metal nanocavity light sources integrated with passive waveguide components

Metal nanocavity light sources integrated with passivewaveguide componentsCitation for published version (APA):Dolores Calzadilla, V. (2016). Metal nanocavity light sources integrated with passive waveguide components.Eindhoven: Technische Universiteit Eindhoven.

Document status and date:Published: 14/04/2016

Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.

Download date: 05. Jun. 2020

https://research.tue.nl/en/publications/metal-nanocavity-light-sources-integrated-with-passive-waveguide-components(a36e7650-3335-45c6-8218-a9964b7b4f56).html

Metal nanocavity light sources

integrated with passive

waveguide components

PROEFSCHRIFT

ter verkrijging van de graad van doctor

aan de Technische Universiteit Eindhoven,

op gezag van de rector magnificus prof.dr.ir. F.P.T. Baaijens,

voor een commissie aangewezen door het College voor Promoties,

in het openbaar te verdedigen op

donderdag 14 april 2016 om 16:00 uur

door

Vıctor Manuel Dolores Calzadilla

geboren te Tlalnepantla de Baz, Mexico

Dit proefschrift is goedgekeurd door de promotoren en de samenstelling van de

promotiecommissie is als volgt:

voorzitter: prof.dr.ir. A.C.P.M. Backx

1e promotor: prof.dr.ir. M.K. Smit

2e promotor: prof.dr. A. Fiore

leden: Prof.Dr. J. Leuthold (ETH Zurich)

prof.dr.ir. D. van Thourhout (Universiteit Gent - IMEC)

dr.ir. H. de Waardt

dr. R. Oulton (Imperial College London)

dr. J.J.G.M. van der Tol

Het onderzoek of ontwerp dat in dit proefschrift wordt beschreven is uitgevoerd in

overeenstemming met de TU/e Gedragscode Wetenschapsbeoefening.

The research presented in this thesis was supported by the European Community’s

Seventh Framework Program through the NAVOLCHI project (288869), and was

carried out in the Photonic Integration group, at the Department of Electrical

Engineering of the Eindhoven University of Technology.

Metal nanocavity light sources integrated with passive waveguide components,

by Vıctor Manuel Dolores Calzadilla

A catalogue record is available from the

Eindhoven University of Technology Library

ISBN: 978-90-386-4037-2

Copyright c© 2016 Vıctor Manuel Dolores Calzadilla

Typeset using LATEX.

Printed by Gildeprint Drukkerijen, the Netherlands.

Cover design by V.M. Dolores Calzadilla based on the Light Festival GLOW 2015.

“Viva la vida”

Frida Kahlo

Summary

Title of dissertation: Metal nanocavity light sources integrated

with passive waveguide components

Photonic integrated circuits are promising to further expand the data communica-

tion systems as well as to meet future bandwidth requirements. Moreover, this is a

flexible technology that will enable new applications in a variety of fields. Photonic

integrated platforms using different material systems have been proposed and are

under continuous research and development. This thesis aims at contributing to

the development of photonic integration at a device level. The experimental work

has been carried out in III-V layer stacks on a silicon substrate. Research on the

following key devices is presented: III-V/Si nanoscale light sources with metal

cavity, a metal grating coupler for efficient chip-to-fiber optical coupling, and a

polarization rotator device. These devices will increase the functionality of future

photonic integrated platforms. Furthermore, the developed fabrication technology

is relevant for other metal-based nanophotonic devices.

Nanoscale light sources using metal cavities to achieve strong light confinement

have been proposed to enable high integration density, efficient operation at low

energy/bit and ultra-fast modulation, which would make them attractive for future

low-power optical interconnects. Research on both lasers and light-emitting diodes

was carried out, as well as the experimental demonstration of the first metal-

cavity nanopillar LED coupled to a waveguide on silicon. The cavity consists of

a metal-coated III-V semiconductor nanopillar which funnels a large fraction of

spontaneous emission into the fundamental mode of an InP waveguide bonded

to a silicon wafer. The device shows on-chip external quantum efficiency in the

10−4− 10−2 range at tens of µA current injection levels, which greatly exceeds the

performance of any waveguide-coupled nanoscale light source integrated on silicon

in this current range. Furthermore, direct modulation experiments reveal sub-

nanosecond electro-optical response with the potential for multi-Gbps modulation

speeds.

i

In the case of grating couplers, they allow for optical coupling between photonic

circuits and optical fibers in a vertical manner. They are used in a variety of

applications, for example, on-wafer characterization and packaging. They are also

attractive for optical interconnect systems that make use of chip-to-fiber and chip-

to-chip coupling schemes. A metal grating coupler consisting of a buried metal

grating and a metal mirror is proposed. According to modeling results, a non-

apodized design provides a fiber-to-chip coupling efficiency at 1.55 µm up to 73%,

whereas apodized designs show efficiencies as high as 89%, with a 3 dB bandwidth

of 61 and 78 nm, respectively. The experimental realization of the nonapodized

design resulted in 54% coupling efficiency and bandwidth of 61 nm. An important

advantage is that the coupling efficiency is independent from the underlying layer

stack, enabling its use in diverse applications without compromising its perfor-

mance. For example, a thick buffer is of interest in III-V/Si membranes since it

will allow thermal isolation between the photonic circuit and an underlying CMOS

chip.

Finally, a polarization converter device based on single-mode single-polarization

waveguides is proposed, which is compatible with both InP-membranes on silicon

and silicon photonics. The output waveguide of the device does not support the

undesired polarized mode, therefore high conversion efficiency is guaranteed by

design (conversion efficiency > 99.9% for a length of 40 µm). Furthermore, since

it is based on an adiabatic transition, it displays negligible insertion loss and large

bandwidth (> 150 nm for conversion efficiency larger than 99%). The fabrication

process is also proposed, which can be done in the same two-etch lithography steps

required for standard passive components, therefore it can be easily integrated in

a generic process. In view of its ultra-high performance and relatively simple

fabrication, it represents a practical solution for on-chip polarization conversion.

Contents

Summary i

Contents iii

1 Introduction 11.1 Light and photonic integration . . . . . . . . . . . . . . . . . . . . . 11.2 Motivation and overview of metal-cavity light sources . . . . . . . . 31.3 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Membrane integrated photonics 92.1 Introduction to membrane photonics . . . . . . . . . . . . . . . . . 92.2 Existing IMOS building blocks . . . . . . . . . . . . . . . . . . . . . 122.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Nanoscale light sources 153.1 Design of metal-cavity light sources . . . . . . . . . . . . . . . . . . 15

3.1.1 Optical modeling of waveguide-coupled metal cavities . . . . 153.1.1.1 Plasmonic Fabry-Perot laser cavity . . . . . . . . . 15

Basic principles of a Fabry-Perot cavity . . . . . . . . 15Plasmonic cavity structure . . . . . . . . . . . . . . . 18Modal properties . . . . . . . . . . . . . . . . . . . . 18Facet reflectivity and waveguide coupling . . . . . . . 20

3.1.1.2 Metallo-dielectric laser nanocavity . . . . . . . . . 22Basic principles of a nanocavity . . . . . . . . . . . . 22Metallo-dielectric cavity structure . . . . . . . . . . . 24Quality factor and waveguide coupling . . . . . . . . 25Differential efficiency and threshold conditions . . . . 27

3.1.1.3 Metallo-dielectric light-emitting diode cavity . . . . 273.1.2 Electrical modeling of III-V layer stack . . . . . . . . . . . . 313.1.3 Thermal modeling . . . . . . . . . . . . . . . . . . . . . . . 333.1.4 Small-signal frequency response . . . . . . . . . . . . . . . . 343.1.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

iii

Contents

3.2 Fabrication technology of metallo-dielectric nanopillar cavities cou-pled to waveguides . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.2.1 Adhesive bonding of III-V layer stacks to silicon . . . . . . . 373.2.2 Waveguides processing with positive resist (ZEP) . . . . . . 383.2.3 Waveguides processing with negative resist (HSQ) . . . . . . 403.2.4 Pedestal nanopillars . . . . . . . . . . . . . . . . . . . . . . 423.2.5 Silver deposition and treatment . . . . . . . . . . . . . . . . 433.2.6 Silver-based ohmic contacts . . . . . . . . . . . . . . . . . . 443.2.7 Process flow of waveguide-coupled nanopillars . . . . . . . . 45

3.3 Characterization of nanopillar LEDs . . . . . . . . . . . . . . . . . 523.3.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . 533.3.2 Static characteristics . . . . . . . . . . . . . . . . . . . . . . 553.3.3 Dynamic characteristics . . . . . . . . . . . . . . . . . . . . 573.3.4 Surface recombination and passivation . . . . . . . . . . . . 593.3.5 Perspectives for improvement . . . . . . . . . . . . . . . . . 613.3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4 Grating couplers 654.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654.2 Dielectric grating couplers . . . . . . . . . . . . . . . . . . . . . . . 674.3 Metal grating couplers . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.3.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.3.2 Fabrication and characterization . . . . . . . . . . . . . . . . 74

4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

5 Polarization rotator 775.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775.2 Design of a high performance polarization rotator . . . . . . . . . . 795.3 Fabrication proposal . . . . . . . . . . . . . . . . . . . . . . . . . . 835.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

6 Conclusions and outlook 85

A Main patterning processes 89

B Small-signal frequency response 91

C Semiconductor single-mode rate equations model 93

Bibliography 95

List of abbreviations 105

Acknowledgements 107

iv

Contents

List of research products 111

Curriculum vitae 117

v

Chapter 1

Introduction

1.1 Light and photonic integration

Technology has always driven the life style from the nomadic tribes to the modern

civilization. In particular, light technologies have impacted on the way we perceive

the macrocosm and the microcosm and therefore shape our world-view. Not to

mention that, just like water, light is a fundamental ingredient of life as we know it.

The modern society would simply not be possible without light-based technologies.

Science and technology of light have been possible thanks to a vast number of

people all around the world who have carried out experiments, discovered phe-

nomena, and transferred the knowledge to new generations. A significant progress

in the understanding of the nature of light happened after Isaac Newton discovered

that white light is composed of different colors in the 17th century. Later, several

scientists carried out individual work on electricity and magnetism, which was

later unified by James Clerk Maxwell in 1873 in its ”A treatise on electricity and

magnetism”. Then, Heinrich Hertz demonstrated that the electromagnetic waves

predicted by J.C. Maxwell indeed existed and thus an impressive development

in electromagnetic engineering occurred in which Nikola Tesla deserves special

mention since his inventions were fundamental for the subsequent technological

progress.

In the beginning of the 20th century, the work of Max Planck (on the black-body

radiation) and Albert Einstein (on a quantum-based description of the photoelec-

tric effect) converged towards a new theory that describes the quantization of the

electromagnetic field. The light amplification by stimulated emission of radiation

(LASER) was demonstrated later by Theodore Harold Maiman in 1960, which

1

Chapter 1. Introduction

Figure 1.1: (a) Analogy between basic building blocks in electronic ICs andPICs. (b) PIC for fiber Bragg sensing applications [4] of 4.5×4 mm2 next to apaper clip. The chip contains a variety of photonic components based on the

interconnection of basic building blocks.

relied on the stimulated emission phenomenon discovered theoretically by Albert

Einstein a few decades earlier. After this milestone, the term ”photonics” started

to flourish, which is known today as the science and technology of generating,

controlling, and detecting photons, which are quanta of electromagnetic energy.

The importance of photonics in the progress of modern society has been recognized

by several Nobel Prizes, including the prize awarded for the invention of lasers

(1964), and more recently for the invention of efficient blue light-emitting diodes

(LEDs) (2014) [1]. Furthermore, 2015 has been proclaimed by UNESCO as the

International Year of Light ”to raise awareness of how optical technologies promote

sustainable development and provide solutions to worldwide challenges in energy,

education, agriculture, communications and health” [2].

It is believed that the 21st century will depend on photonics as much as the 20th

century depended on electronics. Internet itself is possible due to a combination

of electronic and photonic components and systems. According to the perspective

of Bell Labs [3], the invention of internet triggered two universal technological

revolutions: the information and telecommunications revolution (1985-2000) and

the cloud and mobile revolution (2000-2015), which comprise the beginning of a

third industrial revolution [3].

Photonic integration, which was defined at an early stage as ”the integration of

a large number of optical devices on a small substrate, so forming an optical

circuit reminiscent of the integrated circuit in microelectronics” [5], is expected

to have an important role in this transition [3]. In the same way as electronic

2


integrated circuits (ICs) are based on basic building blocks, photonic integrated

circuits (PICs) are also based on such blocks which cascaded and connected in

different topologies build up more complex photonic components. Figure 1.1a

shows a simple analogy between electronic and photonic basic building blocks and

Fig. 1.1b shows an example of a PIC for fiber Bragg grating sensing applica-

tions which contains several photonic components such as semiconductor optical

amplifiers, photodetectors, multi-mode interferometers, multi-mode reflectors and

an arrayed waveguide grating. Just as the technological growth of electronic ICs

follows Moore’s law [6], doubling their complexity every two years, it is expected

that PICs experience a similar exponential growth in the coming decades [7] after

technology standardization is achieved (from the design to the fabrication process)

[8].

The work in this thesis aims at contributing (at a device level) to the development

of a PICs platform based on III-V semiconductors on a silicon substrate for fu-

ture compatibility with electronics. Research on both, passive and active devices

is presented, with focus on metal-cavity nanosocale light sources. The passive

components comprise a metal grating coupler and a polarization rotator.

1.2 Motivation and overview of metal-cavity light sources

The development of high-density optical interconnects with reduced energy con-

sumption have been identified as one of the major challenges in future computing

and communication systems [9]. A new generation of photonic devices, integrated

within or on top of a CMOS chip, is therefore needed, featuring unprecedented lev-

els of integration density, speed and energy efficiency [7][10]. For the light sources,

the most promising performance on silicon has been achieved by ring lasers [11][12],

based on III-V active layers bonded on silicon. These hybrid III-V/Si devices have

however a relatively large footprint (several tens of µm2) and present a power

consumption far exceeding the requirements of future interconnects, which de-

mand a total data-link system energy ∼ 100 fJ/bit and optical output device

energies on the scale of 10 fJ/bit [9]. On the other hand, in the last decade, a

new class of nanophotonic light sources has emerged which use either photonic

crystals or metallic layers to achieve strong optical confinement, eventually lead-

ing to ultralow-threshold lasing and other interesting effects, such as spontaneous

emission enhancement via the Purcell factor [13]. While low-threshold photonic

3


crystal lasers on InP have been demonstrated [14], their device footprint also re-

mains relatively large and integration on silicon has been demonstrated only very

recently [15].

Metal-cavity nanolasers [16][17][18][19] are promising in terms of ultrahigh inte-

gration density, excellent cooling properties and potentially ultra-fast modula-

tion [20]. Electrically pumped metal-coated semiconductor nanolasers were firstly

demonstrated in 2007 operating at cryogenic temperatures [16], and Fabry-Perot

plasmonic lasers were achieved shortly after [21]. The design and fabrication of

nano- and micron size cavities improved and eventually lasing at room temperature

was achieved, initially with optical pumping [17], and later under pulsed electrical

injection [22] and continuous wave operation [23]. These devices have shown lasing

threshold levels ranging from a few µA at low temperatures [16] to a few mA at

room temperature [22][23], although output power has not been reported.

For their application in PICs, such sources must be efficiently coupled to waveguide-

based components. In this respect, a few waveguide-coupling schemes have been

theoretically proposed [24][25][26], however there has been no experimental demon-

stration of practical implementations. Waveguide coupling, modulation properties,

output power and integration on silicon have not been experimentally reported.

Coupling to a waveguide is particularly challenging because it represents an ad-

ditional loss channel for the resonant cavity thereby demanding higher gain from

the active medium in order to reach lasing.

In this context, the use of nanoscale light-emitting diodes (LEDs) instead of

lasers for on-chip communication systems requiring low power consumption has

been suggested [13], and indeed submicrometer LEDs have been demonstrated

[28][29][30][27]. Unlike lasers, LEDs do not exhibit a threshold and can therefore

be efficient at low current injection levels, do not require low-loss cavities and are

therefore less sensitive to fabrication imperfections. Furthermore, nanoscale LEDs

can emit in a single spatial mode as nanolasers, and theoretical studies suggest

that those with significant Purcell-enhanced spontaneous emission can reach mod-

ulation bandwidths as high as lasers [18]. Despite the progress in the field, the

reported devices are not practical for photonic integration because they lack a

low-loss output waveguide and show low external quantum efficiencies in the order

of 10−7, for the case of a plasmonic nano-LED [27].

Figure 1.2 shows a timeline of the progress in the field of metal-cavity light sources.

It shows only demonstrated devices that are relevant for photonic integration to

the eyes of the author of this thesis, i.e. devices based on semiconductors with

4


Figure 1.2: Timeline showing scanning electron microscope (SEM) photosof metal-cavity light sources relevant for photonic integration. Most of thestructures are shown before the deposition of the metal layers. (a) First metal-cavity laser operating under electrical injection at cryogenic temperatures in anintermediate regime between dielectric and plasmonic confinement [16]. Thescale bar represents 103 nm. (b) Electrically pumped Fabry-Perot plasmoniclaser operating at cryogenic temperatures. The scale bar represents 1 µm [21].(c) Metallo-dielectric laser working at room-temperature under optical pumping[17]. (d) Metal-cavity laser with continous wave operation at room-temperatureunder electrical injection [23]. (e) Electrically driven plasmonic nano-LED cou-pled to plasmonic waveguide [27]. The scale bar represents 200 nm. (f) Metallo-dielectric nanopillar LED coupled to a low-loss waveguide and integrated on a

silicon substrate [this thesis].

a diode structure to allow electrical pumping (except for Fig. 1.2c). The time-

line considers both, lasers and LEDs, since there are not fundamental issues that

may prevent these LED structures to turn into lasers in the near future after

design and fabrication improvements. Additional efforts in the field have been

done in pedestal nanolasers [18], in which an undercut is fabricated around the

active medium in order to increase the cavity quality factor, and metallo-dielectric

cavities using Al2O3 [31] whose thermal conductivity is expected to improve their

performance by efficiently dissipating the heat through its cladding.

In this thesis work, research on metal-based nanocavity light sources is presented.

While the initial aim was to demonstrate waveguide-coupled nanolasers integrated

on a silicon substrate, our fabricated devices showed increased cavity losses which

5


prevented lasing operation. However waveguide coupling was achieved. Thus, we

demonstrated the first nanopillar LED device on a silicon substrate working at

telecommunications wavelengths (1.55 µm), coupled to an InP-membrane waveg-

uide (shown in Fig. 1.2f). The device showed nW (tens of nW) measured output

power at ∼ 100 µA (∼ 10 µA) current injection levels, at room and low temper-

ature, respectively, as well as on-chip external quantum efficiency (EQE) ranging

from 10−4 to 10−2 at room temperature and 9.5 K, respectively, achieved at ultra-

low pumping current level (µA). This exceeds the performance of any waveguide-

coupled nanoscale light source integrated on silicon in this current range. Dynamic

characterization measurements revealed sub-nanosecond electro-optical response

and it was confirmed that such fast modulation is possible due to a strong non-

radiative recombination. The reported data shows the potential of metal-cavity

nanopillar LEDs for efficient low-power interconnects operating at Gb/s data rates.

1.3 Thesis outline

The research work reported in this thesis is divided in the five chapters described

below. Conclusions are presented at the end of each chapter.

Chapter 1 provides an introduction to photonics and highlights its importance

in modern society. It also introduces the concept of photonic integration

and describes the state-of-the-art in the field of metal-cavity nanoscale light

sources.

Chapter 2 describes the fundamentals of membrane-based integrated photonics

with focus on the indium-phosphide membranes on silicon (IMOS) photonic

platform. It also provides an insight to mature silicon photonics integrated

platforms.

Chapter 3 presents the core research of this thesis. It discusses the design of

metal-cavity nanoscale lasers and LEDs in the IMOS approach, concerning

electrical, thermal and optical modeling, with emphasis on the latest. The

fabrication technology for their realization is presented in detail, consisting

of the bonding of III-V on Si-wafers, the fabrication of passive elements (e.g.

waveguides), the fabrication of pedestal nanopillars, the deposition and treat-

ment of silver, the development of silver-based ohmic contacts, and surface

passivation studies. Finally, the characterization of the nanopillar devices is

presented and discussed. Although lasing threshold was not reached, clear

6


LED operation was observed with high coupling efficiency to the waveguide.

Direct modulation experiments are also described.

Chapter 4 reports the modeling and experimental results of a new kind of metal

grating coupled which incorporates both, a metal grating and a metal reflec-

tor to decouple optically the waveguide layer from the substrate. This results

in a highly efficient chip-to-fiber coupling and a design that is independent

from the buffer/substrate layer stack. Experimental results are also briefly

presented.

Chapter 5 describes the design of a polarization rotator for membrane photonics

based on single-mode single-polarization waveguides achieved by mode cut-

off. It carries out the polarization rotation in an adiabatic transition and

shows overall ultra-high theoretical performance. Furthermore, the device

fabrication is relatively simple and is highly tolerant to fabrication errors.

Chapter 6 briefly summarizes the main results of the research reported in this

thesis and mentions possible routes to explore in the near future as a contin-

uation of this work.

7

Chapter 2

Membrane integrated photonics

This chapter presents an introduction to the concept of membrane-based photonic

circuits. It will shortly discuss the main approaches: silicon photonics (SiPh), the

heterogeneous integration of III-V and silicon photonics, and InP membranes on

silicon (IMOS), with special focus on the latest.

2.1 Introduction to membrane photonics

Membrane photonic circuits are developed and investigated by a number of aca-

demic and industrial research groups around the globe interested in the high den-

sity integration of photonic and CMOS electronic circuits. Here, the word ”mem-

brane” has to be understood as a thin photonic semiconductor layer stack on top

of a low-index substrate, usually a silicon substrate covered with a silicon dioxide

layer. The photonic layer could then be driven by the underlying electronics layer

in order to carry out the routing of light signals. Using this layer stack, high-

density photonic integration and intimate electronic/photonic integration will be

possible, which is expected to be a solution for the current bandwidth bottleneck

in electrical interconnects.

Figure 2.1 shows a schematic of a photonic membrane on a CMOS circuit. As can

be observed, metal connections run between the electronic and photonic circuitry

across planarization and thermal isolation layers, which either drive the active pho-

tonic components (e.g. lasers and modulators) or carry electrical signals generated

from incident light signals (e.g. in photodetectors).

9

Chapter 2. Membrane integrated photonics

Figure 2.1: Representation of a photonic membrane circuit on top of an elec-tronics plane. The top thick layer is meant to be a heat sink metal layer. Image

courtesy of D. Heiss.

A few alternatives to carry out the photonic-electronic integration have been re-

searched during the last years, which are based either on fully CMOS compatible

processes (as in the case of SiPh) or in the bonding of additional semiconductor

III-V layer stacks on silicon substrates to provide active functionality. The main

bonding techniques used for this purpose are molecular bonding and adhesive

bonding with Benzocyclobutene (BCB) [12].

In molecular bonding, typically a SiO2 layer is deposited on the two wafers to

bond (Si and III-V) and then chemical mechanical polishing (CMP) is used to

get sub-nanometer flatness which allows for van der Waals interaction [12]. This

technique requires extreme surface quality control and high temperature processing

steps. Its use in dissimilar layer stacks represents a challenge due to the thermal

expansion coefficient mismatch, however, low temperature (< 400 C or even room

temperature) bonding has been proposed to address this problem [12].

Adhesive bonding with BCB is more tolerant to surface defects and, as a key

advantage, it also provides a better thermal isolation between the two circuit planes

since it has a lower thermal conductiviy than SiO2. The drawbacks of using BCB

are that its thickness can be well controlled only when using thin layers (e.g. 50

nm) and also out-gassing problems can arise during high-temperature processes

if air is trapped between the surfaces during the bonding process as discussed in

section 3.2.1.

Nowadays the most popular approaches to implement membrane photonic cir-

cuits are: SiPh, the heterogeneous integration of III-V and silicon photonics, and

10


IMOS. Here, one should understand ”heterogeneous integration” as the wafer-

scale processing of wafers which contain photonic layer stacks of different material

systems on top of each other. Figure 2.2 shows a representative cross section of

each photonic platform. In every case a thin semiconductor layer stack acts as

the waveguiding layer and a low-index bottom buffer layer as optical substrate

which can be SiO2, BCB or a combination of them. Such a layer stack system

(semiconductor-dielectric) has a high refractive index contrast and therefore en-

ables sub-micrometer waveguide structures. The devices and circuits described in

this thesis correspond to the IMOS approach.

