Simulation and Design ...... 96

5.3 Fabrication and Testing ...... 99

5.4 Conclusion ...... 103

Chapter 6 ITO Ring Resonators for Electrooptic Modulation ...... 105

6.1 Optical Index Tunability of ITO ...... 109

6.2 Theoretical Study of Optical Ring Resonators ...... 111

6.3 Numerical Study of Optical Ring Resonators ...... 121

6.4 Device Fabrication and Testing...... 123

6.5 Conclusion and Results ...... 129

Chapter 7 Outlook and Conclusion ...... 130

7.1 Bibliography ...... 132

ix List of Figures

Figure 2.1: 2D atomic layered materials offer a variety of band structures ...... 10

Figure 2.2: Characterization of CVD-grown WS2 ...... 14

Figure 2.3: The dispersion relation of SPPs ...... 17

Figure 2.4: Comparison of optical properties of conventional plasmonic materials...... 18

Figure 2.5: Optical characterization of ITO ...... 22

Figure 3.1: light absorption in materials ...... 28

Figure 3.2: The solar spectrum...... 29

Figure 3.3: 2D materials can cover a very wide range of the electromagnetic spectrum. 30

Figure 3.4: The complex refractive index for a trilayer MoS2 ...... 31

Figure 3.5: Abundance of MoS2 on earth is from reference ...... 32

Figure 3.6: Light absorption in 100 nm bulk semiconductors and stacked MoS2...... 33

Figure 3.7: Spiral cell structures...... 35

Figure 3.8: Light absorption in spiral cell ...... 39

Figure 3.9: Spectral current density of spiral solar cell...... 42

Figure 3.10: Effect of hBN thickness on the absorption efficiency ...... 44

Figure 3.11: Polarization depedance of absorption efficiency ...... 45

Figure 3.12: vertical and horizontal illumination of solar cell ...... 46

Figure 3.13: Multi-physics approach, combining optical and electrical simulation ...... 48

Figure 3.14: Generated carrier density in to import to device simulator ...... 49

Figure 3.15: Ultra-light and highly efficient photovoltaics...... 50

Figure 4.1: Optical characterization of CVD-grown WS2 ...... 58

Figure 4.2: Absorption and PL of WS2 at room temperature ...... 58

x Figure 4.3: Cold cavity response...... 60

Figure 4.4: Cold cavity response map for dimer nanodisc antennas ...... 63

Figure 4.5: Electric field intensity enhancement...... 65

Figure 4.6: Electric field distribution profile for monomer antennas...... 69

Figure 4.7: Electric field distribution profile for dimer antennas...... 70

Figure 4.8: Quality factor of nanoantenna ...... 73

Figure 4.9: Photon generation rate...... 74

Figure 4.10: Purcell factor of layered materials under plasmonic cavities...... 78

Figure 4.11: Quenching of the radiative component of Purcell factor ...... 80

Figure 4.12: Fabrication steps of optical antenna-WS2 emitter ...... 83

Figure 4.13: Fabricated optical antenna devices...... 85

Figure 4.14: Schematic of custom-built micro PL setup ...... 86

Figure 4.15: Image of custom-built integrated microscope ...... 87

Figure 4.16: Enhancement of photoluminescence (PL) emission ...... 89

Figure 5.1: The smart window platform...... 94

Figure 5.2: Optimization of grating parameters...... 99

Figure 5.3: High contrast grating prototype ...... 101

Figure 5.4: High contrast grating optical performance measurement...... 102

Figure 5.5 Spectral guiding efficiency and power conversion efficiency ...... 103

Figure 6.1: Critical coupling effect in Reservoir Coupling Electro-Optic Modulator .... 107

Figure 6.2: Ring resonator, showing important key spectral performance metrics ...... 113

Figure 6.3: Experimental normalized transmission function of ring resonators...... 114

Figure 6.4: Sensitivity of ring resonators to minor process variations ...... 116

xi Figure 6.5: Comparison of a) interactivity cavity, b) coupling cavity...... 117

Figure 6.6: Comparison of three theoretical type of ring modulation mechanisms...... 118

Figure 6.7: Required change of effective refractive index to obtain extinction ratio. .... 120

Figure 6.8: Flowchart of simulation steps and optimization of geometrical parameters. 123

Figure 6.9: The CAD layout used for exposure and alignment ...... 124

Figure 6.10: Optical image of the device at each major fabrication steps ...... 125

Figure 6.11: Fabrication steps of the ITO reservoir coupling modulator...... 126

Figure 6.12: Development pipeline and measurement and verification setup...... 128

xii List of Tables

Table 2.1 Bandgap and band type of most popular 2D material...... 11

Table 3.1: Electrical properties of MoS2 for device simulation ...... 49

Table 4.1: Summary of simulation and experimental data result...... 90

Table 6.1: ITO carrier concentration levels and effective indices ...... 110

Table 6.2: Summary of experimental transfer function of ring resonator ...... 115

xiii Abbreviations

1L One layer

2D two dimensional

Al Aluminum

Ag Silver

Au Gold

BP Black Phosphorus

c-band Communication band

cm Centimeter

CMOS Complementary Metal Oxide Semiconductor

Cu Copper

CVD Chemical Vapor Deposition

DFT Density Function Theory

DOS Density of States

eV Electron Volt

EBL Electron Beam Lithography

EOM Electro-Optic Modulator

FDTD Finite Difference Time Domain

FWHM Full Width Half Maxima

Gr Graphene

hBN Hexagonal Boron Nitride

ITO Indium Tin Oxide

xiv LED Light Emitting Diode

meV milli electron volt

MOS Metal Oxide Semiconductor

MoS2 Molybdenum Disulfide

MoSe2 Molybdenum Diselenide

MoTe2 Molybdenum Ditelluride

MSE Mean Squared Error

NA Not Applicable

NIR Near Infrared

nm nanometer

PIC Photonic Integrated Circuits

PL Photoluminescence

PMMA Poly Methyl Metha Acyrlate.

Q factor Quality factor

RF Radio Frequency

Si Silicon

SiO2 Silicon Dioxide

SP Surface Plasmon

SPP Surface Plasmon Polariton

sub-흀 Sub Wavelength

TMD Transition-Metal Dichalcogenides

UV Ultra Violet

WS2 Tungsten Disulfide

xv WSe2 Tungsten Diselenide

xvi Chapter 1 Introduction 1.1 Objective

Light absorbing, emitting, and modulating devices are used widely in modern technologies, processes and applications such as sensing, energy generation, lighting, and telecommunication. A fundamental limitation for active light emitting devices exists due to the inherently weak interaction between light with matter.

This naturally leads to inefficient and bulky photonic devices that have slow device speeds, require higher power consumption, and result in uneconomical system cost.

Light absorption mechanism plays an important role in a broad range of optoelectronic components such as photodetectors, solar cells, optical modulators, or lasers. Although a few layers of layered materials are visually transparent, they are promising sunlight absorbers due to their large absorption per thickness and high density of states. Classically the amount of light absorbed in flat photovoltaic cells is proportional to the film thickness. However, creating an increased optical path length significantly reduces the amount of required photoactive materials.

This can be achieved via light management architectures such as planar metamaterial light-directing structures, Mie scattering surface nanostructures, or metal-dielectric-metal waveguides.

Nanoscale light-emitting structures are desired in a broad range of fields including displays, sensors, and optical interconnects. Currently, solid-state light- emitting diodes are based on direct band gap compound semiconductors such as gallium arsenide or gallium nitride that require epitaxy to grow and are difficult to integrate with silicon technology. Moreover, for nanoscale devices with increased

1 surface to volume ratio, these materials suffer from high parasitic surface velocity recombination rates that limit the internal quantum efficiency. Historically, organic and metal oxide materials have been explored for flexible electronics; However, their low charge mobility (µ) limited their prospects for their practical applications.

Transition-metal dichalcogenides (TMD) offer a mechanically flexible semiconductor material platform with a direct bandgap in optical regime when thinned to the monolayer limit. Mobility of TMD materials is comparable with silicon thin film transistors while being two orders of magnitude thinner with high strain tolerance for flexible electronic application. Additionally, they offer high quantum-yield and the potential for pure excitonic states discussed in this work in detail. As such, TMDs are attractive materials for novel nanoscale optical emitters and optoelectronic devices.

Oxide (ITO) is a CMOS compatible material that shows strong electro-optic effect.

Creating a high quality factor micro ring cavity can increase the light matter interaction and reduce the active device foot print for high speed and energy efficient integrated photonic components.

The main objective of this dissertation is to provide solutions to the following questions:

2 • What are the option for a scalable integration of atomically thin material with

lithographically fabricated devices. We investigate direct CVD growth and

electron beam lithography for device integration. Next, we identify the

thickness, Raman scattering, and the optical bandgap of TMD flakes and

optical index of ITO thin films with nondestructive optical techniques.

• Is it possible to enhance light absorption in 2D materials to be compared to

bulk semiconductor solar cells? Although absorption in 2D materials is

limited by their intrinsic atomic scale, the relative absorption per length in

them is extraordinarily high. We do this by taking advantage of their

mechanical flexibility to increase the optical path in this material. We further

investigate advantages and disadvantage of using 2D materials compared to

conventional solar cells.

• What are the limitations of using 2D materials as optical light sources and

how we can enhance their emission? We develop a test-bed approach to

answer this question and examine a highly emissive and stable TMD

materials and the two popular types of plasmonic particle cavities, namely:

patch disk antenna and dimer gap antennas.

• What are the opportunities in integration of ITO thin films in the silicon

photonic platforms? I will investigate the viability of integration of thin film

ITO materials with Silicon ring resonators for application as on-chip electro-

optic modulator.

1.2 Problem Statement

3 A drawback of nanoscale material for optic and photonic application is the low modal overlap with any optical field originating from the atomic thickness of the monolayer material and the fundamentally weak light-matter interaction. This can be enhanced via resonant (cavity, surface-plasmon-resonance) and non-resonant

(waveguide dispersion, metamaterials, index tuning) systems. Nanoantenna falls into the former category and they can synergistically act to a) increase the absorption cross- section thereby enhancing the pump efficiency, b) accelerate the internal emission rate via the Purcell factor through the nanoscale optical mode of the antenna, and c) improve emission out-coupling to free-space via impedance matching (transformer action). As such, optical antennas increase the excitation rate while simultaneously enhancing the local density of states (DOS) in the emission process, which modifies the spontaneous emission rate (Purcell effect). Hence, these optical antennas behave as electromagnetic cavities that strongly modify spontaneous emission of fluorescence in the spatial and spectral proximity [1]. As such, we apply plasmonic nanoantenna to enhance light emission from emissive 2D materials. Plasmonic antennas are unique cavities; a good antenna has a low quality (Q)-factor and is, thus, an effective radiator.

However, the light-matter interaction enhancement quantifier, or Purcell factor, is proportional to the ratio of Q/V, where V is the optical mode volume. Given the small plasmonic optical mode, relatively low Q factors ensure a decent antenna function, while Purcell factor value of a few 100 is obtainable. Compared to photonic high-Q cavities, plasmonic antennas enable a simultaneous absorption and PL enhancement.

Antenna-enhanced light-emitters have short radiative lifetimes and can have a deep subwavelength optical mode, thus opening the possibility of creating ultrafast,

4 nanoscale emitters [2].

For absorption and photocurrent extraction, previously vertical heterostructure stack of Gr/MoS2 materials has been investigated [3, 4]. These heterostructures can utilize effective combination of good solar spectrum absorption of TMDs and high mobility of graphene for carrier collection. This combination allows carrier collection by direct vertical charge transfer rather than lateral diffusion; hence improving photocurrent collection and reduce interlayer recombination losses. We investigate a three-dimensional, non-bulk solar cell structure based on a stack of 2D materials each with a functional purpose; by rolling a stack of Gr/MoS2/hBN. This structure is named the ‘spiral cell’, and it creates an absorbing cylinder forming a light-concentrating optical cavity as we discuss below. The spiral cell is motivated by deploying the mechanical flexibility of 2D materials to enable a multilayer solar cell without the necessity to contact each of the layers separately. Hexagonal boron nitride (hBN) is employed as barrier layer since it is an isomorph of graphene, optically transparent, electrically an insulator with a wide bandgap of about 5.9 eV, and it reduces the traps of MoS2 layers indicated by improved mobilities. Furthermore, a trilayer MoS2 with a bandgap of 1.6 eV was chosen over a monolayer since it has a more suitable band gap for solar absorption, and bulk behavior is not expected to emerge beyond stacking of

3-4 monolayers, whereas mobility improves with the number of layers. We contrast two absorbers, namely, the rolled-up Gr/TMD/hBN “spiral” cell, and a metal-cladded and metal inner-post “core-shell” cell. The reason behind studying both is that as to separate the effect of the material (i.e. TMD) absorption from any optical nano-cavity effects the core-shell device might exhibit. For the latter design, the stack rolls around

5 a core metallic rod and is then coated by another metallic shell, where both metals are the electrical cell’s contacts. The core metallic rod with a low work function

(aluminum) is in contact with the graphene, whereas the shell contact with a higher work function (gold) contacts to the MoS2 layer towards establishing selective contacts.

Photonic Integrated Circuits (PICs) is an emerging technology that uses semiconductor wafers to integrate passive and active photonic components along with electronic circuitry on a single chip. Compact electro-optic modulators are desired for optical interconnects and photonic circuitry. Silicon is the material of choice for integrated photonic technologies because they benefit from compatibility with the mature CMOS technology. However, Silicon shows a weak electro-optic effect; therefore, classical silicon based electro-optical modulators (EOM) are bulky due to large scale of an optical photon and weak electro-optic effect in silicon. Compact electro-optic modulators are desired for optical interconnects and photonic circuitry not only to reduce the device foot print, but also to increase their speed and power efficiency. To improve modulation speed and power consumption of EOMs we need to minimize their capacitance. There are two solution creating high performance

EOMs:

• CMOS compatible materials with stronger electro-optic effect [5].

• using optical cavities to localize light to reduce the device footprint [6].

Here we combine both solutions and combine ITO which is a CMOS compatible material with significant change in their optical constants by carrier injection with a micro ring resonator cavity to localize light and enhance light matter interaction in 10

6 -nm thin sheets of ITO.

1.3 Summary of Main Contributions

Through this dissertation I have demonstrated enhancement of light matter interaction using optical resonators to reduce the device footprint in optoelectronic devices to enable on chip optical communication, nanoscale light sources and portable and spectrally selective photovoltaics. The contributions of this project are:

1. Proposed and designed the Spiral Solar Cell concept that in principle enables

creating a multijunction solar cell using semiconductor materials with tunable

bandgaps.

2. Demonstrated a roadmap for spectral selectivity of plasmonic nanoantenna for

absorption and emission enhancement for 2D materials.

3. Demonstrated a spectrally selective solar window to generate electricity only

form the invisible near infrared portion of the solar spectrum while allowing the

visible light to light up the building.

4. Proposed and demonstrated a new mechanism of electro-optic modulators using

two layers of ITO material for reservoir coupling of micro ring resonators.

1.4 Organization of the Dissertation

This dissertation has been presented in seven chapters:

Chapter 1 begins with a high-level overview of the background and objectives of this work followed by listing specific contribution of this dissertation research to the field of optoelectronic.

Chapter 2 describes the state-of-the art of nanoscale materials such as monolayer

7 2D materials (i.e. graphene, hBN, MoS2, WS2), 10~20 nm thin transparent conductive oxide ITO, and plasmonic materials at nanoscale. Our work uses high-quality WS2 single-layer islands prepared by chemical vapor deposition (CVD) as the emissive 2D material. Using material obtained by a scalable deposition technique forwards relevance of our finding for future transition technological implementation and mass production.

We finely controlled sputtering and ion beam cluster for deposition of thin-film ITO material. Then various electrical and optical characterization techniques are used to examine the synthesized nanomaterials.

Chapter 3 describes the fundamental limitation in light absorption of an atomically thin sheet of material and propose a novel concept to form a core-shell optical micro- cavity from 2D materials to enhance broadband light absorption in solar cells made of them.

Chapter 4 discusses the limitation in spontaneous emission from semiconductors and particularly the case for 2D materials. We formulate a test bed approach to enhance the light-matter interaction of 2D materials in patch and gap mode plasmonic nanoantennas to enhance their photoluminescence.

Chapter 5 proposes a novel concept to guide the NIR portion of the solar spectrum to the edges of windows in skyscrapers and automobiles using optical high contrast grating. We then place a solar cell at the edge of the window to absorb the guided (near infrared) NIR spectrum.

Chapter 6 proposes a novel concept to modulate telecommunication wavelength using a sub-흀 CMOS compatible electro-optic modulator on coupling region of a microring resonator.

8 Chapter 2 Optoelectronic Properties of Nanoscale Materials

Novel materials enable novel application and assist to overcome bottlenecks of current technologies. This chapter starts with a discussion about the use of 2D materials for application in light absorption and light emission. We discuss various material deposition, etch, and metrology techniques used for 2D materials in this dissertation.

This chapter then discusses dispersion in Plasmonic materials and their application in enhancing light matter interaction. Finally, we introduce thin film ITO as CMOS compatible material with a strong electro-optic effect. We discuss various deposition and metrology techniques available.

2.1 2D Transition Metal Dichalcogenides

The field of atomically thin 2-dimensional (2D) materials enabled study of

material system in quantum confinement in one direction. 2D has demonstrated

non-classical phenomena; for instance, the well-studied carbon material, graphene,

is characterized by an absence of a band gap for monolayers [7]. This is relevant

for photon absorption since unlike semiconductor materials this band structure is

spectrally not band-edge limited, thus enabling broadband absorption.

Molybdenum and tungsten-based Transition-Metal-Dichalcogenides (TMD), do

have a bandgap in the visible range of 1-2 eV (Table 2.1) [8, 9, 10], and have been

found to be thickness, stress and composition-bandgap tunable [11]. The 0.8 eV

bandgap of black phosphorous (BP) [12] and molybdenum telluride (MoTe2) [13]

matches that of the telecom c-band (1550 nm), and hence could find applications

9 in silicon photonics industry. Moreover, the extensive attention to 2D material

research resulted in Van der Waals heterostructures that combine metallic (ex.

graphene), insulating (ex. hBN) and semiconducting (ex. MoS2) 2D materials to

form functional all-2D devices with atomic thickness in a wide range of

applications [3, 4, 14, 15].

Figure 2.1: 2D atomic layered materials offer a variety of band structures towards designing highly functional photo-absorption and photo-emission devices.

TMD is a family of 2D materials that are constructed by the formula MX2 (M = metal e.g. Mo or W; X = semiconductor e.g. S, Se, or Te) and are structured such that each layer consists of three atomic planes: a lattice of transition metal atoms

10 sandwiched between two lattices of chalcogenides. There is strong covalent bonding between the atoms within each layer and adjacent layers are held together by weak van der Waals forces. Taking a closer look at monolayer TMDs reveals some exceptional properties relating to interactions with light; quantum confinement in perpendicular direction to the plane of 2D material causes the band gap of MoS2, for instance, to increase from an indirect band gap of 1.2 eV for bulk material to a direct band gap of

1.8 eV for monolayers, which is accompanied by a 104 fold photoluminescence (PL) enhancement [8]. This property could lead to a high quantum-yield emissive semiconductor material of nanoscale dimensions and pure excitonic states [16, 17, 18]

. These are some of the major reasons why TMDs are attractive materials for novel nanoscale optical emitters and optoelectronic devices.

Table 2.1 Bandgap and band type of most popular 2D material. [8-18]

Additionally, the thickness-dependent bandgap tunability of MoS2 has enabled tunable photo-detectors for absorption in different wavelengths, where single- and double-layer MoS2 absorbs green light, while triple-layer absorbs in the red visible spectrum [11]. Despite being atomically thin, TMDs are promising solar spectrum absorbers originated from a bandgap between 1 and 2 eV, and their rich Van Hove singularity peaks in their density of states [3]. Because of their atomic thickness, optical and electrical properties of TMDs greatly alter by perturbations such as number of layers, physical stress, and coming in contact with other materials. For instance, an induced physical stress on MoS2 reduces the crystal symmetry, leading to significant shifts in the energy bands [19]. Such controlled bandgap engineering can introduce new opportunities for device engineering and unanticipated applications.

The direct band gap of monolayers of several chemically, structurally and electronically similar 2H phased semiconducting TMDs such as MoS2, MoSe2, WS2, and WSe2 has shown light emission in the visible and near-infrared spectrum (Table

2.1). The spatial confinement of carriers to a 3-atom thin physical plane and the weak dielectric screening in atomically-thin materials lead to high oscillator strengths and strong coulomb interactions between the excited electron in conduction band and remaining hole in the valence band even at room temperature [20, 21, 22]. Thus, the exciton dominates the optical properties of 2D materials and consequently large binding energies for monolayer TMDs are expected and 500 meV has been observed for single layer WS2 [22]. Beside excitons, charged trions can also be excited in the

12 presence of residual excess charge carriers. These quasiparticles consist either of two electrons and one hole (A-), or one electron and two holes (A+). Electrostatic gating therefore modifies the spectral weight of charge-neutral excitonic species in TMDs

[22, 18]. Moreover, given the large binding energy of the excitons, the formation of states consisting of two excitons (biexciton) is possible in TMDs, whose photoluminescence (PL) emission is red-shifted due to additional binding energy [21].

Our sample consists of WS2 monolayer flakes directly grown by CVD method on 100 nm of thermal SiO2 on Si wafer and their thickness was verified by optical contrast in bright field microscopy (Figure 2.2), and via observing an intensity change of the interlayer phonon mode A1g [23, 24, 25] of its Raman spectra (Figure 2.2). We characterize our emitter material via Raman and PL spectroscopy, as well as differential reflectance for WS2 flakes on a SiO2/Si substrate. Next, we perform a multi-Lorentzian curve fitting on each spectrum to obtain the composition Raman shift peaks attributed to the different vibrational modes in monolayer and fewlayer WS2

-1 (Figure 2.2). The strongest peak at ~354 cm is attributed to the in-plane E2g1 and second order vibrational mode 2LA(M), and the peak at ~418 cm-1 to the out of plane vibrational mode A1g(G). Our measured Raman modes match well with our theoretically (DFT) calculated Raman modes (see methods section for further details).

The A1g(G) mode blue shifts, and its relative intensity to the in-plane vibrational mode component increases, with an increasing number of WS2 layers. This is expected, due to stronger force caused by van der Waals interactions among the layers [25].

Figure 2.2: Characterization of CVD-grown WS2 at room temperature. (a) Optical images of as-grown WS2 on 100 nm SiO2 on silicon substrate. (inset) comparison of PL emission of monolayer and multilayer WS2 (b) Raman spectra of as-grown WS2 excited by 532 nm laser. (c) Normalized Raman spectra of as-grown monolayer, bilayer WS2. (d) PL spectrum (black curve) and absorption (red curve) spectra of as-grown WS2 monolayers, and PL spectrum (blue curve) of as-grown WS2 after deposition of few nm alumina.

The emission spectrum of WS2 shows a dependence on flake thickness, resulting in drastic increase in emission quantum efficiency on monolayer WS2 films indicating the indirect-to-direct bandgap transition upon multi-layer to monolayer scaling (Figure 2.2). PL spectra of WS2 monolayers are excitonic in nature, exhibiting strong emission corresponding to A excitonic absorption at 633 nm (~1.96 eV).

