Raaij T.T.L. Van - MSC Thesis Thesis - MAP

Eindhoven University of Technology
MASTER
Plasmonics and 2D Materials for Micro-LED Applications
van Raaij, Thomas-Jan T.L.
Award date:
2023
Link to publication
Disclaimer
This document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Student
theses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the document
as presented in the repository. The required complexity or quality of research of student theses may vary by program, and the required
minimum study period may vary in duration.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and 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
Department of Applied Physics
Plasmonics and 2D Materials for Micro-LED

Applications
by
T.T.L. van Raaij
MSC THESIS
Assessment committee Graduation

Member 1 (chair): prof.dr. J. Gómez Rivas Program: Applied Physics
Member 2: prof.dr.ir. O.J. Luiten Capacity group: Photonics and Semiconductor Nanophysics
Member 3: prof.dr. A. Fiore Supervisor: prof.dr. J. Gómez Rivas
Advisory member 1: drs. Mohamed S. Abdelkhalik Date of defense: May 1, 2023
Advisory member 2: dr. M. Liang Student ID: 1318551
Study load (ECTS): 60
Track: Nano, Quantum and Photonics
The research of this thesis has been carried out in collaboration with Lumileds Germany GmbH.
This thesis is confidential for a period of 5 years until the publication date May 1, 2028.
This thesis has been realized in accordance with the regulations as stated in the TU/e Code of Scientific
Conduct.
Disclaimer: the Department of Applied Physics of the Eindhoven University of Technology accepts no
responsibility for the contents of MSc theses or practical training reports.
Abstract
In this thesis, we investigated the possibility of exploiting surface lattice resonances (SLRs) for enhanced
light extraction efficiency (LEE) in micro-LEDs. We fabricated hexagonal arrays of nanoparticles (NPs)
onto a InGaN/GaN light-emitting diode (LED) and covered them with TiO2 to ensure a homogeneous
dielectric environment around the NPs. During atomic layer deposition (ALD), a conformal layer formed
around the NPs. The resulting structure was modelled into a finite element simulation software (COMSOL
Multiphysics), after which a factor 2.1 enhancement for electroluminescence (EL) (452 nm) and a factor
3.08 for photoluminescence (PL) (440 nm) were calculated, when compared to the same micro-LED without
NP arrays. These results appear to compare well to our experiments, where enhancements of 2.42 ± 0.08
and 3.38 ± 0.005 were measured, respectively. From back-focal plane (BFP) images that were measured
in a Fourier microscopy setup, no clear directionality in the emission was obtained. Focused ion beam
(FIB) images showed that the actual period of the NP arrays was considerably smaller than intended, due
to an error that occurred in fabrication. Directionality had, however, been shown before in samples with
correct lattice periods. The absence of directionality was therefore attributed to the absence of SLRs due
to the incorrect lattice period. The observed enhancement, on the other hand, was then concluded to be a
consequence of scattering from the NPs in combination with the presence of the conformal layer.
Furthermore, an alternative to etching through the p-GaN, which is a difficult process for micro-LEDs in
particular, was investigated. In a process called diode burning, it was proposed that upon applying a high-
voltage pulse to a system of two anti-parallel diodes, the breakdown region of the diode in reverse bias
can be reached, such that it burns to a short-circuit. Without the resistance of the diode in reverse bias, we
hypothesized to be able to operate the diode in reverse bias at low operating voltages, comparable to when
etching would be applied. We successfully showed that applying a short pulse width (10 ms), high-voltage
(25 V) signal results in burning through the p-GaN, allowing for EL measurements at voltages as low as 2.5
V.
Moreover, the possibility of integrating two-dimensional (2D) transition metal dichalcogenides (TMDs),
WS2 specifically, into micro-LEDs for efficient white light emission was investigated. Monolayers of WS2
were obtained through mechanical exfoliation from a bulk crystal. Using a confocal microscope and a
spectrometer, the peak wavelength of the PL was measured to be approximately 617 nm. In the same
setup, the absorption was measured to be 16.7% for the C-exciton. In addition, two different van der Waals
(vdW) heterostructures were fabricated using a stamping microscope setup. Both samples were artificial
bilayers, with the one sample hexagonal boron nitride (hBN) was used for spatial separation while for the
other Al2 O3 was used. These samples were then characterized in terms of PL, absorption, lifetime and
efficiency. An increase in absorption of 1.7 percentage point was observed for a bilayer of WS2 , spatially
separated by hBN when compared to a monolayer, while a decrease was observed for the Al2 O3 sample.