In SiPh, the top layer of a silicon on insulator (SOI) wafer is processed to fabricate

mostly passive components. A popular technique to create SOI wafers consists of

the following stages [32]: (1) two initial silicon wafers A and B are available, (2)

oxidation of wafer A to create the SiO2 layer, (3) H+-ion implantation induces

the formation of an in-depth weakened layer, (4) cleaning and bonding of wafer

A onto the non-oxidized wafer B, (5) cleavage at the weakened layer splits off

wafer A, and finally (6) wafer B undergoes annealing and polishing. The passive

functionality of SiPh is consolidated and current efforts focus on the inclusion of

active components [33], which are not straightforward because silicon is an indirect-

bandgap material (i.e. it cannot provide gain at 1.55 µm). The key advantages of

SiPh are the available high-quality large wafers (300 mm), the use of silicon fabs

and the potential low cost [34].

The heterogeneous integration targets the use of SiPh for passives and III-V for

active components. In this approach, a III-V die is bonded onto a SiPh wafer

containing passive waveguide components [12]. In this III-V die the active de-

vices are fabricated. The generated/detected light is coupled between the silicon

waveguides and the III-V layer by means of horizontal tapers to push the photonic

modes down or up.

Figure 2.2: Photonic membrane approaches. Representative cross sectionsof both, active and passive waveguide structures are presented. (a) SiPh. (c)

Heterogeneous integration of III-V and Si-photonics. (b) IMOS.

11


The IMOS platform aims for the use of only III-V for the photonics plane and

silicon for the electronics plane [35]. Both active and passive components are fab-

ricated in III-V semiconductors, thereby avoiding problems of inefficient coupling

structures and reducing the fabrication complexity (since all photonic processing

is done on a single layer stack).

Among these photonic membrane approaches, SiPh has received most of the at-

tention, investments and R&D effort world-wide. In fact, this technology is often

included in the roadmap of the major electronic companies and it is expected to

be the natural choice for large volume production due to the available CMOS pro-

cessing infrastructure. Furthermore, SiPh has reached the maturity level to offer

Multi-Project Wafer (MPW) runs [36], which in the case of compound semiconduc-

tors is only possible in pure generic III-V platforms at the moment [37]. The MPW

runs allow the fabrication of multiple photonic chips for different users (often for

different applications) on the same wafer in order to share the production costs

among the users. Such MPW’s are not new, and indeed it was the approach fol-

lowed by the integrated electronics industry at an early stage to promote standard

design rules and share processing costs [34].

Currently, membrane-based photonic MPW’s have an extensive portfolio of passive

building blocks including some active components as well (e.g. photodiodes and

modulators [36]). Although light sources are not yet offered, significant progress

has been achieved during the last years, including the recent demonstration of

InP distributed feedback lasers integrated monolithically (by epitaxial growth) on

silicon working under optical pumping at room temperature [38]. Other promising

approaches to integrate lasers on SiPh focus on Ge-based lasers [11] and the trans-

fer printing of III-V layers on Si [39]. Despite the significant efforts, a scheme to

incorporate reliably efficient lasers with electrical injection on SiPh using a CMOS

production line remains a challenge.

2.2 Existing IMOS building blocks

The IMOS platform is in continuous development, and different components have

been designed, fabricated and characterized. Table 2.1 summarizes the compo-

nents that have been experimentally demonstrated and their corresponding per-

formance figures. As a reference, the values in parenthesis correspond to the

performance of the building blocks available in a state-of-the-art SiPh integrated

platform.

12


Component Characteristics Reference

Straight waveguide Loss = 2.5 dB/cm (*2.5 dB/cm) [40]

Multi-mode interfer-ence coupler (1x2)

Insertion loss = 0.6 dB (*< 0.5 dB) [41]

Ring resonator(7 µm radius)

Quality factor = 15500Coupling constant = 7 · 10−3

[35]

Dielectric gratingcoupler

Insertion loss of 6.6 dB (*2.5 dB) [This thesis]

Metal gratingcoupler

Insertion loss = 2.7 dB3-dB bandwidth = 61 nm

[42]

Polarizationconverter

Length = 10 µmConversion efficiency > 99%Insertion loss < 1.2 dB

[43]

Wavelengthdemultiplexer(Echelle grating)

Device footprint = 0.25 mm2

Interchannel cross-talk = 18 dBInsertion loss = 1.8 dB

[44]

Wavelengthdemultiplexer(AWG)

Device footprint = 0.2 mm2

Interchannel cross-talk = 10 dBInsertion loss = 10 dB

[Internal report]

Laser source(pulsed)

Output power = 1 mWThreshold current = 100 mAElectro-optical efficiency = 3%

[45]

Photodiode DC responsivity > 0.7 A/W (*0.6 A/W)3-dB bandwidth > 67 GHz (*> 50 GHz)Dark current < 200 nA (*< 50 nA)Open eye at 54 Gb/s

[Internal report]

NanoLED source Cavity footprint <1 µm2

Output power = 60 nWOn-chip ext. quantum efficiency ∼ 1%

[This thesis]

Table 2.1: Demonstrated IMOS components and their corresponding perfor-mance at 1550 nm. The values in parenthesis (*) correspond to the specificationsof IMEC’s platform (ISIPP25G) [46]. Disclaimer: the IMOS building blockscorrespond to the best experimental results whereas the SiPh building blocksare mature and commercially available through MPW’s based on 200 mm SOIwafers. Moreover, IMEC’s platform contains more building blocks than theones mentioned in this table, including modulators, waveguide crossings and

polarization splitting grating couplers [46].

2.3 Conclusions

Membrane photonic circuits represent a promising approach to merge electronic

and photonic integrated technologies. At TU/e, the IMOS platform has been

proposed and much progress has been demonstrated.

13


Although most of the achieved devices are passive components, recent efforts (in-

cluding this thesis work) have resulted in the demonstration of active components

such as lasers, photodiodes and nanoLEDs. The integration of these devices is

paving the way for the consolidation of the IMOS platform and is expected to

allow complex membrane ASPICs in the near future.

14

Chapter 3

Nanoscale light sources

This chapter presents the development of nanoscale light sources (lasers and LEDs)

with metal-cavity coupled to a membrane-based waveguide. In section 3.1, the

device design is discussed from an electrical, thermal and optical point of view

with special emphasis on the optical properties. The fabrication technology that

was developed is described in detail in section 3.3. Finally, the characterization

results of the fabricated metal-cavity nanopillar LEDs are presented and discussed

in section 3.3.

3.1 Design of metal-cavity light sources

We designed two types of laser cavities based on the same layer stack: a plas-

monic Fabry-Perot cavity and a metallo-dielectric nanocavity, described in sec-

tions 3.1.1.1 and 3.1.1.2, respectively. They have been reported in [25, 47–49].

Since lasing was not observed in the fabricated devices, section 3.1.1.3 shows mod-

eling results of the metallo-dielectric nanocavity operating as a LED. Sections

3.1.2, 3.1.3 and 3.1.4 report the electrical and thermal modeling, as well as the

small-signal frequency response, respectively.

3.1.1 Optical modeling of waveguide-coupled metal cavities

3.1.1.1 Plasmonic Fabry-Perot laser cavity

Basic principles of a Fabry-Perot cavity Lasers with Fabry-Perot (FP) cavity

are among the most common type of lasers due to their relatively simple design and

15

Chapter 3. Nanoscale light sources

fabrication tolerance. In general, a Fabry-Perot (FP) laser cavity is formed by two

mirrors with an active medium in between as shown in Fig. 3.1, which typically

supports many longitudinal modes. A pumping mechanism provides energy to

the active medium, which excites the cavity modes initially through spontaneous

emission of light. As soon as the active medium has enough optical gain, the

mode(s) with lowest loss reach lasing. In general, both gain and mode loss are

wavelength dependent. A FP-cavity is characterized then, by two mirrors with

reflectivities R1/2, a mode propagation loss α, and the confinement factor of the

mode Γ within the active medium. In semiconductor lasers, the pumping is done

by current injection and single-mode operation is desired which is achieved by

introducing additional spectral filters (e.g. gratings).

Figure 3.1: Schematic of a Fabry-Perot cavity laser with length L, and reflec-tivity R1 = 100%, R2 < 100%. The active medium is represented in blue andthe generated light in red. For easy visualization, the power levels P1 and P2

are displayed in different locations, however they should be considered to be inthe same spatial point.

The reflectivity determines the amount of incident power that reflects back into

the same optical mode upon interaction with the mirror. In general, a mirror pro-

duces also light scattering and therefore the reflectivity calculation often requires

2D or 3D finite-difference time-domain (FDTD) simulations and an overlap inte-

gral with the mode of interest. For cases with a simple waveguide geometry an

approximation can be made with the Fresnel equations. For example, in photonic

integration the reflectivity of a deep etched (or cleaved) waveguide can be esti-

mated as R = [∆n/(neff + nout)]2 for the case of perpendicular incidence, where

neff and nout are the mode effective refractive index and refractive index of the

outside medium (typically air) and ∆n is their difference.

Since the waveguide structure in FP-type cavities does not vary in the longitudi-

nal direction, a transveral waveguide cross section is enough to study the cavity

modes by means of a two-dimensional mode solver. For such a cross section we

calculate the complex propagation constant β = β + j2α, where β = neff

2πλ

with

16


effective refractive index of the mode neff , and mode propagation loss α. The fac-

tor 12

comes from the fact that the propagation constant is related to the electric

field amplitude, whereas the propagation loss is related to the optical power. The

propagation loss α determines the attenuation of the power along the propagation

direction (e.g. z) according to P2 = P1e−α[m−1]z, where by substituting the defini-

tion αdB[dB/m] = 1L

10 log P1

P2, it leads to the relation αdB[dB/cm] = 4.34α[cm−1]

which is often useful.

The spatial overlap of the optical mode with the active medium is another impor-

tant aspect, since it determines the net gain experienced by the mode, i.e. the

modal gain gm = Γg, where g is the material gain and Γ is the confinement factor

defined as

Γxy =

∫∫active

|E|2dxdy∫∫total

|E|2dxdy, (3.1)

where E is the full vectorial electric field. All these laser cavity parameters are

considered in the so-called laser threshold condition, which can be derived from

an analysis of the FP cavity. If a mode experiences enough gain so that its power

does not change after one round trip (see Fig. 3.1), then

P2

P1

= R1R2e−2αLe−2gmL = 1, (3.2)

from which it follows that

gΓ = α +1

2Lln

1

R1R2

, (3.3)

sometimes written as gthΓ = α + αm, where α stands for propagation loss, αm

for mirror loss and gth is the threshold material gain required to reach lasing.

In this context, the optical efficiency (fraction of power generated that leaves

the cavity) is defined as ηo = αm/(αm + α). Moreover, assuming unit internal

quantum efficiency (fraction of current above threshold which results in stimulated

emission), the differential quantum efficiency can be calculated as ηd = ηcηo where

ηc is the coupling efficiency (fraction of power leaving the cavity that couples to the

waveguide). The differential quantum efficiency is then the number of photons out

per injected electron in a measured power-current characteristics above threshold.

17


These relations are used later to calculate the performance of the FP plasmonic

laser.

Plasmonic cavity structure In this section, the design of a Fabry-Perot plas-

monic cavity on a III-V membrane bonded to a silicon substrate is presented,

which we reported in [47].

The laser structure proposed is shown in Fig. 3.2, which is based on previously

reported plasmonic lasers [50]. However, it is different in the fact that it is coupled

to a waveguide. The laser consists fundamentally of a MISIM (metal-insulator-

semiconductor-insulator-metal) waveguide forming a Fabry Perot resonator cou-

pled to an InP-membrane waveguide. The top n-contact and the lateral p-type

contact provide the electrical pumping for the InGaAs active medium, which has

a bandgap of 1.65µm.

Figure 3.2: (a) Schematic of a Fabry-Perot plasmonic laser in a III-V mem-brane on silicon. The material refractive index at 1.55 µm is shown in parenthe-sis. The index value for InGaAs corresponds to the transparency regime. (b)Longitudinal cross section showing a standing wave inside the FP cavity and

waveguide coupling.

As it can be seen in Fig. 3.2, there is a thin insulating layer of Si3N4 between

the semiconductor layer stack and the metal cladding, which serves to insulate

electrically the structure horizontally and therefore allow a top-down current flow.

The quaternary InGaAsP (Q1.25) acts as the ohmic contact layer for the p-contact.

The back side of the Fabry Perot cavity is completely terminated by metal to

achieve a strong reflection, whereas the front side has an open facet to avoid

metals on the output waveguide which otherwise would allow the excitation of

plasmonic modes.

Modal properties Figure 3.3a depicts the MISIM cross section with the main

parameters and Fig. 3.3b shows the spatial mode distribution of the hybrid surface

18


plasmon polariton mode with lowest loss in the cavity. Due to its plasmonic nature,

it has a dominant Ex component and is mainly confined within the insulation layer,

which leads to a very poor overlap with the active region.

Figure 3.3: (a) MISIM cross section showing the main design parameters. (b)Mode distribution |E|2 of the hybrid plasmonic mode at 1.55µm with lowestoptical loss. Blue: low intensity. Red: high intensity. (c) Normalized one-dimensional intensity distribution across the waveguide core (i.e. along thedashed line in Fig. 3.3b) for different thicknesses td of the insulation layer and

h = 0 nm.

The propagation loss and confinement factor as a function of the width w, can be

observed in Fig. 3.4 for different values of the post height h. This loss is due to the

absorption by the metal and it decreases when the core is wider because the mode

is increasingly confined in the InGaAs core, i.e. the confinement factor increases

as shown in Fig. 3.4b. The confinement factor is defined as described in Eq. 3.1

and determines the modal gain. It is low because the plasmonic mode has its

maximum amplitude at the interface between the metal and the dielectric as seen

in 3.3a. A higher overlap of the mode with the InGaAs core could be achieved

by decreasing the insulation thickness td [51] as shown in the one-dimensional

intensity distribution across the ridge in Fig. 3.3c, although the modal loss would

also increase.

The post height h has also an influence on the loss and the confinement factor due

to the presence of the BCB substrate, which has a lower refractive index than the

semiconductors and therefore contributes to the vertical confinement in the active

region. For this reason, when the post height is increased, the mode extends more

in the InP post and both the modal loss and the confinement in the active region

decrease.

19


Figure 3.4: (a) Mode propagation loss at 1.55µm due to metal absorption.(b) Confinement factor at 1.55µm as a function of the MISIM width w.

As can be seen in Fig. 3.4, a wider MISIM structure leads to a lower propagation

loss and a higher confinement factor, however dielectric modes with better charac-

teristics could exist for such wide structures and then the lasing mode would not

be plasmonic anymore. Since the target of this device design is to minimize the

cavity size while operating with a plasmonic mode at room temperature, we have

fixed the width to 200 nm. This is near the minimum width for this structure to

be able to operate at room temperature (i.e. reach threshold gain under electrical

injection), although using long cavities exceeding several tens of micrometers are

necessary as discussed below.

Facet reflectivity and waveguide coupling The aforementioned structure was

investigated with 3D FDTD simulations in order to calculate the reflectivity of

its facets and the outcoupling to the dielectric waveguide. As observed in Fig.

3.2, the laser is coupled to an IMOS waveguide with a thin InGaAsP layer on

top. Although the laser core and the waveguide are not vertically aligned, a

significant coupling has been obtained. This is possible due to the quaternary

layer, which has an intermediate refractive index between InGaAs (laser core)

and InP (waveguide), which allows a vertically extended laser mode that overlaps

better with the waveguide mode.

Figure 3.5 shows the reflectivity and coupling coefficient as a function of wave-

length for an open facet and different post heights. The first remarkable charac-

teristic is the high reflectivity of around 0.6 at 1.55µm, which is twice the Fresnel

reflectivity for an interface given by InGaAs and air, considering perpendicular

incidence. This strong reflection is due to the large effective mode index mismatch

20


between surface plasmon modes and free propagating modes in air, and it was

found to increase even more for very narrow MISIM structures. In addition, the

reflectivity at the backside of the laser cavity is also shown, which is assumed to

have a metal coating. It is high (R = 98%) and hardly varies for different values

of h.

Figure 3.5: (a) Reflectivity of the FP cavity with w = 200 nm, for an openfacet with post height h and a metal termination with h = 0 nm. (b) Couplingefficiency of the FP cavity with the waveguide for w = 200nm showing ηc = 0.46

at 1.55 µm for h = 100 nm.

The reflectivity shows only a small increase on increasing the post height because

it is determined mainly by the reflection between the MISIM structure and air.

On the other hand, the coupling efficiency continuously decreases and becomes

zero for a very high post. Since our main interest is to design an efficient laser, a

post height of 100 nm would be suitable since it maximizes the reflectivity without

compromising the coupling efficiency. Using a higher post would lead to a decrease

in the laser efficiency. As shown in the 2D longitudinal cross section of Fig. 3.2b,

the high reflectivity creates a standing wave inside the cavity, whereas a fraction

of the transmitted power is coupled into the dielectric waveguide.

Finally, Fig. 3.6 shows the threshold gain as a function of cavity length (i.e.

varying mirror loss) and the corresponding differential quantum efficiency. For

L = 50 µm, gth = 1796 cm−1 with only 5% efficiency, which may be achieved at

room temperature under a high injected carrier density above 6 · 1018 cm−1 [52].

Alternatively, a threshold gain of gth = 2580 cm−1 and an efficiency of 17% are

obtained when considering L = 10 µm. However such a high material gain is

expected to be achievable only at cryogenic temperatures. Higher material gain

can be obtained at lower temperatures due to a higher density of available state

pairs for radiative recombination.

21


Figure 3.6: (Left axis): Threshold gain calculated with Eq. 3.3 for a cavitywith w = 200 nm and h = 100 nm (i.e. α = 0.16dB/µm, Γ = 0.23, R1 = 0.65,R2 = 0.98, ηc = 0.46). (Right axis): Differential quantum efficiency calculated

as ηd = ηcηo.

3.1.1.2 Metallo-dielectric laser nanocavity

Basic principles of a nanocavity Differently from FP cavities, in a 3D-confined

nanocavity no separation between longitudinal and transverse directions can be

made, which leads to modes with all three polarization components. Also, the

losses cannot be expressed in terms of loss rates per unit length, but rather in the

more general form of loss per unit time (or quality factor). Examples of nanocav-

ities are photonic crystal cavities and metallo-dielectric nanocavities which are

addressed in this thesis work. In such a cavity, the metallic cladding provides high

reflectivity but also contributes to the intra-cavity losses due to metal absorption.

A cross section of a metallo-dielectric cavity is shown in Fig. 3.7.

In view of the complexity of a nanocavity, 3D FDTD simulations are usually em-

ployed to obtained its relevant properties, such as cavity quality factor Q, resonant

wavelength λres, confinement factor Γ, and group index ng.

A general method to obtain the cavity quality factor consists in placing a number

of electric or magnetic dipoles inside the cavity with random positions, orienta-

tions (i.e. polarization), and phase, which excite all possible modes inside the

cavity. During the time-domain simulation, the electromagnetic field is moni-

tored in a number of points within the cavity. After the simulation finishes the

time-dependent electromagnetic field is Fourier-transformed and the resonance

22


Figure 3.7: Schematic representation of the different power loss rates. γmetalis the loss due to metal absorption, γsub is the loss due to radiation to the

substrate and γwg is the loss due to coupling to the waveguide.

frequencies fres with FWHM ∆f are found, from which the quality factor is calcu-

lated as Q = fres/∆f . Since a metallo-dielectric nanocavity has two loss channels

(absorption and radiation), the quality factor can be written as Q−1 = Q−1abs+Q

−1rad.

Furthermore, the total power loss rate in this nanopillar laser can be written

as the sum of the loss into the metal, the radiation into the substrate and the

useful coupling to the waveguide as γtotal = γmetal + γsub + γwg, which is depicted

in Fig. 3.7. The optical efficiency for this cavity can be calculated as ηo =

(γsub + γwg)/γtotal, and the coupling efficiency as ηc = γwg/(γsub + γwg). Therefore,

monitoring the power (energy flux) that leaves the cavity, the one that couples

into the waveguide mode, and the power absorbed by the metal, allows for a

determination of such efficiencies. The power escaping from the cavity and the

one coupled to the waveguide can be obtained in a time-domain simulation by

placing proper monitors at the respective interfaces, whereas the power absorbed

can be calculated after FDTD simulations as Pabs = −0.5ω|E|2Imε [53].

The confinement factor Γxyz can be calculated similar to Eq. 3.1 by integrating over

the three dimensions, and the group index can be either calculated with a mode

solver or approximated as the effective refractive index neff in a low-dispersion

structure. Finally, the threshold gain of a mode at resonant wavelength λres is

given by [52]

gth =2πng

λresQΓxyz. (3.4)

23


Metallo-dielectric cavity structure The combination of dielectric and metallic

layers can lead to strong optical confinement with relatively low loss, and has been

used to demonstrate metallo-dielectric nanolasers operating at room-temperature

under optical pumping [17]. However, efficient coupling to a waveguide has not

been demonstrated yet. In Ref. [24] the coupling of a III-V metallo-dielectric

nanopillar laser to a Si/SiO2 waveguide was theoretically proposed. In this section,

a metallo-dielectric cavity laser coupled to a waveguide on a III-V membrane

bonded with BCB to silicon is described and studied by means of 3D FDTD

simulations which were reported in [25, 48, 49].

Figure 3.8: Schematic of the metallo-dielectric laser coupled to an InP-membrane waveguide. The refractive index of each material at 1.55 µm isshown in parenthesis. The inset shows |E|2 in logarithmic scale where the

cavity-waveguide coupling is visible.

The proposed laser structure is shown in Fig. 3.8. The laser consists of a semicon-

ductor nanopillar that stands on top of a thin InP waveguide and is covered with a

SiO2 layer which insulates it from a metallic cladding. A lateral p-contact is electri-

cally connected to the pillar through a highly doped quaternary (InGaAsP) layer.

The metallic cladding acts itself as the n-contact allowing a top-down current flow.

For simplicity, Fig. 3.8 does not show the ohmic contact layers Au/Pt/Ti, however

they were included in the simulation model in order to account for their optical

loss.

Figure 3.9a shows the geometry of the cavity. The influence of the thickness t of

the SiO2 insulation layer, the undercut s and the bottom post cladding h on the Q-

factor and optical efficiency of the laser are discussed below. The undercut, which

is the same in both x and z directions, was introduced to improve the quality factor

while maintaining a short bottom post. The simulated quality factor of the cavity

is shown in Fig. 3.9b as a function of the undercut and bottom post height. A

24


Figure 3.9: (a) Schematic of the cavity with dimensions in nanometers. Theoptimization parameters are shown in blue. (b) Q-factor as a function of un-dercut and post height for an insulation thickness t = 175 nm. The inset showsthe modulus squared of the electric field distribution of the TE polarized mode

in the xy plane across the center of the cavity.

higher bottom post leads to an increased Q-factor due to a reduced coupling with

the waveguide. Likewise, a larger undercut improves the Q-factor. An undercut

of 60 nm was considered as a good compromise between high Q-factor and ease of

fabrication. Therefore, such value was chosen for the simulation results presented

in the following.

Quality factor and waveguide coupling For the optimization of the insulation

thickness t of the cavity, a bottom post height h = 700 nm was initially chosen,

since it provides sufficient isolation from the waveguide as suggested in Fig. 3.9b.

Additionally, the pillar laser is considered to be symmetric along x and z (i.e.

the length and width of the active medium are both 300 nm). As observed in

3.10a, there is an optimum insulation thickness (t = 175 nm) where the Q-factor

is maximum. This is because metal losses are high for a thin insulation layer,

whereas the radiation losses increase for a thick insulation because the metallic

confinement decreases and the cavity mode overlaps more with radiation modes.

The resonant wavelength for this optimum thickness is around 1.4 µm, nevertheless

it will be shown later that this can be increased to 1.55 µm by changing the aspect

ratio of the cavity.