We further measure the differential reflectance of monolayer WS2 to obtain its absorption spectrum in the visible spectrum range and find the A excitonic absorption

14 peak at 631 nm (~1.96 eV). Excitonic emission from monolayer WS2 on SiO2/Si substrate demonstrates a very small stokes shift of ~2nm (less than 0.01 eV) between positions of the band maxima of the absorption and emission spectra (Figure 2.2). The emission is relatively narrow with a spectral full width at half the maxima (FWHM) of about 17 nm. For quantum-well structures, the Stokes shift and FWHM are indicators of interfacial quality. For instance, the magnitude of Stokes shift in monolayer TMD is found to increases with doping concentration [26]. Thus, the narrow emission spectra whose FWHM is comparable to thermal energy at room temperature along with small Stokes shift indicate the high quality of WS2 flakes as demonstrated here.

Excitonic absorption peaks A and B which arise from direct gap transitions at the K point are found, respectively, at 631 nm (1.96 eV) and 522 nm (2.38 eV) for monolayer

WS2 (Figure 2.2d). The energy difference between the A and B peaks of approximately

420 meV is an indicator for spin-orbit coupling in monolayer WS2. We find an additional peak near 460 nm, which arises from the DOS peaks in the valance band and conduction band.

2.2 Plasmonic Materials

Classical optical components are limited by the light cone of the linear dispersion relation: 푐 푘 = 휔√휇휖 = 휔 푛

휖휇 푛 = √ 휖0휇0

15 1 푐 = √휖0휇0

푐: Speed of light in vacuum

2휋 푘 = : the propagation constant 휆

휔 = 2휋휈: photon angular frequency n: refractive index [unitless number] c: speed of light in vacuum = 3 x 108 [m/s]

휖: permittivity of medium [F/m]

-12 휖0 : free space permittivity = 8.854 x 10 [F/m]

휇 : permeability of medium [H/m]

-7 휇0: permeability of free vacuum = 4p x 10 [H/m]

Light is an electromagnetic wave and the interaction of an electromagnetic wave with material interfaces, such as dielectric-dielectric, dielectric-semiconductor, or dielectric-metal, is governed by solving Maxwell’s equation for that boundary condition that depends on charge densities and permittivities. Permittivity of metallic materials at higher frequencies than their plasma frequency 휔푝 (optical or UV frequencies depending on the material) start to become transparent and behave more like a dielectric. Surface plasmon polaritons (SPP) are a natural solution of Maxwell’s equations at the metal-dielectric interface when frequency dependent permittivity of the metal and the dielectric material have opposite signs [27]. At this interface it is possible to have normal electric field components at the interface and a transverse magnetic field that lead to generation of surface charges.

Figure 2.3: The dispersion relation of SPPs suggest they can lead into generation of high k-vector and subwavelength optical particles far beyond the diffraction limit of light in dielectric media.

The dispersion curve of SP mode suggest that it always has a larger momentum than a free space photon at the same frequency.

휔 휖푑휖푚 푘푆푃푥 (휔) = √ 푐 휖푑 + 휖푚

휔 휖2 √ 푚(푑) 푘푆푃푧 (휔) = 푐 휖푑 + 휖푚

When permittivity of the metal is negative an equal to the positive permittivity of the dielectric material at the interface, the k-vector approaches infinity. The surface plasmon is quasiparticle and can be understood as a quantized collective oscillation of surface electrons along the interface of a plasma-like medium at optical frequencies.

The dispersion relation of SPPs suggest they can lead into generation of high k-vector and subwavelength optical particles far beyond the diffraction limit of light in dielectric media. Deep sub wavelength SPP makes a new generation of nanoscale

17 optical components a possibility [28, 29, 30, 31, 32, 33, 34, 35].

Popular materials for visible plasmonic applications are Au, Ag, Al, and Cu.

The dielectric properties in these metals cause a plasmon resonance in the visible spectrum (Figure 2.4). To qualify for SPP application, optical losses in the metal should be small to make light propagation along the surface possible.

Figure 2.4: Comparison of optical properties of conventional plasmonic materials for application in optical frequencies.

We plot the optical and electrical properties of popular plasmonic materials in

18 Figure 2.4. This plot shows that the permittivity of these materials can be expressed well by Drude model which we will discuss in detail in the following section. The complex permittivity and the refractive index of materials are related by:

2 2 휖1 = 푛 + 푘

휖2 = 2푛푘

Gold and copper are ideal for the red and NIR spectrum. Aluminum is a popular choice for the blue and UV spectrum. Silver shows better plasmonics properties than all other classical plasmonic materials at visible frequencies. However chemical stability and mechanical robustness is a concern for application of silver. Exploring new class of plasmonic materials is an active field of research [36]. In Chapter 3 and

4 we use plasmonic devices made of Au.

2.3 Indium Tin Oxide

Indium tin oxide (ITO) is a transparent conducting oxide in industries that manufacture displays, touch panels, defrosting windshields, fiber optics, organic LEDs, solar panels, infrared reflection coating, and even in electromagnetic and radio frequency interference shielding. Since it is one of few materials that offer transparency and conductivity, it is used to coat all sort of transparent materials that do not conduct electrons.

Moreover, it can operate at temperatures up to 1400 °C and can be used in harsh environments.

ITO can be considered a low electron density metal or a degenerate semiconductor with a bandgap of 3.5 eV and a large number of electron carriers. When ITO is biased as one electrode of a capacitor, it can achieve the three known states of accumulation,

19 depletion, or inversion. For example, in accumulation condition of a MOS configuration, free carriers are accumulated in the interface of the ITO and the oxide, thus changing the carrier concentration. The optical property of the active material therefore changes dramatically, resulting in strong optical tunability. Therefore ITO has potential for application in electro-optic modulators. We study their application as a phase modulator on a micro ring cavity in integrated photonics (Chapter 1Chapter 6). ITO should be deposited for a low-loss state (i.e. ON-state in modulation terminology) to achieve a high modulation depth. Because 1) high loss could negatively affect the quality factor of the micro ring resonator and 2) raising the carrier density via a voltage bias increases optical loss shifting the material into the absorption state [36].

To utilize ITO for nanophotonic devices, a detailed knowledge and parameter interdependencies relating to the optical and electrical properties must be known. ITO shows large invariances when deposited under different conditions [37] which makes it a difficult material to work with. We have previously reported on the material processing and focused on RF sputtering methods [36]. Here we focus on showing optical and electrical properties of RF sputtered ITO with process conditions of 25 W sputtering power,

10 sccm Argon flow, and a changing oxygen flow. Then we characterize the optometric properties of each deposition samples.

Ellipsometry is an optical technique to measure the dielectric properties (complex refractive index) of thin films. It directly measures the change of polarization upon reflection and compares it to a model. With appropriate modeling, it is then used to characterize refractive index, roughness, thickness, composition, doping concentration, and other material properties. The polarization change is quantified by the amplitude ratio,

20 Ψ, and the phase difference, Δ. Since the signal depends on the thickness and optical index, ellipsometry can be a universal tool for contact free determination of optical constants and thickness of thin film materials. By analysis of the change in polarization of light, ellipsometry can provide information about layers that are thinner than the wavelength of light down to a single atomic layer.

푟푝 휌 = = tan(Ψ) 푒𝑖Δ 푟푠

• 휌= complex reflectance ratio

• 푟푝= normalized reflected p-polarized component

• 푟푠= normalized reflected s-polarized component

• Ψ= amplitude component

• Δ= phase difference

Although it is common to characterize ITO films via optical measurements, accurate results are difficult to obtain due to the graded microstructure of the film introducing uncertainty in a single value of refractive index in ITO.

Figure 2.5: Optical characterization of ITO (a) measured transmission spectra of 25 nm thick ITO (b) Measured amplitude component (Ψ) and phase difference (Δ) of complex reflectance ratio as a function of wavelength using ellipsometry technique. (b,d) The fit of measured complex reflectance ratio to wavelength dependent refractive index for varying oxygen flow in the sputtering chamber (n is the real part and 휅 is the imaginary part) with an MSE of ~1.

Here we report our procedure and details of the performed ellipsometry tests. We measure the amplitude component (Ψ) and phase difference (Δ) of complex reflectance ratio as a function of wavelength using ellipsometry technique and fit our data to wavelength dependent complex refractive index of the material. The fitting procedure could be done either with a Cauchy model or a double Lorentzian approach. We choose the Cauchy model to fit the data with a Mean Squared Error (MSE) of about ~1. Most well-known models assume the sample is composed of a small number of discrete, well- defined layers that are optically homogeneous and isotropic. Violation of these assumptions requires more advanced fitting techniques. Ellipsometry characterization of

22 optical constant of ITO shows a low loss complex index of refraction (n = 2.15 + i 0.015) at telecom wavelength (1550 nm) for this deposition condition. The main difficulty of RF deposition at room temperature is low electrical conductivity in order of 1500 Ω due to insufficient recrystallization of ITO at low temperatures leading to the poor structural and electrical properties. Post annealing ITO can positively affect both optical and electrical properties of the material. The temperature budget of the annealing process step is in the order off 400 C.

Tunability of refractive index of ITO will be discussed in greater length in Chapter

Drude Model

The optical response (permittivites) in metals is dominated by the collective oscillations of free electrons in metal. Drude model is a purely classical model of electron gas that decsribes the coliision of freely moving electrons. Using Drude model we can estimate the permittivity 휖̃ = 휖′ − 𝑖휖′′ of nobel metals and dielectrics with metallic response (ITO at NIR regime) with a good approximation:

2 휔푝 푛̃2 = 휖̃ = 휖 − ∞ 휔(휔 + 𝑖훾) where 휖∞ is the high-frequency dielectric constant or background permittivity, ω is the

1 angular frequency of the illuminating light, 훾 = is the carrier scattering rate (i.e. 휏 collision frequency) and scales with imaginary part of the dielectric function, and 휔푝 is the unscreened plasma angular frequency defined by:

2 2 푁푐푞 휔푝 = ∗ 휖0푚

23 Here 푁푐 is the carrier concentration, q the electronic charge, 휖0 the permittivity of vacuum, and 푚∗ is the conductivity effective carrier mass. From this relationship we realize that the plasma frequency of materials dpends on carrier concntration 푁푐. In plasmonic metals the carrier concetration is large (~1023 cm-3) electrical tuning of carrier concetration is not possible. However, in ITO materials the plasma frequency can be tuned by changing the carrier concetration (~1019-1020 cm-3). The electron

|푞|휏 mobility 휇 and 휏 are related by 휇 = . Having smaller real part for permittivity would 푚∗ be advantagous, however this would be hard to achieve in metals due to their large

∗ carrie rconcentration. For ITO the conductivity effective mass, 푚 is taken as 0.35푚0, where 푚0 is the free electron mass.

Chapter 3 Fabry-Perot Cavity in Spiral Solar Cells

Solar energy represents one of the most abundant and yet least harnessed sources of renewable energy. Solar constant, which is the total energy flux incident on a unit area perpendicular to a beam outside the Earth's atmosphere, is about

1367[W/m2]. The choice of a semiconductor material with optimal properties for a specific application, the complexity of the required fabrication technology, reliability of its performance, all together determine the final cost of the technology development. Semiconductor electronics including photovoltaic industry is widely dominated by silicon at present. Because silicon had significantly less development complexity compared to other materials in the first place. High costs and investment reliability has prevented any alternative semiconductor material from entering the commercial market for decades except for high-frequency GaAs and very specific optical applications. However there has been a lot new discovery and excitement in material science and novel optoelectronics including 2D materials, plasmonics, light management structures, and perovskites. The recent isolation of two-dimensional

(2D) materials [7, 39] has provided opportunities to form new class of material systems that are attractive candidates for solar energy conversion applications. In this chapter we explore light absorption in semiconductor 2D TMD materials and propose a new class of micro structure that can enable light wave and multijunction solar cell made of a single 2D semiconductor material.

25 3.1 Introduction

Films of semiconductor transition metal dichalcogenide (TMD) have long been considered for photovoltaic devices, due to their large optical absorption, which is greater than 107푚−1across the visible range, meaning that 95% of the light can be absorbed by a 300-nm film [3]. Their monolayer has been recently shown to have extraordinary optical properties in two-dimensional quantum confinement regime and has drawn a lot of attention in the optoelectronic community. 2D materials, such as graphene and TMD, represent the ultimate scaling of material thickness and are of great interest for applications in atomically thin, flexible and transparent optoelectronics. Unlike traditional semiconductors, layered semiconductors show thickness dependent optical properties [Table 2.1]. For instance: bulk semiconductor

TMD has indirect band-gap, but it shows direct band-gap when exfoliated to few layers (3 to 5 layers) or monolayer. Value of the band-gap also increases as number of layers decrease. They also show an extraordinary high photon absorption relative to their thickness. Strong absorption in monolayer TMDs is due to the dipole transition between localized d orbital contributing to visible absorbance and excitonic coupling of such transitions [4]. Fourth, presence of critical points called

Van Hove singularities that generate in 2D or 1D leads to strong light-matter interaction [3].

However, light absorption in 2D materials is limited due to their intrinsic atomic thickness. Therefore, light path engineering in 2D materials become essential to make their application feasible.

26 3.2 Solar Spectrum Absorption of 2D Materials

Similar to other electromagnetic waves, when propagating light is incident on a medium, a portion of the light is reflected at the interface of the incidence

(Reflection) and the rest passes through an enters the second medium. Depending on wavelength of the light and thickness and type of material of the medium, a portion of light gets absorbed (Absorption) and another portion of it passes through

(Transmission) (Figure 3.1). The speed of propagating light in a material depends on the refractive index of the material: 푐 푣 = 푛

푣: speed of light in medium

푣: speed of light in vacuum

푛: refractive index of light

In addition to change of speed and wavelength, when light passes through a medium, some part of it will always be attenuated. This can be conveniently considered by defining a complex refractive index as:

푛̃ = 푛 + 𝑖휅

푛̃: complex refractive index

휅: extinction coefficient

According to Beer-Lambert law, the absorption of light in a material depends on the absorption coefficient of the material and light path length in the material.

퐴 = 1 − 푒−훼푙

퐴: absorptance

푙: beam path length

훼: absorption coefficient

27 Absorption coefficient is often expresses in the unit of [cm-1] and is calculated by:

4휋휅 훼 = 휆

Therefore, it is possible to accurately estimate light absorption in materials by measuring their extinction coefficient. The extinction coefficient can be measured using a variety of techniques most important of which are ellipsometry technique and transmission-absorption measurements. Using this equation, we can also easily confer that the thicker and absorbing material is or if the propagating light is oblique, then the absorption in active absorber layer is higher.

Figure 3.1: light absorption in materials a) reflection, absorption, and transmission upon incidence of light on a semiconductor material. b) dependence of light energy absorption on the bandgap of the semiconductor material.

The refractive index and extinction coefficient of materials are wavelength dependent. This is also called dispersion. It means certain materials are better at absorbing certain wavelength of light (spectral sensitivity). In the case of semiconductors, if the photon energy associated with the frequency (퐸 = ℎ휈) is

28 smaller than the bandgap of the semiconductor, the extinction coefficient will be negligible, and it won’t be absorbed by the semiconductor.

Figure 3.2: The solar spectrum is broadband from UV radiation to NIR regimes. The power from the visible portion of the sunlight is about 43 % of the total energy of the solar spectrum [39].

For a semiconductor if the photon energy of the radiation is larger than the bandgap then the photon can be absorbed by the semiconductor. The family of

TMDs contains four stable 2D semiconductor crystals of molybdenum disulfide

(MoS2), molybdenum diselenide (MoSe2), tungsten disulfide (WS2), and tungsten diselenide (WSe2) at ambient conditions with optical bandgaps in the visible to NIR range of 1~2 eV. These set of materials in combination with metallic graphene and insulating hBN can form a complete material system for optoelectronic applications.

Figure 3.3: 2D materials can cover a very wide range of the electromagnetic spectrum.

Despite being atomically thin, hence visually transparent, TMDs are promising sunlight absorbers originated from their rich Van Hove singularity peaks in their density of states [3]. In other words, 2D materials absorb an order of magnitude more sunlight compared to classical materials such silicon or gallium arsenide in the same amount of material thickness. Ellipsometry and absorption data show that a single layer of TMDs can absorb up to ~7% of visible light. These high light absorption scales up to 3 layers in MoS2 [41, 42].

Figure 3.4: The real and imaginary part of the complex refractive index for a trilayer MoS2 [41]. The refractive index values were fitted in an FDTD simulator.

We can accurately estimate spectral and angle resolved light reflection, transmission, and absorption in materials by knowing their refractive index and extinction coefficient. The experimental data [42] can be fitted using a multi- coefficient model to use in numerical simulations or Fresnel equations.

3.3 Spiral Solar Cell

Previously TMD/Gr vertical heterostructures were studied to improve photocurrent extraction [4, 3, 43, 44] in photovoltaic and photodetection

31 applications. These heterostructures can utilize effective combination of extraordinary solar spectrum absorption of some TMDs such as MoS2 and superior mobility of graphene (i.e. ~200,000 cm2/Vs for suspended graphene [45]) in

Schottky barrier solar cells. Although a few layers of both graphene and MoS2 are visually transparent, they are promising sunlight absorbers due to their large absorption per thickness and high density of states.

Figure 3.5: There is enough Mo atoms on earth to cover it over a billion times with monolayer MoS2. The data for abundance of elements on earth is from reference [46] .

One concern regarding use of a material for photovoltaic application is their abundance and availability. This is because there is need for large surface area of photovoltaics because the maximum normal surface irradiance of solar spectrum is at approximately 1000 W/m2 at sea level on a clear day. Molybdenum sulfide is a known material in other industries such as application in industrial lubrication. They are stable materials at ambient condition and show stability at high temperatures

32 (350oC). Here we calculate that since there is 1 molybdenum atom per million silicon atoms on the earth, there should be 3.6 × 1043 molybdenum on earth.

Knowing the lattice constant of MoS2 to be 0.31 nm 6.13 × 1033 atoms would be enough to cover the surface area on earth. This corresponds to covering earth 6 billion times with monolayer molybdenum sulfide.

Classically the amount of light absorbed in flat photovoltaic cells is proportional to the film thickness. However, creating an increased optical path length significantly reduces the amount of required photoactive materials. This can be achieved via light management architectures such as planar metamaterial light- directing structures, Mie scattering surface nanostructures, metal-dielectric-metal waveguide or semiconductor-dielectric-semiconductor slot waveguides [47].

Figure 3.6: Light absorption in 100 nm bulk semiconductors and a 100 nm of stacked MoS2 layers. (a) Numerical simulation setup compares light absorption in 100 nm traditional bulk semiconductor materials against 100 nm of stacked MoS2. For practical devices it is challenging to extract carrier from individual layers.

33 Atomically thin layered materials are better light absorbers because of their higher extinction coefficient as discussed previously. Initially, we compare light absorption (1-3 [eV]) of 100 nm thick bulk silicon, gallium arsenide, and MoS2 against 100 nm of stacked trilayer MoS2 (Figure 3.6). These results show that significant absorption of layered materials in quantum confinement regime thickness can be extrapolated to highly absorbing thicker absorbers that can compete with popular photovoltaic materials. The challenge however will be extracting carriers from each of these layers individually. Therefore, there is a need for a different device geometry that could extract carriers from each layer at the same time.

Here we investigate a three-dimensional, non-bulk solar cell structure based on a variety of stacked 2D materials each with a functional purpose (Figure 3.7); by rolling a stack of graphene, semiconducting MoS2, and an electrically insulating 2D material. This structure creates an absorbing cylinder forming a light-concentrating optical cavity as we discuss below. This structure is motivated by deploying the mechanical flexibility of 2D materials to enable a multilayer solar cell without the necessity to contact each of the layers separately. The electronic barrier layer needs to be flexible, wide bandgap, and optically transparent insulator to prevent electron and holes generated in a MoS2/Gr heterojunction from recombining in their adjacent stacked layer. Hexagonal boron nitride (hBN) is employed as barrier layer since it is an isomorph of graphene, optically transparent, electrically an insulator with a wide bandgap of about 5.9 eV, and it reduces the traps of MoS2 layers indicated by improved mobilities [19]. Furthermore, a trilayer MoS2 with a bandgap of 1.6 eV

34 [42] was chosen over a monolayer since bulk behavior is not expected to emerge beyond stacking of 3-4 monolayers [4, 8], whereas mobility improves with the number of layers [19].

Figure 3.7: (a) Schematic of “spiral cell structure”; parameters, d, l, and t stand for diameter of the roll, length of the cylindrical structure, and thickness of the hBN layer. (b) “Core- shell structure” of the spiral solar cell. Back reflector is connected to the core contact. Gold (aluminum) is chosen as a shell (core) selective contact. (c) Planar MoS2/Gr solar cell converting vertically incident photons into electron-hole pairs.

In this study we contrast two absorbers, namely, the rolled-up Gr/TMD/hBN

“spiral” cell (Figure 3.7a), and a metal-cladded and metal inner-post “core-shell” cell

(Figure 3.7b). The reason behind studying both is that as to separate the effect of the material (i.e. TMD) absorption from any optical nano-cavity effects the core-shell

35 device might exhibit. For the latter design, the stack rolls around a core metallic rod and is then coated by another metallic shell, where both metals are the electrical cell’s contacts. The core metallic rod with a low work function (aluminum) is in contact with the graphene, whereas the shell contact with a higher work function

(gold) contacts to the MoS2 layer towards establishing selective contacts. In our experiment, we numerically investigate light absorption deploying finite difference time-domain (FDTD) techniques. Comparing this spiral design to previously reported TMD based photovoltaic cells and thick (1 µm) planar MoS2 solar cells, we obtain a relative absorption enhancement which serves as a reference.

3.4 Numerical Methods

To obtain the absorption efficiency and spectral current density of the spiral cell, we first perform 3D FDTD simulations to solve Maxwell’s equations as a function of time and then execute a Fourier transformation. This strategy was selected for efficiency; the time-domain method covers a wide frequency range in a single simulation run.

A multi-coefficient model was used to represent the complex refractive index of a trilayered MoS2 [42], the Graphene monolayer [48], and hBN [48]. Here, the general approach is to send a broadband, normally incident plane wave pulse (300 to 800 nm) on both spiral structures (Figure 3.7). For angles of incidence larger than zero, the spiral cell is tested under both transverse electric (TE) and transverse magnetic (TM) polarization. Results for an unpolarized source can then be calculated by averaging simulation results of these two orthogonally polarized

36 beams (TE and TM). The length of spiral structure studied are between a reasonably small size (0.5 µm) for heavily meshed (0.5 nm × 0.5 nm) simulations, and the length for which absorption approaches its maximum values (3 µm); and their diameter range from 100 nm to 2 µm as the effect of hBN thickness on spiral cell is examined. We analyze an isolated single spiral cell and planar cell for which a perfectly matched layer and periodic boundary condition were selected, respectively, in the x and y direction. Note, a perfectly matched layer is assumed in propagation direction to absorb any back reflected waves. Furthermore, a power monitor surrounds the entire cell to obtain the net flow out of the simulation domain in all directions (Pout). The light source is placed inside the power analysis volume, right below the structures and parallel to the surface of incidence, and the absorption

(A) is obtained by A= 1-Pout. The absorption per unit volume can be calculated from the divergence of the pointing vector:

2 푃푎푏푠 = −0.5 푅(⃗∇⃗ . 푃⃗ ) = −0.5 휔⌈퐸(휔)⌉ 휀"(휔)

where 휔 is the angular frequency, 휀”(휔) the imaginary part of the permittivity, and |퐸(휔)|2 the electric field density. Hence, to calculate the spatial and frequency function of the absorption, we only need to know electric field intensity and imaginary part of the complex refractive index. For a solar cell, of relevance is the current density (J), which requires knowledge of the optical generation rate (G), which is given by

푃 |퐸(휔)2|휀"(휔) 퐺(휔) = 푎푏푠 = ℏ휔 2ℏ

from which we obtain the spectral current density

37 휆 퐽 (휆) = 훼(휆)퐼푄퐸(휆)Γ (휆) 퐴푀 1.5 ℎ푐 퐴푀 1.5

where 훼 is the absorption efficiency, 휆 is the wavelength, and Γ is the spectral irradiance of the ASTM 892 standard or AM 1.5 G solar condition, corresponding to air mass of 1.5 for a 37° tilted surface at one sun solar intensity (data taken from

[40]). Note, the internal quantum efficiency (IQE) is assumed to be unity for calculating the short circuit current (JSC) is obtained by integrating JSC(λ) over the wavelength range of 300 – 800 nm, namely

휆 퐽 = ∫ 퐸푄퐸(휆) Γ (휆) 푑휆 푠푐 ℎ푐 퐴푀1.5

3.5 Results and discussion

The results of the spectral scan display resonant-like fringes in the absorption spectrum, which we associate to cavity resonances as discussed below (Figure 3.8b).