Although the PL in the bilayer area was clearly higher than in the monolayer areas of the heterostructure, the
absolute number of counts in the bilayer was only higher than directly after stamping the monolayer onto
the LED for the hBN sample. From this we concluded that for the Al2 O3 sample, the expected increase in
PL in the bilayer can not compensate for the amount of defects introduced upon stamping using the current
setup and the increased dielectric screening. We concluded that hBN is more suitable to be used as a spacer
to achieve enhanced PL and absorption in artificial bilayers. A layer of hBN was also proposed to be used
for encapsulation of the multilayer structure, to fabricate green quantum dots (QDs) on top for efficient
white LED emission from micro-LEDs in the future. Lastly, a method for calculating the external quantum
efficiency (EQE) was described that allows for the analysis of the radiative and non-radiative components
of the measured carrier lifetimes in multilayered 2D materials.
1
List of abbreviations
2D Two-dimensional
ALD Atomic layer deposition
BFP Back-focal plane
EL Electroluminescence
EQE External quantum efficiency
FWHM Full width at half maximum
hBN Hexagonal boron nitride
IQE Internal quantum efficiency
LED Light-emitting diode
LEE Light extraction efficiency
LSP Localized surface plasmon
MQW Multiple quantum well
NP Nanoparticle
PDMS Polydimethylsiloxane
PL Photoluminescence
QD Quantum dot
SLR Surface lattice resonance
TMD Transition metal dichalcogenide
vdW van der Waals
2
Contents
1 Introduction 5
2 Theoretical background 9
2.1 Light-emitting diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Radiative recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Non-radiative recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.3 Lifetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.4 White LEDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Colorimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Surface lattice resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 2D materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.1 Transition metal dichalcogenides . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.2 WS2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.3 Van der Waals heterostructures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Methodology and setups 25

3.1 Sample fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1.1 InGaN/GaN LED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1.2 Nanoparticle fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Finite element method simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Photoluminescence and Fourier plane imaging setup . . . . . . . . . . . . . . . . . . . . 28
3.4 Diode burning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5 Electroluminescence and Fourier plane imaging setup . . . . . . . . . . . . . . . . . . . . 30
3.6 Mechanical exfoliation of WS2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.6.1 Stacking of atomically thin layers . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.7 Confocal microscopy setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.7.1 Photoluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.7.2 Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.7.3 Fluorescence-lifetime imaging microscopy . . . . . . . . . . . . . . . . . . . . . 35
3.8 Integrating sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4 Results and discussion 39

4.1 COMSOL simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 Power dependent photoluminescence and enhancement . . . . . . . . . . . . . . . . . . . 40
4.3 Diode burning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.4 Power dependent electroluminescence and enhancement . . . . . . . . . . . . . . . . . . 43
4.5 Back-focal plane imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.6 Photoluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.7 Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.8 External quantum efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.9 Lifetime measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.10 Color gamut . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5 Conclusion and outlook 57
3
CONTENTS
Bibliography 62
Appendices 63
A Appendix 64
A.1 COMSOL simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
A.2 Power dependent EL and enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
A.3 Back-focal plane imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
A.4 Photoluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
A.5 Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
A.6 External quantum efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
A.7 Lifetime measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
A.8 Color gamut . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
B MATLAB scripts 72
B.1 Power dependent PL and enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
B.2 Color gamut . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4
TU/e - Department of Applied Physics
1 Introduction
The emergence of new technologies such as smart wearable devices, augmented reality (AR), and virtual
reality (VR) displays, coincides with an increasing demand for high-resolution displays that are small in
size.[1][2] Light-emitting diodes (LEDs) have proven to be very suitable for display applications, based on
their high luminance levels and high efficiency.[3] For LEDs to be effectively applied applied into these
novel applications, high resolution and efficiency need to be preserved upon reducing the size. Notably,
micro-LEDs have emerged as a promising technology that could overcome the limitations of directly emis-
sive LED displays.