Once the optimum insulating thickness has been found, the bottom post height is

varied in order to increase the coupling to the waveguide. The optical efficiency

of the cavity was defined earlier as ηo = (γsub + γwg)/γtotal (ratio of total radiated

25


Figure 3.10: (a) Q-factor and resonant wavelength as a function of the in-sulation thickness, with h = 700 nm. (b) Q-factor and optical efficiency for a

varying bottom post height, with t = 175 nm.

power to total lost power) and is shown in 3.10b as a function of post height. As

expected, the optical efficiency increases at the expense of a drop in Q-factor when

decreasing the post height.

A symmetric cavity has been considered until now, however the coupling efficiency

can be increased by breaking the symmetry of the cavity, which can be realised by

elongating the cavity in the z-direction [24]. The coupling efficiency ηc is defined

as the ratio between the power coupled to the waveguide (in both directions) and

the total radiated power, ηc = γwg/(γsub + γwg). Figure 3.11 shows an increase

in coupling efficiency and resonant wavelength for an elongated cavity, while the

Q-factor decreases. The desired lasing wavelength of 1.55 µm is achieved for a

cavity length of 400 nm.

Figure 3.11: (a) Q-factor and resonant wavelength for an elongated cavitywith active medium length lz. (b) Improved optical and coupling efficiencies by

elongating the cavity. Both graphs use t = 175 nm and h = 400 nm.

26


Differential efficiency and threshold conditions In Fig. 3.12, the differential

quantum efficiency, defined earlier as ηd = ηoηc, and threshold gain are shown for

the case of the asymmetric cavity. The threshold gain is calculated with Eq. 3.4.

According to our simulations, the confinement factor in the cavity varies linearly

from 0.4 to 0.28 for a cavity length varying from 300 nm to 450 nm.

In summary, the optimized cavity has an insulation thickness of 175 nm to maxi-

mize the Q-factor, a bottom post height of 400 nm and a cavity length of 400 nm

for a resonant wavelength of 1.55 µm, which results in a relatively low threshold

gain of 815 cm−1 and a differential quantum efficiency of 0.16.

Figure 3.12: Lasing threshold gain and differential quantum efficiency as afunction of the cavity length.

3.1.1.3 Metallo-dielectric light-emitting diode cavity

We also performed the modeling of a light-emitting diode similar to the struc-

ture reported in the previous section but with no undercut and assuming only

spontaneous emission, in order to interpret the characterization results reported

in section 3.3.

The schematic structure is shown in Fig. 3.13a and the fabricated device in Fig.

3.29. The pillar is covered with a SiO2 layer and then encapsulated with a silver

cladding to form a metallo-dielectric cavity which supports an optical mode around

1.55 µm and suppresses a large fraction of the remaining radiation modes around

the InGaAs active region, thereby making the spontaneous emission coupling to the

mode more efficient. Fig. 3.14a shows the distribution of a mode with dominant in-

plane polarization (i.e. parallel to the substrate) that is well confined in the active

medium and has high coupling efficiency with the fundamental quasi-transverse-

electric (TE) mode of the waveguide shown in Fig. 3.14c, as discussed later. The

27


cavity modes in the nanopillar are evanescently coupled to a 450 nm wide InP

waveguide, which is connected to a grating coupler for characterization purposes.

The design of this grating coupler is reported in section 4.2 (Fig. 4.5).

Figure 3.13: (a) Schematic representation of the nanopillar LED device cou-pled to an InP-waveguide, on a silicon substrate. The layer stack from top tobottom is: n-InGaAs(100nm) /n-InP(350nm) /InGaAs(350nm) /p-InP(600nm)/p-InGaAsP(200nm) /InP(250nm) /SiO2 /BCB /SiO2 /Si. (b) Calculatedspontaneous emission power coupled into the fundamental TE mode of thewaveguide, as a function of the dipole wavelength. The plot shows the caseof an ideal cavity (i.e. low loss silver and vertical sidewalls), a nonideal cav-ity (i.e. lossy silver and 2 sloped sidewalls due to etching), and a nanopillar

without metal cladding and vertical sidewalls.

We used three-dimensional FDTD simulations for the optical design of the metal-

cavity nanopillars LEDs. In order to calculate the spontaneous emission power

coupled into the waveguide, we performed a series of 135 simulations, where we

vary the position and orientation (x, y, z) of a single electric dipole emitting in the

range 1.4 - 1.6 µm within the active medium InGaAs, and monitor the power cou-

pled to the fundamental TE mode of the waveguide. As the cavity is symmetric

along and perpendicular to the waveguide (i.e. in x and y directions) we only place

dipoles in one quarter of the structure on a grid of 80 nm at 3 positions in x,y direc-

tion and 5 positions in z direction. The spontaneous emission coupling efficiency

was calculated by normalizing the power coupled into the waveguide fundamental

mode to the total power emitted by a dipole placed inside the nanopillar cavity.

The calculated waveguide-coupled emission was averaged over all positions and

dipole orientations to account for the average non-coherent emission describing

the spontaneous emission.

We first considered an ideal cavity with a 1.5 µm high nanopillar with horizontal

cross section of 325×325 nm2, vertical sidewalls, 175 nm thick SiO2 cladding, and

encapsulated in silver. Refractive index values for Ag were taken from [54]. The

28


average power emitted by dipoles in the InGaAs region and coupled to the waveg-

uide, normalised to the power emitted in an unstructured bulk medium, is plotted

on a logarithmic scale as a function of wavelength in Fig. 3.13b (green curve).

Clear emission resonances are present at 1568 nm and 1482 nm, corresponding to

the mode of interest with large waveguide coupling efficiency, and a vertical res-

onance (i.e. vertically resonating between the top of pillar and the bottom SiO2

buffer layer), respectively. The profiles of the electric field squared of these cavity

modes are shown in Figs. 3.14a and 3.14b.

Figure 3.14: |E|2 spatial distribution of resonant modes along the waveguidedirection for the ideal cavity case (left and middle), and waveguide fundamentalmode (right). (a) Resonance at 1568 nm with dominant in-plane polarizationand high waveguide coupling efficiency. (b) Vertical resonance at 1482 nm. (c)Quasi-TE waveguide fundamental mode excited by the cavity resonance at 1568

nm. Blue: low intensity. Red: high intensity. Silver is shown in dark blue.

We note that the emission is strongly suppressed out of resonance, while the emis-

sion rate in the modes is slightly enhanced as compared to the value in the bulk,

e.g. it can be observed in Fig. 3.13b (green curve) that the emission reaches values

larger than one even after waveguide-coupling. This is known as the Purcell effect,

which describes the enhancement of the spontaneous emission rate of an emitter

in a cavity as compared to its emission rate in an homogeneous medium [13]. The

Purcell enhancement factor was estimated by dividing the total power emitted

by a dipole in the cavity to the emitted power of a dipole in homogeneous In-

GaAs [55]. Indeed, from simulations we calculate an average Purcell enhancement

of 12 for the mode of interest at 1568 nm in the ideal case of a monochromatic

dipole, however we stress that this estimation is not realistic since homogeneous

and inhomogeneous broadening of the emitters are not taken into account in the

FDTD cavity model. When taking into account broadening effects we expect a

reduction of this factor to a value close to one, i.e. no Purcell enhancement is

29


expected in the real structure. The reduction is of about ∆λemitter/∆λcavity [56]

(when ∆λcavity ∆λemitter as in our bulk InGaAs case), for which we assume

∆λemitter ∼ 30 − 50 nm (distribution of carriers in bands convoluted with homo-

geneous broadening in the bulk active region), and a calculated ∆λcavity ∼ 2.75

nm.

The combination of moderate spontaneous emission into the mode and strong

suppression of the emission into other modes, as enabled by the metallo-dielectric

cavity, is the key approach to obtain efficient funneling of spontaneous emission

into a guided mode. For comparison, the orange curve in Fig. 3.13b reports the

power coupled into the waveguide in the case where no metal is present around

the pillar, showing very poor coupling. By dividing the power emitted in the

waveguide by the total power emitted by the dipoles (not shown), we estimate an

average ”spontaneous emission coupling efficiency” (fraction of emitted photons

which couple to the waveguide) of ∼ 15% around the resonance at λ = 1568 nm.

This would correspond to the efficiency of an ideal device where all recombination

is radiative and the emitters are all spectrally resonant with the mode. We note

however that the emission rate into the mode strongly depends on the spatial and

polarization matching between the dipole and the modal field. For dipoles well

coupled to the mode the spontaneous emission coupling efficiency is calculated

to be 29%, which we attribute mainly to the product of the optical efficiency

(fraction of generated photons in the nanopillar mode that leave the cavity) and the

waveguide coupling efficiency (fraction of photons leaving the cavity that couple to

the waveguide mode), as the spontaneous emission into other modes is negligible

at resonance. However, a part of the active volume is not well coupled to the

mode, making the coupling of spontaneous emission to the waveguide less efficient

on average.

We also considered the influence of nonidealities on the emission. Fig. 3.13b shows

the emission into the waveguide for a cavity with sloped sidewalls under a 2 angle

as observed in the fabricated structures. Additionally, we employed another widely

used Ag material model [57], which exhibits higher losses. The result is shown in

Fig. 3.13b (blue line). Due to the sloped sidewalls, the mode overlap with the InP

material increases which causes a blue shift of the peaks to 1467 nm and 1550 nm.

The power emitted in the vertical resonance is not changed significantly, while

the emission of the mode of interest around 1.55 µm drops which we attribute

mainly to a reduction of the spontaneous emission rate into the mode due to the

loss in quality factor. The corresponding average spontaneous emission coupling

efficiency is reduced to about 7% for the mode of interest.

30


3.1.2 Electrical modeling of III-V layer stack 1

Electrical simulations were carried out with a self-consistent Poisson solver in order

to obtain the gain-current characteristics of the diode and to be able to determine

the lasing threshold current. A detailed description of such simulations can be

found in [48]. Table 1 shows the semiconductor layer stack considered for the

simulations.

Material Thickness(nm) Doping (1/cm3)

n-InGaAs 50 1 · 1019

n-InP 200 5 · 1018

n-InP 100 1 · 1018

i-InGaAs 350 n.i.d.

p-InP 100 3 · 1017

p-InP 100 5 · 1017

p-InP 100 1 · 1018

p-InGaAsP 100 2.4 · 1019

Table 3.1: Semiconductor layer stack considered for electrical and thermalsimulations.

Carrier recombination via Auger (7 · 10−29 cm6s−1 [58]), radiative (0.98 · 10−10

cm3s−1 [58]) and surface recombination (5 · 104 cm s−1 [59]) are considered in the

model. The dependence of the Fermi levels in valence and conduction bands as a

function of current density can be found through simulations. This allows us to

calculate the optical material gain at a temperature of 300 K [60]. Figure 3.15

shows a typical band diagram structure of the diode. The resulting gain spectra

are presented in Fig. 3.16a for current densities ranging from 20 to 200 kA/cm2.

The current-voltage characteristics of the diode are also plotted in Fig. 3.16b. The

device has a total resistance of 2.25 kΩ. This is dominated by a combination of the

p-side (contact and sheet) resistance (∼ 400Ω), where the current is transported in

a 100 nm thick quaternary layer on top of the waveguide, the p-doped region of the

nanopillar with high resistance (∼ 1000Ω) due to the reduced cross section, and

the ohmic contact on the n-doped side of the pillar (∼ 850Ω), where we assume a

contact resistance of 1 ·10−6Ωcm2. When driving a current through the device the

high resistive regions contribute to heat generation as shown in the inset of Fig.

3.18.1The electrical modeling was carried out in cooperation with Dr. D. Heiss. V.D.-C. proposed the

layer stack and D.H. carried out the simulations.

31


Figure 3.15: Band diagram of a double-heterostructure diode with the layerstack of Table 3.1, under a forward bias voltage of 1.5 V. Electrons and holes areinjected from the n- and p-type regions, respectively. The carriers are confinedin the intrinsic InGaAs region with bandgap energy Eg = 0.748 eV (1.65 nm)

where high radiative recombination occurs.

Figure 3.16: (a) Material gain of InGaAs for different current densities. (b)Material gain at 1.55 µm and diode voltage as a function of current density and

injection current.

The material gain at 1550 nm is plotted in Fig. 3.16b as a function of the current

density. The threshold gain of 815 cm−1 found from the optical simulations is

reached with a current density of 100 kA/cm2 corresponding to a threshold current

Ith = 120 µA for the nanolaser with an active cross section of 300x400 nm2.

If no self-heating is considered, the optical output power grows linearly with the

drive current I as Pout = ηd(hc/qλ) (I − Ith) as it is plotted in Fig. 3.17. Here,

ηd = 0.16 is the differential quantum efficiency and Ith = 120 µA is the threshold

current for an emission wavelength of λ = 1.55 µm. In this linear model, an

optical output power of nearly 40 µW is reached for a current of 425 µA and

a voltage of 1.98 V corresponding to a wall plug efficiency of 4.8%. In a real

32


laser device, the self-heating produces gain clamping, thereby limiting the output

power. Additionally, since the heat dissipation in realistic packaging is limited, a

compromise between integration density and available optical power will need to

be found.

Figure 3.17: Waveguide-coupled output power of the nanopillar laser withIth = 120 µA and ηd = 0.16.

3.1.3 Thermal modeling 2

In order to get an insight in the thermal effects in the nanopillar structure induced

by Joule heating, we performed three-dimensional thermal simulations with COM-

SOL. We investigated both, an ideal case that considers densely integrated devices

with the calculated series resistance (2.25 kΩ) and active cooling from the top and

the case of a more realistic (single) device with higher resistance (up to 30 kΩ)

and no cooling. The latest case is more similar to the fabricated and characterized

devices reported in section 3.3.

Despite the sub-micrometer footprint of the nanopillar structure, a practical device

requires a much larger area for heat spreading and dissipation, which is problematic

for high density photonic integration. Figure 3.18a shows the temperature increase

in the laser core as a function of the drive current for one laser per 800, 1500, and

3000 µm2. We assumed packaging with a high performance heat sink as described

in Ref. [61] with a junction-to-ambient heat transfer coefficient of 7000 W/(m2

K) through the top. As expected, there is a quadratic dependence with current

injection. The smaller the footprint, the faster the temperature increase.

2The thermal modeling was carried out in cooperation with Dr. D. Heiss. D.H. did the simulationsof the densely integrated devices, and V.D.-C. performed the simulations of the single device with highresistance and no active cooling).

33


Figure 3.18: (a) Laser temperature increase as a function of drive current.The inset shows a colour plot of the temperature distribution in the cavity,where the highest temperature can be observed at the highly resistive p-InP atthe bottom of the pillar. White: high temperature. Orange: low temperature.(b) Laser temperature increase for different BCB thickness and pillar resistance.

Figure 3.18b shows a more realistic scenario. In this case, neither convective cool-

ing nor surface-to-ambient radiation has been assumed. Only the temperature of

the substrate bottom was fixed to 300 K. Additionally, a total high series resis-

tance across the nanopillar was assumed, ranging from 1 kΩ and 30 kΩ, and a BCB

thickness between 50 nm and 700 nm. As can be observed in Fig. 3.18b, a thick

BCB layer limits the heat conduction towards the substrate which is alleviated

when decreasing its thickness. Unfortunately, a thin BCB leads to a low yield

of the fabrication process due to the BCB outgassing reported in section 3.2.1.

Furthermore, we also varied the metal cladding thickness between 500 nm and 1.5

µm to investigate its influence on heat sinking, however no significant change in

temperature was observed.

3.1.4 Small-signal frequency response

The bandwidth of a directly modulated laser can be found from the analysis of

its response to a small-amplitude modulated signal. In general, the bandwidth

is determined by the cavity design, material properties and pumping conditions,

which determines the relaxation oscillation frequency beyond which the response

rapidly decreases. The response is described by the transfer function H(ω) of the

laser system, whose derivation assumes a sinusoidal current modulation and can

be found elsewhere (e.g. [60]) as

34


H(ω) =ω2R

ω2R − ω2 + jωγ

, (3.5)

where ωR is the so-called relaxation resonance frequency and γ is a damping fac-

tor. The relaxation resonance is directly proportional to the photon density and

inversely proportional to the photon life time.

Figure 3.19: Small signal frequency response of the metallo-dielectricnanolaser, for different values of the waveguide-coupled output power.

We calculated the small-signal frequency response with Eq. 3.5 using the param-

eters of the metallo-dielectric nanolaser listed in Appendix B. We stress that the

use of these standard expressions for the parameters governing the frequency re-

sponse is not fully justified for nanolasers (the Purcell enhancement of spontaneous

emission is for example not properly taken into account), so that the calculated

values are only indicative. Figure 3.19 shows the expected performance of the de-

signed nanolaser for a direct modulation scheme. In case of 50 µW output power,

f3dB ∼ 60 GHz. As Fig. 3.19 suggests, the relaxation oscillation shifts to higher

frequencies for higher output power levels. However in practice the maximum gain

(output power) is limited by the ability to remove the heat from the device, this

effect is known as thermal roll-off.

3.1.5 Conclusions

Two types of laser cavities with similar layer stack were designed, an FP cavity

and a metallo-dielectric nanocavity. The modeling results show that a metallo-

dielectric nanolaser offers several advantages over a plasmonic laser. An optical

mode in the metallo-dielectric cavity has a lower optical loss because it is mainly

35


confined in the semiconductor rather than in the metal, and its footprint remains

small due to the metal confinement. In order to compare their performance, their

corresponding threshold gain and differential quantum efficiency were calculated

and presented in Figs. 3.6 and 3.12. It can be concluded that the metallo-dielectric

nanolaser shows a better performance because of its relatively low lasing thresh-

old gain which can, in principle, be achieved at room temperature under current

injection, and its higher differential quantum efficiency in combination with a sub-

micrometer footprint.

We also modeled the metallo-dielectric nanocavity operating as a LED. It was

shown that operating in the spontaneous emission regime, the light source is ca-

pable of percentage-level on-chip external quantum efficiency.

Furthermore, it was also confirmed that a thin BCB bonding layer allows for a more

efficient heat dissipation to the substrate. Therefore, a thin bonding is preferred

from the design point of view, although this is technologically more challenging

as discussed later in section 3.2.1. Finally, the calculation of the small-signal

frequency response of the metallo-dielectric nanolaser indicates that an electro-

optical bandwidth of a few tens of GHz can be expected for tens of µW output

power.

3.2 Fabrication technology of metallo-dielectric nanopillar

cavities coupled to waveguides

Although novel semiconductor technology has been developed in the field of metal-

cavity nanoscale light sources, very little processing information has been revealed

through publications (mainly [16][62][63]). In this contribution we provide an ex-

tensive description of the technology developed to fabricate our waveguide-coupled

metal-cavity III-V nanopillars on silicon substrates and highlight the challenges to

overcome in the near future.

The section is organized as follows. First, the particular technological develop-

ments are presented, which are related to the bonding of III-V waveguide layers

on silicon, the processing of IMOS waveguides, the fabrication of pedestal pillars

to improve their Q-factor, as well as the deposition and treatment of silver for

reduced optical loss and contact resistance. Afterwards, the detailed process flow

to fabricate metal-coated nanopillars coupled to waveguides is presented.

36


3.2.1 Adhesive bonding of III-V layer stacks to silicon

The integration of III-V layer stacks on silicon has been researched during the

last decade to provide Si-photonics with lasers and efficient modulators. The most

widely used techniques for the integration of different semiconductor layer stacks

are molecular bonding (relying on Van der Waals forces) and adhesive bonding with

BCB [12] as revised in section 2.1. While molecular bonding is highly sensitive to

the wafer surface quality, the tolerance of adhesive bonding strongly depends on

the BCB thickness.

For the heterogeneous integration of III-V and SOI, typically a thin bonding layer

(e.g. 50 nm thin BCB) is required to allow vertical optical coupling between InP

and Si waveguides [64]. The disadvantage of using a thin bonding is that air in-

clusions trapped within the bonding layers can result in severe outgassing that

damages the III-V membrane during processing steps with temperatures higher

than 300 C (at which the bonding process is performed). This issue has been

previously reported and a solution has been proposed, which consists on the cre-

ation of III-V islands by patterning and etching down trenches (i.e. channels) in

the III-V layer stack down to the bonding layers [65]. The gases generated during

high temperature processing can then escape to the ambient thought the channels.

Figure 3.20: (a) The photo shows four quarters (of a two inch wafer) BCB-bonded to a Si-substrate after cleaving and InP-substrate removal. (b) Focus ionbeam (FIB) cut across the III-V/Si to visualize the bonding layer stack. Whenusing a thick-bonding recipe, the BCB thickness was found to vary between 300

nm to 1.5 µm.

Alternatively, we employed a thick-bonding recipe in order to be tolerant to the

III-V wafer irregularities and prevent outgassing without increasing the fabrication

complexity of the devices. We observed that the outgassing problem is alleviated

37


by using a thick BCB (a few hundred nanometers) probably because the generated

gas has more room to propagate laterally within the bonding layers. Furthermore,

a thick BCB provides thermal isolation between the photonic III-V layer stack and

the underlying substrate which could contain hot electronics in the near future.

The thick-bonding recipe developed uses the following steps:

1. Both Si and III-V are prepared. The III-V wafer is cleaned with an O2

plasma and protective layers are selectively wet etched until reaching the

semiconductor layers that serve as membrane. The Si wafer is cleaned with

a H2O:NH4OH:H2O2 RCA-1 solution.

2. A SiO2 layer is deposited on both wafers with PECVD (50 nm and 1 µm on

III-V and Si, respectively).

3. A BCB layer is spun on the Si-substrate and partially cured at 200C.

4. The III-V wafer (or dies) is (are) placed upside-down on top of the BCB-

coated Si wafer.

5. The actual bonding is done in vacuum using a force of 620 N per two inch

wafer.

6. After the bonding, the Si-substrate can be cleaved if needed.

7. Finally, the InP-substrate is removed by selective wet etching using an HCl:H2O

(4:1) solution at elevated temperature (e.g. 35C) for a fast etching. The

result is shown in Fig. 3.20a.

3.2.2 Waveguides processing with positive resist (ZEP) 3

The ZEP520A resist has been validated in several fabrication runs demonstrating

passive components [41]. Since it is a positive resist, only trenches are e-beam

exposed and etched, thereby allowing the mechanical protection of the patterned

circuits. This process was used to fabricate the grating couplers of the nanopillar

LEDs described in 3.2.7.

Figure 3.21 shows a schematic representation of the process. Only one lithography-

etching step is shown, however it can be repeated several types if more etching

levels are required (e.g. to fabricate grating couplers). It is worth mentioning that

3The experimental results shown in this section were carried out by V.D.-C., however the fabricationtechnology was developed during the former years of IMOS.

38


the process starts with a bonded III-V layer stack on silicon; which is described in

detail in section 3.2.1.

Figure 3.21: Process flow to fabricate passive components with ZEP520 resist.(a) Initial layer stack including InGaAs cap layer. (b) Sample after depositionof Si3N4 and spinning of ZEP resist. (c) Sample after e-beam exposure, ZEPdevelopment and dry etching of Si3N4. (d) Sample after removal of ZEP mask.

(e) Sample after waveguides etching and removal of Si3N4 hardmask.

At first, the semiconductor surface has to be cleaned. For this, an O2-plasma is

needed followed by a diluted phosphoric acid dip. Then, the InGaAs cap layer is

selectively removed by wet etching using a H2O:H2SO4:H2O2 solution. After this,

a 50nm thick Si3N4 layer for use as hardmask is deposited by Plasma-Enhanced

Chemical Vapor Deposition (PECVD). On top, ZEP520A resist is spun at 5000

rpm during 40 seconds to achieve a 300nm thickness. Two baking steps are per-

formed, the first one going from 100C to 150C during 4 minutes and the second

one at 200C for 2 minutes.

Once the sample is prepared, it is e-beam exposed using a dose of ∼38 µC/cm2 for

2.5 µm wide trenches to define 450 nm wide waveguides. After exposure, develop-

ment is performed with n-Amyl acetate during 1 minute, followed by rinsing in a

MIBK-Isopropanol (89:11) solution. Finally a post-baking step is done at 150C

(2 minutes) for the resist re-flow in order to reduce the pattern roughness [66].

The resist pattern is then transferred to the Si3N4 hardmask layer by Reactive-Ion

Etching (RIE) using CHF3. Then the e-beam resist is removed with 30 minutes

of O2 plasma, and the sample is cleaned with diluted phosphoric acid followed by

a 1% HF dip.