We also find a higher visibility for the core-shell cell (with the metal cladding) compared to the spiral design (pure dielectric), which can be understood from the optical field confinement inside the structure. The resonances themselves suggest the cylinder spiral structure resembling a nanowire, and hence exhibiting Fabry-Perot cavity behavior [46, 47]. We confirm the latter via (i) investigating the modal features of this cavity (Figure 3.8c, d), and (ii) analyzing their frequency profile (Figure 3.8c).

Regarding (i), the transverse (x-y direction, i.e. x in Figure 3.8a) mode profile indicates a dipole for larger wavelength, which turns into quadruples and doubled-quadruples for blue shifting the resonance wavelength (6 to 1 in Figure 3.8c, d). In addition the

38 cavities’ standing waves can be seen in the cross-sectional-longitudinal modal profile

(i.e. x’ in Figure 3.8a), where the mode spacing decreases with wavelength (Figure

3.4c,d).

Figure 3.8: (a) Schematic of spiral cell (left) and core-shell structures (right). Horizontal and vertical cross sections labeled by X and X’ are monitors that record the power profiles shown in (c) and (d). (b) Absorption efficiency of spiral cell and core- shell structure. (c) Power profile of absorption efficiency peaks from 1 to 6 for spiral cell structure correspond to 344, 375, 441, 550, 627, and 744 nm wavelength respectively. (d) These wavelengths correspond to wavelength of peaks of the core- shell structure at 340, 388, 455, 550, 640, and 750 nm.

The higher Q-factors observed of the core-shell cell relative to the spiral cell are clearly visible in the crossectional mode profiles as distinct power density lopes (Figure 3.8b).

The apparent longitudinal focusing effect might be connected to a changing (i.e. increasing) local effective index as experienced by the wave traveling in positive z.

39 Regarding (ii), analyzing the resonance peak-spacing from Figure 3.8b and relating them to the inverse of the cavity length allows to test the Fabry-Perot cavity hypothesis via:

−1 휆2 푑푛 Δ휆 = (푛 − 휆 ( )) 2퐿 푑휆

−1 푑푛 푛 = (푛 − 휆 ( )) 푔 푑휆

where L is the length of the spiral cell (cavity), n is the effective refractive index, 푛푔 is the group index and 휆 is the wavelength (Figure 3.9c). Finding the results along a straight line confirms that the core-shell cell is a nanoscale cavity. This discussion suggests that the spiral structure behaves like a circular dielectric waveguide where the end facets act as reflecting mirrors [52, 53]. Neglecting dispersion, the expression for the mode spacing simplifies to:

2푛퐿 푚 = 휆 where m is the mode number (an integer). The spiral cell structure exhibits mode numbers between 5 and 13 that correspond to 9 visible peaks of absorption efficiency

(Figure 3.8b). This is supported by longitudinal mode profiles recorded by power monitors in Figure 3.8c, d, which demonstrates a higher interference visibility for the core shell structure for all monitored wavelengths.

The spectral current density of a 1 µm thick planar MoS2 cell, the spiral cell, and the core-shell structure are compared (Figure 3.9a). We find the spectral current density to be higher for the spiral structure over almost the entire investigated wavelength range. This is mainly because spiral structures have lower surface

40 reflection due to lower effective refractive index of spiral structures compared to planar

MoS2 structure. Core-shell structure has even higher absorption efficiency than the spiral cell, because the back reflector at the end of core-shell structure allows for light to pass twice through the absorber. The overall current densities of the planar cell, spiral cell, and core-shell structure integrated over wavelengths of 300 to 800 nm are

25.5, 29.5, and 37.2 mA/cm2 respectively. These current densities are several times higher than reported values for silicon nanowire or planar silicon solar cell with antireflection coating [54, 55]. The current density of the spiral (core-shell) cell shows

16 (46)% enhancement compared to the planar structure which is expected, because the spectral current density is higher for spiral and core-shell structures over almost the entire wavelength.

Figure 3.9: (a) Spectral current density of all three structures. (b) Optimization of hBN thickness to achieve maximum current density to photoactive material ratio. (c) Longitudinal mode spacing versus inverse nanowire length. The average group index for core-shell, and spiral structures are 5.28 and 3.81 respectively.

42 Note, that although the spiral length (l) was kept constant (1 m) this does not imply that the same amount of photoactive material was used; for example in a particular simulation, the thickness of monolayer graphene, trilayer MoS2, and few layers of hBN are set to 0.5 nm, 2.0 nm and 35.0 nm, respectively; and for the planar cell thickness of MoS2 is 1 m.

Thus, the absorbing materials (graphene and MoS2) occupy only 6% of the total volume of spiral cells. This suggests that the ratio of solar energy absorption to volume of photoactive material was ~7.67:1 compared to a bulk MoS2 photovoltaic cell of the same size. We name this ratio “enhancement” and define it as

훼푠푝𝑖푟푎푙 푐푒푙푙 − 훼푏푢푙푘 푐푒푙푙 푡 + 푡 ÷ 푀표푆2 퐺푟 훼푏푢푙푘 푐푒푙푙 푡푀표푆2 + 푡퐺푟 + 푡ℎ퐵푁 where denotes absorption and t refers to the respective physical layer thickness. This enhancement is proportional to the absorption efficiency of the cell and thickness of hBN layer. However, increasing the thickness of the hBN layer, decreases this absorption efficiency due to a reduction in the amount of absorbing material (Figure 3.6). Hence, to optimize the ratio of absorption enhancement relative to the volume of the photoactive material, the number of core shell structures with different hBN thicknesses are analyzed

(Figure 3.9b, Figure 3.10). For core-shell structures with 5 rings and 500 nm in length, the optimized enhancement is 762% for an hBN of 40 nm thickness. The dielectric field strength of hBN was previously studied to be 7.94 MV/cm [56], which corresponds to a breakdown voltage of 9 volts for a 10 nm thin layer of hBN. This is sufficient to prevent excitons generation to electrically short the cell.

Figure 3.10: Effect of hBN thickness on the absorption efficiency of spiral structure (a) and core-shell structure (b). The number of rings are set to 5 and length of the spirals are 500 nm. In both structures, reducing hBN thickness leads to increase in absorption efficiency.

Because of their atomic thickness, the electronic properties of 2D materials are sensitive to perturbations such as induced by strain via, for instance, a rolling process, as required here.

It has previously been confirmed that a linear redshift of 45-120 meV in the optical bandgap per percent of applied strain for monolayer and bilayer MoS2, and a direct-to–indirect transition of the optical bandgap at applied strains larger than 1% occurs for monolayer and bilayers of MoS2 [19, 41, 57]. We can estimate the strain in the MoS2 layer of spiral structure by:

푡 휎 = 2푟 where is the strain, t the thickness, and r the bending radius of the substrate (hBN).

Here we assume that strain effects for a trilayer is similar to a bilayer of MoS2 and a respective substrate. For a core-shell structure with hBN thickness of 10 nm, the tensile

44 strain values are positive and range from 8% at the first ring to 1% at the tenth ring, since the MoS2 trilayer rolls around the center post. However, in a spiral structure with the same hBN thickness, compressive strain values are negative and range from -17% at the first ring to -5% at the tenth ring. These strain values correspond to the indirect bandgap regime of strain except of the first ring of the spiral structure (i.e. -17% strain), which would make the material effectively a metallic [58, 59].

Figure 3.11: (a) Refractive index of the core-shell spiral structure at wavelength of 450 nm. (b) Brewster angle of the core shell spiral structure with effective index of 1.9 (effective index of structure at 500 nm). Absorption efficiency of the cell with TM (c) and TE (d) polarized source with different incident angles.

In order to verify the performance of the spiral cell at large angles of incidence (i.e. PV cells without tracking system), we compared the light absorption efficiency at varying angle of incidence measured for both transverse electric (TE) and transverse magnetic

(TM) polarizations (Figure 3.11). For normal incidence, the absorption efficiencies are essentially identical for both TE and TM mode and equal to that of unpolarized light. This

45 is because the reflection coefficient at normal angle of incidence is about the same for all polarizations. However, as the angle of incidence increases up to values near the Brewster’s angle, the reflection decreases for p polarized, and increases for s polarized or unpolarized light. This critical angle depends on the effective index of refraction, which has not only dispersive effects, but is also affected by the mode in which light propagates through the structure. The effective index used in Figure 3.11b is 1.9 which corresponds to effective index of the core shell structure at about 550 nm. The effective index of the core-shell structure ranges from 1.6 to 2.6 corresponding to wavelengths of 500 nm and 600 nm respectively. Group index of the core-shell structure is about 5.3 which is in range values often observed in plasmonic devices [60].

Figure 3.12: vertical and horizontal illumination of solar cell. (a) Absorption efficiency of spiral cell under vertical and horizontal illumination. (b) Schematic of spiral solar cell structures. It is possible to illuminate and monitor the optical field in this geometry horizontally or vertically (labeled by X and X’ are monitors that record the power profiles). (c) Power profile of absorption efficiency peaks shown from 1 to 3 for spiral cell structures correspond to ~ 350, 550, and 650 wavelengths respectively.

It is also important to note and compare the light absorption performance of spiral cell under horizontal illumination (Figure 3.12). The results of the spectral scan for vertical illumination display resonant-like fringes in the absorption spectrum, which we associate to cavity resonances as discussed below (discussed previously). We find a much less visibility for the horizontal illumination of the spiral cell compared to both spiral and core-shell structures under vertical illumination, which can be understood from the lack of aforementioned optical field confinement inside the vertical facets of the structure. The curvature in spiral cells also helps to focus the light at the center of the spiral cell.

3.6 Conclusion and Outlook

We have investigated a novel photovoltaic absorber and successfully demonstrated that rolling 2D materials into 3D structures can significantly improve their photo absorption compared to atomically thin or even bulk configurations [61, 62, 63, 64, 65,

66]. The spiral solar cell design can be optimized by tuning the hBN thickness to maximize both the absolute absorption (90%), and the absorption relative to the amount of photoactive material used.

We can estimate the power conversion efficiency for the investigated spiral core-sell design using the results obtained in this work. We state a lower and a reasonable limit for

2 this efficiency and use the current density (퐽푆퐶) of 37.2 mA/cm for the cell with length of

1 m and radius of 400 nm. An open circuit voltage (푉푂퐶) of 0.1 V and 0.5 V [4], and fill factor (FF) of 0.3 and 0.7, can be estimated as a lower and reasonable limit, respectively.

2 In both cases an IQE of unity and input power (푃𝑖푛) of 1000 W/m are assumed. The

47 efficiency 휂 is then calculated from

퐹퐹 × 퐽 × 푉 휂 = 푆퐶 푂퐶. 푃푖푛 resulting in an overall power conversion efficiency ranging between 1 to 13 %. The main reason for this medium high efficiency is the low 푉푂퐶 and should be the first target of further improving the cell. One possible way to overcome this barrier is to use either a multi-stack of 2D materials with different bandgap, such as alloying two materials creating a heterojunction. Furthermore, recent efforts on enabling horizontally tunable bandgap

TMDs [67, 68] utilized in this structure could lead to a multijunction photovoltaic cell that also only requires two contacts.

Figure 3.13: Evaluating performance metric for solar cells requires a multi-physics approach, combining optical and electrical simulation

Accurate determination of performance metric for solar cells requires a multi- physics approach, combining optical and electrical stimulation. For future work we can numerically model the power conversion efficiency of these geometries by importing the light generation profile from optical domain simulation to electrical device simulation

48 using continuity and drift-diffusion models.

Figure 3.14: (a) Absorption efficiency of bulk MoS2, spiral cell and core-shell structure. (b) Generated carrier density in a five-layer spiral cell ready to import to device simulator

From optical simulation methods such as FDTD, beside monitoring the optical transmission at edges of a device and the total light absorption, it is possible to accurately model and export spatial optical generation profile within the device geometry (Figure

3.14). This is essential for electrical domain simulation since the active material mobilities and the distance of generated electron/hole pair from contacts determine the external quantum efficiency of the photovoltaic cell. For device simulation a finite element method can import optical generation profiles and material properties to estimate the electrical conversion efficiency.

Table 3.1: Electrical properties of MoS2 for device simulation

Figure 3.15: Two dimensional materials show promise in achieving ultra-light and highly efficient photovoltaics.

Although still at an early stage of development, the known properties of atomically thin material systems for photon conversion are motivating for on-going research. Together with graphene and insulating layered materials such as h-BN, 2D semiconducting materials are an attractive choice for constructing solar cells on flexible and transparent substrates with ultra-thin form factors, and potentially even for mid-to-high efficiency solar cells

(Figure 3.15).

Chapter 4 Light Emission Enhancement of 2D Materials using Optical Antennae

Abstract

Monolayer transition metal dichalcogenides (TMDs) are materials with unique potential for photonic and optoelectronic applications. They offer well-defined tunable direct band gaps in a broad electromagnetic spectral range. The small optical path across them naturally limits the light-matter interactions of these two- dimensional (2-D) materials, due to their atomic thinness. Nanoscale plasmonic antennae offer substantial increase of field strength over very short distances, comparable to the native thickness of the TMD. For instance, it has been demonstrated that plasmonic dimer antennae generate hot-spot field enhancements by orders of magnitude when an emitter is positioned exactly over the middle of their gap. However, 2-D materials cannot be grown or easily transferred, to reside mid-gap of the metallic dimer cavity. Hence, it is not plausible to simply take the peak intensity as the emission enhancement factor. Here we show that the emission enhancement generated in a 2-D TMD film by a monomer antenna cavity rivals that of dimer cavities at a reduced lithographic effort. We rationalize this finding by showing that the emission enhancement in dimer antennae depends not on the peak of the field enhancement at the center of the cavity, but rather from the average field enhancement across a plane located beneath the optical cavity where the emitting 2-

D film is present. We test multiple dimer and monomer antenna geometries and

51 observe a representative 3-fold emission enhancement for both monomer and dimer cavities as compared to the intrinsic emission of chemical vapor deposition (CVD)- synthesized WS2 flakes. This finding suggests facile control and enhancement of the photoluminescence yield of 2-D materials based on engineering of light-matter interactions that can serve as test-bed for their rapid and detailed optical characterization.

4.1 Introduction

An optical antenna converts propagating electromagnetic wave in the optical regime to localized energy electrical current and vice versa. Despite the widespread use of antennas in GHz frequencies used in telecommunication, antenna’s use in the optical regime was hindered due to limitation in nanofabrication technologies. This is now has become possible due to accessibility of nanofabrication technologies such as electron beam lithography and steppers.

Nanoscale on-chip light-emitting structures are desired for a broad range of applications including displays, sensors, and optical interconnects. Transition-metal dichalcogenides (TMDs) are 2-D materials that show photoluminescence in the visible optical spectrum due to the emergence of a direct bandgap at the K point in the Brillouin zone. They also offer the potential for high quantum-yield emission and pure excitonic states [8, 69, 70, 10]. As such, TMDs are attractive materials for novel nanoscale optical emitters and optoelectronic devices [71, 61, 72, 3, 11, 15,

16].

The structural, chemical and electronically similar monolayer semiconducting

52 TMD MoS2, MoSe2, WS2, and WSe2 provide light emission in the visible and near- infrared spectral regions (1.1 eV – 2.0 eV). The spatial confinement of carriers to a

3-atom thin physical plane and the weak dielectric screening in atomically-thin materials lead to high oscillator strengths and strong Coulombic interactions between the excited electron-hole pairs. This results in strong binding energies allowing for observation of excitons at room temperature [73, 74, 75, 21]. In addition to neutral excitons, charged trions can also be excited in the presence of residual excess charge carriers. These quasi particles consist either of two electrons and one hole (A-), or one electron and two holes (A+). Therefore, electrostatic gating modifies the spectral weight of charge-neutral excitonic species in TMDs [17, 18,

20]. Moreover, given the large binding energy of the excitons, the formation of states consisting of two excitons (biexciton) is possible in TMDs, whose photoluminescence (PL) emission is red-shifted due to the additional binding energy

[17, 21, 22].

A drawback of using TMD materials for photonic and optoelectronic applications is the low modal overlap with any externally-applied optical field because of the atomic thickness of the monolayer material and the fundamentally weak light-matter interaction of bosons with fermions. Different strategies to tailor light matter interaction and enhance both absorption and emission of these materials have been developed such as resonant (cavity, surface-plasmon-resonance) and non- resonant (waveguide dispersion, metamaterials, index tuning) systems were utilized to achieve maximum light absorption and emission [76, 77, 78, 79, 80] [81, 82, 83,

84, 85] [86, 87, 88, 89, 90]. Plasmonic nanoantennae fall into the former category

53 and they can synergistically a) increase the absorption cross-section thereby enhancing the pump efficiency, b) accelerate the internal emission rate via the

Purcell effect through the nanoscale optical mode of the antenna, and c) improve emission out-coupling to free-space via impedance matching (transformer action).

As such, optical antennae increase the excitation rate while simultaneously enhancing the local density of states (DOS) in the emission process, which modifies

(here accelerates) the spontaneous emission rate known as the Purcell effect [1].

Hence, these optical antennae behave as electromagnetic cavities that strongly modify spontaneous emission of fluorescence in the spatial and spectral proximity.

Plasmonic antennae are unique cavities; a good antenna has a low quality (Q)-factor and is, thus, an effective radiator. However, the light-matter interaction enhancement quantifier is proportional to the ratio of Q/V, where V is the optical mode volume. Given the possibility for sub-diffraction limited plasmonic optical mode volumes, the inherently low Q factors ensure a decent antenna function, while high Purcell factors of ~10’s to 100’s are obtainable [83, 91, 92]. Compared to photonic high-Q cavities, plasmonic antennae allow for simultaneous high absorption and PL emission, if quenching is prevented. Antenna-enhanced light- emitters have short radiative lifetimes and can have a deep subwavelength optical mode, thus opening the possibility of creating ultrafast, nanoscale emitters [2].

Recently, multiple TMD-plasmonic hybrid nanostructures have been investigated. However, a detailed study probing the actual origin of light emission enhancement in these hybrid nanostructures have not yet been reported. A common claim in previous studies is that the gap-mode of the plasmonic dimer’s field

54 enhancement enabling high Purcell factors. Plasmonic dimers, which are nanoscale structures consisting of two metallic nanoparticles separated by a small gap, support hybridized plasmon resonances because of the capacitive coupling between the plasmon modes of each nanoparticle. For a quantum emitter (e.g. <20 nm quantum dot) placed inside this gap, the coupling strongly localizes charges at the junction between the two nanoparticles, giving rise to large field enhancements at the center of the feed-gap of the dimer antenna. However, for emitters that are not comparable in size to a quantum dot in all three dimensions, such as a TMD films, the emitter must be offset from the antenna either below or on top of the antenna and, thus, is unable to take the advantage of the highest electromagnetic field enhancement at the hotspot. For TMD emitters with zero distance from the plasmonic antennae, the strong field gradients of the point dipole source can efficiently excite lossy multipolar modes of the antenna which are mostly dark or weakly coupled to the radiation field and, therefore, convert the electromagnetic energy mostly into heat

[90]. To avoid the emitter being quenched (i.e. coupling to lossy plasmonic surface waves) the emitter should be separated from the metallic nanoparticle by a distance previously reported to be about ~8 to 10 nm [91, 92, 93]. Before we fabricated

TMD-nanocavity systems, we modeled a broad spectrum of dimer and monomer antenna configurations and find that – within the limitations imposed by the 2-D geometry of the film and the need for a separator - the maximum field enhancement of a dimer antenna relative to a monomer disc is only about two-fold. For such designs, we observe that the pertinent cavity field enhancement in the dimer case does not originate from the gap between the metal particles, but from each monomer

55 disc. We then further compare the TMD emission enhancement of monomer vs. dimer antennae relative to intrinsic PL emission.

Our work uses WS2 single-layer islands prepared by chemical vapor deposition

(CVD). CVD offers an, in principle, scalable source of TMD material for future technological implementation of this approach. Prior work [83, 87, 89] on plasmonic antennae utilized material obtained through exfoliation. The CVD process is the preferred method for synthesizing TMD materials due to the pristine monolayer quality of the materials including a high-level of process control as readily demonstrated by the semiconductor industry.

Although in atomically thin layered TMDs, atoms are confined in a plane, the electric field originating from dipole charges in the 2-D crystals have both in-plane as well as out-of-plane components. Moreover, the large surface to volume ratio in

2-D materials enhances the significance of surface interactions and charging effects.

Thus, the dielectric permittivity mismatch between the 2D semiconductor film, the surrounding environment, and an induced strain from capping material can intricately affect the electronic and optoelectronic properties of low dimensional materials [94].

In this study, we demonstrate that spontaneous emission of atomically thin WS2 film coupled to 4 different plasmonic nano-cavity design can be substantially enhanced up to 3 times compared to the intrinsic emission of CVD-grown WS2 flakes. We show that fluorescent enhancement of nano antenna coupled 2-D materials unlike quantum dots is an areal average effect rather than a hot-spot like effect and discuss how the On- and Off-resonance plasmonic of the TMD film’s

56 excitonic luminescence is susceptible to electric field intensity variations caused by surface plasmons.

Samples studied here consist of WS2 monolayer flakes grown directly by CVD on 100 nm of thermal SiO2 on a Si wafer. The bright field microscopy image in

Fig.1a shows the bare substrate appearing purple, the single-layer material as dark blue and thicker material regions as lighter blue areas. To characterize our emitter material, micro-Raman, micro-PL and differential reflectance spectra were taken on

WS2 flakes on a SiO2/Si substrate at room temperature. We confirm the thickness of the WS2 material by the appropriate difference in the intensity of the interlayer phonon mode A1g between material identified as multilayer and single-layer

(Figure1 b) as well as a corresponding difference in PL intensity [23, 24]. We further analyze the Raman signal by a multi-Lorentzian fitting of all recognizable features

- in both monolayer and few layer WS2 (Figure 4.1b). The strongest peak at ~354 cm

1 1 is attributed to the in-plane E2g and second order vibrational mode 2LA(M), and

-1 the peak at ~418 cm to the out-of-plane vibrational mode A1g(G). Raman modes derived from Density Functional Theory (DFT) calculations confirm the Raman mode assignment (see methods). The A1g(G) mode blue shifts, and its relative intensity to the in-plane vibrational mode component increases, with an increasing number of WS2 layers. This is expected due to a stronger interlayer force caused by van der Waals interactions [24]. The emission spectrum of WS2 shows a dependence on flake thickness: monolayer WS2 flakes exhibit a drastically increased emission quantum efficiency indicating the indirect-to-direct bandgap transition upon multilayer to monolayer scaling [25] (Figure 4.1a).