In industry, the so-called pixel pitch, that is the distance from the center of one pixel to the adjacent one, is
an important parameter for the resolution. Industry’s smallest pixel pitch to date is 0.625 mm.[4] A simple
calculation then teaches us that in order to achieve the modern standard of 4K (3840 × 2160) resolution,
a minimum diagonal display size of well over 100 inches would be required. Given the fact this is con-
siderably larger than the most common TV sizes, it is immediately apparent that the current limitation of
directly emissive LED displays is the resolution at small display sizes. There is an increasing demand from
industry for a solution to this limitation to be applied in emerging technologies such as smart wearable de-
vices, and AR/VR displays.[1][2] This demand for small size, high-resolution displays, in turn, has driven
the development of new technologies such as micro-LEDs. Micro-LEDs are LED devices with an emission
area no larger than 50 µm2 .[3] When combining micro-LED pixels into an array, a micro-LED display can
be created. Micro-LEDs are expected to become the leading technology in future display technology, out-
competing conventional liquid-crystal displays (LCD) and organic light-emitting diode (OLED) displays.
Specifically, micro-LEDs excel in exhibiting a high contrast, short response time, low power consumption
and a long lifetime.[1] The starting point of this thesis will be a micro-LED with multiple quantum wells
(MQWs) made of InGaN/GaN, of which the schematic cross section is given in figure 1.1.
Figure 1.1: Schematic cross section of the InGaN/GaN micro-LED, used as a starting point for this thesis.
The overall efficiency of LEDs is limited by the value of the light extraction efficiency (LEE), to which
micro-LEDs are no exception. The low LEE mainly originates from a large difference in refractive index at
the interface of the semiconductor (nGaN = 2.45) and air (nair = 1), as well as from internal absorption
and reflection of the generated light by the contacts.[5] Based on these refractive indices, the critical angle θc
can then be calculated using Snell’s Law, which dictates the largest possible angle of incidence before total
internal reflection will occur. Typically, LED structures consist of several layers with high refractive index,
5
essentially functioning as a waveguide, which potentially supports a large number of guided modes.[5] In
practice, this means only a small amount of the generated light within the LED will be able to escape the
device, while the trapped light will eventually be converted to heat.[6][7]
In large area devices, a profiled sapphire substrate is commonly used for achieving omnidirectional reflec-
tion, increasing the LEE of the LED. This method, however, can not be applied into micro-LEDs due to
the feature dimensions of the profiled sapphire being in the order of a few microns. This is quite large
when compared to the size of the micro-LED, allowing only for a small number of features to be within
the micro-LED area. [8] Alternatively, etching through the sidewalls of the micro-LED for creating so-
called mesa sidewalls, which serve as microreflectors, has shown to enhance the light outcoupling. [9] On
the other hand, mesa-etching increases dangling bonds and defects at the sidewalls, which serve as non-
radiative recombination centers. [10] These non-radiative recombination centers, in turn, negatively affect
the performance of the LED in terms of efficiency.