Finally, the semiconductor is etched using a CH4-H2 chemistry in an Inductively

Coupled Plasma (ICP) RIE process. The sample is cleaned one more time with O2

39


Figure 3.22: Characterization of waveguides and grating couplers fabricatedwith ZEP520A resist. Each measured point represents the loss of a waveguidewith varying length plus two grating couplers used for light in- and out-coupling.After a linear fit, a waveguide propagation loss of 6.2 dB/cm and a gratinginsertion loss of 6.6 dB are obtained. The inset shows a SEM picture of the

waveguide cross section.

plasma and diluted phosphoric acid and the hardmask is removed with Buffered

Hydrofluoric Acid (BHF). Optionally, several cycles of O2-plasma plus diluted

phosphoric acid can be applied to reduce the sidewall damage introduced by the

etching process.

Figure 3.22 shows typical characterization results of passive components (i.e. waveg-

uides and dielectric grating couplers). Several pairs of grating couplers connected

with waveguides of varying length were fabricated. Their insertion loss was mea-

sured in transmission and plotted as a function of the waveguide length. After a

linear fit, the waveguide propagation loss can be extracted from the slope, whereas

the intercept with the vertical axis corresponds to two times the grating insertion

loss. It is worth commenting that the lowest waveguide loss (2.5 dB/cm) has been

obtained recently using a C60/ZEP mixed resist which reduces the e-beam induced

roughness [40].

3.2.3 Waveguides processing with negative resist (HSQ)

In order to increase the fabrication flexibility of the IMOS devices, a process flow to

fabricate passive components with the negative resist HSQ was developed within

this thesis work. This process was used to fabricate the waveguides coupled to

nanopillar LEDs described in 3.2.7. In this process technique, the HSQ resist is

40


directly used as hardmask and therefore no additional dielectric layers (SiO2 or

Si3N4) are required.

Figure 3.23: Process flow to fabricate passive components with HSQ resist.(a) Initial layer stack including InGaAs cap layer. (b) Sample after spinning ofHSQ resist. (c) Sample after e-beam exposure, development of HSQ and dry

etching of waveguides. (d) Sample after removal of HSQ hardmask.

Similar to the ZEP-based process flow, the wafer is cleaned in the first steps with

an O2-plasma and diluted phosphoric acid. Then, the InGaAs cap layer is removed

with the same selective wet etching step.

After that, O2 plasma is used to oxidize the InP surface thereby promoting the

adhesion of HSQ. The HSQ resist is spun at 4000 rpm for 1 minute and then

two 2-minute baking steps are done, the first one at 150C and the second one at

220C. This results in a 100 nm thick HSQ layer. The last step for the sample

preparation consists on depositing 7.5 nm of gold to prevent excessive charging

during e-beam lithography. If such gold layer is not deposited, e-beam focusing

on the sample is more challenging.

After the sample has been prepared, it is e-beam exposed using a dose of 600µC/cm2

for 450µm wide waveguides. Then the gold layer is selectively etched away using

KI-I2 for 30 seconds and the HSQ resist is developed with MaD531S at 60C.

Finally, the HSQ hardmask selectivity is enhanced with oxygen plasma. This

mechanism has been suggested to increase the resistance of HSQ to the semi-

conductor etching chemistry [67]. In our experiments, we obtain a mask erosion

below 0.5 nm/min when using O2 at 50 W during 10 minutes for such enhance-

ment. Later, the pattern is etched into the InP layer with a RIE process using a

CH4-H2 chemistry at 70 nm/min. Afterwards, the sample is cleaned from poly-

mers with O2-plasma and diluted phosphoric acid, and the HSQ is removed with

41


BHF. The same technique described earlier in the ZEP process flow can be used

to reduce the sidewall damage.

3.2.4 Pedestal nanopillars

For a high-quality factor of the nanopillar cavity it is important that the sidewalls

are close to vertical [62]. In practice, a small angle is introduced as a result of

the dry etching process which is typically a few degrees (we achieved 1.8 in our

devices reported in section 3.3). This pushes the cavity mode a bit downwards

leading to increased leakage and a drop in the cavity quality factor. It is, therefore,

important to decrease the sidewall angle by process optimization.

In order to relax the sidewall verticality tolerance, the optical confinement can be

improved by introducing an undercut above and below the active medium. This

results in an increase of quality factor as discussed in section 3.1.1.2, which makes

the cavity more tolerant to the sidewall angle.

Figure 3.24: Pedestal nanopillars fabricated with selective wet etching. TheInP posts below and above the active region are etched with H2O:H3PO4:HCl(2:4:1). (a) Cross section along the (011) direction. (b) Cross section alongthe (011) direction. (c,d) Collapsed pillars as a result of over-etching; only the

InGaAs contact layer and bulk active core are left.

We developed a simple and reproducible process to fabricate pedestal nanopillars.

After nanopillar dry etching (see Fig. 3.27a), a wet etching step that selectively

etches the InP posts can be performed. Typical InP etching solutions (selective

against InGaAs and InGaAsP) are H3PO4:HCl and HCl:CH3COOH, for which

we found an undercut etch rate of 17 nm/s and 83 nm/s at 23C, respectively.

For a slower etch rate that allows an accurate etch control, water dilution can be

employed in a volume ratio H2O:H3PO4:HCl=2:4:1 resulting in an average etch

rate of 4 nm/s.

42


Figure 3.24 shows pedestal pillars with sub-micrometer side length. As can be

observed, the etch rate slightly depends on the doping type and the crystal ori-

entation (i.e. it is anisotropic). Furthermore, a negative slope is created in the

(011) direction, whereas a positive slope is formed in the (011) direction. Figures

3.24c and 3.24d show over-etched pillars where the InP posts have been completely

etched.

3.2.5 Silver deposition and treatment

Silver is well known as the metal of choice for nanoscale devices based on metal-

optics or plasmonics due to its relatively low loss at 1.55 µm, which usually offers

the best device performance. However, in many cases gold is used for a practi-

cal device implementation. For example, gold was used to demonstrate the first

metal-coated nanopillar laser [16] and plasmonic phase and Mach-Zehnder modu-

lators [68] [69]. This is simply because gold-based processing technology is better

developed, and also because silver oxidates when exposed to air. Here we report

technology development for improving silver deposition, annealing and oxidation

prevention.

Figure 3.25: SEM photos of silver deposited by thermal evaporation on IMOSsamples (a) with SiO2 coating (b) with SiO2 and RTA post-treatment at 350C for 60 seconds and (c) with SiO2 and RTA post-treatment at 400 C for 30

seconds.

We investigated thermal evaporation of silver with a fixed deposition rate of 2

A/s. As can be seen in Fig. 3.25a, during the deposition it will typically form

grains of a few tens of nanometers (average size of 30 nm in our case). Such a

size is well below 1550 nm and could produce two types of optical loss: through

Rayleigh scattering and through electron scattering in the plasmonic regime due

to a reduced mean free path of electrons (typically tens of nanometers) limited by

43


the grain size. In order to reduce the metal loss in our metal-cavity nanopillar,

grains larger than the wavelength in the medium are required.

In order to increase the grain size, thermal annealing was carried out. After the

silver deposition, we performed RTA at different conditions and optimized the

process. As shown in Figs. 3.25b and 3.25c, the grains grow larger than 1 µm

for temperatures higher than 350C at annealing times longer than 30 seconds. It

is worth mentioning that such annealing conditions are similar to the annealing

required for Ag/Ge ohmic contacts as revised in the next section, which provide

low optical loss and low contact resistance at the same time.

3.2.6 Silver-based ohmic contacts 4

The metallization used in active optoelectronic devices has a great influence on

their performance. A bad metallization can lead to high series resistance, which

has a detrimental impact on the device characteristics. For example, it will limit

the switching speed through the RC time constant, and it will result in high

threshold lasers due to Joule heating, thereby limiting the device wall-plug effi-

ciency. Furthermore, in nanoscale devices the proximity of electrical contacts to

the optical mode causes additional loss by metal absorption.

In membrane-based photonic circuits, the metallization layers are deposited on a

smaller contact area and closer to the optical mode than in traditional photonic

circuits. This leads to an increase of contact resistance and also to a higher optical

loss. As an example, n-type ohmic contacts based on Au/Ge/Ni are widely used in

InP photonic circuits, however this scheme is not attractive for membrane photon-

ics since both, gold and nickel, have a high optical absorption at 1.55 µm. Because

of these reasons, the development of a new metallization scheme is required.

We developed a novel Au/Ag/Ge n-type metallization which can be used either on

InP or on InGaAs (the scope of this thesis). Germanium is used for two reasons:

it acts as an adhesion layer between SiO2/III-V and Ag, and it provides additional

doping by diffusing into the semiconductor contact layer when performing anneal-

ing [70]. Both Ge and Ag have a relatively low optical loss at 1.55 µm compared

to typical metallization layers. The gold layer is used to prevent the oxidation of

silver.4The development of silver-based ohmic contacts was done in cooperation with L. Shen. V.D.-C.

proposed the metallization scheme and carried out the experiments presented in this section, whereasL.S. performed the detailed studies reported in [70] [71].

44


Ge-thickness Annealing Specific contact resistance(nm) (Temperature, time) (Ω· cm2)

4 No 7 · 10−6

2 350 C, 30 s 6 · 10−8

2 400 C, 15 s 1.3 · 10−7

15 350 C, 30 s 4 · 10−7

15 400 C, 15 s 1.8 · 10−7

Table 3.2: Specific ohmic contact resistance of Au/Ag/Ge contact on n-InGaAs(N = 1 · 1019cm−3) .

We performed a series of experiments to determine the specific contact resistance

of this n-type contact by means of the circular transfer length method (circular-

TLM) as described in [72]. We fabricated the circular-TLM test structures with a

single lift-off process, measured the current-voltage characteristics and did the data

analysis to extract the specific contact resistance. Table 3.2 shows the experiments

with recipes that are compatible with the processing of metal-cavity nanopillars in

IMOS, i.e. temperatures that also result in a silver grain size increase (as presented

in section 3.2.5), using a thin Ge layer and a relatively short annealing process to

avoid outgassing from the bonding layers. Specific contact resistance in the range

of 10−8 to 10−7 Ω· cm2 can be obtained.

In order to demonstrate its superior performance also in the optical domain, more

general and extensive studies on this metallization scheme were carried out sep-

arately on highly n-doped InP [70] and InGaAs [71], which confirmed that the

contact provides both low optical loss and low contact resistance and, further-

more, it does not spike [70]. Therefore, we believe this is the ideal n-type ohmic

contact for metal-based membrane devices.

3.2.7 Process flow of waveguide-coupled nanopillars

The fabrication process requires a variety of techniques, such as optical and elec-

tron beam lithography (EBL), plasma-enhanced chemical vapor deposition (PECVD)

of dielectrics, reactive ion etching (RIE) processes, wet-chemical etching, thermal

and electron-beam evaporation of metals, rapid thermal annealing, etc. The defi-

nition of the pillar, waveguide and grating coupler is carried out by EBL in three

different steps and the rest of the process is done by optical lithography.

An overview of the fabrication process flow to fabricate metal-coated nanopillars

on top of waveguides, which are connected to grating couplers, in III-V membranes

45


on a silicon substrate is presented in Fig. 3.26. Appendix A contains a detailed list

of the main processing recipes followed for the fabrication of the nanopillar LEDs

described in this thesis. In the following, the process flow is described in 10 stages

which are depicted in Fig. 3.26. The description is supported by experimental

results shown in Figs. 3.27 and 3.28.

1. At first, the III-V layer stack is expitaxially grown on an InP substrate by

metal-organic chemical vapor deposition (MOCVD) and adhesively bonded

46


Figure 3.26: Process flow to fabricate metal-cavity nanopillars coupled towaveguides in III-V membranes on silicon. The longitudinal and transversalcross sections are shown (i.e. along and perpendicular to the waveguide, respec-

tively).

47


to a silicon substrate with benzocyclobutene (BCB) [12]. After bonding, the

InP substrate is selectively removed by wet etching using H2O:HCl (1:4).

2. The definition of the pillar is done using hydrogen silsesquioxane (HSQ) neg-

ative resist, which is well known for its usage in high resolution lithography

[73], in combination with HPR504 to form a bilayer resist scheme for etch-

ing structures with high aspect ratios (i.e. the nanopillar) [74]. First, a

SiO2 (450nm) hardmask is deposited with PECVD, and then HPR540 (450

nm) and HSQ (100 nm) resists are spun. After e-beam lithography, HSQ

development is done with MaD531S and the pattern is transfered to the

HPR layer with an O2-plasma reactive-ion etching (RIE). Then, the pat-

tern is transferred to SiO2 with a CHF3 chemistry, and finally the semicon-

ductor nanopillar is etched with an inductively coupled plasma (ICP) RIE

process based on methane-hydrogen (CH4:H2). The ICP-RIE typically pro-

vides higher etching anisotropy that RIE, which translates into more vertical

sidewalls. The experimental result is shown in Fig. 3.27a.

3. Later, a new HSQ resist layer is spun while keeping the previous SiO2 on top

of the nanopillar, and e-beam patterned as shown in Fig. 3.27b, using again

MaD531S as developer. The HSQ selectivity against semiconductor etching

is enhanced by treating it with an oxygen plasma [67]. Then, the HSQ resist

pattern is used as hardmask to dry etch the waveguide while protecting the

pillar with its original SiO2 hardmask. Fig. 3.27c shows the result after

removal of the masking layers.

4. An optical lithography is performed using AZ4533 resist and MaD531S de-

veloper (see Fig. 3.28a), in order to expose the waveguides and wet etch

their top InGaAsP cladding with H2O:H2SO4:H2O2 (10:1:1), which would

cause strong waveguide loss due to intraband absorption when it would not

be removed. Fig. 3.28b shows the result.

5. The next step is to fabricate the grating coupler. For this, a Si3N4 layer

is deposited with PECVD and then ZEP520A resist is spun. The resist is

e-beam patterned and developed with n-Amyl acetate; then, the pattern is

transferred into the Si3N4 hardmask with a CHF3 based RIE.

6. In order to protect the nanopillar during the grating etching into the semi-

conductor, an optical lithography is carried out using once more AZ4533

resist and MaD531S developer (see Fig. 3.28c). The pattern is baked hard

to withstand a CH4:H2 RIE etch with which the grating coupler is etched into

48


the InP layer. The result is shown in Fig. 3.28d after resist and hardmask

removal.

Figure 3.27: Scanning electron microscope (SEM) images showing the fab-rication process with focus on the nanopillar. (a) Image after the etching ofthe nanopillar. (b) After the development of HSQ resist to act as waveguidehardmask. (c) After waveguide etching and removal of hardmask layers. (c)After waveguide etching and removal of masking layers. (d) After deposition ofSiO2 cladding. (e) After etching of SiO2 from the pillar top to expose the n-doped InGaAs contact layer. (d) After deposition of Ag/Ge and rapid thermal

annealing.

7. Once the nanopillar, waveguide and grating coupler are fabricated, a SiO2

layer (175 nm) is conformally deposited with PECVD as shown in Fig. 3.27d,

to serve as dielectric cladding for the cavity and for electrical insulation of the

diode structure. Then, Au/Ti (40 nm/50 nm) adhesion pads are deposited

around the pillar with a lift-off process, whose purpose is to promote the

adhesion of silver later on (see Figs. 3.28e and 3.28f).

8. In order to contact electrically the n-InGaAs layer, the SiO2 must be removed

from the pillar top. For this, MaN440 resist (thicker than the nanopillar

height) is spun on the chip, cured (95C, 5min) and slowly developed (etched

back) with an MaD531S:H2O (2:1) solution at a rate of 7 nm/s, until the

top of the nanopillars is exposed. Then, the SiO2 is etched only from the

nanopillar top with RIE (CHF3:O2). The result is shown in Fig. 3.27e after

resist removal.

49


9. The next steps consist of cleaning the InGaAs contact layer with a short wet

etch and depositing metals to form the n-contact as well as the metal-cavity.

For this, Ag/Ge (>300 nm/4 nm) layers are deposited all over the wafer

sample by thermal evaporation and treated with rapid thermal annealing

(RTA) at 350 C for 30 seconds to (1) provide a low contact resistance n-

type ohmic contact and (2) to increase the silver grain size, thereby reducing

its optical loss (see Fig. 3.27f). Sections 3.2.5 and 3.2.6 provide a detailed

description of these technology developments. Afterwards, a 100 nm thick Au

layer is sputtered to prevent silver oxidation. Then, an optical lithography

is performed with AZ5433 resist and MaD531S developer in order to protect

the n-contact pad and the nanopillar (see Fig. 3.28g) regions during wet

etching of the metal layers Au/Ag with a KCN-based solution. Figure 3.28h

shows the result after the photoresist has been removed.

10. Finally, a last optical lithography step is required to fabricate the p-contact

pad with lift-off using MaN440 resist and MaD531S developer (see Figs. 3.28i

and 3.28j). For this, firstly the SiO2 is removed with buffered hydrofluoric

etch (BHF) to have access to the p-InGaAsP contact layer and then a stan-

dard Au/Pt/Ti (200 nm/75 nm/60 nm) metallisation is performed. Although

its use is very common due to its relatively low contact resistance on InGaAs,

unfortunately it does not provide an ohmic contact on p-InGaAsP (this de-

vice case), which was the reason of the high series resistance observed in the

diodes (see Fig. 3.31a).

Following the aforementioned process flow, we fabricated a batch of devices as

shown in Fig. 3.29. The complete characterization is presented in detail in section

3.3. As the cavity resonance frequency is very sensitive to nanopillar size, we fab-

ricated a series of devices with slightly different symmetric cross sections between

300×300 nm2 and 400×400 nm2. The experimental results of the static and dy-

namic characterization presented in section 3.3, were obtained on nanopillars with

side length of 350 nm and 340 nm, respectively.

50


51


Figure 3.28: Photos showing the fabrication process with special focus on thelarge scale patterns (i.e. the optical lithography steps), (a) after optical lithog-raphy and development to expose the Q1.25 cladding on top of the waveguides,(b) waveguide edge after wet etching of Q1.25 from the top of the InP-waveguideand removal of AZ4533, (c) with the grating pattern defined in a Si3N4 layerwhile protecting the pillar with AZ4533 resist, (d) after dry etching of the grat-ing coupler into the semiconductor and removal of the AZ4533 protection resistand Si3N4 hardmask, (e) after optical lithography and development of MaN440resist to deposit the adhesion layers Au/Ti, (f) after deposition of Au/Ti ad-hesion layers by lift-off around the nanopillar, (g) after optical lithography anddevelopment of AZ4533 resist to protect the pillar and n-contact region in or-der to wet etch the Au/Ag layers, (h) after etching the metal layers Au/Ageverywhere except in the device region and removal of photoresist, (i) after op-tical lithography and wet etch of SiO2 cladding to access the Q1.25 layer inorder to do the p-contact metallization, and (j) after deposition of Au/Pt/Ti

metallization layers by lift-off.

3.3 Characterization of nanopillar LEDs 5

In this section, we report the characterization results of the fabricated metal-cavity

light source described in section 3.3. Although we could not drive the device above

lasing threshold, we observed clear LED operation with a high coupling efficiency

to the output waveguide. This is the first metal-cavity nanopillar LED coupled to

a waveguide and integrated on a silicon substrate. The device was characterized

through a grating coupler and shows on-chip external quantum efficiency in the

10−4-10−2 range at tens of µA current injection levels, which is orders of magni-

tude higher than the best results (10−7) reported earlier [27]. Furthermore, direct

modulation experiments reveal sub-nanosecond electro-optical response.

The section is organized as follows. Section 3.3.1 describes the experimental setup

used for the characterization of the nanopillar LEDs. Section 3.3.2 reports the

5The characterization was carried out in cooperation with Dr. F. Pagliano, Dr. B. Romeira and Dr.S. Birindelli.

52


Figure 3.29: (a) False-colored scanning electron microscope (SEM) imageshowing the fabricated device structure before metallization. The nanopillarlies on top of a waveguide connected to a grating coupler. (b) Metal-coatednanopillar after silver evaporation and rapid thermal annealing, prior to thesputtering of gold (100 nm) to prevent silver oxidation. (c) False-colored image

of the device after metallization, showing the electrical contacts.

static measurements of the LED devices, comprising current-voltage and light-

current characteristics, as well as the emission spectrum. Section 3.3.3 describes

electro-optical modulation experiments, both in a pulsed regime and with a con-

tinuous modulation signal. Next, section 3.3.4 presents studies on surface recom-

bination and passivation of the nanopillars in order to support the interpretation

of the dynamic characterization. Finally, the perspectives for improvement and

conclusions are presented in sections 3.3.5 and 3.3.6, respectively.

3.3.1 Experimental setup

The electrical characterization of the nanopillar LEDs (current-voltage character-

istics) was carried out with a Keithley sourcemeter and using a pair of individual

electrical probes to contact the devices.

For the electro-optical characterization we used a confocal micro-electroluminescence

(µEL) setup and different detection techniques, which are depicted in Fig. 3.30.

Both sample holder, piezo stage and electrical probes were placed inside a liquid

helium cryostat chamber allowing electro-optical measurements down to ∼ 9 K.

The devices were contacted electrically by 40 GHz high-speed probes with ground-

signal-ground configuration (100 µm pitch). For the static characterization, a D.C.

53


voltage signal was used to bias the devices, whereas a high-frequency bias-T (0.2

MHz - 12 GHz) was employed to provide both an A.C. and a D.C. drive signal for

the dynamic characterization. The A.C. signal was provided with a pulse generator

(Agilent 81134A or Anritsu MP1701A).

Figure 3.30: General schematic of setup [75] used for the characterizationof nanopillar LEDs. The different measurement techniques employed are de-picted by the light paths (a), (b) and (c). (a) The measurement of the opticalpower was done with a photodetector (depicted by the dashed line). (b) Theelectroluminescence spectrum was measured with an InGaAs detector camera.(c) The time-resolved electroluminescence was measured with a time-correlated

single-photon counting (TCSPC) module.

The light emitted from the waveguide-coupled LED device was collected in free-

space from the device’s integrated grating coupler using a high numerical aperture

(NA) objective (50x, NA = 0.42). For the case of the light-current characteristics,

we measured the optical power directly with a photodetector after the microscope

objective (see Fig. 3.30a).

In order to measure the electroluminescence spectra, the image formed by the

objective was projected onto a single mode optical fiber (SMF-28J). The fiber

collected the light from a limited circular area (∼ 1.7 µm diameter) of the grating

coupler and guided it to a Horiba FHR 1000 monochromator equipped with an

InGaAs cooled detector camera as depicted in Fig. 3.30b.

Finally, the fast dynamic characteristics of the LED devices were measured with

a time-correlated single-photon counting (TCSPC) module for time-resolved mea-

surements as depicted in Fig. 3.30c. The fiber guided the light to a single photon

counting detector. Here, a Scontel superconducting single photon detector (SPD)

housed inside a cryogenic dipstick at 1.5 K and working at a bias current Ib ∼ −19

54


µA with quantum efficiency of ∼ 20%, was used to measure the temporal decay.

Prior to detection, a wavelength filtering (FWHM 50 nm centered at 1550 nm)

was used to select the signal of interest and filter out the unwanted signal. The

SPD was connected to a correlation card (PicoHarp 300) controller, i.e. a time-to-

digital converter. This controller measures the time between the excitation pulse

(the start signal arriving from the trigger signal of the pulse generator) and the

arrival of a luminescence photon at the SPD (stop signal). A histogram of these

arrival times is then constructed corresponding to the time-dependent output in-

tensity of the electrically pumped nanopillar LED.

3.3.2 Static characteristics

The measured electrical characteristics of the nanopillar diodes are shown in Fig.

3.31a. The diodes exhibited a constant reverse-bias current of 0.5 µA and showed

a high series resistance of about 22 kΩ due to the high p-contact resistance since

this contact was not annealed. We measured contact resistances of ∼ 7 × 10−4

Ωcm2 and ∼ 5× 10−7 Ωcm2 for the p- and n-contact, respectively.

Figure 3.31: (a) Current-voltage (I-V) diode characteristics. The arrows indi-cate the bias points used to obtain the spectra. (b) Emission spectrum at room

temperature for different bias currents.

Figure 3.31b shows the spectrum at different bias points (indicated by arrows in

the I-V curve of Fig. 3.31a). We observed a carrier-induced blue shift of up to

0.1 nm/µA. For low pumping conditions, only the low-frequency cavity resonance

is visible due to the better overlap with the luminescence spectrum of the bulk

InGaAs, whereas an additional resonance shows up at 1425 nm for higher pumping

currents. The spectral position of the two peaks qualitatively compares well with

the resonances in Fig. 3.13b(blue line). The model does not provide a complete

55


quantitative description of the measured spectra because it does not consider the

carrier distribution, the homogeneous broadening, and the change in refractive

index due to current injection and local temperature.