Figure 4.1: Optical characterization of CVD-grown WS2 at room temperature. (a) Optical images of as-grown WS2 on 100 nm SiO2 on silicon substrate. (inset) Comparison of PL emission of monolayer and multilayer WS2 (b) Room-temperature Raman spectra from a monolayer WS2 flake, including Lorentzian peak fits for the 532-nm laser excitation.

We further measure the differential reflectance of monolayer WS2 to obtain its absorption spectrum in the visible spectrum range and find the A excitonic absorption peak at 631 nm

(~1.96 eV). Excitonic emission from monolayer WS2 on SiO2/Si substrate demonstrates a very small stokes shift of ~2 nm (less than 0.01 eV) between the band maxima of the absorption and emission spectra (Figure 4.2b).

Figure 4.2: Characterization of CVD-grown WS2 at room temperature. (a) Normalized Raman spectra of as-grown monolayer, bilayer WS2 . (b) PL spectrum (black curve) and absorption (red curve) spectra of as-grown WS2 monolayers, and PL spectrum (blue curve) of as-grown WS2 after deposition of few nanometers of alumina.

58 The emission spectrum of monolayer WS2 is relatively narrow with a full width at half maximum (FWHM) value of about 17 nm. For quantum-well structures, the Stokes shift and FWHM are indicators of interfacial quality. For instance, the magnitude of the Stokes shift in monolayer TMD increase with doping concentration. Thus, the narrow emission spectra, whose FWHM is comparable to thermal energy at room temperature in combination with small Stokes shift indicates high quality of the WS2 flakes. The absorption spectrum of monolayer WS2 exhibits two excitonic absorption peaks A and B that arise from direct gap transitions at the K point; these are located at 631 nm (1.96 eV) and 522 nm (2.38 eV) respectively (Figure 4.2b). The energy difference between the A and

B peaks is approximately 420 meV, revealing the spin-orbit splitting in monolayer WS2.

4.2 Mie Scattering

When a quantum emitter interacts with the local fields of an optical antenna, the coupled system has a larger absorption cross-section compared to an isolated emitter leading to an optical concentration effect that enhances the effective pump intensity. To reveal such far- field cold-cavity effect and to find the resonances for the monomer and dimer nanoantennae, we analytically describe their spectral scattering efficiency and absorption loss response by dividing both the absorption cross section area and scattering cross section area of each individual antenna by its geometrical area:

퐶푒푥푡 퐶푠푐푎푡+퐶푎푏푠 푄푒푥푡 = 푄푠푐푎푡 + 푄푎푏푠 = = (1) 퐶푔푒표 퐶푔푒표

where 푄푒푥푡, 푄푠푐푎푡, 푄푎푏푠 (퐶푠푐푎푡, 퐶푎푏푠) are extinction, scattering, and absorption efficiency

(cross section), respectively. 퐶푔푒표 is the geometrical cross section of the antenna which

59 depends on its radius. We note that 푄푎푏푠 is a parasitic part of the cold-cavity and needs to be minimized, and it should not be evaluated as the emitter absorption where an enhancement is preferred as long as the system operates in the linear regime. The resonance wavelength and scattering cross section of the monomer antenna is a function of the permittivity of the plasmonic material and the dimension of the antenna. We have chosen gold for fabrication of our antennae due to its Drude-like response for wavelengths above

600 nm.

Figure 4.3: Cold cavity response. Absorption loss (Qabs) (a) and far field scattering efficiency (Qscat) (b) mapping of monomer nanodisc antennae for radius range of 50 nm to 200 nm (The black and white points represent our fabricated antennae for emission and excitation wavelengths respectively. Absorption loss (c) and far field scattering efficiency (d) for dimer nanoantenna of single 75 nm radius and in a gap sweeping range from --150

60 nm (overlapping charge transfer mode) to 150 nm (gap plasmon mode). The points represent our fabricated 75 nm dimer antenna. The scale bar shows the ratio for absorption or scattering cross section to geometrical cross section of each type of antenna. The dashed lines are guides to the eye.

Our modeling of the monomer antenna with a 75 nm radius reveals a scattering efficiency of about 1.8 and 6.3 at excitation wavelength of 532 nm and emission wavelength of ~640 nm, respectively, in which the emission of the WS2 is in resonance with the resonance of the cavity (Figure 4.3b). Since the monomer is electromagnetically a simple dipole under excitation, we observe the expected monotonic resonance redshift with increasing dimension of the monomer particle (dashed lines), while the discrepancy from a linear trend can be explained by dispersion. The spectral regions and antenna dimensions to avoid absorption losses are a) near the blue frequencies in the visible, and b) small monomers

(<75 nm) (Figure 4.3a). Further we find that the dimer antennae fall into two categories depending on whether the interparticle distances (i.e. gap) is positive (true dimer), or negative (lumped dimer) (Figure 4.3c). For the dimer antennae, the radius of each discs is kept at constant values of 75 nm, 100 nm, and 200 nm and the interparticle distance in between them is swept from negative to positive values, where the negative values of the gap represent overlapping dimer discs. Comparing the spectral resonances, we find that the lumped dimers effectively behaves as monomers with a metal particle diameter about equal the total length of the dimer (푟푑𝑖푚푒푟 ≈ 2 × 푟푚표푛표푚푒푟) (black dashed line, Figure 4.3d).

Although the scattering efficiency of each dimer antenna is less than the scattering efficiency of its corresponding monomer antenna with same radius, the scattering cross section is larger (~2x) relative to that of monomer antennae. As expected, the resonance for dimers with large positive gaps approaches that of isolated monomer.

61 For comparison we additionally study the far-field, cold-cavity effect for dimer nanoantennas with larger radii of 100 nm and 200 nm. We analytically describe their spectral scattering efficiency and absorption loss response. The dimer antenna with a 75 nm radius and 25 nm gap has a scattering efficiency of about 1.7 and 5.5 at excitation and emission wavelengths of 532 and 640nm, respectively. The

Scattering efficiencies reduce to 1.4 and 4.5 at a radius of 100 nm radius and a gap of 55 nm; and to 1.1 and 2.2 for dimers with 200 nm radius and 80 nm gap (Figure

4.4).

Figure 4.4: Cold cavity absorption loss (Qabs) and scattering efficiency (Qscat) mapping of dimer nanodisc antennas for radii of 75 (a, b), 100 (c, d), and 200 nm (e, f), respectively. We sweep the interparticle gap from negative (overlapping charge transfer mode) to positive values (gap plasmon mode). Denoted points represent our fabricated 75 nm dimer antenna, where the white dots correspond to excitation wavelength and black dots to the emission. The scale bar is the ratio of scattering cross section to geometrical cross section

63 of each type of antenna.

4.3 Field Enhancement

Moreover, the emission intensity profile of an emitter in an optical-cavity- antenna is governed by:

퐼 = (퐼 ) (휂 )(휂 ) 표푢푡 푝푢푚푝 푒푓푓 𝑖푛푡 표푐

|퐸|2 ∬ 푑푠 |퐸 |2 = [퐼 . ( 0 )] [퐹 . 푄퐸] 0 푝 휆−퐸푚𝑖푡 퐶푔푒표 휆−푃푢푚푝

where 퐼표푢푡 is the outgoing emission from the system, 퐼푃푢푚푝 is the effective incoming laser pump intensity onto the monolayer WS2 emitter, 휂𝑖푛푡 is the internal photon generation process of the emitter equal to spontaneous emission rate enhanced by the Purcell factor, 휂표푐 is the out-coupling efficiency of emitted photons from the monolayer WS2 via the cavity-antenna system into free space above the sample where the light is collected at the microscope objective, represented here as the quantum efficiency of the antenna.

Figure 4.5: Electric field intensity enhancement (|E|/|E0|) comparison of dimer and monomer antennae showing comparable enhancement at TMD position, which is separated by a spacer layer to avoid quenching. a) Schematic of the purposed optical antenna types for PL enhancement of monolayer TMDs. b) Side view of the electric field intensity magnitude enhancement distribution in a 4 nm and 25 nm gapped 75 nm radii dimer antenna and a 75 nm radius monomer antenna, respectively from top to bottom. c) Comparison of electric field intensity enhancement for the same antennae shown in top view and two different z-normal plane positions at the midpoint of the antenna and at the z-normal plane where TMD layer is positioned. The maximum value, the averaged value over the area of the beam spot size of the simulation, and the averaged value over the area 2 of geometrical antenna cross-section of (|E|/|E0|) is reported for each case.

Regarding the internal photon generation enhancement process of the TMD- cavity system, we focus on the near field enhancement and Purcell product in Eqn.

65 2. The overall fluorescence enhancement of monolayer WS2 by the plasmonic optical antenna can be expressed as the product of excitation rate enhancement, the spontaneous emission probability enhancement (Purcell effect), and the outgoing portion of the spontaneous emission:

퐼표푢푡 훾푒푚 훾푒푥푐 훾푟 = 0 = 0 0 퐼0 훾푒푚 훾푒푥푐 훾푟

where the ‘0’ denotes the intrinsic value, 훾푒푚 is the enhanced and intrinsic fluorescence rate, 훾푒푥푐 is the excitation rate at the excitation wavelength of 532 nm, and 훾푟 is the radiative decay rate of the emitter at emission wavelength of 640 nm.

Since the emitter is excited optically, the excitation rate enhancement is then proportional to the ratio of the squared electric field of the emitter with the optical cavity and without the cavity:

0 2 2 훾푒푥푐⁄훾푒푥푐 = |퐸| ⁄|퐸0|

It is tempting to consider the peak field enhancement of a dimer antenna as the anticipated enhancement of the emission. This is however not an accurate interpretation of the actual experiment; because the spot size of our pump laser is significantly larger than the physical area of the antennae, the excitation and hence the photon generation is not a local, hot-spot like effect, but rather originates from an average across the pump beam radius. In order to obtain a) an accurate emission enhancement originating from a ~0.8 µm large pump diameter, and b) a complete picture of the nature of the electric field enhancement distribution in the presence of either monomer or dimer antennae, we calculate the spatial distribution of the electromagnetic field profile at the location of the monolayer WS2 and, for comparison, at the cross section of the optical antenna for both the excitation and

66 emission wavelengths (Figure 4.5) (see methods). The dimer metallic nanoparticles separated by a small gap (4 nm in Figure 4.5b) supports hybridized plasmon resonances because of the capacitive coupling between the plasmon modes of each nanoparticle. Thus, the often-cited high peak-field enhancement (here 60x) is observed only if a few-atom small point emitter would be positioned precisely inside the gap center Figure 4.5c). Even if this would be achieved (e.g. using dye molecules), the signal could not be collected from this hot-spot only, because even the highest resolution near-field light collectors such as a NSOM averages its signal over an area of hundreds of square nanometers [94]. Since the 2-D TMD can neither be placed inside the gap, nor right underneath the metal nanoparticle (to avoid quenching), the only logical position would be to place it below or above the antenna, separated by a thin spacer (Figure 4.5a). Thus, when we measure the field enhancement at the position of a TMD flake residing at an optimized length of 8 nm beneath the metal nanoparticle, the peak field enhancement is only 3.4-fold (Figure

4.5c). We observe a similar trend for the 100 nm and 200 nm radius dimer antennae

(Figure 4.6, Figure 4.7). It is thus not plausible to simply take the peak intensity as the emission process enhancing value. One can conclude that for non-point emitters such as the 2-D WS2 flake, it becomes necessary to define an averaged excitation field enhancement factor such as by integrating the emission over either the physical area of the antenna,

|퐸|2 ∬ 2 푑푠 훾푒푥푐 |퐸0| 0 = 훾푒푥푐 푆

or, more accurately, over the pump beam area, where 푆 is either the geometrical

67 cross section of the antenna or the area of the beam. The peak intensity inside a small (4 nm) gap dimer of 60, drops to 1.5 (0.8) when averaged over the antenna

(beam) area at the unphysical mid-gap dimer position. For the same antenna, the

3.4x enhancement at the TMD plane underneath the cavity drops to 1.9 (0.9) when averaged over the antenna (beam) (Figure 4.5c). This shows that no actual excitation rate enhancement is expected for small gap dimers when TMDs are sitting at a quenching-safe distance away from the antenna. Interestingly, when the gap is increased from 4 to 25 nm (as studied here), the average enhancements for the antenna (beam) increases by 21% (53%) compared to the narrower hot-spot gap dimer case. We also note that the monomer with the same radius offers the highest antenna and beam enhancement (Figure 4.5c, bottom left corner); with the simpler fabrication of monomers over dimers, these results suggest that monomers are equally well-performing to enhance 2D material PL.

Figure 4.6: Electric field distribution profile of the emitter at the proximity of the monomer nanodisc optical antenna corresponding to two different radii of 75 nm (a, b, c, d), and 100 nm (e, f, g, h). The two left columns correspond to the z-normal cross-section of the device at the WS2 material layer. The columns on the right correspond to the x cross-section of the device. The dashed white lines represent the position of the antennas.

To obtain a comprehensive picture of the electric field enhancement distribution in the presence of the dimer (Figure 4.7) or the monomer antenna (Figure

4.6), we calculate the spatially resolved electromagnetic field profile at the location of the monolayer WS2 film (left two columns), as well as at a cross section across the optical antenna for both the pump and emission wavelength.

Figure 4.7: Electric field distribution profile of the emitter at the proximity of the dimer nanodisc optical antenna corresponding to three different radii of 75 nm (a, b, c, d), 100 nm (e, f, g, h), and 200 nm (i, j, k, l). The two left columns correspond to the z cross-section of the device at the WS2 material layer. The columns on the right correspond to the x cross-section of the device. The dashed white lines represent the position of the antennas.

A comparison of the far field (scattering efficiency of disc antenna) and near- field spectra (electric field enhancement) of individual Au nanodiscs (Figure 4.4,

Figure 4.6, Figure 4.7) shows that the far-field scattering efficiency peaks at a larger nanoparticle size than the near-field intensity enhancement. Consequently, one

70 should acknowledge that maximum scattering efficiency is not synonymous with highest near-field enhancement as claimed also in previous studies [95, 96, 97].

Also, often in optics we seek the highest possible quality factor (Q factor) for highest possible light matter interaction; but in a high-performing antenna we seek the opposite because we desire radiation losses [2, 98].

Our theoretical near-field and far-field study reveals that the observed PL enhancement of monolayer WS2 is due to localized electric field and local density of states at both excitation and emission wavelength enhancing both the excitation and emission rate (Figure 4.3-Figure 4.7). However, since the pump is Stokes-shifted from the emission an antenna could be designed to enhance either process selectively provided Q’s are sufficient for this spectral filtering to occur.

4.4 Quality Factor and Purcell Effect

During the study of Mie scattering and the electric field enhancement in plasmonic nanoantennas we observed that the quality factor of each disk nanoantenna increases with reducing radius of the cavity both in our measurement and Mie description of our cavity (Figure 4.8). We explain this result with Wheeler’s limit as the lower bound of the quality factor of the cavity:

3 휆 3 푄 > ( ) ( ) 4휋2 푎

where 푎 is the length of the antenna. Next, we provide a brief derivation of this limit starting with defining the quality factor as the ratio of stored energy in the cavity per energy lost in the cavity (radiated power in the case of an antenna):

71 energy stored 푄 = 2휋 energy lost per cycle

The stored energy in the resonator is calculated by:

퐼 2 2 1 푄 (휔) 퐸푠푡푟 = = 2 퐶 휖0푎

Where Q is the stored charged and C is the molecular Capacitance ≈ 휖0푎.

Next, we use the Larmor formula to calculate the total power radiated by non- relativistic point charge as it accelerates:

2 2 1 휔 푎 2 푃푟푎푑 = 3 퐼 6휋휖0 푐

Hence, the radiate energy per ½휋푓 cycle is given by:

2 1 휔푎 2 퐸푟푎푑 = 3 퐼 6휋휖0 푐

Using the stored energy and radiated energy per cycle we can calculate the quality factor:

3 2 3 퐸푠푡푟 6휋휖0푐 퐼 3 휆 푄 = 2휋 = 3 3 2 = 2 ( ) 퐸푟푎푑 휖0휔 푎 퐼 4휋 푎

While the quality factor, Q, of antenna is a function of impedance of an antenna structure, there is also a fundamental relationship between Q and the size of the antenna. The minimum achievable Q for an antenna of volume V is:

3 휆3 푄 > ( ) 푟푎푑 4휋2 푉

Figure 4.8: Quality factor of nanoantenna cavities reduces with increasing radius. The FWHM for spectral extinction of monomer antenna with radius values of 75, 100, and 136 nm, correspond to 132, 260 and 470 nm FWHM and 8, 3 and 2 Q factors, respectively (a,b). c) Quality factor for local surface plasmon resonance (LSPR) in gold. d) The FWHM for spectral extinction of dimer antenna with radius values of 75 and 100 correspond to 201 and 301 nm FWHM and 3 and 2 Q factors, respectively.

One consequence of Wheeler’s limit is that the bandwidth of an antenna decreases, and the quality factor increases as the size of an antenna decreases [2,

98]. We observe this effect in experimental results where quality factor of the nanoantennae increase with scaling down their radius. Often in optics we seek higher quality factors, but in antenna we seek radiation losses. Wheeler’s limit

73 shows the lowest Q factor that can be achieved in an antenna.

Figure 4.9: Photon generation rate. Spatial map of the quantum efficiency (a, c) and enhancement in the total radiative rate of Fp×QE (b, d) for a dipole emitting at a z normal plane corresponding to the position of TMD in the fabricated device. The dashed white lines represent the position of the dimer antennae.

The photon generation rate, here defined as the product between the quantum efficiency and the Purcell factor, is equal to the quantitative radiative decay rate

훾푟 enhancement ( 0) (Figure 4.8b, d): 훾푟

훾푟 훾푟 훾푠푝 퐺 = 0 = 퐹푝 × 푄퐸 = ( ) × ( 0) 훾푟 훾푠푝 훾푟

where 훾푠푝 is the spontaneous emission rate of the emitter at emission

74 wavelength of ~640 nm, 퐹푝 is the Purcell factor, and QE is the quantum efficiency defined as the portion of spontaneous emission coupled out of the cavity into the free space.

We obtain the generation rate at the emission wavelength of the monolayer WS2

(~640 nm) positioned under a 75 nm radius monomer or dimer nanocavity compared to an as-grown WS2 flake (Figure 4.9a, b) (see methods). The maximum attained generation rate for a dimer of 75 nm radius and 25 nm gap is ~39 times, and the averaged value over the area underneath the geometrical cross-section of the dimer antenna is about ~10 times the intrinsic radiative rate (Figure 4.9d). The QE of the fluorescence process is estimated by the ratio of the radiated power measured in the far-filed to the total power injected by the emitter (Figure 4.9a, c):

훾 푃 푄퐸 = 푟푎푑 = 푟푎푑 훾푟푎푑 + 훾푙표푠푠+훾푛푟 푃푟푎푑 + 푃푙표푠푠

However, the decay rate of the excitons to non-radiative channels (훾nr) such as phonons is not an EM process and is not captured in our FDTD simulation; hence, it is taken as zero. In this approximation, the average QE of the emitter sitting below the dimer cavity is ~39%.

4.5 Analytical and Numerical Method

Reflectance, in-plane and out-of-plane electric field intensities, Purcell factor, scattering, and absorption spectra were calculated. Optical constants of WS2 were obtained from Liu et al [100]. The refractive index of silica, silicon, alumina, and gold were used directly from Palik as well as Johnson and Christy [101, 102].

We used the Lumerical three-dimensional finite-difference time-domain

(FDTD) solver for all Maxwell equation calculations. For scattering effect calculations, we employed a total field scattered field (TFSF) method. The plane wave was launched normally from the top of monomer or dimer antennae. Two

(virtual) power-flow monitors (six detectors each) were positioned inside and outside the TFSF source, surrounding the antenna, to measure the absorption and scattering cross sections, in order. The power flow analysis calculates the net absorbed power and scattered power from the particle. For Purcell factor calculations, a dipole source was used (Section 4.5). In all other problems, a normal incident broadband plane wave was implemented. To ensure that scattered light does not return to the simulation region a Perfect Matching Layer (PML) was applied as boundary condition in all directions.

For Purcell factor and Quantum efficiency calculations, we sampled the plane at the level of the 2-D TMD sheet with a dipole source. We calculated the radiated power by a set of (virtual) field power monitors as one transmission box surrounding only the dipole source and another transmission box far away from the source but surrounding both the emitter and cavity. The expression deployed in this method to evaluate the Purcell factor is Fp = Emitted-Power (f) / Source-Power (f), where f is the optical frequency. The Source-Power returns the power that the dipole would radiate in a homogeneous material. The Emitted-Power is the power transmitted out of a boxed area surrounding the dipole source due to the dispersive materials used in the cavity.

76 The Purcell factor is expressed by [1]:

3 3 푄 휆푐 퐹푃 = 2 ( ) 4휋 푉푚 푛

To calculate the Purcell factor using this equation, the quality factor (푄) and mode volume (푉푚) of the cavity must be calculated first and plugged into the equation. We confirm that using this method overestimates the Purcell enhancement in case of our device, because the simulation will take the hot spot of the cavity as the maximum electric field (E) and accordingly measure the confinement of the modal fields using:

1 푉 = ∫ (퐸2 > ) 푑푉 푚 퐸2

This results in small modal volumes (V) of the order of ~10-24, or

휆 3 푉 ~0.001 ( ) due to hot spot of the cavity. Although, the resulting quality factor 푚 푛 simulation is the reasonable number of ~6, this result in unrealistic Purcell factor of

~35,000. This is clearly not the case for 2D material in our device, since the TMD is positioned separated from the hot spot of a dimer. In addition, there are different methods to define structure dependent modal volume, which may result in uncertainty in Fp values calculated using this method.

Modeling TMD materials in FDTD

For the optical absorption, electric field enhancement, and device resonance simulations, we model the 2D material as a 1 nm thin box with refractive index (n, k) obtained from Liu et al [100]. For simulation of the emission from 2D materials,

Purcell enhancement and quantum efficiency calculations, the PL emission of the

77 WS2 layers can be described by an isotropic distribution of dipoles lying in the 2D material plane.

Figure 4.10: Comparison of the in-plane and out-of-plane orientations on FDTD simulation of the Purcell factor of layered material in a dimer cavity (a) Schematic of the dimer nanoantenna, showing the orientation of the two dipole orientations swept along the plane of the representative WS2 material (b) top view of the same device; labeled marks from H1 to H9 represent the position (c) numerical values for the radiative components of the 0 Purcell factor (훾푟⁄훾푟 = 퐹푝 × 푄퐸) in the in-plane and (d) out-of-plane orientation.

It has been shown previously that in-plane orientation of exciton species are dominant in the photoluminescence from MoS2 [103]. However, in WS2 addition to neutral excitons, charged trions and biexcitons can also be excited in the presence of residual excess charge carriers. These quasi particles consist either of two electrons and one hole (A-), one electron and two holes (A+), or two excitons

(biexciton) is possible in TMDs (Section 4.1). The physical orientation of such

78 excitonic states for WS2 has not been studied. Although, in atomically thin layered

TMDs, atoms are confined in a plane, the electric field originating from charges in the 2-D crystals have both in-plane as well as out-of-plane components. Moreover, the large surface to volume ratio in 2-D materials enhances the significance of surface interactions and charging effects. Thus, the dielectric permittivity mismatch between the 2D semiconductor film, the surrounding environment, and an induced strain from capping material can intricately affect the electronic and optoelectronic properties of low dimensional materials [94]. Therefore, we model 2D material by sweeping the position of a dipole source for both in-plane and with an out-of-plane orientation over 9 positions and compare here (Figure 4.10). In the center of the dipole, the in-plane dipole orientation yields a higher radiative rate than the out-of- plane dipole. However, on average, the Purcell factor for out-of-plane dipole orientation (~7) is higher than the average Purcell factor for the in-plane dipole orientation (~4). These average values are representatives of the comparison between the two types of orientation across 9 sampling point and should not be considered an accurate estimate. In the manuscript, we show a 900-sampling average for the out-of-plane dipole and the average value of the Purcell factor in that case approaches ~10.