Figure 3.32a shows the light-current (L-I) curves at different temperatures and

Fig. 3.32b shows the corresponding on-chip external quantum efficiency (EQE),

ηQE. The maximum measured output power is nearly 4 nW at room-temperature

(RT) and increases to 60 nW at 9.5 K. At cryogenic temperatures, the optical

emission reaches saturation at lower injection currents because the non-radiative

recombination rate is considerably lower and therefore the radiative efficiency is

higher. The experimental results are fitted adequately with a rate equations model

(see Appendix C). Since thermal effects are not considered in the model, it does

not fit the thermal roll-off at high injection levels.

Figure 3.32: (a) Light-current characteristics of the metal-cavity LEDs atdifferent temperatures (9.5K and 75 K: left axis; 297 K: right axis) including anumerical model fit (dashed curves). (b) Calculated external quantum efficiency

for the on-chip waveguide-coupled power.

For the calculation of EQE (total number of photons coupled to the waveguide

divided by number of injected electrons), we multiplied the measured power by

a factor of 5.6 to account for the fact that we are collecting from only one side

of the waveguide, the grating out-coupling efficiency (∼ 50%) and the measured

collection efficiency (71%) of the setup. The setup collection efficiency determines

the percentage of light emitted from the sample (within the objective NA) that

reaches the detection system. Figure 3.32b shows an ηEQ which varies from ∼ 10−4

(at RT) to ∼ 10−2 (at 9.5 K). The low-temperature EQE is smaller than the

average spontaneous emission coupling efficiency of ∼ 0.07 calculated with realistic

parameters used for the blue line in Fig. 3.13b. This may be related to the fact

that, even at low temperature, the spectral distribution of carriers in the bands

56


does not perfectly match the mode spectrum, and presumably to a Ag loss higher

than the one reported in Ref. [57]. The latter effect is also confirmed by the fact

that the experimental linewidth of the resonance at low temperature (25 nm) is

larger than the calculated value of 14 nm.

The difference between room- and low-temperature efficiency is largely due to

the effect of nonradiative recombination, dominant at RT as observed from time-

resolved photoluminescence experiments (see section 3.3.4). Additional contri-

butions to the efficiency reduction may come from the broader thermal carrier

distribution and from an increased attenuation coefficient in Ag. The change in

Ag optical loss with temperature is confirmed by a larger linewidth of 42 nm

observed at room temperature.

3.3.3 Dynamic characteristics

Time-resolved electroluminescence experiments were carried out to determine the

modulation speed characteristics of our devices with the setup depicted in Fig.

3.30. We directly modulated the nanopillar LEDs using a pulse pattern generator

with a periodic pulse train of 80 MHz with pulse widths varying from 1 ns to 100

ps. Fig. 3.33a shows the detected modulated optical output when the device was

D.C.-biased with a current of 11 µA and modulated with a small amplitude pulse

(i.e. in the small signal regime).

The on-off switching time of the device is in the sub-nanosecond regime, and

the switch-off can be fitted with a single exponential decay curve allowing to

estimate a lifetime of τ = 289± 3 ps. The results in Fig. 3.33a can be compared

with the slow exponential decay time (τbulk ≈ 2.7 ns) measured from the time-

resolved photoluminescence (TRPL) of the InGaAs bulk material, in a test sample

prepared with an identical epitaxial layer stack employed in the fabricated LEDs

(see section 3.3.4). The results indicate that minority carrier lifetimes in the metal-

cavity nanopillar are much shorter than the nanosecond-range lifetimes in the

bulk material which otherwise would limit the nanopillar LED frequency response

to a 3-dB modulation bandwidth ∼ 60 MHz (f3dB = 1/2πτbulk). The lifetimes

measured by TRPL on nanopillars in the test sample of comparable dimensions

to the nanopillar LED devices show recombination lifetimes varying from 324 ps

to 184 ps for pillars with side lengths of 0.59 µm and 0.388 µm (see Fig. 3.34a),

respectively, which compare well with the sub-nanosecond recombination lifetimes

measured in Fig. 3.33a for a nanopillar LED. These results, together with the

57


negligible Purcell enhancement factor estimated in the design section, confirm

that the main physical process explaining the fast carrier dynamics measured in

the nanopillar LEDs is non-radiative recombination due to surface states in the

pillars with sub-micrometer cross section.

Figure 3.33: (a) Time-resolved electroluminescence showing the electro-opticalmodulation response of the nanopillar LED to electrical pulses with widths of 1ns and 100 ps at room temperature. (b) Direct modulation at room temperaturewith a pulsed pattern with 50% duty cycle at 2 and 5 GHz. The device wasmodulated with a peak-to-peak voltage signal of 1.4 V on a D.C. bias of 0.5 V.

The experimental electro-optical modulation results are also in good agreement

with the numerical results (dashed curves in Fig. 3.33a) employing a rate equations

model that takes into account both radiative and non-radiative recombination

processes (see Appendix C), further confirming that the sub-nanosecond response

measured in the device is dominated by a high surface recombination.

Although the efficiency of the LEDs is limited by such a non-radiative process, it

allows spontaneous emission light sources to operate at modulation bandwidths

well beyond the limitations of the slow radiative recombination process of the

semiconductor material. In order to show the potential of the nanopillar LEDs for

modulation at GHz frequencies, we directly modulated the device using a pulse

pattern generator in a Return-to-Zero modulation scheme with a periodic pulse

train having a 50% duty cycle at repetition rates ranging from 2 GHz to 5 GHz

(corresponding to a pulse width of 250 ps and 100 ps, respectively). Fig. 3.33b

presents the optical output showing that the nanopillar LED output replicates

the injected on-off periodic bit sequences. The bit stream has clearly resolvable

off-pulses even at frequencies higher than 5 GHz. At 2 GHz the optical response

has a modulation depth of 96%, whereas it decreases to 42% at 5 GHz. This is

a typical behaviour in directly modulated LEDs following a single-pole response

58


of the form H(f) = A0/(1 + jf/f3dB) (A0 is the static electro-optical conver-

sion efficiency), where the modulation depth decreases monotonically for higher

modulation frequencies.

3.3.4 Surface recombination and passivation

In order to confirm that non-radiative recombination is indeed the main process

that explains the sub-nanosecond carrier lifetimes discussed in section 3.3.3, we

performed a systematic study of the carrier dynamics in single InP-InGaAs-InP

nanopillars. We fabricated and probed optically nanopillars with square cross

section and varying side length d (as shown in inset of Fig. 3.34b), using micro

photoluminescence (µPL) and TRPL spectroscopy.

The nanopillars were covered with a 175 nm thick SiO2 cladding similar as used

in the nanopillar LEDs. Since these nanopillars were not covered by the metal

cladding, we could probe the free carrier dynamics optically allowing us to perform

a comprehensive analysis of the optical quality of the InGaAs surface in terms

of carrier lifetimes. Figure 3.34a shows the TRPL curves for nanopillars with

different side length, indicating minority carrier lifetimes in the nanopillars in the

order of hundreds of picoseconds for pillars with sizes similar to the nanopillar

LEDs. Such lifetimes are shorter than the nanosecond-range lifetime of the bulk

and comparable with the sub-nanosecond response measured in Fig. 3.33a.

Figure 3.34: (a) Comparison between the photoluminescence signal of nanopil-lars and bulk InGaAs. The legend indicates the corresponding nanopillar sidelength d. (b) Carrier lifetimes determined by the TRPL experiment for differentpillar side length. A surface recombination velocity S ≈ 105 cm s−1 is obtained

by means of a linear fitting.

59


On the basis of the measured lifetimes determined by the TRPL measurements,

the surface recombination velocity S can be estimated by the following equation:

1

τPL=

1

τbulk+

2S

d(3.6)

where τPL is the effective carrier lifetime, τbulk is the carrier lifetime in the bulk

material, and d is the side length of the pillar cross section. From the fitting with

Eq. 3.6, we estimate a surface recombination velocity of around 105 cm s−1 for

the pillars with surface areas below 1 µm2. The estimated values are one to two

orders of magnitude higher than the measured values of InGaAs lattice-matched

to InP with improved passivation techniques [76], indicating that the surface of

the nanopillars has been strongly affected by the process.

We did a further investigation to determine the cause of the high surface recom-

bination. For this, we measured the PL intensity of a sample after each relevant

processing step. Figure 3.35 shows the comparative results for different pillar

dimensions. It shows the relative PL intensity after dry etching, after surface

cleaning steps and, finally, after the deposition of the SiO2 cladding. For the

cleaning, we employed 10 cycles of the following three steps: (1) O2-plasma, (2)

1% HF and (3) H3PO4:H2O (1:10) for 2 minutes followed by UPW rinsing.

Figure 3.35: Relative comparison of the InGaAs PL peak intensity of nanopil-lar test structures at different processing stages: (1) after dry etching, (2) aftersurface cleaning and (3) after deposition of 175 nm thick SiO2 with PECVD.

It can be observed that the PL intensity improves significantly after the sidewall

cleaning. Nevertheless, after the dielectric deposition, it drops drastically to an

intensity that is just about two times higher than the PL intensity after dry etching.

Therefore, the SiO2 deposition with PECVD at 300C was identified as the cause

of the high surface recombination.

60


An improved passivation could be achieved by replacing the SiO2 cladding by Si3N4

and optimizing the thermal annealing conditions for recovering the surface damage

induced by its deposition with PECVD [63]. A different reported solution consists

in performing two depositions: a thin passivating Si3N4 layer at low temperature

and a thick high quality Si3N4 for optical confinement at high temperature [16].

Due to a limited time for process development, these techniques were not explored

in this thesis work.

3.3.5 Perspectives for improvement

In order to improve the cavity quality factor to a value which allows lasing, we de-

signed and fabricated micron-size cylindrical cavities supporting whispering gallery

modes. In a larger cavity, the loss due to the metallic confinement is reduced and

therefore the cavity quality factor is improved. Additionally, we also considered the

fabrication of an undercut around the active medium (as describe in section 3.2.4)

to further increase the Q-factor as well as the tolerance to the sidewall angle. This

resulted in calculated Q-factors of ∼ 1150 (λ = 1533 µm) and ∼ 2095 (λ = 1583

µm) for diameters of 1 µm and 1.25 µm, respectively, having a 35% confinement

factor within the active medium. Figure 3.36a shows a photo of the fabricated

structure with expected higher quality factor at an intermediate processing step.

Figure 3.36b shows the spatial profile of the whispering gallery modes. In this

design, the coupling with the waveguide fundamental mode decreases to ∼ 3%

and ∼ 1% for the cavities with 1 µm and 1.25 µm diameter, respectively. This is

because (1) the waveguide is as wide as the pillar and therefore it is multi-mode

and (2) the whispering gallery mode has maximum intensity close to the pillar

perimeter whereas the intensity of the fundamental waveguide mode is maximum

at its center.

The fabrication run that we did went very successful up to the last step, where it

failed because the metal contacts broke due to stress, so unfortunately this design

could not be tested. If the improved cavity can be operated in the lasing regime,

we expect significantly higher output powers and modulation speeds (compared

to the nanopillar LED performance).

3.3.6 Conclusions

In summary, we present the first waveguide-coupled nanopillar LED device with

metal-cavity, which is fabricated in a III-V layer stack bonded to a silicon substrate.

61


Figure 3.36: (a) SEM photo of the cylindrical micro-pillar with undercut of∼ 100 nm and covered by a 100 nm thick SiO2. (b) Horizontal cross sectionof the cavity showing the simulated |E|2 spatial profile of a whispering gallerymode with index (b) M = 4 for a pillar diameter of 1 µm and (c) M = 5 for a

pillar diameter of 1.25 µm. Blue: low intensity. Red: high intensity.

The device shows the highest on-chip external quantum efficiency (10−4 to 10−2 for

room-temperature and 9.5 K, respectively) reported so far and output power (i.e.

nW up to tens of nW) far exceeding state-of-the-art nanoscale LEDs exhibiting

pW output [30]. This percentage-level efficiency at low temperature suggests the

potential of this device to perform very efficiently even at room-temperature if the

non-radiative recombination processes (found to be dominant in our devices) are

suppressed by better passivation techniques.

Furthermore, an improved spatial matching of the active region with the optical

mode will increase the coupling of spontaneous emission into the cavity mode,

thereby increasing the efficiency above the 1% level. We note that the achieved

power level at low temperature (∼ 60 nW measured off-chip) corresponds to over

1300 photons/bit (∼ 168 nW on-chip in a single waveguide direction) at 1 Gb/s,

which is far above the shot-noise limited sensitivity of an ideal receiver (20 pho-

tons/bit for a bit error rate < 10−9, -59 dBm for 1 Gb/s) [77]. We assume a

receiver sensitivity of -41 dBm (i.e. 84 nW received power considering NRZ cod-

ing), which is reasonable since most of the receivers operate about 20 dB away from

the quantum limit to compensate for thermal noise [77]. With the expected low

loss of short-distance interconnects and continuous progress in integrated receivers,

this power level may enable intrachip data communications with an ultracompact

source.

We explored also the dynamic characteristics of the devices and found sub-nanosecond

62


lifetimes, related to surface recombination at the pillar sidewalls. Although the

radiative efficiency is compromised by this effect, it can also be used for fast on-off

switching. We demonstrated this by modulating the nanopillar LEDs with a pulse

pattern generator at frequencies up to 5 GHz. In future designs, an optimized

emitter-cavity spatial matching may result in significant Purcell enhancement,

enabling high modulation speed without compromising the radiative efficiency.

These results are encouraging for future high-density optical interconnect sys-

tems requiring Gbps data rates at ultra-low power consumption, which could be

achieved with arrays of directly modulated integrated nanoscale sources [10].

63

Chapter 4

Grating couplers

This chapter presents the work on grating couplers carried out within the frame-

work of this thesis, which is mostly published in [42, 78]. It comprises typical

dielectric grating couplers and metal grating couplers. In the next section, a gen-

eral introduction to the vertical fiber-to-chip coupling is presented. The second

section introduces the dielectric grating couplers and presents the design of the

grating coupler used for the characterization of the nanopillar LEDs reported in

3.3. Finally, the design and characterization of a novel metal grating coupler is

presented.

4.1 Introduction

Grating couplers are an alternative to end-fire coupling in integrated photonics

to couple light in and out of a photonic circuit. They are used in a variety of

applications, for example, on-wafer characterization and packaging [79], [80]. They

are also attractive for optical interconnect systems that make use of chip-to-fiber

and chip-to-chip coupling schemes [81], [82].

In membrane-based photonic platforms, grating couplers are of special interest as

they allow for mode matching between photonic wires and optical fibers without

the use of lenses and spot-size converters [83]. However there are a few issues

that typically limit their performance, which are: a mismatch between the beam

coupled out of the grating and the optical mode of the fiber, power leakage to the

substrate and free space diffraction.

65

Chapter 4. Grating couplers

Figure 4.1: Representation of the light propagation and scattering in the k-space. A waveguide mode propagates with βx and suffers a periodic perturbationwith kΛ. The semi-circles represent the free-space and substrate radiation modes

above and below the waveguide, respectively.

Figure 4.1 shows a schematic that explains the working principle of such a coupler.

When a guided mode with propagation constant βx propagates along the grating

coupler region, it suffers a perturbation given by the grating coupler k-vector in

opposite direction, which is kΛ = 2π/Λ. This perturbation causes light scattering

with an in-plane component β′x. Since there are no guided modes with such a

small β′x component, the scattered light couples to free space modes (top green

arrow) which add coherently at a diffraction angle θ and substrate radiation modes

(bottom green arrow).

Figure 4.2: Schematic representation of the diffracted power distribution (a)along the grating coupler and (b) transversal to the grating coupler, correspond-ing to the directions indicated by a dashed line in the grating coupler photo ofFig. 4.5b. Pdiffracted represents the spatial distribution of the diffracted opti-cal power above the grating coupler and Pdesired depicts the desired Gaussian

distribution of a single mode fiber.

The mode matching problem in both directions is depicted in Fig. 4.2. In the

longitudinal direction, this issue arises from the fact that the optical power de-

creases exponentially as it propagates along a uniform grating coupler, whereas

66


it has to couple to the single mode fiber whose fundamental mode has a nearly

Gaussian distribution. It has been shown theoretically that an exponential and a

Gaussian distribution show a maximum power coupling of 80% [83] if the coupling

length of the grating (the length over which the field decays to its 1/e value) is

matched to the fiber mode width. The optimum grating length depends on the

grating strength. Regarding the transversal mode matching, the maximum mode

overlap happens for a grating width around 15 µm when considering a fiber core

diameter of 9 − 10µm as shown in Fig. 4.3b. This results from the fact that a

high optical contrast membrane waveguide supports a fundamental mode with a

cos2-distribution instead of Gaussian.

Figure 4.3: Power coupling between the mode distribution of an IMOS waveg-uide with varying width wwg and the fundamental mode of an optical fiber witha fiber core diameter of 9 and 10 µm. The parameter wwg is indicated in Fig.4.2b. The coupling between the mode distributions was calculated with MODE

Solutions (Lumerical) through an overlap integral.

4.2 Dielectric grating couplers

Several designs have been proposed to improve the grating coupler efficiency and

the bandwidth, from shallow etched gratings which are simple in terms of design

and fabrication [84], to fully etched complex gratings based on nanostructures

whose main advantage is their fabrication with a single etching step [85], and

metal-based grating couplers [86–88], which are discussed in the next section.

A critical parameter for the optimization of dielectric grating couplers is the buffer

thickness. Since the grating produces diffraction upwards and downwards, the re-

flection from the buffer-to-substrate interface adds to the initial upward diffraction

67


thereby causing interference. Therefore, the efficiency performance as a function

of the buffer thickness is periodic due to the constructive and destructive interfer-

ence effect as shown in Fig. 4.4. It varies with a periodicity of ∼ 500 nm, which

is about half wavelength in the material, i.e. λ/(2nBCB) for λ = 1550 nm and

nBCB = 1.54. In the IMOS technology, a buffer of around 1850 nm is expected to

maximize the performance of grating couplers as well as to provide enough optical

isolation between the photonic layer and the Si-substrate, however, it might vary

as a result of fabrication imperfections.

Figure 4.4: Upwards chip-to-free space coupling efficiency as a function of thebuffer thickness tbuffer. The parameter tbuffer is indicated in Fig. 4.2b. The

buffer layer stack is SiO2(50nm)/BCB.

Although the dielectric grating couplers are not tolerant to buffer thickness vari-

ations, the nanopillar LEDs reported in section 3.3 were characterized through

this type of grating coupler due to their fabrication simplicity. Due to the strong

light confinement in the nanocavities described in this thesis, their cavity size has

a strong impact on the resonant wavelength (i.e the free spectral range is in the

order of hundreds of nanometers). Therefore fabrication errors produce a signifi-

cant change in the spectral emission and because of that we designed a broadband

grating coupler.

Figure 4.5 shows the simulated grating coupler performance targeted for the char-

acterization of the nanopillar LEDs. Although the actual characterization was

done through free space optics, the simulated chip-to-fiber efficiency is also shown.

The design was performed with Lumerical (FDTD Solutions), for a 250 nm thick

InP-waveguide, a grating period Λ = 680nm, a 50% filling factor, a grating etch

depth of 100 nm and a buffer thickness of 1850 µm. Due to the narrow band-

width of the chip-to-fiber coupling and the uncertainty of the cavity emission,

68


Figure 4.5: (a) Performance of grating coupler designed to characterize thenanopillar LEDs reported in section 3.3. (b) SEM photo of a typical IMOSdielectric grating coupler. The dashed lines indicate the directions depicted in

the schematics of Fig. 4.2.

the nanopillar LEDs were characterized through free-space using a high numerical

aperture (NA = 0.42) lens that collects light diffracted up to an angle of 24, as

described in detail in 3.3.

4.3 Metal grating couplers

Grating couplers based on metallic structures have been proposed during the last

years to improve their coupling efficiency. Figure 4.6 shows general schematics of

some variations of metal-based grating couplers.

Figure 4.6a depicts the case where a metal layer is placed at the bottom of the

buffer layer to reflect the downwards diffraction, which increases the coupling

efficiency up to 78% [86]. Another design consists of a metal grating deposited

on top of a waveguide by lift-off. The metal elements produce a strong coherent

scattering and calculated efficiencies up to 60% with a 1 dB-bandwidth of 40 nm

have been reported [87]. While these designs exhibit better coupling efficiencies

than dielectric gratings, the efficiency is strongly dependent on the buffer layer

thickness below the photonic membrane.

More recently, the design of a buried metal grating was reported [88], which forbids

diffraction to the bottom due to a photonic bandgap effect, understanding the

grating as a one-dimensional metallic photonic crystal. This structure is depicted

in Fig. 4.6b. The coupling efficiency in this approach is independent of the buffer

layer thickness, but it requires a 600 nm thick metal grating. Fabricating metal

69


Figure 4.6: (a) Grating coupler with metal mirror at bottom of buffer layer.(b) Buried metal grating acting as a metallic photonic crystal. (c) Grating

coupler with both, a buried metal grating and a metal mirror.

gratings with the required width below 300 nm in such thick layers is extremely

challenging using standard lift-off or etching processes.

We proposed a novel metal-based grating coupler (depicted in Fig. 4.6c) [42, 78],

which combines the advantages of previously reported devices by using both a

buried metal grating and a metal mirror layer which inhibits power leaking into the

substrate. The inclusion of the metal mirror layer allows for a shallow metal grating

which is more simple to fabricate. An important advantage is that the coupling

efficiency is independent from the underlying layer stack, enabling its use in various

applications. For example, a thin buffer layer is required to achieve optical coupling

for the heterogeneous integration of III-V and silicon photonics, whereas a thick

buffer is of interest for thermal isolation between photonic membranes and CMOS

circuits. The grating design works with maximum efficiency in both schemes.

Furthermore, the thickness independence relaxes the requirements in the bonding

process, simplifying the fabrication technology. In the following, the design of this

metal grating coupler for the IMOS platform is presented.

4.3.1 Design

The design of the metal-grating coupler is carried out for TE polarization by means

of two-dimensional FDTD simulations. Initially, the grating coupler is optimized

for high fiber-to-chip coupling. In a second step, apodization schemes are proposed

to increase the coupling even further. Finally, the device tolerance is studied to

determine the technological challenges for its fabrication.

The proposed device has been designed for a 300 nm thick InP-membrane waveg-

uide. The bonding layer consists of BCB, whose refractive index is 1.54 at 1.55µm.

Figure 4.7 shows a schematic representation of the metal grating coupler design.

It consists of alternating stripes of silver and silicon dioxide below an InP waveg-

uide. In the simulations, we consider 20 periods resulting in a grating length of

70


Figure 4.7: (a) General schematic of the reflective metal grating coupler. (b)Main parameters of the grating.

12.7µm, which is enough to provide a good mode size matching with the fiber

mode. Further, we consider a 200 nm thick metal mirror. The modeling has been

done considering wavelength dependent refractive index models for InP, Si and

SiO2 [57], as well as for Ag [54].

Before performing detailed FDTD-studies we estimate the grating period using

the Bragg condition

sin θ = (neff −mλ/Λ)/nc (4.1)

, where θ is the diffraction angle, neff is the effective index of the mode propagating

along the grating, m is the diffraction order, λ is the wavelength, Λ is the grating

period and nc is the refractive index of the cladding (air in this case).

Using a mode solver, we calculate the mode index of the modes propagating in

the slab regions Air/InP/SiO2 and Air/InP/Ag, which are n1 = 2.7242, and

n2 = 2.5394 + 0.0008i, respectively, for a 300 nm thick InP-membrane. Con-

sidering a 50% filling factor, defined as ff = w/Λ, the grating mode index can be

approximated as neff = Ren1 + n2/2. If we further consider θ = 10, m = 1,

λ = 1.55µm and nc = 1, it follows from the Bragg condition that the grating

period is 631 nm. In the refined 2D FDTD simulations, a grating with period

Λ = 635 nm and depth d = 125 nm result in the highest efficiency at 1.55µm as

discussed in the following. For the simulation, we considered a 9 µm core diam-

eter fiber with index contrast of 0.005 between core and cladding, tilted at 10,

placed 10 µm far from the grating and with an anti-reflective coating for 1.55 µm

71


at the facet. The fundamental fiber mode is excited and the power coupled into

the fundamental mode (TE polarized) of the waveguide is calculated.

In order to couple the light from the InP-membrane to the optical fiber with high

efficiencies, the grating strength has to be tuned to get a good match between

the decaying mode in the grating and the fiber mode. In a uniform grating the

mode will decay exponentially and, consequently, the outcoupled field will also

have an exponential shape. The shape can be modified by modifying the grating

strength along the grating. Figure 4.8a, left axis, shows the coupling efficiency

as a function of the grating depth d for different grating periods. We calculate a

peak efficiency of 73% when d = 175 nm and d = 125 nm, for Λ = 630 nm and

Λ = 635 nm respectively. The latest case is preferred, as it has a shallower grating

depth which will facilitate the fabrication. As a reference, Fig. 4.8a also shows

the total diffraction efficiency of the grating for the reciprocal case in the right

axis, i.e. when the light is diffracted from the waveguide into air. The diffraction

efficiency increases to 92% for d > 200 nm. In this case the residual 8% contains

the following contributions: 2% of the power is reflected back into the fundamental

mode of the waveguide, 2% is lost by metal absorption, 1% is transmitted through

the grating coupler and 3% corresponds to scattering loss at the beginning of the

grating coupler.

Figure 4.8: (a) Left axis: fiber-to-chip coupling efficiency at 1.55 µm as afunction of grating depth d for different grating periods, considering a fillingfactor ff = 50%. Right axis: total upwards diffraction efficiency for the caseof Λ = 635 nm. The inset shows the modulus of the electric-field distributionshowing the coupling from a fiber to an IMOS waveguide at 1.55 µm. Right:Fiber-to-chip coupling efficiency for both, apodized and non-apodized designs,with Λ = 635 nm, d = 125 nm and ff = 50% for the non-apodized design. The

filling factor of apodized designs is shown in Table 4.1.

72


Figure 4.8b (green line) shows the spectral distribution of the coupling efficiency

for the optimum case (Λ = 635 nm and d = 125 nm). The result is in good

agreement with grating couplers theory, which states that the maximum overlap

between an ideal exponential intensity distribution and a Gaussian distribution is

80% [83]. The coupling efficiency can be further increased by apodization of the

grating. Here, the grating strength is varied along the coupler in order to shape

the field decay in such a way that its overlap with the fiber mode is maximized.

Since the power decays exponentially as it propagates along a grating coupler with

constant strength, this can be achieved by varying the coupling strength from weak

to strong along the direction of propagation.

Apodization type Filling factor α-value Efficiency, η

(1) Linear αn 0.022 89%(2) Increasing exponential αn − 1 1.020 89%(3) Asymptotic exponential 1− αn 0.975 89%

Table 4.1: Parameters of different apodization schemes. The correspondinggrating coupler performance is shown in Fig. 4.8

The apodization of the grating can be performed with any parameter that in-

fluences the grating strength. We varied the filling factor because it is easier to

change this parameter rather than the grating depth during fabrication. Table 4.1

summarizes the investigated apodization models, which were optimized for a high

coupling efficiency. All of them consist of a formula to calculate the filling factor

along the grating, where n is the number of the grating period (varying from 1

to 20) and α is the tuning parameter. We optimized the α-values to maximize

the coupling efficiency for the different models. The resulting α–values and peak

efficiencies are shown in 4.1, column 3 and 4, respectively.

A simple linear apodization which increases the filling factor at a rate of 2.2%

per grating period leads to a coupling efficiency of 89%, whereas an increasing

exponential model and an asymptotic exponential apodization result in the same

peak coupling efficiency. The result of both, apodized and non-apodized grating

couplers, are shown in Fig. 4.8b. As it is seen, the resonant wavelength is slightly

red-shifted for the case of apodized gratings due to the fact that we varied only

the filling factor but kept the periodicity constant. It is interesting to note that

these three different apodization models produce a similar result in terms of the

spectral efficiency distribution, what is an indication of the high tolerance of the

apodization of such metal grating coupler.

73


Figure 4.9: Tolerance of the grating coupler on (a) the filling factor and (b)the grating depth. The grating parameters are: Λ = 635 nm, (a) d = 125 nm

and (b) ff = 50%.

Finally, we investigated the tolerance of the non-apodized device on the main

parameters that can compromise the grating efficiency due to fabrication defects.

The filling factor is the most difficult parameter to control, since it depends not

only on the electron-beam exposure parameters, but also on the wafer layer stack

due to the forward and backward electron scattering. Figure 4.9a shows that the

variation of the filling factor will produce a wavelength-shift as well as an efficiency

drop. We observe a reduction by 16% for ff = 60%. On the other hand, there

is a reasonable tolerance on the grating depth as shown in Fig. 4.9b. Negligible

peak-shifts and efficiency variations are observed for a grating depth between 125

nm and 175 nm. Furthermore, we also investigated the tolerance on a fiber height

change of 10 µm, which resulted in a negligible efficiency reduction (less than 1%).

4.3.2 Fabrication and characterization 1

The metal grating coupler was fabricated in the IMOS platform by a double-sided

processing similar to the fabrication process reported in [86]. The fabrication

process flow is depicted in Fig. 4.10a. In a first step, the dielectric grating (here

made of SiO2) is fabricated with e-beam lithography and dry etching on top of the

semiconductor wafer. Afterwards, the metal layer is deposited by evaporation in

a lift-off process, forming both the metal grating itself and the metal mirror in a

single fabrication step. Since silver does not have a good adhesion to InP neither

1The reported work on the metal grating coupler has been done in cooperation with A. Higuera-Rodriguez. V.D.-C. did the design of the couplers, and A.H.-R. did the development of the fabricationprocess and the experimental characterization.

74


to SiO2, the deposition of a thin adhesion layer (few nanometers) is required before

the actual silver deposition, which could be either Ti, Cr or Ge. The latest is the

best choice as it does not introduce a large absorption loss to the device. Later,

adhesive bonding is done on a silicon wafer by BCB [89]. Then, the InP substrate is

removed by selective wet etching, using the waveguide layer as an etch-stop layer.

Finally, the membrane waveguide is e-beam patterned and dry etched. Figure

4.10 shows the experimental results of the device with a chip-to-fiber coupling

efficiency of 54% and 3-dB bandwidth of 61 nm [42]. The difference with respect

to the simulated efficiency (green line in Fig. 4.8b) may be due to a metal loss

higher than expected.

Figure 4.10: Left: Process flow for the fabrication of metal grating couplers.(b) Experimental result of the grating coupler efficiency. The inset shows a

SEM picture of a Focused Ion Beam (FIB) cut along the grating.

4.4 Conclusions

In this chapter we discussed the design of grating couplers. For the case of typ-

ical dielectric IMOS grating couplers, a buffer thickness of 1850 nm was found

to optimize the coupling efficiency, and a grating coupler width of around 15 µm

was found to provide a good transverse match between the waveguide mode and

the fundamental mode of an optical fiber. The design of a broadband (non-fiber

coupled) IMOS grating coupler for the characterization of the nanopillar LEDs

discussed in this thesis was presented, operating with a chip-to-free space diffrac-

tion efficiency above 40% at diffraction angles below 24 for the wavelength range

1.4− 1.6 µm.

75


Furthermore, a novel metal grating coupler design that is independent from the

buffer layer thickness was proposed and optimized for high coupling efficiency with

a single mode optical fiber. The device shows a theoretical coupling efficiency of

73% for a non-apodized design, and an efficiency as high as 89% for an apodized

design with a 3 dB-bandwidth of 61 nm and 78 nm, respectively. The experimental

realization of the non-apodized design resulted in a coupling efficiency of 54%

with a 3-dB bandwidth of 61 nm. In view of its high efficiency, independence

from the buffer thickness and ease of fabrication, we consider this device as very

promising for applications for which power budget is of key importance and layer

stack flexibility is required.

76

Chapter 5

Polarization rotator

We present the design of a novel polarization rotator based on an adiabatic transi-

tion between two single-mode single-polarization (SMSP) waveguides achieved by

mode cut-off. The concept is reported in [90]. The device can be realized in both

photonic platforms, IMOS and SOI. The theoretically predicted performance of

the device is very high: polarization conversion efficiency as high as 99.9%, negli-

gible insertion loss and bandwidth larger than 150 nm for a conversion efficiency

above 99%. Furthermore, its fabrication is relatively simple and can be easily

integrated in a generic process.

Section 5.1 provides a brief introduction to integrated polarization rotators and

proposes the device based on SMSP waveguides. Section 5.2 describes the device

design. The fabrication is discussed in section 5.3 and finally the conclusions are

presented in section 5.4.

5.1 Introduction

In both photonic platforms IMOS and SOI, which have waveguides with high index

contrast, the birefringence is very sensitive to small changes in the waveguide

dimensions [91] and, therefore, the tolerances for the waveguide dimensions are

extremely tight. Because of this, having a photonic platform with polarization

diversity is fundamental to guarantee the proper functionality of PICs regardless

of the input/emitted polarization of light [92]. Furthermore, polarization handling

enables polarization division multiplexing which adds a degree of freedom in the

search for higher bandwidth [93].

77

Chapter 5. Polarization rotator

Several polarization converters (PCs) have been proposed. Generally there are

two kinds of PCs: those based on mode interference and those which rely on

adiabatic mode evolution [94]. The first type of devices are typically short, but

they have a narrow bandwidth and their performance is sensitive to the device

length. Examples of this kind are a compact (< 10 µm) PC demonstrated for

IMOS with a Polarization Conversion Efficiency (PCE) higher than 99% and 1.2

dB insertion loss (IL) [43], as well as a 5 µm long plasmonic-based polarization

rotator (PR) proposed for SOI, with extinction ratio (ER) above 14 dB, IL = 2.1

dB and 3-dB bandwidth larger than 150 nm [95]. Mode evolution converters are

typically broadband and require a long device to achieve adiabatic behavior. For

example, Zhang et. al. [96] demonstrated a PR device with ER = 15 dB and

IL < 1 dB, for a device length of 40 µm.

Figure 5.1 shows a schematic of the polarization rotator proposed in this thesis

based on adiabatic mode evolution. The rotator itself consists of an adiabatic

transition that connects a SMSP TE waveguide with a SMSP TM waveguide.

The use of SMSP waveguides guarantees a high PCE and also makes the device

tolerant in length as there is no energy transfer to other modes as discussed later.

In Fig. 5.1, the buffer layer consists of a combination of SiO2 and a BCB bonding

layer for the case of IMOS, whereas it is only SiO2 for SOI.

Figure 5.1: General schematic of the polarization rotator for IMOS and SOI.The insets show the mode profile (modulus squared of the electric field) at dif-ferent cross sections for λ = 1.55 µm. (a) Quasi-TE mode (b) Hybrid polarized

mode (c) Quasi-TM mode. Blue: low intensity. Red: high intensity.

While most of the devices reported in literature focus only on achieving a high

PCE, our design achieves overall high performance (i.e. ultra-high PCE, very

low IL and a large bandwidth) with a relatively short device length. Since the

proposed device is based on SMSP waveguides, it allows by design (since the

78


undesired polarized mode is cut-off) a fast conversion while keeping an ultra-high

PCE.

5.2 Design of a high performance polarization rotator

For a practical implementation, the polarization rotator should be connected to a

standard waveguide that supports both TE and TM polarizations. Waveguides in

membranes of 300 nm or thickner, which are frequently used in both IMOS and

SOI, always support both polarizations. Figure 5.2 shows a schematic of the full

converter structure, which consists of three sections. In section 1, a TE-polarized

mode is firstly coupled from a dual-polarization waveguide (a) to a SMSP TE

waveguide (c), which is cut-off for the TM mode due to its particular geome-

try. This transition is done in an adiabatic way through a symmetric waveguide

(b). Later, in section 2, the polarization rotation takes place by means of a fully

adiabatic transition through an asymmetric waveguide (d) from the SMSP TE

waveguide (c) to the SMSP TM waveguide structure (e). Finally, in section 3, the

SMSP TM waveguide (e) is tapered into a standard (dual-polarization) waveguide

(f).

Figure 5.2: Schematic of polarization rotator device. Waveguides cross sec-tions: (a) Dual-polarization waveguide. (b) Symmetric transition from dualwaveguide to SMSP TE waveguide. (c) SMSP TE waveguide. (d) Asymmetrictransition from SMSP TE to SMSP TM waveguide. (e) SMSP TM waveguide.

(f) Dual-polarization waveguide.

In this contribution we focus on the design of the rotator itself comprising section

2 in Fig. 5.2. We verified with 3D FDTD simulations that an adiabatic transition

79


(i.e. transmission> 99.9%) in the TE and TM tapers of sections 1 and 3, can be

achieved with lengths of 5 µm(3 µm) and 15 µm(30 µm), respectively, for IMOS

and SOI (numbers between brackets). For this, we considered dual-polarization

waveguides with the dimensions 400× 300 nm2 and 400× 270 nm2, for IMOS and

SOI, respectively.

Using a two-dimensional mode solver, we investigated the waveguide dimensions

(width w and thickness t) required to get SMSP operation around 1.55µm. We

calculated the waveguide dispersion (including material dispersion) for the case

when the waveguide is thin enough as to be cut-off for the TM mode, and for the

case when the waveguide is narrow enough to be cut-off for the TE mode. Figures

5.3 and 5.4 show the effective mode indices of the proposed SMSP TM (nTM-wg)

and SMSP TE (nTE-wg) waveguides, for IMOS and SOI, respectively.

Figure 5.3: Effective mode index of IMOS waveguides. The dashed line indi-cates approximately the mode cut-off limit (i.e. refractive index value of SiO2).(a) Dual-polarization waveguide with dimensions 400× 300 nm2 (b) SMSP TMwaveguide with wTM = 270 nm and tTM = 300 nm (c) SMSP TE waveguide

with wTE = 400 nm and tTE = 200 nm.

As can be observed in Fig. 5.3b, a 270 nm wide IMOS waveguide is cut-off for

the TE mode, whereas it still supports the TM mode. Oppositely, a 200 nm

thick waveguide is cut-off for the TM mode and supports only the TE mode as

it is shown in Fig. 5.3c. A similar situation is found in SOI for a 250 nm wide

waveguide (TE cut-off) and a 190 nm thick waveguide (TM cut-off) as shown in

Figs. 5.4b and 5.4c. The main difference between these photonic platforms is that

the effective mode index difference between TE and TM mode in SOI is larger

than in IMOS due to the higher refractive index of silicon compared to indium

80


phosphide. This characteristic influences the taper length required for adiabatic

conversion as it is discussed later.

Figure 5.4: Effective mode index of SOI waveguides. The dashed line indicatesapproximately the mode cut-off limit (i.e. refractive index value of SiO2). (a)Dual-polarization waveguide with dimensions 400 × 270 nm2 (b) SMSP TMwaveguide with wTM = 250 nm and tTM = 270 nm (c) SMSP TE waveguide

with wTE = 400 nm and tTE = 190 nm.

After defining the two SMSP waveguides in both platforms, we investigated the

performance of the PR as a function of length with three-dimensional FDTD sim-

ulations. In such simulations, we considered the transmission from the SMSP

TE waveguide to the SMSP TM waveguide. Figure 5.5 shows the rotation from

the TE-mode (injected at the left side) to a TM-mode (right side). It is worth

to mention that even when the waveguides work in a cut-off regime, the unde-

sired polarized mode might still be transmitted for a short taper length since the

power in a non-adiabatic transition can be carried by radiation modes through the

substrate. Such undesired transmission is higher for shorter taper lengths.

In Figure 5.6, the polarization conversion efficiency is defined as PCE = 100 ·PTM/(PTM +PTE), where PTM and PTE are the power in the TE and TM modes,

respectively, at the PR output. Additionally the insertion loss is defined as IL =

10 · log(PTEin/PTE + PTM), where PTEin is the power in the TE mode at the PR

input. As can be observed in Fig. 5.6, the PCE constantly increases with the taper

length which confirms the adiabatic behavior, reaching PCE > 99.9% (ER >

30dB) in both platforms. Therefore, if it is sufficiently long, the rotator is tolerant

in length. Previously reported polarization rotators have an oscillatory efficiency

behavior in length due to their multi-mode characteristics [97]. In general, it can

be seen that the SOI-based rotator requires twice the length of the IMOS rotator

81


Figure 5.5: Cross section (top view) of a 10µm long PR showing the dominantelectric field components (a) ReEx and (b) ReEy, during polarization con-version for λ = 1.55µm. The color plots are in arbitrary units and the aspect

ratio was adjusted for easy visualization.

to achieve nearly the same conversion efficiency. This can be explained by the

larger difference in effective mode index between the converted modes in the SOI

waveguides (∆n = 0.41) compared to IMOS (∆n = 0.13) which can be observed

in Figs. 5.3 and 5.4. Therefore, a longer transition is required to achieve adiabatic

behavior. The rotator insertion losses decrease monotonously with the length, they

are negligible (below 0.01 dB) for a length of 40 µm in both photonic platforms.

Figure 5.6: Polarization conversion efficiency and insertion loss as a functionof taper length. Both cases, IMOS and SOI, are depicted

To our knowledge it is the first PR design offering both ultra-high PCE and neg-

ligible insertion loss. This is only possible due to (1) the adiabatic transition and

(2) the SMSP waveguides based on mode cut-off at the rotator ports.

We also carried out broadband FDTD simulations to investigate the bandwidth

82


Figure 5.7: Bandwidth of the PR device for a length of 20 µm (left axis) and40 µm (right axis). The dashed line indicates the level of PCE = 99%. Both

photonic platforms are depicted.

of the device; Fig. 5.7 shows the results. A large bandwidth was obtained, which

is typical for adiabatic transitions. For a 40 µm long rotator, we find a bandwidth

(PCE > 90%) larger than 180 nm and 130 nm, for IMOS and SOI, respectively.

The conversion efficiency for a shorter rotator of 20 µm is also plotted for com-

parison.

Regarding the fabrication tolerance, we notice that the waveguide widths can

be well controlled with lithographic processes since they depend directly on the

mask pattern. Additionally, the TM waveguide thickness, tTM , is well controlled

through the epitaxial growth of the layer stack. However, since accurate etch

depths are difficult to control we identify that the etching step which defines

the TE waveguide thickness is the most critical fabrication step. Therefore, we

investigated the influence of this parameter on the device performance. As can be

seen in Fig. 5.8, the device shows a high tolerance to this etching step in both

photonic platforms, within a tolerance window larger than 80 nm to guarantee

PCE > 99% at 1550 nm. The insertion loss remains below 0.3 dB for such

window. An etching accuracy better than 80 nm can easily be achieved with

current state-of-the-art clean room equipment.

5.3 Fabrication proposal

The PR device presented can be fabricated with standard clean room processes.

In an IMOS platform, the III-V layer stack is bonded to a Si-substrate by adhesive

83


Figure 5.8: Tolerance of the PCE on a fabrication error of ±50 nm aroundthe ideal TE waveguide thickness. Left axis: IMOS platform. Right axis: SOI

platform.

bonding with BCB and then the III-V substrate is selectively wet etched. Then,

an etching of the full waveguiding layer (i.e. tTM etch depth) is done by means of a

first e-beam lithography (EBL) thereby creating waveguides (a), (e) and (f) in Fig.

5.2. Later in a second EBL, the waveguide top can be etched to reach the desired

thickness tTE of the SMSP TE waveguide at the same time as dielectric grating

couplers are etched (waveguide profiles (b) (c) and (d) of Fig. 5.2 are completed

in this step). In a SOI platform, the device can be fabricated in the same way

although without the need of the bonding process. So, if the etching step for

the dielectric gratings can be combined with the etching step for the polarization

rotator, the rotator requires no additional process steps.

5.4 Conclusions

We proposed a polarization rotator for membrane photonic circuits, compatible

with both IMOS and SOI. The rotation is performed in an adiabatic transition

between two SMSP waveguides. The waveguide dimensions are such that the

undesired polarized mode is cut-off at each device port. Simulations predict a high

efficiency of 99.9% and a negligible insertion loss over a bandwidth larger than 150

nm in both photonic platforms, and the smallest device footprint in IMOS. The

device performance was shown to be highly tolerant to the most critical fabrication

step. Its fabrication can be integrated in a generic process in the same two etching

steps required for passive components.

84

Chapter 6

Conclusions and outlook

Research on devices and fabrication technology for photonic integration in III-V

membranes on silicon was reported. The investigated devices are: metal-cavity

nanoscale light sources coupled to waveguides (lasers and LEDs), a novel metal

grating coupler and a high performance polarization rotator. The results reported

in this thesis contribute to the continuous development of photonic integration

technologies, particularly to the IMOS platform proposed at TU/e. However, some

of the results can be applied also to other photonic platforms, for example, the

polarization rotator can be implemented in silicon photonics, whereas the metal

grating couplers can enable vertical coupling in the integration of mature generic

InP on silicon substrates.

We demonstrated the first metal-cavity nanopillar LED coupled to a waveguide on

silicon. The device was characterized through grating couplers and showed up to

60 nW measured output power and relatively high on-chip external quantum effi-

ciency in the range of 10−4-10−2 at room temperature and 9 K, respectively. This

greatly exceeds the performance of previously reported nanoscale light sources.

Furthermore, modulation experiments revealed sub-nanosecond electro-optical re-

sponse with potential for multi-Gbps modulation speeds. Future developments to

reach lasing in this device can focus on:

• Studies on passivation techniques to reduce the surface recombination in

InGaAs found to be significant in the nanopillars. This may enable lasing

or result in nanopillar LEDs with higher efficiency. A passivation technique

for these devices will be useful if (1) it consists of a low refraction index

material able to provide strong dielectric confinement of the mode, so that

metal losses are not increased, and (2) the quality of the passivation remains

85

Chapter 6. Conclusions and outlook

high (or improves) after high temperature processing steps required for the

subsequent annealing of the silver cladding and ohmic contacts.

• The optimization of the ICP-RIE etching of nanopillars to result in more

vertical sidewalls, which decreases the cavity radiation losses. A more vertical

etching can be obtained by process optimization, although it is challenging.

• The fabrication of pillars with an undercut around the active medium to

increase the quality factor as mentioned at the end of Chapter 3. Due to the

slopes introduced by the wet etching, the fabrication of this undercut is only

practical in pillars with ∼ 1 µm diameter or larger.

A novel metal grating coupler was also proposed. This device combines a metal

grating with a metal reflector in order to result in a high performance coupler

(73% and 89% efficiency for a uniform and non-uniform design, respectively),

whose performance is independent from the buffer thickness. The device structure

avoids power leakage into the substrate which is expected to result in highly reliable

grating couplers. Its experimental realization showed 54% coupling efficiency for

a uniform grating. We believe it represents a promising device to be used in

applications that require layer stack flexibility. Further developments of this device

could include:

• The fabrication of the non-uniform design described in this thesis which

should result in increased coupling efficiency.

• The design and experimental realization of focusing metal grating couplers

which combine the advantages of our device with a small footprint.

• The application of this device to beam steering. In this case, additional

electrical contacts for current injection should be fabricated. Two factors are

expected to have a relevant role for the steering: the change in refractive

index due temperature increase and due to high current density in the thin

semiconductor layers.

• The integration of metal grating couplers in the generic InP platforms whose

next largest technological target is their integration with CMOS electronics.

The grating coupler can be fabricated at the last steps of the III-V photonic

chip and before the bonding onto a CMOS chip.

Finally, a polarization converter device was proposed and designed for both IMOS

and SOI platforms. It is based on single-mode single-polarization waveguides and

86

Chapter 6. Conclusions and outlook

according to our design it shows ultra-high performance, namely polarization con-

version efficiency > 99.9%, a bandwidth larger than 150 nm for PCE above 99%

and negligible insertion loss, for a length of 40 µm. It is, to our knowledge, the po-

larization converter for photonic integration with the best theoretical performance.

Furthermore, its fabrication is relatively simple and fully compatible with the fab-

rication of standard passive elements (waveguides and dielectric grating couplers),

therefore we believe it is the ideal integrated polarization converter for both plat-

forms. Its fabrication is underway in a collaborative project and experimental

results may be achieved soon.