Quenching of photoemission

In section 4.1 we discussed that we should separate the emitter from the metallic nanoparticle by a distance to avoid the photoemission being quenched (i.e. coupling to lossy plasmonic surface waves). This has been previously reported to

79 be about ~8 to 10 nm [90, 91, 92]. Here we explore this phenomenon using an FDTD simulation by sweeping the position of a dipole emitter in the plane normal to the antenna device, demonstrating the positive effect of the spacer layer to avoid quenching of the radiative component of the Purcell factor.

Figure 4.11: Quenching of the radiative component of Purcell factor near the plasmonic antenna is demonstrated by sweeping the position of an emitter dipole in the gap of the dipole antenna. (a) Quantum efficiency of the dipole defined by the ratio of the radiated power measured in the far-filed to the total power injected by the dipole emitter (b) Nonradiative and the radiative components of the Purcell factor 0 (훾푟⁄훾푟 = 퐹푝 × 푄퐸).

DFT calculations

Density functional theory (DFT) was used to compute the Raman modes in monolayer WS2. All DFT calculations were performed with the Vienna Ab Initio

Simulation Package (VASP) [104]. The Projected-Augmented Wave (PAW) method [105, 106] and a plane-wave basis set with a kinetic energy cutoff of 350 eV were used. We treated electron exchange and correlation effects with the generalized gradient approximation (GGA) functional of Perdew, Burke, and

80 Ernzerhof (PBE) [107]. An 18181 Monkhorst-Pack [108] k-point grid was used to sample the Brillouin zone. A hexagonal unit cell with one formula unit of WS2 was used. The computational cell was chosen to be 16 Å along the interlayer direction minimizing interlayer interactions.

4.6 Device Fabrication

2D Material Growth method

Single- and Multi-layer Tungsten Disulfide (WS2) was grown via ambient- pressure chemical vapor deposition (CVD) utilizing a tube furnace. The process is a variation of our previously published work on transition metal dichalcogenide materials [68, 109]. The reagents were ammonia meta tungstate (AMT) and elemental sulfur. The process starts by spin-coating a 3.1mmol aqueous solution of

AMT onto a SiO2/Si substrate. The resultant residue serves as the tungsten source for our CVD growth. We place a target substrate directly face-down onto this source substrate and insert this stack into the growth region at the center of the process tube in our tube furnace. An alumina boat containing elemental sulfur is placed inside the process tube far enough upstream of the furnace heating coil so that it just fully melts when the furnace center reaches the peak growth temperature of 850°C.

The temperature ramp for WS2 growth commences after a nitrogen purge of the process tube for 15 minutes at 5 SCFH. After that we ramp the furnace to 500°C so as to decompose the AMT into tungsten trioxide releasing water and ammonia vapor. After 20 minutes the reaction is complete, and we ramp up to the growth temperature of 850°C and remain there for 15 minutes. Subsequently, the furnace

81 can cool naturally to 200°C before the process tube is opened to air and the target substrate retrieved.

Device Integration Methods

After WS2 films were deposited, a ~8 nm of aluminum oxide (Al2O3) was deposited onto the sample as a spacer between the emitter and the plasmonic antennae. Next, we fabricated monomer and dimer optical disk antennae, varying the diameters and gaps, onto the large area WS2 film using electron beam lithography. We then followed with gold deposition using electron beam evaporation. Finally, the gold film is lifted-off using PG remover. Lithographical fabrication of the antennae allows for control over dimensions and position of particles, permitting us to tune the plasmonic resonance of each antenna element.

The nanoantenna were arranged at a center-to-center distance of 4 µm to prevent interference of absorption cross section of individual antennae (Figure 4.13).

Figure 4.12: Fabrication steps of optical antenna-WS2 emitter

The spectral response of nanoantennas strongly depend on their dimensions.

Hence, fabrication of nanoantennas requires precise and reproducible structuring techniques. We fabricate monomer and dimer disk optical antennas with various diameters and gaps onto the large area WS2 film by conventional electron beam lithography followed by gold deposition with an electron beam evaporator, and finally a lift-off process in PG remover. To increase the stability of the fabricated nanostructures during lift-off, a thin layer of chromium (typically < 2 nm) is used as an adhesion layer. Only larger patches of metal survive lift-off without such precautions.

83 The nanoantennas were prepared and arranged in a square array with the center to center distance between the antennas designed to be as large as 4 µm to prevent interference of absorbing cross section of individual antennas.

Before fabricating the antenna, we coat the monolayer WS2 with a thin layer of

Al2O3. Although Atomic Layer Deposition (ALD) is the most precise method for a high-quality dielectric spacer layer, it cannot be initiated on a single layer TMD material. Selectivity of absorption based ALD on multilayer TMD vs monolayer

TMD was previously attributed to significant change in electronic structure when going from multilayer structure to a single layer [20]. Also, we note that general sputtering method gives a smoother deposition profile than e-beam evaporation.

However, we observe on several trials that sputtering on atomically thin material creates defects on the material and in some cases, entirely remove the TMD monolayer. We find e-beam evaporation to be the most benign and ‘gentle’ deposition method for capping layer on monolayer TMDs. However, e-beam deposited oxides are typically of medium quality, but natively grown oxides are typically of high quality (e.g. wet oxidation of silicon for gate oxides). We follow such reasoning here by depositing a thin (~1 nm) layer of aluminum (Al) followed by oxidation to alumina (Al2O3) under ambient conditions resulting in ~ 2 nm of

Al2O3. We iterate the process 4 times to achieve 8 nm of Al2O3 thin film deposition.

The thickness of the film is monitored through quartz crystals during the deposition and estimated by an ellipsometry after the oxidation process.

Figure 4.13: (a) Design of one unit cell of antenna array. Antennas are isolated by 4 micron spacing between them to avoid the effect of individual antennas on their neighbor antennas (b) Optical image the fabricated device (c) SEM image of fabricated optical antenna cavities on monolayer WS2 with their assigned cavity number. Scale Bar is 5 microns. (d) cross section view of a monomer antenna.

4.7 Measurement

Raman, Photoluminescence, and Reflectance Spectroscopy

Raman and PL spectroscopy were performed using custom-built spectrometer equipped with a 532-nm excitation laser and a CCD detector (Figure 4.15). The nonlinear emission spectra were acquired in reflection geometry to a spectrometer including a 532-nm notch filter to reject the pump wavelengths. The spectrometer acquisition parameters were held constant and set to ensure high signal-to-noise

85 ratio for the weakest signal. All measurements were performed in air at ∼25°C and atmospheric pressure.

Figure 4.14: Schematic of our custom-built micro PL setup capable of time resolved photoluminescence measurement. The setup is equipped with continuous wave and pulsed laser and photomultiplier tube (PMT) connected to a time digitizer for lifetime measurements.

To study the lifetime of 2D material emitters and exploring observation of reduction in radiative relaxation lifetime, we used a custom-built micro PL setup capable of time resolved photoluminescence measurement (Figure 4.14). The setup is equipped with continuous wave and pulsed laser and photomultiplier tube (PMT) connected to a time digitizer for lifetime measurements. Our results showed that the lifetime of WS2 is in the order of 700 ps. However, for lifetime measurements of antenna integrated device we were limited by the instrument response at 300 ps.

Figure 4.15: Image of custom-built integrated microscope capable of Raman, reflection and photoluminescence measurement.

Reflectance measurements were performed using a 20W halogen light source by measuring the difference in reflected intensity from the WS2/SiO2 and bare silicon wafer substrate and normalizing this to the substrate-reflected intensity. For optical microscopy, we used white light and a 100x objective lens. Absorption and transmission measurements were performed using a commercial photo-spectrometer.

4.8 Experimental Results and Discussion

87 The PL emission spectrum of monolayer WS2 flake is dominated by the A exciton peak at 633 nm (~1.96 eV) wavelength. After deposition of an Al2O3 spacer layer, the PL emission spectrum of monolayer WS2 slightly redshifts from the pristine value of 633 nm

(~1.96 eV) to 637 nm (~1.95 eV) (Figure 4.16). We attribute this redshift to a combination of strain imparted by the electron beam evaporation of Al2O3 and weak electronic interaction with the Al2O3 film. Following the deposition of the antennae, we find an enhancement of the PL emission of monolayer WS2 flakes. The 75 nm dimer cavity exhibits the largest increase by a factor of 3.2 (2.7) in peak intensity (in integrated PL count) relative to emission of the reference sample (bare monolayer WS2 flake) at the same excitation power density (Figure 4.16). In contrast, monomer and dimer antennae with a radius of 200 nm do not substantially alter the PL peak or intensity of monolayer WS2. We note that some fluctuations for the enhancement values and spectral response shifts can be expected, because the optical properties of the WS2 monolayer are strongly influenced by the nanoantenna surface plasmon that can alter the effective pumping of WS2, generation rate of electron hole pairs, and quantum yield of the emitter system, discussed below.

Figure 4.16: Collected photoluminescence (PL) emission from monolayer WS2 is enhanced when it is placed under a plasmonic monomer or dimer antenna cavity with a resonance close to that of the emission wavelength. Electron beam lithography (EBL) has been used for fine-tuning of the optical antenna dimensions on the scale of 10s of nanometers (see methods). a) PL intensity of CVD-grown monolayer WS2 before and after fabrication of 4 different optical antennae. Insets are SEM images of each type of optical cavity; the scale bar is 400 nm. (b) Full Width Half Maxima (FWHM) and enhancement of integrated PL emission from the as-grown sample and for each optical antenna. c) Dark-field optical image of the 637 nm PL emission from a monolayer WS2 under an optical excitation of 532 nm. Fitting the beam intensity profile shows that 90% of the Gaussian beam power is within 800 nm beam spot size. White scale bar = 5 µm.

To understand the PL enhancement that we observe in our experiments, we systematically simulate excitation rate enhancement and radiative rate enhancement factors for each type of monomer and dimer antenna. We report the peak enhancement values, averaged enhancement over geometrical size of the antenna, and the averaged enhancement effect over the excitation beam. These values are summarized in Table 2.1.

Based on these numerical estimate, we expect PL enhancement over the beam spot

(geometrical cross section) for the 75 nm radius monomer and dimer antenna-cavities as

2.1 (30.0) and 3.8 (23.0) fold respectively. Note the ~10x reduction as a consequence of the mismatch between beam and optimized antenna size.

89 Table 4.1: Summary of simulation and experimental data result.

To compare our measurements to the modeling, we fit the experimental beam intensity profile to a Gaussian function and take the FWHM as the beam spot size (Figure.

4.16c). These experimental results corroborate our computational expectations but show slightly lower values. Another common way of expressing experimental results is to define the normalized experimental enhancement factor (< 퐸퐹푒푥푝 >) as [83]:

퐼푐푎푣 푆푐푎푣 < 퐸퐹푒푥푝 ≥ ⁄ 퐼0 푆0 where 푆푐푎푣 is the area of a cavity, 푆0 is the area of the Gaussian excitation beam, 퐼푐푎푣 is the

PL intensity from the cavity, and 퐼0 is the PL intensity from the as-grown WS2 flake; resulting in an < 퐸퐹푒푥푝 > of 50.9 and 39.7-fold. This evaluation leads to an overestimation of the local enhancement within the cavity from our experimental result, because per our simulation the electric field intensity and Purcell factor distribution directly under nanoantenna are significantly larger than the area around the antenna 푆푐푎푣. Consequently,

90 our findings point to the need for an evaluation of the field enhancement over a much larger area than 푆푐푎푣 in order to deduce realistic material properties from the cavity experiments.

The shape of the PL spectrum in the cavity is conceptually similar to that of the control sample. However, we observe a spectral narrowing in FWHM value from 16.7 nm to 15.7,

13.5, and 14.6 nm respectively for the cases of the 75 nm radius monomer disc antenna, the 75 nm and 100 nm radii dimer antennae, respectively. We note that these effects are not sustained for the PL spectrum of the 200-nm radius dimer antenna for which we observe a red shift by 3.5 nm (relative to the intrinsic emission spectrum of as-grown WS2 flakes), and a band broadening to 21.8 nm. We attribute this observation to convolution of the PL emission spectrum of monolayer WS2 with the fundamental resonance of the 200-nm dimer cavity (Figure 4.4). Supporting modeling indicates narrowing of the emission spectrum with the dot radius, so that for the cases of the 75 nm and 100 nm radii dimers the resonance condition affects narrowing and not broadening of the WS2 emission peak. This results in higher quality factor and sharper spectral response (Figure 4.8), in line with a prediction based on Wheeler’s limit [99]. Our experimental findings highlight that the quality factor of nanoantennae increases as the dimensions of the disc decreases.

In conclusion, we have demonstrated that optical nanoantennae can be used to control the emitting properties of monolayer TMDs [110, 111, 112]. This control was achieved using two types of metallic cavities (monomer vs dimer) at four different sizes.

These emission dynamics were also supported by numerical calculations. In particular, we have demonstrated the fluorescent enhancement of 2-D materials, and unlike quantum dots there is an areal average effect that has to be taken into account. We have also observed band narrowing of the emission response when resonance of the cavity corresponds to

91 emission wavelength of monolayer WS2. Both monomer and dimer nanoantenna architecture are scalable to emission resonances of other members of the TMD family as well including MoTe2 which emits at a telecommunication wavelength in the near infrared.

The demonstrated nano-antenna controlled emission from a monolayer WS2 flake could open a pathway to visible light sources based on lithographically fabricated nano-antennae supporting a variety of opto-electronic applications.

92 Chapter 5 Diffraction Grating for Light Path Engineering

The innovation is a novel coating able to filter and direct electromagnetic radiation over the coated object. We optimize this structure to deliver EM waves that match the band gap of a single junction solar cell while minimizing inefficiencies from solar cell heat up. Thus, the coating acts as a spectral filter and enables relative solar cell cooling leading to high conversion efficiency and increased lifetime. Both increase the economic value of the unit cell. We aim to demonstrate a prototype a Smart Power

Window design (Figure 5.1c). Aforementioned coating comprises a high contrast grating (HCG) at the front of the window which filters the incoming broadband sunlight, while the window glass itself acts as synergistic waveguide. As a result, the power-converting solar cell is only required at the edge of the glass, hence reducing the area thus saving cost. The HCG effectively works as a spectral filter with the added advantage of being a solar concentrator. Such a smart window simultaneously allows for selecting visible light to pass through the window thus reducing lighting needs inside a building, while keeping it cooler (blocks the heating IR frequencies) and directs the near IR radiation to the band-gap optimized solar cell integrated in the window. This enables lower heating of the solar cell (relative cooling) compared to the full broadband exposure, which increases the conversion efficiency and cell lifetime.

5.1 Introduction

A sky scraper covered with glass can trap solar radiation and impose additional heat loads to the cooling systems. This condition is particularly relevant for tropical countries commonly known as Urban Heat Islanding (UHI) effect [113]. UHI causes the temperature

93 rise of an urban area up to 30% compared to the surrounding rural areas [114]. This can adversely affect the population by increased power demands and water shortages. Urban building designs and materials play a major role in UHI [115]. This in-turn increases the energy consumption of the building. Large amount of glass surface area used in skyscrapers shows a great capacity for a rising technology that generates electricity from solar radiation. However, traditional solar panels block or reflect the incoming sunlight. This requires an optical design that is transparent in the visible regime and coverts the invisible solar radiation to energy. Various concentrator technology such as luminescent solar concentrators [116], holographic concentrators [117], micro-optic concentrator [118]. are being actively pursued.

Figure 5.1: The smart window platform a) the spectral filtering component allowing visible portion of light to transmit through and guide the NIR portion of the light. b) the photovoltaic solar cell with a bandgap matching the guided spectrum, c) the integration of high contrast diffraction grating optical component with photovoltaic electricity generating component.

Planar holographic concentrator is a potential technology for improving the power

94 generation of solar cells by 25%. However, this technique lack wavelength selectivity and will also block the visible light. Luminescent solar concentrator uses organic dyes for up/down conversion of the solar radiation to the desired wavelength. In their device geometry, luminescent solar concentrators have the solar cells at the edges and a mirror at the bottom of the unit which blocks the through light transmission, making it undesirable for building integrated photovoltaics. One of the major challenges is the degradation of dyes and lower conversion efficiencies. Planar micro optic concentrator is another class of planar concentrators in which micro lenses and diffusers are the main components for concentrating the radiation onto the solar cells. This technology can achieve a higher concentration than any of its counter technologies however, local heating, bulkiness and inability to filter undesired wavelengths are the limitations. Here, we aim to design a smart window which transmits the visible light but harnesses the NIR radiation in a solar spectrum to convert it to electricity.

High Contrast Grating (HCG) structures are diffraction gratings with a higher refractive index compared to the surroundings environment and substrate. Various applications can benefit from HCGs, such as broadband reflectors, high Q resonators, and optical sensors. We can optimize the grating parameters such as material, pitch, thickness and duty cycle for different spectral response and applications. This proposal focuses on the novel applications of High Contrast Gratings for Smart Power Windows for Building

Integrated Photovoltaics (Figure 5.1).

Smart Power Windows (SPW) using HCGs made of high refractive index material such as silicon nitride, titanium dioxide, or zinc oxide have great potential to reduce heat load, generate electrical power and provide natural lighting, thus contributing to both

95 energy conservation and generation power. SPW works by transmitting the visible light but diffracts the NIR light into the glass substrate using engineered gratings. The diffracted light reaches high concentration at the edges which can be converted to electricity using a

Si or other photovoltaic cells. SPW generates energy with a lesser area of photovoltaic cell as it concentrates radiation. Silicon solar cells are the most commercialized and cost effective with a bandgap of 1.1 eV. Shockley – Queisser limit of solar cells can be overcome if the incident radiation is near the bandgap resulting in electrical conversion efficiencies as high as 60%. In a SPW, radiation in the range of 800-1100 nm is harnessed to generate a power of 20-40 W/m2.

One of the main challenges of high diffraction gratings is the spectral selectivity of the structure. Here we explore a novel HCG structure made of silicon nitride on glass.

Silicon nitride is high index, high performance material that shows low loss and low sensitivity to temperature variations and is widely used in the state of the art CMOS foundries. Apart from the microelectronics industry, Silicon Nitride is also widely used in the automotive industry, mainly for engine parts, which has prompted a significant and continuing cost reduction since the 1990s. Moreover, it can guide light in both visible and near infrared regime, covering the solar spectrum. This concentrator is wavelength selective and tunable to match the bandgap of the absorbing solar cell material. HCG can uniformly distribute the diffracted light over the receiver area without forming a hotspot which improves the life span and efficiency of solar cells.

5.2 Simulation and Design

The high contrast grating coated window consists of a transparent glass substrate with high refractive index coating (silicon nitride in this case) followed by a grating of the

96 coating material (Figure 5.1). The grating is essentially a Bragg grating optimize to diffract the NIR portion of the solar spectrum into the glass window. Therefore, the basic operation of the grating structure is defined by the Bragg condition:

휆 푛 = 푛 sin 휃 + 푚 푒푓푓 푡표푝 퐶 Λ

푛푒푓푓: effective refractive index of grating

푛푡표푝: refractive index of the material on top of the grating (air in this case)

휃퐶: coupling angle; the angle between the incident and the normal vectors

휆: wavelength of the incident light

Λ: the grating period

푚: diffraction mode number

Bragg condition doesn’t give any insight regarding the coupling efficiency of the grating structure. Although there are many equations developed to provide an estimate of coupling efficiencies, none of them is inclusive for all the structural parameters associated with a grating and it is necessary to use computer simulations for an accurate number. Since this is a two-dimensional problem, it is not a very memory intensive problem to solve.

We use numerical methods to optimize the HCG component of the integrated system. Initial device optimization uses Rigorous Coupled Wave Analysis (RCWA) and the final optimization is carried out using Finite Difference Time Domain (FDTD) method.

The device structure is optimized by sweeping parameters such as grating thickness, grating period, grating width and the high refractive index layer thickness (Figure 5.2). We use RCWA at the first step to understand the diffraction efficiencies of various material combination and a preliminary tool for choosing material. Based on the simulation results various material combinations on glass substrate is optimized to yield a better efficiency

97 such as titanium dioxide, silicon nitride, silicon and zinc oxide. RCWA is an approximation method and considers the device as an infinite structure. However, the device is a finite structure in reality. Thus, it is necessary to use FDTD method to simulate a scaled model of the device to obtain accurate guiding efficiencies of the device. We also use FDTD simulation to confirm the guiding at the edge of the substrate. The optimization process initially aims at maximizing the guiding efficiency at both the edge of substrate where we place the photovoltaic cell. We further optimize device parameters to maximize the acceptance angle without affecting the guiding efficiency. From the simulation results

(Figure 5.2), it can be inferred that a shallow etched grating thickness of 250 nm of the total

400 nm Si3N4 thickness with a duty cycle of 0.3 shows the highest guiding efficiency

(ng=%37). Moreover, the guided wavelength and bandwidth can be tuned further to device specification or to match any solar cell material bandgap. Such spectral tuning is achieved by varying the grating parameters.

The Visible Light Transmission efficiency (ηVLT) and guiding efficiency (ηg) are the two key metrics for the performance evaluation of PSC:

700 ∫ 퐼(휆)푇(휆)d휆 휂 = 400 x 100% 700 ( ) 푉퐿푇 ∫400 퐼 휆 d휆

1100 ∫ 퐼푔(휆)d휆 700 x 100% 휂푔 = 1100 ∫700 퐼(휆)d휆 I(λ): Incident spectra Ig(λ): Guided spectrum T(λ): Spectral transmittance

A shallow etched grating is explored to improve the guiding efficiency angular tolerance. R. Magnuson shows that the broadband reflection can be achieved with a zero- contrast grating by elimination of localized reflections and phase changes. Inspiring from

98 this a zero-contrast grating is optimized to obtain higher angular tolerant structure and the simulation results are shown.

Figure 5.2: Optimization of grating parameters to maximize guiding efficiency using full wave analysis a) map of guiding efficiencies for the full duty cycle range and thickness values of 100 to 500 nm. The color bar is guiding percentage. b) the guiding efficiency, transmission, and Reflection for the optimized 40% duty cycle and 250 nm etch depth. The guiding efficiency is 37% from 700 nm to 1100 nm. c) the angular tolerance of two set of optimized grating structures for different incident angles. For power conversion efficiencies, The solar cell fill factor is assumed to be 0.7 with a power conversion efficiency of 50% in 700 to 1100 nm wavelength d) The portion of solar spectrum available for harvesting in the smart window concept.

5.3 Fabrication and Testing

We developed two processes using stepper lithography and large area electron beam lithography (EBL) to fabricated Si3N4 HCG component of the integrated SPW system. The glass substrate after prior cleaning is coated with 400 nm of silicon nitride using PECVD.

We monitor the refractive index of the deposited Si3N4 using ellipsometry (Supplementary

99 Information). We also verify the surface roughness of the deposition using atomic force microscope (AFM).

EBL has the advantage of resolution and flexibility in choice of substrate. Using electron beam lithography an area of 7 mm x 7 mm grating is fabricated on a 2 mm thick glass substrate coated with silicon nitride. The C4 electron beam resist is used and which is spin coated at an rpm of 4000 for 45 seconds followed by baking it at 1200 C for 2 minutes. The E-beam lithography uses an Electron High Tension of 50 keV with an aperture of 30 µm and 1nA current. However, the PMMA photoresist is sensitive in the

Si3N4 Dry etch process. Therefore, instead of using the resist itself as a mask we initially deposit a 30 nm thin Cr layer to be used as an etch mask.

The process fabrication with a stepper tool has the advantage of speed and throughput as it uses only 20 seconds for each exposure and process the whole area in parallel (vs beam rastering in EBL). Since stepper limits the sample dimensions, it must be a 4-inch substrate with a standard flat cut and thickness less than 1 mm. The substrate after PECVD deposition is spin coated with anti-reflection coating and photoresist and pre-baked. Then the sample is loaded for patterning into the stepper lithography tool. After the desired exposure, the sample is developed using MF-26A for 1 minute followed by DI water rinse and dried using a nitrogen gun. The gratings were etched using RIE using the photoresist as the mask.

100

Figure 5.3: High contrast grating prototype a) the optimized mask design for guiding the NIR portion of the solar spectrum, b) top-down optical microscope image, c) cross- sectional AFM, and d) top-down SEM image of the fabricated HCG device

Optical Characterization:

The optical measurements are carried out to obtain transmission, reflection and guiding efficiency. In-house built characterization setup is used as shown in fig. (include transmission reflection and guiding efficiency setup). The results are shown for ebeam sample is shown in figure. The setup uses a halogen lamp from Ocean optics HL 2000 followed by collimating optics with chromatic aberration correction. The guided light is collected using an integrating sphere and fed to the spectrometer.

101

Figure 5.4: High contrast grating optical performance measurement a) schematic of the setup, b) image of the measurement setup. For measuring guiding efficiency the sample is rotated 180 degrees and is remeasured.

As demonstrated in both experimental and simulated results, 2% guiding efficiency can be achieved.

102

Figure 5.5 Spectral guiding efficiency and power conversion efficiency a) measured guiding efficiency, b) Integrated SPW power conversion performance for e beam fabricated grating and silicon solar cell at the edge of HCG.

5.4 Conclusion

We demonstrated visually transparent planar concentrator that have unique potential in generating electricity from building where conventional solar cells are impractical. The proposed dielectric, HCG on the glass is the core of this system, enabling splitting of the incoming spectrum allowing spectral filtering to match the bandgap of the respective PV materials. As such, this PSC system features societal benefits, and its advantages are well

103 suited to become an integral part of future on-site electricity generation, heating and cooling infrastructure for modern buildings. We used lithography to demonstrate the prototype system. The fabricated device shows a conversion efficiency of 2.86% Such device in principle can be fabricated on large area using cost-effective techniques such as

Interference Lithography (IL) and Nano-Imprint Lithography (NIL).

104 Chapter 6 ITO Ring Resonators for Electrooptic Modulation

6.1 Introduction

Photonic Integrated Circuits (PICs) is an emerging technology that uses semiconductor wafers to integrate passive and active photonic components along with electronic circuitry on a single chip. PICs have potential for disruptive impact on the market for on-chip data communication, telecommunication, real-time sensing for health diagnostics, and neuromorphic computing. The global development of data centers and high-performance computing demands optics enabled data transmission [119]. Compact and CMOS compatible electro-optic modulators are desired for integration of optical interconnects and photonic circuitry with the electronics. Therefore, sub wavelength,

CMOS compatible photonics have been a topic of significant research [5, 6].

Silicon is a popular material of choice for integrated photonic technologies because: a) it benefits from compatibility with the mature CMOS technology, b) silicon is transparent at telecom wavelength of 1310 and 1550 nm, c) silicon core has a high refractive index of neff ≈ 3.45 compared to silicon oxide cladding with a refractive index of neff ≈ 1.45 offering compact and densely integrated optics devices. Strip silicon waveguide thickness is 220 nm on top of a 2 µm buried silicon oxide on silicon substrate (Figure 6.1).

The 220 nm thickness of silicon waveguide is to support a single mode signal in a 1550 nm operating frequency.

However, Silicon shows a weak electro-optic effect; therefore, classical silicon based electro-optical modulators (EOM) are bulky due to large scale of an optical photon and weak electro-optic effect in silicon. Mach-Zehnder modulators (MZMs) and ring phase

105 shifting modulators based on plasma dispersion of silicon have been demonstrated and has significantly improved modulator performance metrics [6, 120]. Compact electro-optic modulators reduce the device foot print and capacitance, to increase their speed and power efficiency. To improve modulation speed and power consumption of EOMs we need to minimize their capacitance. There are two solution creating high performance EOMs:

• Other CMOS process compatible materials with stronger electro-optic effect.

• Using optical cavities to localize light in order to reduce the device footprint.

Here we combine both solutions and combine ITO which is a CMOS compatible material with significant change in their optical constants by carrier injection [5, 121] with a micro ring resonator cavity to localize light and enhance light matter interaction [122, 123, 124,

125, 126, 127] (Figure 6.1).

106

Figure 6.1: Critical coupling effect in Reservoir Coupling Electro-Optic Modulator. a) Schematic concept, b) conceptual cross section of the device region forming a two-layer ITO capacitor. Two layers of ITO are separated by a 17 nm interlayer dielectric to form a parallel-plate capacitor, c) Optical image showing the complete device with bus waveguide, ring resonator and two Ti/Au contact pads for each device. Scale bar is 40 µm, d) Schematic of cross section image of the coupling region between the ring and the bus where the light coupling occurs.

In principle there are two types of electro-optic modulation effect on a propagating mode in photonic waveguides:

• Electro optic phase shift: change of the real part of the refractive index

• Electro absorption: change of the imaginary part of the refractive index

The basic idea for electro optic phase shifters is that the real portion of the refractive index of electro-optic materials can be significantly changed by applying an electric field.

In this case, the phase of the propagating mode shifts. Using the optical interference or

107 optical resonance effects, such as in Mach-Zehnder interferometer or micro ring resonators, the phase shift can lead to significant modulation extinction ratios. If the active modulator device is in a ring cavity system, the resonance of the ring resonator shifts and the transmission of the operating bandwidth of the device changes from peak of the resonance to the tail of the spectral response (Figure 6.6).

On the other hand, by significantly changing of the imaginary part of the refractive index, active material enters a lossy state causing the propagating light to be absorbed by the modulator device. There are other types of electro absorption modulation such as Franz-

Keldysh effects or quantum confined Stark effect. This type of modulation mechanism seldom relies on ring resonators or MZI structures, however they could benefit from hybrid plasmonic cavities. If the active device is in a ring cavity system, this causes the quality factor of the resonance to reduce.

Due to Kramers-Kronig relations the real part of optical index of a material relates to its imaginary part:

푐 ∞ 훼(Ω)푑Ω 푛(휔) − 1 = 푃 ∫ 2 2 휋 0 Ω − 휔

Ω: is angular frequency variable running through the whole integration range

푃: Cauchy principal value

4휋휅 훼 = : absorption coefficient 휆 n: refractive index [unitless number]

Therefore, due to Kramers-Kronig relation, practically in every device a large change in real part of optical index is also accompanied by a significant extinction coefficient

(imaginary part of optical index). We study both of these effects independently and

108 together on the modulation depth and bandwidth.

Here, we report on an electrically driven, CMOS compatible, ring resonator coupling modulation mechanism based on tuning of free carriers in Indium Tin Oxide

(ITO). The modulator device consists of two ITO layers separated by a thin oxide layer fabricated on the coupling region of silicon ring resonator.

6.2 Optical Index Tunability of ITO

The aim of an electrooptic modulator is to alter the effective modal index of a propagating optical mode. In this section, we discuss the change of ITO’s refractive index by changing the carrier concentration via a Drude-Lorentz model.

It has been shown that the Drude model accurately predicts the permittivity of ITO at the wavelengths beyond 1 µm [38]. We characterize the corresponding complex refractive index variation of ITO from electrical tuning by the Drude-Lorentz model. The permittivity is given by:

2 휔푝 휀(̅ 휔) = 휀 + 𝑖휀̃ = 휀 − 푟 푟 ∞ 휔(휔 + 𝑖훾)

where 휀∞ is the long-angular-momentum-limit permittivities or background permittivity,

ω is the angular frequency of the illuminating light, 훾 = 1/휏 is the carrier scattering rate(i.e. collision frequency), and 휔푝 is the unscreened plasma angular frequency defined by:

2 2 푛푐푒 휔푝 = ∗ 휀0푚

Here 푛푐 is the carrier concentration, q the electronic charge, 휀0 the permittivity of vacuum, and 푚∗ is the conductivity effective carrier mass. The electron mobility 휇 and 휏 are related

109 by 휇 = |푞|휏/푚∗. the effective free electron mass , 푚∗, can be determined from fitting the electronic band structure to a parabolic function [128]. The conductivity effective mass is taken as 0.35푚0, where 푚0 is the free electron mass [129]. As the drive voltage is increased, free carriers leading to dispersive effects are created and a corresponding net increase in the carrier concentration occurs. Several previous studies have calculated the permittivity of ITO using the experimentally measured reflectance and transmittance, and here a fitting result of Michelotti et. al is choosen whereas 휖∞ and γ depend on the

14 - deposition conditions. In our analysis we have taken 휀∞ = 3.9 and γ = 1.8×10 rad s

1 ,respectively. The plasma frequency, 휔푝 is determined by the carrier density 푁푐, which is equal to the concentration of Sn atoms, and thus can be within the range of 1019–1021 cm-3 depending on the deposition conditions, defect states, and film thicknesses [130]. If the carrier concentration of ITO can be changed from ~1019 to ~1021 cm−3, the index changes dramatically.

ITO and related transparent conducting oxides (TCOs) have recently been explored as plasmonic materials in the optical frequency range [130, 131, 132, 133]. The carrier concentration levels, effective indices and extinction coefficients corresponding to the states of operation are listed in Table 6.1.

Table 6.1: ITO carrier concentration levels, effective indices and extinction coefficients corresponding to the states of operation for the tunable cavity.

ITO Carrier Effective index, 푛̃ States of 푒푓푓 Concentration Real part, operation -3 Imaginary part, 휅푒푓푓 (cm ) 푛푒푓푓 ON 1019 2.26989 2.14288×10-4 OFF 1020 2.25264 0.00253

Similar to Silicon, ITO is able to tune its index via the carrier-dependent Drude model.

110 Yet ITO exhibits a number of significant advantages over Si with respect to the index tuning. Firstly, the carrier concentration in ITO can exceed that of in Si by at least an order of magnitude; the concentration of indium atoms in ITO reaches a few percent, which is significantly above attainable donor concentration in Si. Secondly, the effect of changing the carrier concentration in ITO on its refractive index is more dramatic than in Silicon.

This can be attributed to the higher bandgap and consequently lower refractive index of

ITO compared to that of Silicon. If the change of the carrier concentration δNc (e.g. due to an applied bias) causes a change in the relative permittivity (dielectric constant) δϵ, the corresponding change in the refractive index can be written as δn = δϵ1/2 ~ δϵ/2ϵ1/2, hence the refractive index change is greatly enhanced when the permittivity ϵ is small.

HfO2 is chosen as high-k dielectric material because of its high static permittivity

(휖퐻푓푂2~25) [134]. HfO2 thin films can be reliably grown using atomic layer deposition

(ALD) and can allow inducing high electric fields with relatively low biases [135].

6.3 Theoretical Study of Optical Ring Resonators

Optical ring resonators are promising building blocks for photonic integrated circuits. They can be small, highly efficient, low loss, high speed, and CMOS compatible at the expense of being narrowband [6, 127, 136, 137]. The modulation energy can be only a few fJ/bit for the modulator itself, and <100 fJ/bit including the driver circuit. The downside is that they are very sensitive to manufacturing errors, temperature drift, or laser wavelength change, thus they typically need dynamic and precise tuning, which may potentially require a control circuit and consume additional power.

111 An optical ring resonator consists of straight waveguides coupled to a ring resonator

(Figure 6.2a). When the waveguides are at proximity, the propagating wave in one of the waveguides can couple (evanescent coupling) into the other waveguide. The roundtrip length of the ring waveguide is 퐿푟푡 = 2휋푟푏 + 2퐿퐶, where 푟푏 is the bending radius and 퐿퐶 is the coupling length. The coupling region can be a single point (퐿퐶 = 0, point-coupled resonator), or a straight line elongated the bus (racetrack resonator). Ring resonators with one waveguide bus are called “all-pass” and rings with two waveguide buses are called add-drop rings. Here, we will focus on point-coupled add-drop rings.

The spectral response of ring resonators is the key performance metrics of the passive resonator platform and can be easily derived under continuous wave operation, matching field, negligible back-reflection assumptions:

2 2 2 퐼푝푎푠푠 푟2 푎 − 2푟1푟2푎 cos 휙 + 푟1 푇푝푎푠푠 = = 2 퐼𝑖푛푝푢푡 1 − 2푟1푟2a cos 휙 + (푟1푟2푎)

2 2 퐼푑푟표푝 (1 − 푟1 )(1 − 푟2 )푎 푇푑푟표푝 = = 2 퐼𝑖푛푝푢푡 1 − 2푟1푟2푎 cos 휙 + (푟1푟2푎)

Where 푇푝푎푠푠 ( 푇푑푟표푝) are the ratio of transmission intensity in the pass port (drop port) to the intensity of the input signal, 푟1 and 푟2 (푘1 and 푘2) are self-coupling (cross- coupling) coefficients in the first and second coupling region, 푎 = √푒−훼퐿 is the single pass amplitude transmission which includes both propagation loss in ring and loss in coupler (훼 is the power attenuation constant [1/cm]), 휙 = 훽퐿푟푡 is the roundtrip phase shift (훽 is the propagation constant). The coupling coefficients 푟2 and 푘2 are power splitting ratios and satisfy 푟2 + 푘2 = 1 if the coupling is lossless.

112

Figure 6.2: Ring resonator a) A point-coupled add-drop ring resonator showing various signal ports and coupling points. b) transmission spectrum of an add-drop ring resonator for Pass and Drop ports, showing important key spectral performance metrics of the resonator.

If the propagation length around the ring (퐿푟푡) is an integral number of wavelengths, the field becomes resonant and a strong field builds up in the ring. Therefore, the resonant wavelength can be obtained by:

푛푒푓푓퐿푟푡 휆 = , 푚 = 1,2,3, … 푟푒푠 푚

After propagation around the ring waveguide, some light couples back to the straight waveguide and interferes with the incident light. At resonance, completely destructive interference can be obtained, with no transmitted light. This makes the optical ring resonator an ideal notch filter, blocking the light at the resonant wavelength. This will result in optical transmission spectrum of the bus waveguide to have notches at the ring resonances, and the resonance wavelengths can be shifted by modulating voltages.

From transmission ration equations, the quality factor of the resonances in a ring resonator can be derived as:

휆 휋푛푔퐿√푟1푟2푎 푄 = 푟푒푠 = Δ휆퐹푊퐻푀 휆푟푒푠(1 − 푟1푟2푎)

113 Δ휆퐹푊퐻푀 is the resonance full width at half maxima which is also referred to as the resonance bandwidth and can be estimated by reordering the same equation:

2 (1 − 푟1푟2푎)휆푟푒푠 Δ휆퐹푊퐻푀 = 휋푛푔퐿√푟1푟2푎

Higher quality factors improve the modulation efficiency and requires lowers energy consumption in the device at the cost of reducing the modulation bandwidth. A typical linewidth for silicon ring resonators is about 5-15 GHz. Another important parameter for resonator is the finesse, which is calculated by:

퐹푆푅 휋√푟1푟2푎 퐹𝑖푛푒푠푠푒 = = Δ휆퐹푊퐻푀 1 − 푟1푟2푎

Figure 6.3: Experimental normalized transmission function of ring resonators with different radii. (a) transmission function of rings with radii of 80, 60, and 50 microns in spectral range of 1530 to 1590 nm, corresponding to the broadband EDFA output. The dashed box is magnified in figure (b). The FSR of ring reduces and the quality factor increases by increasing the radius.

FSR is the free spectral range and is defined by the distance between two resonances in the transfer function of a ring resonator and can be approximated by:

114 휆2 퐹푆푅 = 푛푔퐿

This equation suggests that by increasing the length of the ring resonator the free

FSR of the ring would decrease. We tape-out SOI rings with 50,60, and 80 microns diameters and use a broadband source and optical spectrum analyzer to monitor the transmission function of them. These results (Figure 6.3) are normalized to 1 in order to make a fair comparison between performance of rings without considering the variation in coupling ratio variations of grating couplers. The FSR of ring reduces from 4.1 nm to 2.5 nm and the quality factor increases from 756 to 1239 by increasing the diameter from 50 microns to 80 microns (Table 6.2).

Table 6.2: the summary of experimental transfer function of ring resonator with 50, 60, and 80 micron diameter (Figure 6.3)

These results closely match with theoretical values of FSR calculated next. For simplicity we only calculate the FSR values. To estimate the theoretical FSR of rings we can use 휆 = 1550 [푛푚] and 푛푔~4.25. We can calculate the circumference of the ring

(round trip waveguide length) by 퐿 = 2휋푟 which will be equal to ~157, 188, and 251 휇m resulting in FSR values of ~ 4, 3, and 2 nm for ring diameters of 50, 60, and 80 휇m in order.

One of the biggest challenges of using high Q factor ring resonators for modulation application is their sensitivity to minor process variations and temperature. We show this by measuring the transmission function for three of the exact same 50 휇m diameter and three of the exact same 60 휇m diameter ring resonator fabricated side by side on the same

115 chip. This sensitivity in spectral response is one of the limiting factors in their application in photonics integrated circuit which usually requires active tuning to stabilize the transfer function. This is one of the main motivation for ring coupling modulation which is broadband in nature and will be independent of the sensitive and sharp resonance of rings.

Figure 6.4: Sensitivity of ring resonators to minor process variations. (a) layout of fabricated chip for three 60 휇m diameter rings and three 50 휇m diameter ring. (b) Transmission function of the same type of devices show tangible variation in spectral shift in resonance of each type of device.

116

Figure 6.5: Comparison of two types of ring resonator cavities a) interactivity cavity, b) coupling cavity.

Based on the location of an active device on a ring resonator we can categorize ring based devices to two types: a) interactivity device b) coupling device. There are three pure type of modulation mechanism in ring resonator-based modulators based on what transmission response parameter is changed (modulated):

• Loss modulation: where 푎 varies, but 푟 , 푘, and 휙 are constant.

• Index modulation: where 휙 varies in time but 푟 , 푘, and 푎 are constant

• Coupling modulation: where 푟 and 푘 vary, but 휙 and 푎 are constant.

The index modulation is the most effective mechanism for highest extinction ratio since the sharpness of the resonance remains almost unaffected. Therefore, the peak extinction ratio can be as large as the difference between the maximum transmission and the dip of transmission resonance. However, speed of the interactivity index modulation mechanism is essentially limited by the cavity photon lifetime and has a narrow bandwidth. Loss

117 modulation on the other hand, has a larger bandwidth but it suffers from low extinction ratio and is also limited by the cavity photon lifetime. Coupling modulation does not suffer from the same speed limitations as loss or index modulation [138, 139, 140]. Moreover, their bandwidth is not limited to the resonance wavelength of the cavity and can be broadband. However, coupling modulation extinction ratio becomes smaller than index modulation. This comparison between these modulation mechanisms is detailed in Figure

6.6.

Figure 6.6: Comparison of three theoretical type of modulation mechanisms for ITO ring modulator a) fully interactivity index modulation, b) fully interactivity loss modulation, c) fully coupling modulation.

To calculate the transmission function and their characteristics including extinction

118 ratio and quality factors in these two modulation mechanisms, we need to implement change of phase in interactivity modulation and coupling factors in coupling modulation.

For the case index modulation, the change in phase can be modeled by:

2휋 Δ휙 = Δ푛 퐿 휆 푒푓푓

The ring resonance is characterized by its full width at half maximum (FWHM) δλ; and the interactivity phase shift modulation is characterized by the resonance shift Δλ.

Therefore, to achieve large ER in these types of modulation mechanism, one ought to maximize Δλ/δλ.

At critical coupling of the ring (when transmission is at minimum), the resonance linewidth and its quality factor Q, can be derived by:

휆2훼 훿휆 ≈ 푛푔휋

휆 푛푔휋 푄 = ≈ 훿휆 휆훼

Where group index ng is 4.25 at 1550 nm (the geometric parameters determined in Mode solution). Similarly, the resonance shifts due to index change of Δneff is:

Δ푛푒푓푓 Δ휆 = 휆 푛푔

According to these equation, phase change in propagating mode in a waveguide

(Δ휙) will depend on change of refractive index in the electro optic modulator and its modal overlap with the propagation, and how much effect it has on change on the “effective index” of mode (Δ푛푒푓푓). The effective index can be calculated using eigen mode solvers discussed in more details in Chapter 6.4. Depending on the change in effective index the modulator length will be determined 퐿. Here we estimate the device length to acquire

119 maximum extinction ratio for index modulation Figure 6.7.

Figure 6.7: Required change of effective refractive index to obtain the desired change in transmission (extinction ratio) of 1550 nm light in a index modulation device, for various device length of 2,4, and 8 휇m.

The photon lifetime-limited modulation bandwidth cannot exceed the resonance linewidth

δλ; thus, there is a tradeoff between higher bandwidth and higher extinction ratios in phase shift ring modulators.

Here, coupling modulators can offer advantage in higher modulation bandwidth and smaller device footprint that can consume less power per bit of modulation. Moreover,

120 modulators using ITO has also been explored previously [91, 120], but ITO modulator on ring resonators have not been demonstrated to this date.

Performance Metrics

We use link metrics (optical loss, extinction ration (ER), and bandwidth) to determine data rate, SNR, and bit error rate (BER) of the ring resonator.

The optical loss of a modulator is often defined as the loss at a voltage with minimum loss.

6.4 Numerical Study of Optical Ring Resonators

The passive and active photonic component simulations used Lumrical FDTD and eigen mode solver. The FDTD simulations were full three-dimensional simulation in which PML boundary conditions were applied along all axis. The mode simulation were two dimensional simulations, in which eigenmode analysis was performed on the cross-section of the device for calculating the effective mode index.

We simulate the geometric parameters of the silicon waveguide and ring resonator using finite difference eigen mode (FDE) solvers to achieve a desired free spectral range and quality factor.

First, we find the circumference of the ring needed to obtain a desired FSR. We use

Eigenmode solver on the waveguide cross section to calculate the group index.

푑β 푛 ≈ 푐 푔 dω

We see that the group index is 4.25 at 1550 nm. Hence, the desired length of the ring should can be derived from:

121 휆2 퐹푆푅 ≈ 푛푔퐿

Next, we carried out a 3D finite-difference-time-domain (FDTD) simulation to design and optimize passive waveguides as well as the active region of the device (Fig.

2a). In order to accurately simulate the characteristics of the two 16 nm thick ITO layers in the active region, a detailed Drude model is used based on our previous work and covered by a nanometer level fine mesh. The arbitrary dispersion of refractive index of material are fitted using a multi-coefficient model. As the tuning material of the device, the carrier concentration of the ITO layers changes from 1019 cm-3 to 1020 cm-3 when the applied bias voltage swept from 0V (OFF state) to 5V (ON state). And in terms of the light source, a

1550 nm wavelength TM-mode input is injected from the left side of the bus and depending on the bias voltage of the ITO layers, the resonance of the ring will be changed which can be demonstrated by the shifting of the coupling efficiency at the active region (Fig. 2b, 2c).