87

Appendix A

Main patterning processes

Chemistry(Volume ratio)

Etches Etch rate

H3PO4:HCl (4:1) InP ∼ 400 nm/min

H2SO4:H2O2:H2O (1:1:10) InGaAsInGaAsP

∼ 900 nm/min∼ 50 nm/min

H3PO4:H2O (1:10) InP-based oxides Nanometer scale

BHF SiO2

Si3N4

> 200 nm/min> 50 nm/min

HF 1% SiO2

Si3N4

< 30 nm/minN.A.

KCN (Degussa) Ag, Au ∼ 200 nm/min

Table A.1: Main wet etching recipes used for the patterning of semiconductors,dielectrics and metals. All etch rates are for room temperature. N.A. stands

for not available.

Name Chemistry(sccm)

Recipe details Etches Rate

ICP Etch: CH4:H2

(30:70)Descum: O2

(60)

Etch: 100W RF,100W ICP, 6mT,150CDescum: 75C

InPInGaAsInGaAsP

∼ 38 nm/min∼ 20 nm/min∼ 25 nm/min

Nitride RIE CHF3 (60) 50 W, 15 mT SiO2

Si3N4

∼ 18 nm/min∼ 23 nm/min

Polymer RIE O2 (100) 150 W, 100 mT HPR504 ∼ 63 nm/min

Table A.2: Main dry etching recipes used for the patterning of semiconductorsand dielectrics. R.T. stands for room temperature.

89

Appendix A. Main patterning processes

Chemistry Resist

MaD531S HSQMaN440AZ4533

n-Amyl Acetaat ZEP520A

Table A.3: Main developing solutions used for e-beam and optical resists.All developers were used at room temperature, except HSQ development whichwas done at 60C. Such temperature was previously found to provide a betterstructure definition [51]. After development, the samples are water rinsed exceptwhen developing ZEP5020A with n-Amyl Acetaat which has to be rinsed in

MIBK:IPA (89:11).

90

Appendix B

Small-signal frequency response

In section 3.1.4, the small-signal frequency response was calculated according to

[60]

H(ω) =ω2R

ω2R − ω2 + jωγ

, (B.1)

where ωR is the relaxation resonance frequency and γ = γNN +γPP is the damping

factor with γNN = 1/τ∆N + vgaNp and γPP = ΓvgapNp. The relaxation resonance

frequency that can be approximated as ωR ≈vgaNp

τp[60], where τp is the photon

lifetime τp.

B.1. List of parameters

Parameter Value Description

Γ 0.33 Confinement factorvg 8.8235 · 109 cm s−1 Group velocityNp 8.3440 · 1015 cm−3 for Pout = 50 µW Photon densityτ∆N 1.2564 ns Differential carrier lifetimea 4.1118 · 10−16 cm2 ∂g/∂Nap 4.2978 · 10−15 cm2 ∂g/∂Np

Table B.1: Parameters used to calculate the small-signal frequency responseof the metallo-dielectric nanolaser. The parameters were approximated usingstandard formulas available in [60], based on the simulated optical and electricalperformance presented in section 3.1.1.2 for a nanopillar with InGaAs activemedium of dimensions 400×300×350nm3 (length×width×thickness). Np stands

for photon density and N for carrier density.

91

Appendix C

Semiconductor single-mode rate

equations model 1

In order to interpret further the measurement results of Fig. 3.32a and Fig. 3.33a,

we used the following two-level semiconductor single-mode rate equations system

to model the carrier population N and the photon number Nph:

N =ηiI

q−Rr −Rnr (C.1)

Nph = Rr,cav −Rp. (C.2)

This system describes an electrically pumped nanopillar LED with injection cur-

rent I, where q is the electron charge and ηi the injection efficiency. The rate of

stimulated emission is assumed to be negligible and it is not included in the model.

The rates describe the physical processes occurring in the LED including the total

spontaneous emission recombination rate, Rr = BN2/Va, where B is the bimolec-

ular recombination coefficient, and Va the volume of the active material; the ra-

diative decay rate into the cavity mode, Rr,cav = βBN2/Va, where β is the sponta-

neous emission factor; the non-radiative recombination rate, Rnr =SAN

Va+CN3

V 2a

,

includes both surface recombination described by the surface velocity S, and by

the surface area of the active region A, and Auger recombination, described by C;

and the photon loss rate is given by Rp = Nph/τp, where τp =λcQ

2πcis the photon

lifetime which is determined from the cavity Q-factor and the cavity resonance

wavelength λc, where c is the speed of light.

1The numerical fitting was carried out in cooperation with Dr. B. Romeira. B.R. implemented therate equations model.

93

Appendix C. Semiconductor single-mode rate equations model

For the fitting of the L-I characteristics, Fig. 3.32a, the photon population was

converted to an output power P = Nphηhc/τpλc, where h is the Planck’s constant,

and the parameter η was fixed to a value ∼ 4×10−2 and accounts for the waveguide

coupling efficiency, the grating out-coupling efficiency and the setup collection

efficiency. For this fixed value of η, the L-Is showed the best fitting choosing

β = 0.05. In the fitting of the L-I characteristics at low temperature, the non-

radiative recombination rate was neglected and the remaining parameter values

employed are ηi ∼ 0.8, B = 9 × 10−10 cm3s−1, and the measured quality factor

Q = 62. At room temperature, the term Rnr was included and the parameter

values of B = 0.25 × 10−10 cm3s−1 at room temperature [98], C = 9.8 × 10−29

cm6s−1, S = 3.8×104 cm·s−1, and the measured quality factor at room temperature

Q = 37, provided a good agreement with the experimental L-Is. Finally, for the

measured nanocavity presented in Fig. 3.33a, a similar procedure was used to fit

the corresponding modulation properties and L-I characteristics (not reported) at

room temperature. In this case, the parameter S = 2 × 104 cm·s−1 provided the

best fit.

94

Bibliography

[1] All Nobel prizes in Physics. http://www.nobelprize.org/nobel prizes/physics/

laureates, 2015.

[2] International Year of Light. http://www.light2015.org, 2015.

[3] M. K. Weldon. The future X network: a Bell Labs perspective. CRC Press,

2015.

[4] V. Dolores-Calzadilla, D. Pustakhod, X.J.M. Leijtens, and M.K. Smit. Fiber

Bragg grating sensor based on external cavity laser. In 20th Annual Sympo-

sium of the IEEE Photonics Benelux Chapter, 2015.

[5] P. K. Tien. Integrated optics and new wave phenomena waveguides. Reviews

of Modern Physics, 49(2):361–420, 1977.

[6] G. E. Moore. Cramming more components onto integrated circuits. Electron-

ics, 38(144), 1965.

[7] M. Smit, J. van der Tol, and M. Hill. Moore’s law in photonics. Laser and

Photonics Reviews, 6(1):1–13, 2012.

[8] M. Smit, X. Leijtens, H. Ambrosius, E. Bente, J. van der Tol, B. Smal-

brugge, T. de Vries, E. J. Geluk, J. Bolk, R. van Veldhoven, L. Augustin,

P. Thijs, D. D’Agostino, H. Rabbani, K. Lawniczuk, St. Stopinski, S. Tahvili,

A. Corradi, Emil Kleijn, D. Dzibrou, M. Felicetti, E. Bitincka, V. Moskalenko,

J. Zhao, R. Santos, G. Gilardi, W. Yao, K. Williams, P. Stabile, P. Kuinder-

sma, J. Pello, S. Bhat, Y. Jiao, D. Heiss, G. Roelkens, M. Wale, P. Firth,

F. Soares, N. Grote, M. Schell, H. Debregeas, M. Achouche, J.-L. Gentner,

A. Bakker, T. Korthorst, D. Gallagher, A. Dabbs, A. Melloni, F. Morichetti,

D. Melati, A. Wonfor, R. Penty, R. Broeke, B. Musk, and D. Robbins. An

introduction to InP-based generic integration technology. Semiconductor Sci-

ence and Technology, 29(8):083001, 2014.

95

Bibliography

[9] D. A. B. Miller. Optical interconnects to electronic chips. Applied Optics,

49(25):59–70, 2010.

[10] J. Leuthold, C. Hoessbacher, A. Melikyan, S. Muehlbrandt, M. Kohl, C. Koos,

W. Freude, V. Dolores-Calzadilla, M. Smit, I Suarez, J. Martinez-Pastor, E.P.

Fitrakis, and I. Tomkos. Plasmonic communications: light on a wire. Optics

and Photonics News, (May 2013):28–35, 2013.

[11] D. Liang and J. E. Bowers. Recent progress in lasers on silicon. Nature

Photonics, 4(7):511–517, 2010.

[12] G. Roelkens, L. Liu, D. Liang, R. Jones, A. Fang, B. Koch, and J. Bowers.

III-V/silicon photonics for on-chip and intra-chip optical interconnects. Laser

and Photonics Reviews, 4(6):751–779, 2010.

[13] E. M. Purcell. Spontaneous emission probabilities at radio frequencies. Phys-

ical Review Letters, 69(681), 1946.

[14] K. Takeda, T. Sato, A. Shinya, K. Nozaki, W. Kobayashi, H. Taniyama,

M. Notomi, K. Hasebe, T. Kakitsuka, and S. Matsuo. Few-fJ/bit data

transmissions using directly modulated lambda-scale embedded active region

photonic-crystal lasers. Nature Photonics, 7(5):569–575, 2013.

[15] G. Crosnier, A. Bazin, P. Monnier, S. Bouchoule, R. Braive, G. Beaudoin,

I. Sagnes, R. Raj, and F. Raineri. High Q-factor InP photonic crystal

nanobeam cavities for laser emission. In 26th International Conference on

Indium Phosphide and Related Materials (IPRM), pages Mo–B1–1, Montpel-

lier, France, 2014.

[16] M. T. Hill, Y.-S. Oei, B. Smalbrugge, Y. Zhu, T. de Vries, P.J. van Veldhoven,

F. W. M. van Otten, T. J. Eijkemans, J. P. Turkiewicz, H. de Waardt, E. J.

Geluk, S.-H. Kwon, Y.-H. Lee, R. Notzel, and M. K. Smit. Lasing in metallic-

coated nanocavities. Nature Photonics, 1(9):589–594, September 2007.

[17] M. P. Nezhad, A. Simic, O. Bondarenko, B. Slutsky, A. Mizrahi, L. Feng,

V. Lomakin, and Y. Fainman. Room-temperature subwavelength metallo-

dielectric lasers. Nature Photonics, 4(4):1–5, 2010.

[18] J. H. Lee, M. Khajavikhan, A. Simic, Q. Gu, O. Bondarenko, B. Slutsky,

M. P. Nezhad, and Y. Fainman. Electrically pumped sub-wavelength metallo-

dielectric pedestal pillar lasers. Optics express, 19(22):21524–31, 2011.

96

Bibliography

[19] K. Ding and C. Z. Ning. Metallic subwavelength-cavity semiconductor

nanolasers. Light: Science & Applications, 1(7):e20, July 2012.

[20] M. T. Hill and M. C. Gather. Advances in small lasers. Nature Photonics,

8(11):908–918, November 2014.

[21] M. T. Hill, M. Marell, E. S. P. Leong, B. Smalbrugge, Y. Zhu, M. Sun, P. J.

van Veldhoven, E.J. Geluk, F. Karouta, Y.-S. Oei, R. Notzel, C.-Z. Ning,

and M. K. Smit. Lasing in metal-insulator-metal sub-wavelength plasmonic

waveguides. Optics express, 17(13):11107–12, 2009.

[22] K. Ding, M. T. Hill, Z. C. Liu, L. J. Yin, D. Sahin, P. J. Van Veldhoven,

E. J. Geluk, and T. D. Vries. Room Temperature Lasing in Subwavelength

Cylindrical Metallic Cavity under Pulse Electric Injection. In Conference on

Lasers and Electro-Optics 2012, page CTh4M.4, San Jose, CA, USA, 2012.

[23] K. Ding, Z. Liu, L. Yin, M. Hill, M. Marell, P. van Veldhoven, R. Noetzel, and

C. Ning. Room-temperature continuous wave lasing in deep-subwavelength

metallic cavities under electrical injection. Physical Review B, 85(4):1–5, Jan-

uary 2012.

[24] M. K. Kim, A. M Lakhani, and M. C Wu. Efficient waveguide-coupling of

metal-clad nanolaser cavities. Optics express, 19(23):23504–12, 2011.

[25] V. Dolores-Calzadilla, D. Heiss, A Fiore, and M Smit. Waveguide-coupled

nanolasers in III-V membranes on silicon. In 15th International Conference

on Transparent Optical Networks, page We.D6.1, Cartagena, Spain, 2013.

[26] N. Li, K. Liu, V. J. Sorger, and D. K. Sadana. Monolithic III–V on silicon

plasmonic nanolaser structure for optical interconnects. Scientific Reports,

5:14067–9, 2015.

[27] K. C. Y. Huang, M.-K. Seo, T. Sarmiento, Y. Huo, J. S. Harris, and M. L.

Brongersma. Electrically driven subwavelength optical nanocircuits. Nature

Photonics, 8(2):244–249, 2014.

[28] A. Fiore, J. X. Chen, and M. Ilegems. Scaling quantum-dot light-emitting

diodes to submicrometer sizes. Applied Physics Letters, 81(10):1756–1758,

2002.

[29] M. Francardi, L. Balet, A. Gerardino, N. Chauvin, D. Bitauld, L. H. Li,

B. Alloing, and A. Fiore. Enhanced spontaneous emission in a photonic-

crystal light-emitting diode. Applied Physics Letters, 93(14):143102–3, 2008.

97

Bibliography

[30] G. Shambat, B. Ellis, A. Majumdar, J. Petykiewicz, M. A Mayer,

T. Sarmiento, J. Harris, E. E. Haller, and J. Vuckovic. Ultrafast direct mod-

ulation of a single-mode photonic crystal nanocavity light-emitting diode.

Nature Communications, 2(October):539, 2011.

[31] Q. Gu, J. Shane, F. Vallini, B. Wingad, J. S. T. Smalley, N. C. Frateschi, and

Y. Fainman. Electrically pumped etallo-dielectric pedestal nanolasers with

amorphous Al 2O 3 shield. IEEE Journal of Quantum Electronics, 50(7):499–

509, 2014.

[32] G. Celler and M. Wolf. Smart Cut: a guide to the technology, the process,

the products. Technical report, SOITEC, 2003.

[33] R. Soref. The Past , Present , and Future of Silicon Photonics. IEEE Journal

of Selected Topics in Quantum Electronics, 12(6):1678–1687, 2006.

[34] M. Hochberg and T. Baehr-Jones. Towards fabless silicon photonics. Nature

Photonics, 4(8):492–494, 2010.

[35] J. Van der Tol, R. Zhang, J. Pello, F. Bordas, G. Roelkens, H. Ambrosius,

P. Thijs, F. Karouta, and M. Smit. Photonic integration in indium-phosphide

membranes on silicon. IET Optoelectronics, 5(5):218–225, 2011.

[36] EPIXfab. www.epixfab.eu, 2015.

[37] JePPIX. www.jeppix.eu.

[38] Z. Wang, B. Tian, M. Pantouvaki, W. Guo, P. Absil, J. Van Campenhout,

C. Merckling, and D. Van Thourhout. Room Temperature InP DFB Laser

Array Directly Grown on (001) Silicon. Nature Photonics, 9(12):837–842,

2015.

[39] J. Justice, C. Bower, M. Meitl, M. B. Mooney, M. A. Gubbins, and B. Corbett.

Wafer-scale integration of group III–V lasers on silicon using transfer printing

of epitaxial layers. Nature Photonics, 6(9):612–616, 2012.

[40] Y. Jiao, T. de Vries, R.-S. Unger, L. Shen, H. Ambrosius, C. Radu, M. Arens,

M. Smit, and J. van der Tol. Vertical and smooth single-step reactive ion

etching process for InP membrane waveguides. Journal of the Electrochemical

Society, 162(8):E90–E95, 2015.

[41] J. van der Tol, J. Pello, Y. Jiao, D. Heiss, G. Roelkens, H. Ambrosius, and

M. Smit. Photonic integration in indium-phosphide membranes on silicon. In

SPIE Proceedings, volume 8988, pages 89880M–17, 2014.

98

Bibliography

[42] A. Higuera-Rodriguez, V. Dolores-Calzadilla, Y. Jiao, Geluk E.J., D. Heiss,

and M.K. Smit. Realization of efficient metal grating couplers for membrane-

based integrated photonics. Optics Letters, 40(12):2755–2757, 2015.

[43] J. Pello, J. Van der Tol, S. Keyvaninia, and R. van Veldhoven. High-

efficiency ultrasmall polarization converter in InP membrane. Optics letters,

37(17):3711–3713, 2012.

[44] J. Pello, M. Muneeb, S. Keyvaninia, J. J. G. M. van der Tol, G. Roelkens, and

M. K. Smit. Planar concave grating demultiplexers on an InP-membrane-on-

silicon photonic platform. IEEE Photonics Technology Letters, 25(20):1969–

1972, 2013.

[45] Y. Jiao, D. Heiss, L. Shen, S. Bhat, M. Smit, and J. Van der Tol. First

demonstration of an electrically pumped laser in the InP membrane on silicon

platform. In Advanced Photonics Congress, pages IM4B.3–1/3, Boston, 2015.

[46] IMEC’s silicon photonics platform (ISIPP25G).

http://www2.imec.be/content/user/File/NEW/Services/isipp25g leaflet-

jan30.pdf.

[47] V. Dolores-Calzadilla, A. Fiore, and M. K. Smit. Towards plasmonic lasers

for optical interconnects. In 14th International Conference on Transparent

Optical Networks, page Th.A5.7, Coventry, England, 2012.

[48] D. Heiss, A. Fiore, and M. Smit. Design of a waveguide-coupled nanolaser

for photonic integration. In Advanced Photonics Congress, volume 2, page

IM2A.3, 2013.

[49] V. Dolores-Calzadilla, D. Heiss, A. Fiore, and M. Smit. Metallo-dielectric

nanolaser coupled to an InP- membrane waveguide. In 17th Annual Sympo-

sium of the IEEE Photonics Society Benelux Chapter, pages 195–198, Mons,

Belgium, 2012.

[50] M. T. Hill. Status and prospects for metallic and plasmonic nano-lasers [In-

vited]. Journal of the Optical Society of America B, 27(11):36–44, 2010.

[51] M. Marell. Gap plasmon mode distributed feedback lasers. PhD thesis, Tech-

nische Universiteit Eindhoven, 2011.

[52] C.-Y. Lu and S. L. Chuang. A surface-emitting 3D metal-nanocavity laser:

proposal and theory. Optics express, 19(14):13225–44, 2011.

99

Bibliography

[53] Calculation of metal absorption with FDTD.

https://kb.lumerical.com/en/layout analysis pabs simple.html, 2015.

[54] P. B. Johnson and Christy. R. W. Optical Constants of the Noble Metals.

Physical Review B, 6(12):4370–79, 1972.

[55] Y. Xu, J. S. Vuckovic, R. K. Lee, O. J. Painter, A. Scherer, and A. Yariv.

Finite-difference time-domain calculation of spontaneous emission lifetime in

a microcavity. Journal of the Optical Society of America B, 16(3):465–474,

1999.

[56] A. Auffeves, J. M. Gerard, and J. P. Poizat. Pure emitter dephasing: a

resource for advanced solid-state single-photon sources. Physical Review A,

79(5):053838–5, 2009.

[57] E. D. Palik. Handbook of Optical Constants of Solids. Academic Press, New

York, 1998.

[58] C. H. Henry, R. A. Logan, F. R. Merritt, and C. G. Bethea. Radiative

and nonradiative lifetimes in n-type and p-type 1.6 µm InGaAs. Electronics

Letters, 20(9):358–359, 1984.

[59] K. Ding, M. T. Hill, Z. C. Liu, L. J. Yin, P. J. van Veldhoven, and C. Z.

Ning. Record performance of electrical injection sub-wavelength metallic-

cavity semiconductor lasers at room temperature. Optics express, 21(4):4728–

33, 2013.

[60] L. A. Coldren, S. W. Corzine, and M. L. Masanovic. Diode lasers and photonic

integrated circuits. John Wiley & Sons, Inc., 2nd edition, 2012.

[61] B. Guenin. Thermal interactions between high-power packages and heat sinks.

Electronics Cooling, 16(12), 2010.

[62] Q. Gu, B. Wingad, F. Vallini, B. Slutsky, M. Katz, M. P. Nezhad, N. C.

Frateschi, and Y. Fainman. Electrically pumped metallo-dielectric pedestal

nanolasers. In Conference on Lasers and Electro-Optics - Pacific Rim, pages

Wl4–3, Kyoto, Japan, 2013.

[63] K. Ding and C. Z. Ning. Fabrication challenges of electrical injection metal-

lic cavity semiconductor nanolasers. Semiconductor Science and Technology,

28(12):124002, 2013.

[64] S. Keyvaninia, G. Roelkens, D. Van Thourhout, C. Jany, M. Lamponi, A. Le

Liepvre, F. Lelarge, D. Make, G.-H. Duan, D. Bordel, and J.-M. Fedeli.

100

Bibliography

Demonstration of a heterogeneously integrated III-V/SOI single wavelength

tunable laser. Optics Express, 21(3):3784–92, 2013.

[65] S. Stankovic. Hybrid III-V/Si DFB lasers based on polymer bonding technol-

ogy. PhD thesis, Universiteit Gent, 2013.

[66] J. Pello. Building up a membrane photonics platform in indium phosphide.

PhD thesis, Technische Universiteit Eindhoven, 2014.

[67] D. Lauvernier, M. Carette, J.-P. Vilcot, D. Bernard, and D. Decoster. Simple

technological process for the fabrication of optical III-V nanowires integrated

into a benzocyclobutene matrix. ECS Transactions, 3(6):305–309, 2006.

[68] A. Melikyan, L. Alloatti, A. Muslija, D. Hillerkuss, P. C. Schindler, J. Li,

R. Palmer, D. Korn, S. Muehlbrandt, D. Van Thourhout, B. Chen, R. Dinu,

M. Sommer, C. Koos, M. Kohl, W. Freude, and J. Leuthold. High-speed plas-

monic phase modulators. Nature Photonics, 8(February):229–233, February

2014.

[69] C. Haffner, W. Heni, Y. Fedoryshyn, J. Niegemann, A. Melikyan, D. L. Elder,

B. Baeuerle, Y. Salamin, A. Josten, U. Koch, C. Hoessbacher, F. Ducry,

L. Juchli, A. Emboras, D. Hillerkuss, M. Kohl, L. R. Dalton, C. Hafner,

and J. Leuthold. All-plasmonic Mach–Zehnder modulator enabling optical

high-speed communication at the microscale. Nature Photonics, 9(8):525–

528, 2015.

[70] L. Shen, V. Dolores-Calzadilla, C. W. H. A. Wullems, Y. Jiao, A. Millan-

Mejia, A. Higuera-Rodriguez, D. Heiss, J. J. G. M. van der Tol, H. P. M. M.

Ambrosius, G. Roelkens, and M.K. Smit. Low-optical-loss, low-resistance

Ag/Ge based ohmic contacts to n-type InP for membrane based waveguide

devices. Optical Materials Express, 5(2):393, January 2015.

[71] L. Shen, P. J. van Veldhoven, Y. Jiao, V. Dolores-Calzadilla, J. J. G M.

van der Tol, G. Roelkens, and M. K. Smit. Ohmic contacts with ultra-low

optical loss on heavily doped n-type InGaAs and InGaAsP for InP-based

photonic membranes. Optical Materials Express (submitted), 2015.

[72] M. C. J. C. M. Kramer. Fabrication and Characterization of Semiconductor-

Metal contacts. PhD thesis, Technische Universiteit Eindhoven, 2000.

[73] A. E. Grigorescu and C. W. Hagen. Resists for sub-20-nm electron beam

lithography with a focus on HSQ: state of the art. Nanotechnology,

20(29):292001, July 2009.

101

Bibliography

[74] F. C. M. J. M. van Delft, J.P. Weterings, A. K. van Langen-Suurling, and

H. Romijn. Hydrogen silsesquioxane/novolak bilayer resist for high aspect

ratio nanoscale electron-beam lithography. Journal of Vacuum Science &

Technology B, 18(6):3419, 2000.

[75] F. Pagliano. Dynamic control of the spontaneous emission of single quantum

dots in photonic crystal cavities. PhD thesis, 2014.

[76] M. Boroditsky, I. Gontijo, M. Jackson, R. Vrijen, E. Yablonovitch, T. Krauss,

Chuan-Cheng Cheng, A. Scherer, R Bhat, and M. Krames. Surface recom-

bination measurements on III–V candidate materials for nanostructure light-

emitting diodes. Journal of Applied Physics, 87(7):3497–3504, 2000.