Based on the simulation result, 6.2 dB extinction ratio is able to be achieved at the output side of the bus with 0.6 dB insertion loss.

122

Figure 6.8: Flowchart of simulation steps and optimization of geometrical parameters.

6.5 Device Fabrication

Initially we developed three types of passive platforms for various projects and material systems, namely, Al2O3 on MgF2 substrate, Si3N4 on silicon substrate, and silicon on insulator platform. For the purpose of ITO coupling modulator we used the SOI substrate. The utilized SOI substrates has 220 nm of silicon on top of 2 microns of silicon oxide box on top of standard silicon wafer. The passive waveguides and grating coupler were patterned using electron beam lithography (EBL). We also utilized an external multi project wafer project for passive SOI platforms.

123

Figure 6.9: The CAD layout used for exposure and alignment of different material layers. The device region at the coupling region is a sweep from 1 to 8 microns. And distance between contact pads were kept at 180 microns to match the distance between our RF probes.

To fabricate the active device, a parallel plate ITO capacitor was designed and fabricated over the SOI waveguide near the coupling region. We developed the CAD layout using K-Layout software and imported each layer to the EBL software for exposure (Figure

6.9). Before each step we cleaned the sample by soaking it in acetone, rinsing with isopropanol (IPA) and baking at 100 oC for few minutes. Initially naked passive components were coated with 10 nm of thermal atomic layer deposition (ALD) of HfO2.

We then spin coated 250 nm of A4 PMMA, baked the spin coated photoresist at 200 oC for

2 minutes before e-beam exposure.

124

Figure 6.10: Optical image of the device at each major fabrication steps. (a) The first ITO deposition before the lift-off process, (b) the first metallization, (c) the second ITO exposure and deposition before lift-off, (d) the completed device after second round of metallization. The ITO layers are 15 nm thin and optically transparent.

We used a Raith Voyager EBL system to accurately align and expose the photoresist. We then developed the exposed PMMA using MIBK/IPA solution. Next, we deposited a 15 nm layer of ITO using room temperature sputtering and performed a lift-off process in acetone to get rid of the photoresist and the ITO deposited on top of the photoresist. Although sputtering is a conformal deposition technique, we confirm that the lift-off was successful (Figure 6.10). We deposited ITO layers using an optimized sputtering recipe to reduce the extinction coefficient of the material to avoid losses and reduction of device quality factors. Next, we repeated the EBL process for metallization, this time by

125 depositing 50 nm of Ti/Au using electron beam evaporator which is a non-conformal deposition technique and ideal for lift-off process.

Figure 6.11: fabrication steps of the ITO reservoir coupling modulator (a), list of materials processed in the fabrication process (b), and a schematic of final active device (c).

Sputtering is the most widely used techniques for the deposition of TCO films.

Typically, magnetron sputtering processes are performed at high substrate temperatures

(200∘C), as these allow the best results in terms of layer transparency and conductivity to be obtained. However, several applications, for example, solar cells and devices on plastics, require a low deposition temperature, as higher temperatures would damage the underlying

126 either electronic device structures or substrate itself. This challenging task has been investigated by several workers where the RF technique was mostly adopted. The main difficulty of RF deposition at room temperature is due to insufficient recrystallization of

ITO at low temperatures leading to the poor structural and electrical properties.

ITO sputtering targets typically consist of 10% SnO2 to 90% In2O3 by weight. ITO has been widely adopted by the solar industry as a transparent electrode due to both its high optical transmittance yet high electrical conductivity. ITO has been found to be a difficult material to work with due to its processing invariances. An electron beam can be a more reproducible method for low resistivity and high transparency films. The deposition of ITO films by the CVD method generally faces difficulties due to a lack of volatile and thermally stable source materials.

In the next step, we used atomic layer deposition (ALD) to deposit 17 nm thin Al2O3 to separate the two ITO layers to form the capacitor. The thickness of deposited oxide layers was also confirmed by ellipsometry or reflectometer. Two contact pads for each ITO layer was defined by e-beam lithography followed by e-beam evaporation of Ti/Au with a thickness of 10/40 nm.

6.6 Measurement Setup

We demonstrate electro-optic tuning of optical transmission in coupling modulator.

A TM grating coupler guides the incoming fiber couple tunable telecom laser or broadband

EDFA signal into the waveguide. We use 3-paddle fiber polarization controller to select between TE and TM input polarization. The propagation mode after travelling through device is then coupled out from another TM grating coupler into a tapered lens fiber and

127 sent to an optical power meter or optical spectrum analyzer (OSA). Figure 6.4 shows a typical transmission spectrum of a ring resonator on OSA. To study the dc response of the reservoir coupling modulator, an electrical bias was applied using a Keithley Source

Measure Unit (SMU) across the lower ITO arm and the grounded ITO layer above using two electrical probes.

Figure 6.12: Development pipeline and measurement and verification setup, (a) the design development flow from physical modeling down to layout. (b) Photograph of the fiber coupled electro-optic measurement setup used. (c, d) schematic of the measurement setup.

Upon applying a voltage, the refractive index of ITO changes the self-coupling and cross coupling coefficients of the ring resonator at the coupling region and modifies its quality factor, resulting in a broadband change in transmission output. We also use the

SMU unit to read the current flowing though the device to note the breakdown voltage across the ALD Al2O3 gate oxide.

128

6.7 Results and Conclusions

We have demonstrated the first reservoir coupling ITO modulator by tuning the coupling factor of a silicon micro ring cavity using a sub wavelength ITO active capacitor.

Figure 6.13: a) FDTD simulation results for on state of the coupling region of the reservoir modulator b) Transmission optical power for various applied DC voltages on the ITO coupling modulator showing 3dB extinction ratio and 5V swing voltage.

Factors such as the device footprint, modulation depth, modulation speed, and power efficiency of EOMs are the key performance metrics that must be taken into account. The figure of merit (FOM) for an electro-optic modulator is given by:

퐸푅 퐹푂푀 = 퐿푒푛푔푡ℎ × 푉표푙푡푎푔푒

The length of the coupling modulator device is very small (~1 µm) compared to conventional modulators. This lead to a significant FOM of 0.6 [db/µm-1V-1].

129

Chapter 7 Outlook and Conclusion

Concluding this dissertation, we have shown how application of optical cavities and beam path engineering lead to sub wavelength novel optoelectronic device physics. In particular the reduction of device footprint in plasmonic cavities can enhance optical absorption/emission processes. Furthermore, we have demonstrated that sub wavelength electro-optic modulators can be fabricated on coupling region of a micro-ring resonator towards compact high-bandwidth modulators. We also demonstrated that using optical techniques we modify the beam path of solar spectrum to match photon energy with semiconductor bandgap and enable novel concepts such as smart power window.

Optoelectronic characteristics for atomically thin TMD material was studied using Raman and photoluminescence spectroscopy. A platform to tune spectral scattering of gap mode and patch mode plasmonic antennas to match optical bandgap of TMD materials was developed and verified experimentally for the case of monolayer WS2. Here we show that effect of relative position of 2D sheets to the plasmonic antenna. We found out that the emission enhancement generated in a 2D TMD film by a monomer antenna cavity rivals that of dimer cavities at a reduced lithographic effort. We rationalize this finding by showing that the emission enhancement in dimer antennae depends not on the peak of the field enhancement at the center of the cavity, but rather from the average field enhancement across a plane located beneath the optical cavity where the emitting

130 2D film is present.

A new type of optical micro cavity leveraging mechanical flexibility of 2D materials was proposed to enhance light absorption in atomically thin material to rival that of classical bulk materials. Additionally, we proposed a method to effectively extract carriers from the multilayer spiral solar cell using only two contact point in the center post and the outer shell of the cavity (core-shell).

We combined the concept of spectral filtering in high optical index grating structures with silicon solar cells to demonstrate SPW that works by transmitting the visible light while diffracts the NIR light into the glass substrate. The diffracted light reaches high concentration at the edges which can be converted to electricity using a Si or other photovoltaic cells. SPW generates energy with a lesser area of photovoltaic cell as it concentrates radiation.

Finally, we demonstrated a CMOS compatible sub wavelength electro-optic modulator for integrated photonics platform that benefits from sensitivity of the coupling region in high Q factor microring cavities. We demonstrate that although the extinction ratio in coupling modulators is limited compare to interactivity mode ring modulators, their operating bandwidth is not limited by the cavity linewidth.

131 Bibliography

[1] E. M. Purcell, H. C. Torrey, and R. V. Pound, "Resonance Absorption by Nuclear Magnetic Moments in a Solid," Phys. Rev., vol. 69, no. 37, 1946. [2] Michael S. Eggleston, Kevin Messer, Liming Zhang, Eli Yablonovitch and Ming C. W, "Optical antenna enhanced spontaneous emission," proceedings of the national academy of sciences, vol. 112, no. 6, 2015. [3] L. Britnell1, R. M. Ribeiro2,3, A. Eckmann4, R. Jalil5, B. D. Belle5, A. Mishchenko1, Y.- J. Kim1,6, R. V. Gorbachev, T. Georgiou, S. V. Morozov, A. N. Grigorenko, A. K. Geim, C. Casiraghi, A. H. Castro Neto, K. S. Novoselov, "Strong Light-Matter Interactions in Heterostructures of Atomically Thin Films," Science, vol. 340, no. 6138, 2013. [4] Marco Bernardi, Maurizia Palummo, and Jeffrey C. Grossman, "Extraordinary Sunlight Absorption and One Nanometer Thick Photovoltaics Using Two-Dimensional Monolayer Materials," Nano Letters, vol. 13, no. 8, 2013. [5] Volker J. Sorger, Norberto D. Lanzillotti-kimura , Ren-min Ma , Xiang Zhang, "Ultra- compact silicon nanophotonic modulator with broadband response," Nanophotonic, 2012. [6] Qianfan Xu, Bradley Schmidt, Sameer Pradhan & Michal Lipson, "Micrometre-scale silicon electro-optic modulator," Nature, vol. 435, 2005. [7] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, A. A. Firsov, "Electric Field Effect in Atomically Thin Carbon Films," Science, vol. 306, no. 5696, 2004. [8] Kin Fai Mak, Changgu Lee, James Hone, Jie Shan, and Tony F. Heinz, "Atomically Thin MoS2: A New Direct-Gap Semiconductor," Phys. Rev. Lett., vol. 105, no. 136805, 2010. [9] Qing Hua Wang, Kourosh Kalantar-Zadeh, Andras Kis, Jonathan N. Coleman & Michael S. Strano, "Electronics and optoelectronics of two-dimensional transition metal dichalcogenides," Nature Nanotechnology, vol. 7, no. 699, 2012. [10] Namphung Peimyoo, Jingzhi Shang, Chunxiao Cong, Xiaonan Shen, Xiangyang Wu, Edwin K. L. Yeow, and Ting Yu, "Nonblinking, Intense Two-Dimensional Light Emitter: Monolayer WS2 Triangles," ACS Nano, vol. 7, no. 12, 2013. [11] Oriol Lopez-Sanchez, Dominik Lembke, Metin Kayci, Aleksandra Radenovic & Andras Kis, "Ultrasensitive photodetectors based on monolayer MoS2," Nature Nanotechnology, vol. 8, 2013. [12] Vy Tran, Ryan Soklaski, Yufeng Liang, and Li Yang, "Layer-controlled band gap and anisotropic excitons in few-layer black phosphorus," PHYSICAL REVIEW B, vol. 89, no. 235319, 2014. [13] Claudia Ruppert, Ozgur Burak Aslan, and Tony F. Heinz, "Optical Properties and Band Gap of Single- and Few-Layer MoTe2 Crystals," Nano Letters, vol. 14, no. 11, 2014. [14] F. Withers, O. Del Pozo-Zamudio, A. Mishchenko, A. P. Rooney, A. Gholinia, K. Watanabe, T. Taniguchi, S. J. Haigh, A. K. Geim, A. I. Tartakovskii & K. S. Novoselov, "Light-emitting diodes by band-structure engineering in van der Waals heterostructures," Nature Materials, vol. 14, 2015. [15] Rui Cheng, Dehui Li, Hailong Zhou, Chen Wang, Anxiang Yin, Shan Jiang, Yuan Liu, Yu Chen, Yu Huang, and Xiangfeng Duan, "Electroluminescence and Photocurrent Generation from Atomically Sharp WSe2/MoS2 Heterojunction p–n Diodes," Nano Letters, vol. 14, no. 10, 2014.

132 [16] Weihuang Yang, Jingzhi Shang, Jianpu Wang, Xiaonan Shen, Bingchen Cao, Namphung Peimyoo, Chenji Zou, Yu Chen, Yanlong Wang, Chunxiao Cong, Wei Huang, and Ting Yu, "Electrically Tunable Valley-Light Emitting Diode (vLED) Based on CVD-Grown Monolayer WS2," Nano Letters, vol. 16, no. 3, 2016. [17] Jingzhi Shang, Xiaonan Shen, Chunxiao Cong, Namphung Peimyoo, Bingchen Cao, Mustafa Eginligil, and Ting Yu, "Observation of Excitonic Fine Structure in a 2D Transition-Metal Dichalcogenide Semiconductor," ACS Nano, vol. 9, no. 1, 2015. [18] X. H. Wang, J. Q. Ning, Z. C. Su, C. C. Zheng, B. R. Zhu, L. Xie, H. S. Wu and S. J. Xu , "Photoinduced doping and photoluminescence signature in an exfoliated WS2 monolayer semiconductor," RCS Advances, vol. 6, 2016. [19] Gwan-Hyoung Lee, Young-Jun Yu, Xu Cui, Nicholas Petrone, Chul-Ho Lee, Min Sup Choi, Dae-Yeong Lee, Changgu Lee, Won Jong Yoo, Kenji Watanabe, Takashi Taniguchi, Colin Nuckolls, Philip Kim, and James Hone, "Flexible and Transparent MoS2 Field-Effect Transistors on Hexagonal Boron Nitride-Graphene Heterostructures," ACS Nano, vol. 7, no. 9, 2013. [20] Gerd Plechinger, Philipp Nagler, Julia Kraus, Nicola Paradiso, Christoph Strunk, Christian Schüller, and Tobias Korn, "Identification of excitons, trions and biexcitons in single-layer WS2," Phys. Status Solidi RRL, vol. 9, no. 8, 2015. [21] Yumeng You, Xiao-Xiao Zhang, Timothy C. Berkelbach, Mark S. Hybertsen, David R. Reichman & Tony F. Heinz, "Observation of biexcitons in monolayer WSe2," Nature Physics, vol. 11, 2015. [22] Ziliang Ye, Ting Cao, Kevin O’Brien, Hanyu Zhu, Xiaobo Yin, Yuan Wang, Steven G. Louie & Xiang Zhang, "Probing excitonic dark states in single-layer tungsten disulphide," Nature, 2014. [23] X. H. Wang, J. Q. Ning, C. C. Zheng, B. R. Zhu, L. Xie, H. S. Wu and S. J. Xu, "Photoluminescence and Raman mapping characterization of WS2 monolayers prepared using top-down and bottom-up methods," J. Mater. Chem. C, vol. 3, 2015. [24] Ayse Berkdemir, Humberto R. Gutiérrez, Andrés R. Botello-Méndez, Néstor Perea-López, Ana Laura Elías, Chen-Ing Chia, Bei Wang, Vincent H. Crespi, Florentino López-Urías, Jean-Christophe Charlier, Humberto Terrones & Mauricio Terrones, "Identification of individual and few layers of WS2 using Raman Spectroscopy," Scientific Reports, vol. 3, 2013. [25] Weijie Zhao†§, Zohreh Ghorannevis, Leiqiang Chu, Minglin Toh, Christian Kloc, Ping- Heng Tan, and Goki Eda, "Evolution of Electronic Structure in Atomically Thin Sheets of WS2 and WSe2," ACS Nano, vol. 7, no. 1, 2013. [26] Yi-Hsien Lee, Xin-Quan Zhang, Wenjing Zhang, Mu-Tung Chang, Cheng-Te Lin, Kai-Di Chang, Ya-Chu Yu, Jacob Tse-Wei Wang, Chia-Seng Chang, Lain-Jong Li, Tsung-Wu Lin, "Synthesis of Large-Area MoS2 Atomic Layers with Chemical Vapor Deposition," Advanced Materials, vol. 24, no. 17, 2012. [27] J.R. Sambles, G.W> Bradbery, Fuzi Yang, "Optical excitation of surface plasmons: an introduction," Contemporary Physics, vol. 32, no. 3, 1991. [28] Junichi Takahara, Suguru Yamagishi, Hiroaki Taki, Akihiro Morimoto, and Tetsuro Kobayashi, "Guiding of a one-dimensional optical beam with nanometer diameter," Optics Letters, vol. 22, no. 7, 1997. [29] Kh.V.Nerkararyan, "Superfocusing of a surface polariton in a wedge-like structure," Physics Letters A, vol. 237, no. 1, 1997. [30] H. Raether, Surface Plasmons on Smooth and Rough Surfaces and on Gratings, Springer,

133 1988. [31] Stefan A. Maier and Harry A. Atwater, "Plasmonics: Localization and guiding of electromagnetic energy in metal/dielectric structures," Journal of Applied Physics, vol. 98, 2005. [32] W. L. Barnes, "Surface plasmon–polariton length scales: a route to sub-wavelength optics," Journal of Optics A, 2006. [33] Dmitri K. Gramotnev & Sergey I. Bozhevolnyi, "Plasmonics beyond the diffraction limit," Nature Photonics, vol. 4, 2010. [34] E. N. Economou, "Surface Plasmons in Thin Films," Phys. Rev., vol. 182, no. 539, 1969. [35] J. J. Burke, G. I. Stegeman, and T. Tamir, "Surface-polariton-like waves guided by thin, lossy metal films," Physical Review B, vol. 33, 1986. [36] Gururaj V. Naik , Vladimir M. Shalaev , and Alexandra Boltasseva, "Alternative Plasmonic Materials: Beyond Gold and Silver," Advanced Materials, vol. 25, 2013. [37] Zhizhen Ma, Zhuoran Li, Ke Liu, Chenran Ye, Volker J. Sorger, "Indium-Tin-Oxide for High-performance Electro-optic Modulation," vol. 4, no. 1, 2015. [38] F. Michelotti, L. Dominici, E. Descrovi, N. Danz, and F. Menchini, "hickness dependence of surface plasmon polariton dispersion in transparent conducting oxide films at 1.55 휇m," Optics Letters, vol. 34, no. 6, 2009. [39] K. S. Novoselov , D. Jiang , F. Schedin , T. J. Booth , V. V. Khotkevich , S. V. Morozov , and A. K. Geim, "Two-dimensional atomic crystals," Proc. Natl. Acad. Sci. U.S.A., vol. 102, no. 10451 , 2005. [40] "ASTM G173-03 Reference Solar Spectral Irradiance," NREL, [Online]. Available: http://rredc.nrel.gov/solar/spectra/am1.5/ASTMG173/ASTMG173.html. [41] Keliang He, Charles Poole, Kin Fai Mak, and Jie Shan, "Experimental Demonstration of Continuous Electronic Structure Tuning via Strain in Atomically Thin MoS2," Nano Letters, vol. 13, no. 6, 2013. [42] Chanyoung Yim, Maria O'Brien, Niall McEvoy, Sinéad Winters, Inam Mirza, James G Lunney, Georg S Duesberg, "Investigation of the optical properties of MoS2 thin films using spectroscopic ellipsometry," Applied Physics Letters, vol. 104, 2014. [43] Marco M. Furchi, Andreas Pospischil, Florian Libisch, Joachim Burgdörfer, and Thomas Mueller, "Photovoltaic Effect in an Electrically Tunable van der Waals Heterojunction," Nano Letters, vol. 14, no. 8, 2014. [44] Wenjing Zhang, Chih-Piao Chuu, Jing-Kai Huang, Chang-Hsiao Chen, Meng-Lin Tsai, Yung-Huang Chang, Chi-Te Liang, Yu-Ze Chen, Yu-Lun Chueh, Jr-Hau He, Mei-Yin Chou & Lain-Jong Li, "Ultrahigh-Gain Photodetectors Based on Atomically Thin Graphene- MoS2 Heterostructures," Scientific Reports, vol. 4, 2014. [45] K.I. Bolotina, K.J. Sikesb, Z. Jianga, M. Klimac, G. Fudenberga, J. Honec, P. Kima, H.L. Stormera, "Ultrahigh electron mobility in suspended graphene," Solid State Communications, vol. 146, no. 9-10, 2008. [46] Gordon B. Haxel, James B. Hedrick, and Greta J. Orris, "Rare Earth Elements—Critical Resources for High Technology," US Geological Survey , 2002. [47] Albert Polman & Harry A. Atwater, "Photonic design principles for ultrahigh-efficiency photovoltaics," Nature Materials, vol. 11, no. 174-177, 2012. [48] V. G. Kravets, A. N. Grigorenko, R. R. Nair, P. Blake, S. Anissimova, K. S. Novoselov, and A. K. Geim, "Spectroscopic ellipsometry of graphene and an exciton-shifted van Hove peak in absorption," Phys. Rev. B, vol. 81, no. 155413 , 2010.

134 [49] S. L. Ren, A. M. Rao, P. C. Eklund and G. L. Doll, "Ellipsometric study of boron nitride thin‐film growth on Si(100)," Applied Physics Letters, vol. 62, no. 1760, 1993. [50] X. Duan, Y. Huang, R. Agarwal, & C.M. Lieber, "Single-nanowire electrically driven lasers," Nature, vol. 421, no. 241–245, 2003. [51] Rupert F. Oulton, Volker J. Sorger, Thomas Zentgraf, Ren-Min Ma, Christopher Gladden, Lun Dai, Guy Bartal & Xiang Zhang, "Plasmon lasers at deep subwavelength scale," Nature, vol. 461, no. 629-632, 2009. [52] Jinyao Tang, Ziyang Huo, Sarah Brittman, Hanwei Gao & Peidong Yang, "Solution- processed core–shell nanowires for efficient photovoltaic cells," Nature Nanotechnology, vol. 6, no. 568–572, 2011. [53] J. F. C. S.-K. K. R. W. D. D. C. B. H.-G. P. a. C. M. L. Thomas J. Kempaa, "Coaxial multishell nanowires with high-quality electronic interfaces and tunable optical cavities for ultrathin photovoltaics," Proceedings of National Academy of Science, vol. 109, no. 5, 2012. [54] J. Grandidier, D. M. Callahan, J. N. Munday, and H. A. Atwater, "Light Absorption Enhancement in Thin-Film Solar Cells Using Whispering Gallery Modes in Dielectric Nanospheres," Advanced Materials, vol. 23, 2011. [55] Sun-Kyung Kim, Kyung-Deok Song, Thomas J. Kempa, Robert W. Day, Charles M. Lieber, and Hong-Gyu Park, "Design of Nanowire Optical Cavities as Efficient Photon Absorbers," ACS Nano, vol. 8, no. 4, 2014. [56] Gwan-Hyoung Lee, Young-Jun Yu, Changgu Lee, Cory Dean, Kenneth L Shepard, Philip Kim, James Hone, "Electron tunneling through atomically flat and ultrathin hexagonal boron nitride," Applied Physical Letters, vol. 99, no. 243114 , 2011. [57] Hiram J. Conley, Bin Wang, Jed I. Ziegler, Richard F. Haglund, Sokrates T. Pantelides, and Kirill I. Bolotin, "Bandgap Engineering of Strained Monolayer and Bilayer MoS2," Nano Letters, vol. 13, no. 8, 2013. [58] Won Seok Yun, S. W. Han, Soon Cheol Hong, In Gee Kim, and J. D. Lee, "Thickness and strain effects on electronic structures of transition metal dichalcogenides," Physical Review B, vol. 85, 2012. [59] Emilio Scalise, Michel Houssa, Geoffrey Pourtois, Valery Afanas’ev, and André Stesmans, "Strain-Induced Semiconductor to Metal Transition in the Two-Dimensional Honeycomb Structure of MoS2," Nano Research, vol. 5, no. 1, 2012. [60] Martin T. Hill, Milan Marell, Eunice S. P. Leong, Barry Smalbrugge, Youcai Zhu, Minghua Sun, Peter J. van Veldhoven, Erik Jan Geluk, Fouad Karouta, Yok-Siang Oei, Richard Nötzel, Cun-Zheng Ning, and Meint K. Smit, "Lasing in metal-insulator-metal subwavelength plasmonic waveguides," Optics Express, vol. 17, no. 13, 2009. [61] Mohammad H Tahersima and Volker J Sorger, "Enhanced photon absorption in spiral nanostructured solar cells using layered 2D materials," Nanotechnology, vol. 26, no. 34, 2015. [62] Volker J. Sorger, Mohammad Tahersima, and Ke Liu, "Strong Light-Matter-Interactions in Nanoscale and 2D Material Systems," in OSA Novel Optical Materials and Applications , Boston, 2015. [63] Mohammad H Tahersima, Volker J Sorger, "2D materials for photon conversion and nanophotonics," in SPIE Infrared Remote Sensing and Instrumentation , San Diego, 2015. [64] Mohammad H Tahersima, Volker J Sorger, "Strong Photon Absorption in 2-D Material- Based Spiral Photovoltaic Cells," in Cambridge University Press, Boston, 2016. [65] Svetlana V Boriskina, Martin A Green, Kylie Catchpole, Eli Yablonovitch, Matthew C Beard, Yoshitaka Okada, Stephan Lany, Talia Gershon, Andriy Zakutayev, Mohammad H

135 Tahersima, Volker J Sorger, Michael J Naughton, Krzysztof Kempa, Mario Dagenais, Yuan Yao,, "Roadmap on optical energy conversion," Journal of Optics, vol. 18, no. 7, 2016. [66] Shuai Sun, Mohammad Tahersima, Volker J Sorger, Ke Liu, "Physical Scaling Laws of Nanophotonics: Case Photon Conversion," in OSA Optical Nanostructures and Advanced Materials for Photovoltaics, Germany, 2016. [67] Chunming Huang, Sanfeng Wu, Ana M. Sanchez, Jonathan J. P. Peters, Richard Beanland, Jason S. Ross, Pasqual Rivera, Wang Yao, David H. Cobden & Xiaodong Xu, Lateral heterojunctions within monolayer MoSe2–WSe2 semiconductors, vol. 13, 2014. [68] John Mann, Edwin Preciado, David Barroso, and Ludwig Bartels, "2-Dimensional Transition Metal Dichalcogenides with Tunable Direct Band Gaps: MoS2(1–x)Se2x Monolayers," Advanced Materials, vol. 26, no. 9, 2014. [69] Andrea Splendiani, Liang Sun, Yuanbo Zhang, Tianshu Li, Jonghwan Kim, Chi-Yung Chim, Giulia Galli and Feng Wang, "Emerging Photoluminescence in Monolayer MoS2," Nano Letters, vol. 10, no. 4, 2010. [70] Humberto R. Gutiérrez, Nestor Perea-López, Ana Laura Elías, Ayse Berkdemir, Bei Wang, Ruitao Lv, Florentino López-Urías, Vincent H. Crespi, Humberto Terrones, and Mauricio Terrones, "Extraordinary Room-Temperature Photoluminescence in Triangular WS2 Monolayers," Nano Letters, vol. 13, no. 8, 2013. [71] Fengnian Xia, Han Wang, Di Xiao, Madan Dubey & Ashwin Ramasubramaniam, "Two- dimensional material nanophotonics," Nature Photonics, vol. 8, 2014. [72] Zhizhen Ma, Mohammad H. Tahersima, Sikandar Khan, Volker J. Sorger , "Two- Dimensional Material-Based Mode Confinement Engineering in Electro-Optic Modulators," IEEE Journal of Selected Topics in Quantum Electronics, vol. 23, no. 1, 2017. [73] Kin Fai Mak, Keliang He, Changgu Lee, Gwan Hyoung Lee, James Hone, Tony F. Heinz & Jie Shan, "Tightly bound trions in monolayer MoS2," Nature Materials, vol. 12, 2013. [74] Zefang Wang, Liang Zhao, Kin Fai Mak , and Jie Shan, "Probing the Spin-Polarized Electronic Band Structure in Monolayer Transition Metal Dichalcogenides by Optical Spectroscopy," Nano Letters, vol. 17, no. 2, 2017. [75] Alexey Chernikov, Timothy C. Berkelbach, Heather M. Hill, Albert Rigosi, Yilei Li, Ozgur Burak Aslan, David R. Reichman, Mark S. Hybertsen, and Tony F. Heinz, "Exciton Binding Energy and Nonhydrogenic Rydberg Series in Monolayer," Physical Review Letters, vol. 113, no. 076802, 2014. [76] Xuetao Gan, Yuanda Gao, Kin Fai Mak, Xinwen Yao, Ren-Jye Shiue, Arend van der Zande, Matthew E. Trusheim, Fariba Hatami, Tony F. Heinz, James Hone, and Dirk Englund, "Controlling the spontaneous emission rate of monolayer MoS2 in a photonic crystal nanocavity," vol. 103, 2013. [77] Yu Ye, Zi Jing Wong, Xiufang Lu, Xingjie Ni, Hanyu Zhu, Xianhui Chen, Yuan Wang & Xiang Zhang, "Monolayer excitonic laser," Nature Photonics, vol. 9, 2015. [78] Zhuan Zhu, Jiangtan Yuan, Haiqing Zhou, Jonathan Hu, Jing Zhang, Chengli Wei, Fang Yu, Shuo Chen, Yucheng Lan, Yao Yang, Yanan Wang, Chao Niu, Zhifeng Ren, Jun Lou, Zhiming Wang, and Jiming Bao, "Excitonic Resonant Emission−Absorption of Surface Plasmons in Transition Metal Dichalcogenides for Chip-Level Electronic−Photonic Integrated Circuits," ACS Photonics, vol. 3, no. 5, 2016. [79] Michael S. Eggleston, Sujay Desai, Kevin Messer, Surabhi Madhvapathy, Jun Xiao, Xiang Zhang, Eli Yablonovitch, Ali Javey, Ming Wu, "Enhanced spontaneous emission from an optical antenna coupled WSe2 monolayer," in OSA CLEO, San Jose, 2015. [80] Hossein Taghinejad, Mohammad Taghinejad, Alexey Tarasov, Meng-Yen Tsai, Amir H.

136 Hosseinnia, Hesam Moradinejad, Philip M. Campbell, Ali A. Eftekhar, Eric M. Vogel, and Ali Adibi, "Resonant Light-Induced Heating in Hybrid Cavity-Coupled 2D Transition- Metal Dichalcogenides," ACS Photonics, vol. 3, no. 4, 2016. [81] B. Mukherjee, N. Kaushik, Ravi P. N. Tripathi, A. M. Joseph, P. K. Mohapatra, S. Dhar, B. P. Singh, G. V. Pavan Kumar, E. Simsek & S. Lodha, "Exciton Emission Intensity Modulation of Monolayer MoS2 via Au Plasmon Coupling," Scientific Reports, vol. 7, 2017. [82] Hossein Taghinejad, S. Hamed Shams-Mousavi, Yongji Gong, Mohammad Taghinejad, Ali A. Eftekhar, Pulickel Ajayan, Ali Adibi, "Lattice Plasmon Induced Large Enhancement of Excitonic Emission in Monolayer Metal Dichalcogenides," Plasmonics, vol. 12, 2016. [83] Gleb M. Akselrod, Tian Ming, Christos Argyropoulos, Thang B. Hoang, Yuxuan Lin, Xi Ling, David R. Smith, Jing Kong, and Maiken H. Mikkelsen, "Leveraging Nanocavity Harmonics for Control of Optical Processes in 2D Semiconductors," Nano Letters, vol. 15, no. 5, 2015. [84] Serkan Butun, Sefaattin Tongay, and Koray Aydin, "Enhanced Light Emission from Large- Area Monolayer MoS2 Using Plasmonic Nanodisc Arrays," Nano Letters, vol. 15, no. 4, 2015. [85] Johannes Kern, Andreas Trügler, Iris Niehues, Johannes Ewering, Robert Schmidt, Robert Schneider, Sina Najmaei, Antony George, Jing Zhang, Jun Lou, Ulrich Hohenester, Steffen Michaelis de Vasconcellos, and Rudolf Bratschitsch, "Nanoantenna-Enhanced Light– Matter Interaction in Atomically Thin WS2," ACS Photonics, vol. 2, no. 9, 2015. [86] Jingzhi Shang, Xiaonan Shen, Chunxiao Cong, Namphung Peimyoo, Bingchen Cao, Mustafa Eginligil, and Ting Yu, "Observation of Excitonic Fine Structure in a 2D Transition-Metal Dichalcogenide Semiconductor," ACS Nano, vol. 9, no. 1, 2015. [87] Bumsu Lee, Joohee Park, Gang Hee Han, Ho-Seok Ee, Carl H. Naylor, Wenjing Liu, A.T. Charlie Johnson, and Ritesh Agarwal, "Fano Resonance and Spectrally Modified Photoluminescence Enhancement in Monolayer MoS2 Integrated with Plasmonic Nanoantenna Array," Nano Letters, vol. 15, no. 5, 2015. [88] Wenjing Liu, Bumsu Lee, Carl H. Naylor, Ho-Seok Ee, Joohee Park, A. T. Charlie Johnson, and Ritesh Agarwal, "Strong exciton-plasmon coupling in MoS2 coupled with plasmonic lattice," Nano Letters, vol. 16, no. 2, 2016. [89] Sina Najmaei, Adnen Mlayah, Arnaud Arbouet, Christian Girard, Jean Léotin, and Jun Lou, "Plasmonic Pumping of Excitonic Photoluminescence in Hybrid MoS2-Au Nanostructures," ACS Nano, vol. 8, no. 12, 2014. [90] Udai Bhanu, Muhammad R. Islam, Laurene Tetard & Saiful I. Khondaker, "Photoluminescence quenching in gold - MoS2 hybrid nanoflakes," Scientific Reports, vol. 4, 2014. [91] Chang-Hee Cho, Carlos O. Aspetti, Michael E. Turk, James M. Kikkawa, Sung-Wook Nam & Ritesh Agarwal, "Tailoring hot-exciton emission and lifetimes in semiconducting nanowires via whispering-gallery nanocavity plasmons," Nature Materials, vol. 10, 2011. [92] Volker J. Sorger, Nitipat Pholchai, Ertugrul Cubukcu, Rupert F. Oulton, Pavel Kolchin, Christian Borschel, Martin Gnauck, Carsten Ronning, and Xiang Zhang, "Strongly Enhanced Molecular Fluorescence inside a Nanoscale Waveguide Gap," Nano Letters, vol. 11, no. 11, 2011. [93] Pascal Anger, Palash Bharadwaj, and Lukas Novotny, "Enhancement and quenching of single-molecule fluorescence," Physical Review Letters, vol. 96, 2006. [94] Yang Cui, Run Xin, Zhihao Yu, Yiming Pan, Zhun-Yong Ong, Xiaoxu Wei, Junzhuan

137 Wang, Haiyan Nan, Zhenhua Ni, Yun Wu, Tangsheng Chen, Yi Shi, Baigeng Wang, Gang Zhang, Yong-Wei Zhang, Xinran Wang, "High-Performance Monolayer WS2 Field-Effect Transistors on High-κ Dielectrics," Advanced Materials , vol. 27, no. 35, 2015. [95] Ali Sobhani, Adam Lauchner, Sina Najmaei, Ciceron Ayala-Orozco, Fangfang Wen, Jun Lou, and Naomi J. Halas, "Enhancing the photocurrent and photoluminescence of single crystal monolayer MoS2 with resonant plasmonic nanoshells," Applied Physics Letters, vol. 104, no. 031112, 2014. [96] Bo Yan, Svetlana V. Boriskina, and Björn M. Reinhard, "Design and Implementation of Noble Metal Nanoparticle Cluster Arrays for Plasmon Enhanced Biosensing," Journal of Physical Chemistry C, vol. 115, no. 50, 2011. [97] Paolo Biagioni, Jer-Shing Huang and Bert Hecht, "Nanoantennas for Visible and Infrared Radiation," Reports on Progress in Physics, vol. 75, no. 2, 2012. [98] Alireza Bonakdar and Hooman Mohseni, "Impact of optical antennas on active optoelectronic devices," Nanoscale, vol. 6, 2014. [99] H. Wheeler, "Fundamental Limitations of Small Antennas," in Proceedings of the IRE, 1947. [100] Hsiang-Lin Liu, Chih-Chiang Shen, Sheng-Han Su, Chang-Lung Hsu, Ming-Yang Li, and Lain-Jong Li, "Optical properties of monolayer transition metal dichalcogenides probed by spectroscopic ellipsometry," Applied Physics Letters, vol. 105, no. 201905 , 2014. [101] E. D. Palik, Handbook of optical constants of solids, Waltham, MA: Academic Press, 1998. [102] P.B. Johnson, R.W. Christy, "Optical Constants of the Noble Metals," Physical Review B, vol. 6, no. 4370, 1972. [103] Jon A. Schuller, Sinan Karaveli, Theanne Schiros, Keliang He, Shyuan Yang, Ioannis Kymissis, Jie Shan & Rashid Zia, "Orientation of luminescent excitons in layered nanomaterials," Nature Nanotechnology, vol. 8, 2013. [104] G. Kresse and J. Furthmüller, "Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set," Physical Review B, vol. 54, no. 11169, 1996. [105] P. E. Blöchl, "Projector augmented-wave method," Physical Review B, vol. 50, no. 17953 , 1994. [106] G. Kresse, "From ultrasoft pseudopotentials to the projector augmented-wave method," Physical Review B, vol. 59, 1999. [107] John P. Perdew, Kieron Burke, and Matthias Ernzerhof, "Generalized Gradient Approximation Made Simple.," Physics Review Letters, vol. 77, no. 3865 , 1996. [108] Hendrik J. Monkhorst and James D. Pack, "special points for Brillouin-zone integrations," Physical Review B, vol. 13, no. 5188, 1976. [109] Iori Tanabe, Michael Gomez, William C. Coley, Duy Le, Elena M. Echeverria, Gordon Stecklein, Viktor Kandyba, Santosh K. Balijepalli, Velveth Klee, Ariana E. Nguyen, Edwin Preciado, I-Hsi Lu, Sarah Bobek, David Barroso, Dominic Martinez-Ta, "Band structure characterization of WS2 grown by chemical vapor deposition," Applied Physics Letters, vol. 108, no. 252103, 2016. [110] Mohammad H Tahersima, M Danang Birowosuto, Zhizhen Ma, William C Coley, Michael D Valentin, Sahar Naghibi Alvillar, I-Hsi Lu, Yao Zhou, Ibrahim Sarpkaya, Aimee Martinez, Ingrid Liao, Brandon N Davis, Joseph Martinez, Dominic Martinez-Ta, Alison Guan, Aria, "Testbeds for Transition Metal Dichalcogenide Photonics: Efficacy of Light Emission Enhancement in Monomer vs Dimer Nanoscale Antennae," ACS Photonics, vol. 4, no. 7, 2017.

138 [111] Mohammad H Tahersima, Muhammad Danang Birowosuto, Zhizhen Ma, William C Coley, Michael Valentin, I-Hsi Lu, Ke Liu, Yao Zhou, Amy Martinez, Ingrid Liao, Brandon Davis, Joseph Martinez, Sahar Naghibi, Dominic Martinez-Ta, Allison Guan, Ariana E Nguyen, Cesa, "Optical antenna enhanced spontaneous emission from CVD-grown monolayer WS 2," in CLEO: Applications and Technology, San Jose, 2017. [112] Mohammad H Tahersima, Muhammad Danang Birowosuto, Zhizhen Ma, William C Coley, Michael Valentin, Sahar Naghibi Alvillar, Cesare Soci, Ludwig Bartels, Volker J Sorger, "Light Emission Enhancement of 2D Materials in Monomer vs. Dimer Nanoantennae," in Laser Science, Washington, 2017. [113] T. R. Oke, "The energetic basis of the urban heat island," Quarterly Journal of the Royal Meteorological Society, vol. 108, no. 455, 1982. [114] [Online]. Available: https://www.epa.gov/heat-islands. [115] Rizwan Ahmed Memon, Leung YC Dennis, and L. I. U. Chunho, "A review on the generation, determination and mitigation of Urban Heat Island," Journal of Environmental Sciences, vol. 20, no. 1, 2008. [116] L. H. Slooff, E. E. Bende, A. R. Burgers, T. Budel, M. Pravettoni, R. P. Kenny, E. D. Dunlop, A. Büchtemann, "A Luminescent Solar Concentrator with 7.1% power conversion efficiency," Physica Status Solidi, vol. 2, no. 6, 2008. [117] Vadim Karagodsky, Forrest G. Sedgwick, and Connie J. Chang-Hasnain, "Theoretical analysis of subwavelength high contrast grating reflectors," Optics Express, vol. 18, no. 16, 2010. [118] Katherine Baker, Jason Karp, Justin Hallas, Joseph Ford, "Practical implementation of a planar micro-optic solar concentrator," in Nonimaging Optics: Efficient Design for Illumination and Solar Concentration, San Diego, 2012. [119] M. Glick, "Optical Interconnects in Next Generation Data Centers: An End to End View.," in Optical Interconnects for Future Data Center Networks. Optical Networks, New York, Springer, 2013. [120] Ansheng Liu, Richard Jones, Ling Liao, Dean Samara-Rubio, Doron Rubin, Oded Cohen, Remus Nicolaescu & Mario Paniccia, "A high-speed silicon optical modulator based on a metal–oxide–semiconductor capacitor," Nature, vol. 427, 2004. [121] Chenran Ye, Sikandar Khan, Zhuo Ran Li, Ergun Simsek, Volker J Sorger, "λ-Size ITO and Graphene-Based Electro-Optic Modulators on SOI," IEEE Journal of Selected Topics in Quantum Electronics, vol. 20, no. 4, 2014. [122] William A. Zortman, Anthony L. Lentine, Douglas C. Trotter, and Michael R. Watts, "Low- voltage differentially-signaled modulators," Optics Express, vol. 19, no. 27, 2011. [123] Po Dong, Shirong Liao, Dazeng Feng, Hong Liang, Dawei Zheng, Roshanak Shafiiha, Cheng-Chih Kung, Wei Qian, Guoliang Li, Xuezhe Zheng, Ashok V. Krishnamoorthy, and Mehdi Asghari, "Low Vpp, ultralow-energy, compact, high-speed silicon electro-optic modulator," Optics Express, vol. 17, no. 25, 2009. [124] William A. Zortman, Douglas C. Trotter, Anthony L. Lentine, Gideon Robertson, Alex Hsia, and Michael R. Watts, "Monolithic and Two-Dimensional Integration of Silicon Photonic Microdisks With Microelectronics," IEEE Photonics, vol. 4, no. 1, 2012. [125] G. Li, "High-efficiency 25Gb/s CMOS ring modulator with integrated thermal tuning," in IEEE Xplore, London, 2011. [126] Xi Xiao, Xianyao Li, Hao Xu, "44-Gb/s Silicon Microring Modulators Based on Zigzag PN Junctions," IEEE Photonics Technology Letters, vol. 24, no. 19, 2012. [127] J. C. Rosenberg, W. M. J. Green, S. Assefa, D. M. Gill, T. Barwicz, M. Yang, S. M. Shank,

139 and Y. A. Vlasov, "A 25 Gbps silicon microring modulator based on an interleaved junction," Optics Express, vol. 20, no. 24, 2012. [128] Scott H. Brewer, Stefan Franzen, "alculation of the electronic and optical properties of indium tin oxide by density functional theory," Chemical Physics, vol. 300, no. 1-3, 2004. [129] F. Neumann, Y. A. Genenko, C. Melzer, S. V. Yampolskii, and H. von Seggern, "Self- consistent analytical solution of a problem of charge-carrier injection at a conductor/insulator interface," Physical Review B, vol. 75, no. 205322, 2007. [130] Alok P. Vasudev, Ju-Hyung Kang, Junghyun Park, Xiaoge Liu, and Mark L. Brongersma, "Electro-optical modulation of a silicon waveguide with an “epsilon-near-zero” material," Optics Express, vol. 21, no. 22, 2013. [131] Kaifeng Shi, Riaz R. Haque, Bingyin Zhao, Runchen Zhao, and Zhaolin Lu, "Broadband electro-optical modulator based on transparent conducting oxide," Optics Letters, vol. 39, no. 17, 2014. [132] P.R. West S. Ishii G.V. Naik N.K. Emani V.M. Shalaev A. Boltasseva, "Searching for better plasmonic materials," Lasers and Photonics Review, 2010. [133] M. A. Noginov1, Lei Gu1, J. Livenere, G. Zhu, A. K. Pradhan, R. Mundle, M. Bahoura, Yu. A. Barnakov, and V. A. Podolskiy, "Transparent conductive oxides: Plasmonic materials for telecom wavelengths," Applied Physics Letters, vol. 99, 2011. [134] S. Govindarajan, T. S. Boscke, P. Sivasubramani, P. D. Kirsch, B. H. Lee, H.-H. Tseng, R. Jammy, U. Schroder, S. Ramanathan, and B. E. Gnade, "Higher permittivity rare earth doped HfO2 for sub-45-nm metal-insulator-semiconductor devices," Appl. Phys. Lett., vol. 91, no. 062906 , 2007. [135] H. Kim, P. C. McIntyre, and K. C. Saraswat, "Effects of crystallization on the electrical properties of ultrathin HfO2 dielectrics grown by atomic layer deposition," Appl. Phys. Lett., vol. 82, no. 1, 2003. [136] W. Bogaerts, P. De Heyn, T. Van Vaerenbergh, K. De Vos, S. Kumar Selvaraja, T. Claes P. Dumon, P. Bienstman, D. Van Thourhout, R. Baets, "Silicon microring resonators," Laser and Photonics Review, vol. 6, 2011. [137] Yunhong Ding, Xiaolong Zhu, Sanshui Xiao, Hao Hu, Lars Hagedorn Frandsen, "Effective Electro-Optical Modulation with High Extinction Ratio by a Graphene−Silicon Microring Resonator," Nanoletters, vol. 15, 2015. [138] W. D. Sacher, W. M. J. Green, S. Assefa, T. Barwicz, H. Pan, S. M. Shank, Y. A. Vlasov, and J. K. S. Poon, "Coupling modulation of microrings at rates beyond the linewidth limit," Optics Express, vol. 21, no. 8, 2013. [139] Yunchu Li, Lawrence S. Stewart, and P. Daniel Dapkus, "High speed silicon microring modulator employing dynamic intracavity energy balance," Optics Express, 2012. [140] Wesley D. Sacher, Eric J. Zhang, Brett A. Kruger, and Joyce K. S. Poon, "High-speed laser modulation beyond the relaxation resonance frequency limit," Optics Express, 2010.

140