[77] G. P. Agrawal. Fiber-optic communications systems. Third edit edition, 2002.

[78] V. Dolores-Calzadilla, D. Heiss, and M. Smit. Highly efficient metal grating

coupler for membrane-based integrated photonics. Optics letters, 39(9):2786–

9, 2014.

[79] B. W. Snyder and P. A. O’Brien. Developments in packaging and integration

for silicon photonics. In Proc. SPIE - Reliability, Packaging, Testing, and

Characterization of MOEMS/MEMS and Nanodevices XII, volume 8614, page

86140D, San Francisco, CA, USA, 2013.

[80] J. Hofrichter, W. M. J. Green, F. Horst, S. Assefa, M. Yang, B. Offrein,

and Y. Vlasov. Grating couplers as optical probe pads in a standard CMOS

process. In 8th IEEE International Conference on Group IV Photonics, pages

127–129, London, England, 2011.

[81] J. Yao, X. Zheng, G. Li, I. Shubin, H. Thacker, Y. Luo, K. Raj, J. E. Cun-

ningham, and A.V. Krishnamoorthy. Grating-coupler based low-loss optical

interlayer coupling. In IEEE International Conference on Group IV Photon-

ics, pages 383–385, London, England, 2011.

[82] J. Kang, Y. Atsumi, T. Sifer, Y. Hayashi, T. Amemiya, N. Nishiyama, and

S. Arai. Inter-layer grating coupler with metal mirrors for 3D optical inter-

connects. In Conference on Lasers and Electro-Optics - Pacific Rim, pages

MM1–5, Kyoto, Japan, 2013.

[83] T. Tamir and S. T. Peng. Analysis and design of grating couplers. Applied

Physics, 14:235–254, 1977.

102

Bibliography

[84] D. Vermeulen, S. Selvaraja, P. Verheyen, G. Lepage, W. Bogaerts, P. Absil,

D. Van Thourhout, and G. Roelkens. High-efficiency fiber-to-chip grating

couplers realized using an advanced CMOS-compatible silicon-on-insulator

platform. Optics express, 18(17):18278–83, 2010.

[85] R. Halir, P. Cheben, S. Janz, D.-X. Xu, I. Molina-Fernandez, and J. G.

Wanguemert-Perez. Waveguide grating coupler with subwavelength mi-

crostructures. Optics Letters, 34(9):1408, April 2009.

[86] F. Van Laere, G. Roelkens, M. Ayre, J. Schrauwen, D. Taillaert, D. Van

Thourhout, T. F. Krauss, and R. Baets. Compact and highly efficient grat-

ing couplers between optical fiber and nanophotonic waveguides. Journal of

Lightwave Technology, 25(1):151–156, 2007.

[87] S. Scheerlinck, J. Schrauwen, F. Van Laere, D. Taillaert, D. Van Thourhout,

and R. Baets. Efficient, broadband and compact metal grating couplers for

silicon-on-insulator waveguides. Optics express, 15(15):9625–30, July 2007.

[88] P.-T. Lin, C.-Y. Wu, and P.-T. Lee. Buried metal grating for vertical fiber-

waveguide coupling with high directionality. In Advanced Photonics Congress,

page IM4B.6, Washington, D.C., USA, 2012. Osa.

[89] G. Roelkens, D. Van Thourhout, and R. Baets. Ultra-thin benzocyclobutene

bonding of III–V dies onto SOI substrate. Electronics Letters, 41(9):4–5, 2006.

[90] V. Dolores-Calzadilla, J. J. G. M. Van der Tol, and M. K. Smit. Polarization

rotator with high performance for membrane integrated photonics. Optics

Letters (to be submitted), pages 1–4, 2016.

[91] G. T. Reed, G. Z. Mashanovich, W. R. Headley, B. Timotijevic, F. Y. Gardes,

S. P. Chan, P. Waugh, N. G. Emerson, C. E. Png, M. J. Paniccia, A. Liu,

D. Hak, and V. M .N. Passaro. Issues associated with polarization inde-

pendence in silicon photonics. IEEE Journal of Selected Topics in Quantum

Electronics, 12(6):1335–44, 2006.

[92] T. Barwicz, M. R. Watts, M. A. Popovic, P. T. Rakich, L. Socci, F. X.

Kartner, E. P. Ippen, and H. I. Smith. Polarization-transparent micropho-

tonic devices in the strong confinement limit. Nature Photonics, 1(1):57–60,

January 2007.

[93] D. Dai and J.E. Bowers. Silicon-based on-chip multiplexing technologies and

devices for Peta-bit optical interconnects. Nanophotonics, 3(4-5):283–311,

January 2014.

103

Bibliography

[94] M. R. Watts and H. A. Haus. Integrated mode-evolution-based polarization

rotators. Optics Letters, 30(2):138–140, 2005.

[95] J. N. Caspers, M. Z. Alam, and M. Mojahedi. Compact hybrid plasmonic

polarization rotator. Optics Letters, 37(22):4615–4617, 2012.

[96] J. Zhang, M. Yu, G.-Q. P. Lo, and D.-L. Kwong. Silicon-waveguide-based

mode evolution polarization rotator. IEEE Journal of Selected Topics in

Quantum Electronics, 16(1):53–60, 2010.

[97] B. Troia, F. De Leonardis, M. Lanzafame, T. Muciaccia, G. Grasso, G. Gian-

noccaro, C. E. Campanella, and V. M. N. Passaro. Design and optimization of

polarization splitting and rotating devices in silicon-on-insulator technology.

Advances in OptoElectronics, 2014, 2014.

[98] E. Zielinski, H. Schweizer, K. Streubel, H. Eisele, and G. Weimann. Excitonic

transitions and exciton damping processes in InGaAs/InP. Journal of Applied

Physics, 59(6):2196–2204, 1986.

104

List of abbreviations

PIC Photonic integrated circuit

IC Integrated circuit

CMOS Complementary metal-oxide-semiconductor

SiPh Silicon photonics

IMOS Indium-phosphide membranes on silicon

SOI Silicon on insulator

MPW Multi-project wafer

AWG Arrayed waveguide grating

FP Fabry-Perot

MISIM Metal-insulator-semiconductor-insulator-metal

FDTD Finite-difference time-domain

LASER Light amplification by stimulated emission of radiation

LED Light-emitting diode

III-V In1−xGaxAsyP1−y compound semiconductors

(lattice-matched to InP)

Q1.25 InGaAsP

BCB Benzocyclobutene

EBL Electron beam lithography

ZEP Electron-beam resist ZEP520A (positive tone)

HSQ Electron-beam resist FOX XR1541-6% (negative tone)

(Hydrogen silsesquioxane)

HPR Optical resist HPR504

AZ Optical resist AZ4533

BHF Buffered hydrofluoric acid

MOCVD Metal organic chemical vapor deposition

PECVD Plasma-enhanced chemical vapor deposition

RIE Reactive-ion etching

ICP Inductively coupled plasma

CMP Chemical mechanical polishing

105

List of abbreviations

RTA Rapid thermal annealing

SEM Scanning electron microscope

FIB Focused ion beam

NA Numerical aperture

SMF Single mode fiber

RT Room-temperature

PL Photoluminescence

EL Electroluminescence

EQE External quantum efficiency

TCSPC Time-correlated single-photon counting

SPD Single photon detector

TRPL Time-resolved photoluminescence

TLM Transfer length method

FWHM Full width at half maximum

TE Transverse electric

TM Transverse magnetic

PC Polarization converter

SMSP Single-mode single-polarization

PCE Polarization conversion efficiency

ER Extinction ratio

IL Insertion loss

106

Acknowledgements

I would like to thank all the people I met during my PhD adventure with whom I

shared a piece of my life in a personal or professional manner. I’ve met wonderful

people from all around the world who enriched life.

First, I would like to thank my supervisors. Meint, I learned too much from your

vision and passion for technology, as well as from your experience as a manager

and leader, I am happy I was able to work with you directly. Andrea, thank you

for providing the physical touch to our research and being tough but fair, I enjoyed

a lot working with you.

I continue with the Photonic Integration (PhI) group. Kevin, I wish you success

as chairman and leader of the group, I am sure PhI will contribute significantly to

the future of the PICs technology. Jos, it was a pleasure to cooperate with you in

the development of membrane devices. Erwin, thank you for keeping OLA a nice

lab to work and also for solving my practical questions on equipment. Xaveer, my

contact with you was typically limited to software issues, however I was happy to

work in FBG sensing based on your previous work. Huub, thanks for keeping the

high standards in the clean room and sharing your valuable knowledge on technol-

ogy. Tjibbe, thanks for always giving sharp replies to questions, and of course for

keeping the PhI tradition at Het Walhalla. What are we without traditions? Noth-

ing! Erik Jan, thanks for sharing your knowledge in nanotechnology. Jeroen, your

long explanations sometimes made me think twice and avoid processing mistakes,

thanks for that. Barry, you are the master of the clean room, thanks for sharing

your knowledge and wisdom. I promise to read Under the Volcano sometime soon!

Robert, thanks for always be willing to help.

Dominik, it was a good experience to cooperate with you on nanoscale lasers, I

appreciate your remarkable input on the work presented in this thesis. Gilardi,

it was an honor to share office with you since our first week in TU/e, and to

have black-humor funny conversations! Rui and Sylvester, it was fun to quench

107

Acknowledgements

our thirst for drinks around the city center sometimes. Yuqing, our cooperation

in membrane devices was quite successful, good luck in consolidating the IMOS

platform.

Manuela and Dima (crazy), thank you for our discussions on processing technology

at the beginning of my project. That was definitely a good starting point for me.

Domenico, thanks for playing relaxing melodies during the afternoons in our office

at the Pontentiaal. Rui (Zhang) and Dima, thanks for the invitation to join the

boxing (fight) club, which was useful to remove the stress. Elton, thanks for

being part of our amazing Californian road trip! Alonso, fue un placer continuar

compartiendo vivencias amigo, buena suerte con lo que viene! Weiming, thank you

for always be willing to listen and help! Dima (normal), thanks for sharing your

knowledge on characterization of PICs, it was fun to do the sensing experiments.

Longfei, it was a pleasure to cooperate with you in the development of the silver

contacts. Srivathsa, thanks for our interesting conversations after office hours.

Marija, thank you for organizing diverse get-together events. Valeria, do not for-

get to watch Twin Peaks, it is coming! Dan, thank you for cooking that amazing

korean food. Florian, thank you for all memories you captured with your lens.

Vadim, it was fun to share office with you, good luck with the lasers. Stefanos,

Marc, Jorn, good luck with your PhD project! Valentina, Monica, Perry, Kasia,

Luc, Mike, Daan, Antonio, Nicola, Simone, Hadi, Jolanda, thank you for promot-

ing a friendly environment.

Bruno and Francesco, thank you for the support and useful discussions at the

latest stage of my PhD project. I still remember our excitement and jokes (about

the powerful nanolasers) in the laboratory during those cold winter evenings.

I also would like to acknowledge the colleagues of the NAVOLCHI project: Juerg,

Manfred, Martin, Argishti, Sascha, Claudia, Dries, Ioannis, Christoforos, Thymios,

Juan, Isaac, Zeger. It was a rewarding experience to collaborate with you in this

challenging project.

Ramon, gracias por ser mi primer mentor en optica en el Instituto de Ingenierıa,

y tambien por siempre estar dispuesto a escuchar mis disyuntivas profesionales.

A mis papas, gran parte de lo que soy se los debo a ustedes. Mama, gracias por

esforzarte para que tuvieramos educacion y etica, gracias por amarme. Papa, tuve

todo lo que necesite gracias a tu arduo trabajo y responsabilidad. Raul y Katy,

me siento feliz de haber crecido junto a ustedes y haber compartido nuestra ninez

haciendo tonterias. Los quiero!

108

Acknowledgements

Aura, me siento afortunado de poder compartir mi vida contigo. Gracias por hacer

de nuestra vida una experiencia formidable. Gracias por llenarme de felicidad!

Victor

109

List of research products

Patent applications

• V. M. Dolores Calzadilla, A. Higuera Rodriguez, D. Heiss, “Metal grating

coupler for membrane-based integrated photonics”, USA provisional patent

application filed 61/979111, 2014.

Journal publications

• V. Dolores-Calzadilla, B. Romeira, F. Pagliano, S. Birindelli, A. Higuera-

Rodriguez, P.J. van Veldhoven, M. Smit, A. Fiore, and D. Heiss, “Waveguide-

coupled nanopillar metal-cavity light emitting diodes on silicon”, Nature

Communications (submitted), 2016.

• V. Dolores-Calzadilla, J.J.G.M. van der Tol, M.K. Smit, “Polarization ro-

tator with high performance for membrane integrated photonics”, Optics

Letters (submitted), 2016.

• L. Shen, P.J. van Veldhoven, Y. Jiao, V. Dolores Calzadilla, J.J.G.M. van der

Tol, G.C. Roelkens, and M.K. Smit, “Ohmic contacts with ultra-low optical

loss on heavily doped n-type InGaAs and InGaAsP for InP-based photonic

membranes”, Optical Materials Express (submitted), 2015.

• A. Higuera-Rodriguez, V. Dolores Calzadilla, Y. Jiao, E. J. Geluk, D. Heiss,

and M. K. Smit, “Realization of efficient metal grating couplers for membrane-

based integrated photonics”, Optics Letters, 40(12), 2755-2757, 2015.

• L. Shen, V. Dolores Calzadilla, C.W.H.A. Wullems, Y. Jiao, A.J. Millan

Mejia, A. Higuera Rodriguez, D. Heiss, J.J.G.M. van der Tol, H.P.M.M.

Ambrosius, G.C. Roelkens, and M.K. Smit, “Low-optical-loss, low-resistance

Ag/Ge based ohmic contacts to n-type InP for membrane based waveguide

devices”, Optical Materials Express, 5(2), 393-398, 2015.

111


• V. Dolores Calzadilla, D. Heiss, and M. K. Smit, “Highly efficient metal

grating coupler for membrane-based integrated photonics”, Optics Letters,

39(9), 2786-2789, 2014.

• J. Leuthold, C. Hoessbacher, S. Muehlbrandt, A. Melikyan, M. Kohl, C.

Koos, W. Freude, V. Dolores-Calzadilla, M. K. Smit, I. Suarez, A. Martin, J.

Martinez Pastor, E.P. Fitrakis, and I. Tomkos, “Plasmonic communications:

light on a wire”, Optics and Photonics News, 24(5), 28-35, 2014, Invited

paper.

• R. Gutierrez-Castrejon, V. Dolores-Calzadilla, and M. Duelk, “Gain-controlled

semiconductor optical preamplifier for the 100 Gbit/s 40 km ethernet re-

ceiver”, Applied Optics, 48(25), F82-F89, 2009.

Conference and symposium proceedings

• V. Dolores-Calzadilla, D. Pustakhod, X.J.M. Leijtens, and M.K. Smit, “Fiber

Bragg grating sensor based on external cavity laser”, Proceedings of the 20th

Annual Symposium of the IEEE Photonics Benelux Chapter, 8-9 February

2016, Brussels, Belgium.

• B. Romeira, V. Dolores-Calzadilla, D. Heiss, F. Pagliano, S. Birindelli, P. J.

van Veldhoven M.K. Smit, and A. Fiore, ”Dynamic characteristics of electri-

cally pumped waveguide-coupled metal-cavity nanoLEDs”, NanoCity2015,

5-6 October, 2015, Amersfoort, The Netherlands, Session D1 Beyond Moore,

Talk.

• V. Dolores-Calzadilla, D. Heiss, B. Romeira, F. Pagliano, P.J. van Veldhoven,

A. Fiore, and M. Smit, “Integrated metal-cavity nanoLEDs in III-V mem-

branes on silicon”, The 11th Conference on Lasers and Electro-Optics Pacific

Rim, 24-28 August 2015, Busan, Korea, 27J2-2.

• B. Romeira, V. Dolores-Calzadilla, D. Heiss, F. Pagliano, S. Birindelli, P.J.

van Veldhoven, M.K. Smit, and A. Fiore, “Dynamic characteristics of electri-

cally pumped waveguide-coupled metal-cavity nano-LEDs and nanolasers”,

European Semiconductor Laser Workshop, 24-25 September 2015, Madrid,

Spain, Session ID.2-5.

• L. Shen, L., P.J. van Veldhoven, Y. Jiao, V.M. Dolores-Calzadilla, J.J.G.M.

van der Tol, G.C. Roelkens, and M.K. Smit, “Low Optical Loss Ohmic Con-

tacts on Heavily Doped N-type InGaAs for InP Membrane Devices”, The

112


27th International Conference on Indium Phosphide and Related Materials

(IPRM), 28 June - 2 July 2015, Santa Barbara, USA, (pp. 177-178).

• A. Higuera-Rodriguez, V. Dolores Calzadilla, Y. Jiao, E. J. Geluk, D. Heiss,

and M. K. Smit, “Membrane-based high efficiency metallic grating couplers

for integrated photonics”, The 11th Conference on Lasers and Electro-Optics

Pacific Rim, 24-28 August 2015, Busan, Korea, 27J2-4.

• B. Romeira, V. Dolores-Calzadilla, D. Heiss, M. Smit, and A. Fiore, “High-

speed dynamics of electrically modulated semiconductor nanolasers”, Nanoc-

ity 2014, 27-28 October 2014, Utrecht, The Netherlands. Poster contribution.

• V. Dolores-Calzadilla, A. Millan-Mejia, J.J.G.M. van der Tol, and M. Smit,

“Diffraction-supressed adiabatic tapers for photonic circuits”, Proceedings of

the 19th Annual Symposium of the IEEE Photonics Benelux Chapter, 3-4

November 2014, Enschede, The Netherlands, (pp. 103-106).

• D. Heiss, A. Higuera-Rodriguez, V. Dolores-Calzadilla, A. Fiore, and M.

Smit, “Design of an efficient photonic crystal beam laser”, Proceedings of

the 19th Annual Symposium of the IEEE Photonics Benelux Chapter, 3-4

November 2014, Enschede, The Netherlands, (pp. 161-164).

• L. Shen, C.W.H.A. Wullems, P.J. van Veldhoven, V. Dolores-Calzadilla, D.

Heiss, J.J.G.M. van der Tol, M.K. Smit, and H.P.M.M. Ambrosius, “Char-

acterization of Ge/Ag ohmic contacts for InP based nanophotonic devices”,

Proceedings of the 19th Annual Symposium of the IEEE Photonics Benelux

Chapter, 3-4 November 2014, Enschede, The Netherlands, (pp. 177-180).

• A. Higuera-Rodriguez, V. Dolores Calzadilla, D. Heiss, and M. Smit, “Fab-

rication of an efficient metal grating coupler for membrane-based integrated

photonics”, Proceedings of the 19th Annual Symposium of the IEEE Photon-

ics Benelux Chapter, 3-4 November 2014, Enschede, The Netherlands, (pp.

51-54).

• D. Heiss, A. Higuera-Rodriguez, V. Dolores-Calzadilla, A. Fiore, and M.

Smit, “Design of an efficient photonic crystal beam laser”, European Semi-

conductor Laser Workshop, 18-19 September 2014, Paris, France, CK-3-1.

• D. Heiss, A. Higuera-Rodriguez, V. Dolores-Calzadilla, A. Fiore, M. Smit,

“Towards efficient densely integrated lasers”, Proceedings of the 16th Inter-

national Conference on Transparent Optical Networks (ICTON 2014), 6-10

July 2014, Graz, Austria, (pp. Mo.D2.3-1/4), Invited paper.

113


• M. Smit, E. Bente, V. Dolores-Calzadilla, and D. Heiss, “Lasers in generic

photonic foundry platforms”, 24th International Semiconductor Laser Con-

ference, 7-10 September 2014, Palma de Mallorca, Spain, Plenary talk.

• V. Dolores-Calzadilla, D. Heiss, and M. K. Smit, “Nanometallic lasers for

optical interconnects”, Proceedings of the 18th OptoElectronics and Com-

munications Conference/Photonics in Switching (OECC/PS 2013), 30 June

- 4 July 2013, Kyoto, Japan, SWB2-6, Invited paper.

• D. Heiss, V. Dolores-Calzadilla, A. Fiore, and M. K. Smit, “Design of a

waveguide-coupled nanolaser for photonic integration”, Integrated Photonics

Research, Silicon and Nanophotonics (IPRSN), 14-17 July 2013, Rio Grande,

USA, (pp. IM2A.3).

• V. Dolores-Calzadilla, D. Heiss, A. Fiore, and M. K. Smit, “Waveguide-

coupled nanolasers in III-V membranes on silicon”, Proceedings of the 15th

International Conference on Transparent Optical Networks (ICTON 2013),

23-27 June 2013, Cartagena, Spain, (pp. We.D6.1-1/4), Invited paper.

• V. Dolores-Calzadilla, E. J. Geluk, T. de Vries, E. Smalbrugge, P. J. van

Veldhoven, H.P.M.M. Ambrosius, D. Heiss, A. Fiore, M. K. Smit, “Fabrica-

tion technology of metal-cavity nanolasers in III-V membranes on silicon”,

Proceedings of the 18th Annual Symposium of the IEEE Photonics Benelux

Chapter, 25-26 November 2013, Eindhoven, The Netherlands, (pp. 243-246).

Poster.

• V. Dolores Calzadilla, D. Heiss, A. Fiore, and M. K. Smit, “Metallo-dielectric

nanolaser coupled to an InP-membrane waveguide”, Proceedings of the 17th

Annual Symposium of the IEEE Photonics Society Benelux Chapter, 29-30

November 2012, Mons, Belgium, (pp. 195-198), Poster.

• A. Melikyan, M. Sommer, A. Muslija, M. Kohl, S. Muehlbrandt, A. Mishra,

V. Dolores Calzadilla, Y. Justo, J. P. Martınez-Pastor, I. Tomkos, A. Scan-

durra, D. Van Thourhout, Z. Hens, M. K. Smit, W. Freude, C. Koos, and

J. Leuthold, “Chip-to-chip plasmonic interconnects and the activities of EU

project NAVOLCHI”, Proceedings of the 14th International Conference on

Transparent Optical Networks (ICTON 2012), 2-5 July 2012 , Warwick,

United Kingdom, (pp. Th.A5.1-1/3).

114


• V. Dolores-Calzadilla, A. Fiore, and M. K Smit, “Towards plasmonic lasers

for optical interconnects”, Proceedings of the 2012 14th International Con-

ference on Transparent Optical Networks (ICTON), 2-5 July 2012, Warwick,

United Kingdom, (pp. Th.A5.7-1/4).

• F. Widulle, V. Dolores Calzadilla, S. Shukla, and B. Kleemann, “Nanometrol-

ogy of molded sub-µm structures by means of light scattering”, Proceedings

of the 113th Annual Meeting of the German Branch of the European Optical

Society, 29 May - 2 June 2012, Eindhoven, The Netherlands, Talk A27.

• F. Widulle, V. Dolores Calzadilla, E. Gridneva, H. Wegendt, B. Kleemann,

and A. Heinrich, “Nanometrology of Periodic Nanopillar Arrays by Means

of Light Scattering”, 9th International Conference on Multi-Material Micro

Manufacture, 9-11 October 2012, Vienna, Austria.

115

Curriculum vitae

Vıctor Manuel Dolores Calzadilla was born on

March 14, 1986, in Mexico. In 2009, he grad-

uated of B.Eng. in Telecommunications with

honors from the National Autonomous Univer-

sity of Mexico, with a project on optical soli-

tons in non-linear optical fibers. Later, he was

awarded a scholarship to pursue a M.Sc. in

Photonics at the Friedrich-Schiller-Universitat

Jena, Germany, where he obtained the degree

with a thesis on the optical metrology of nanos-

tructures by light scattering carried out at Carl

Zeiss AG. From 2012 to 2015, he worked on

a PhD project at the Eindhoven University of

Technology, the Netherlands, within the Photonic Integration (PhI) group, where

his researched focused on metal-cavity nanoscale light sources in III-V semicon-

ductors on silicon. Additionally, he also researched on metal grating couplers,

polarization converters and fiber Bragg grating sensors. Since 2016, he is with the

Department of Photonic Components at the Fraunhofer Heinrich Hertz Institute,

in Berlin.

117

metal nanocavity light sources integrated with passive waveguide components

Documents