Another issue with GaN LEDs is the fact that their emission is Lambertian, that is, it follows a cosine
emission law. [11] The limited directionality associated with Lambertian emission, limits the application of
LEDs into technologies where directional emission is required.[6][12] Ideally, the emission pattern should
be tailored, allowing for an increase in emission in the desired direction. Furthermore, this ought to be
achieved without the need for secondary optical elements like reflectors, which would render the final
device bulky and would require extra steps in fabrication.[13][14]
The first topic of this thesis involves a single solution to the two aforementioned problems in the form of
plasmonic nanoparticle (NP) arrays. These NPs support coherent oscillations of their free electrons, known
as localized surface plasmon (LSP) resonances.[15] Moreover, under the right conditions, the resonances
of the individual NPs couple to the in-plane diffraction orders of the array, also known as Rayleigh anoma-
lies (RAs).[16] By choosing the appropriate NP size and shape, as well as array period, the NP array may
support collective plasmonic resonances known as surface lattice resonances (SLRs).[17] In sufficiently
large NP arrays, these resonances allow for enhancement of the absorption as well as the electric field in
the vicinity of the NPs. Consequently, the introduction of metasurfaces of metal (Ag, Al or Au) NPs, al-
lows for control over the photonic environment of the MQWs, which can be exploited for increasing light
outcoupling.[18] Moreover, by tailoring the array period, coherent scattering from periodically arranged
NPs can be exploited to achieve beaming, i.e. directional emission, into defined directions of small solid
angle in the far-field.[19] Both the principle of NP enhanced LEE and the principle of NP induced direc-
tional emission will be discussed in more detail in the theory section of this thesis. Optical measurements of
the photoluminescence (PL) and back-focal plane (BFP) imaging will be performed for samples both with
without NPs, to observe enhancement and reshaping of the MQWs emission.
The characterization of many aspects of the LED device can be performed optically. From an application
point of view, however, it is important that the proposed solution is compatible with electrical excitation. In
the proposed device, both the positive and the negative contact are fabricated at the top, that is the anode
side of the device. After the current flows through the diode in the forward direction once, it needs to
flow back through the diode in the reverse direction in order to reach the negative contact. Commonly,
an etching process is applied to the p-side of the junction and the MQWs to overcome the high resistance
of the reverse biased diode. This etching process proves to be difficult and time-consuming, though. In
this work, we propose an alternative to etching through the p-side of the junction, in the form of diode
burning. It is hypothesized that when high-voltage pulses are applied, we reach the breakdown region of
the diode in reverse bias. After burning the reverse diode, the LED in forward bias will operate at the same
low voltages as when etching is done. This novel approach will be tested as a proof of concept, and the
process will be characterized in terms of applied voltage and pulse width. Afterwards, measurements of
electroluminescence (EL) and BFP imaging can be performed for LEDs both with and without NP arrays
for comparison. The observed emission enhancement can then also be compared to the results of the
measurements of its optical counterpart.
The second topic of this thesis will focus on obtaining efficient white light emission in the proposed micro-
LEDs. For obtaining white LED emission, which is a requirements for many applications, there are two
main options. The first option includes the mixing of the primary colors, by individually driven red, green
and blue LEDs.[20] The resulting color, as perceived by the human eye, will then be determined by the
fraction in which each color is present, and can be calculated by means of the color gamut.[11] The lack
of efficient green LEDs restricts this method for efficient white light generation.[21] The second option
6
involves a blue, (InGaN/GaN for example, see figure 1.1) LED, and one or more down-converting materials,
also known as phosphors. Phosphors absorb part of the blue emission from the LED, and re-emit at a longer
wavelength. The part of the blue light that is not absorbed, together with the emission of the phosphors,
can then be tuned to be perceived as white light by the human eye.[22] Current phosphor-converted LEDs
(pc-LEDs) commonly use rare-earth ions, which have a low absorption cross section in the blue region of
the electromagnetic spectrum. This causes a relatively large amount of material to be required in order
to achieve sufficient absorption.[23] Consequently, there is a demand for small size but highly absorbing
down-converting materials to be applied in micro-LEDs.
With this in mind, two-dimensional (2D) transition metal dichalcogenides (TMDs) are a promising candi-
date for fulfilling this demand of small size, highly absorbing down-converters. TMDs have unique optical
and electrical properties and have therefore drawn considerable scientific attention, both from a theoretical
and from an application point of view.[24] Contrary to its bulk counterpart, TMD monolayers are direct
band gap semiconductors, allowing for applications in light emission. In addition, the properties of rela-
tively high absorption, small size and the possibility of creating heterostuctures without lattice mismatch,
make 2D TMDs a potentially suitable candidate as down-converting material in future pc-LEDs.[25][26]
WS2 monolayers for example, display strong PL at a band gap energy of approximately 2 eV (∼ 620
nm).[27] Therefore, the material could be used for the red light emission of the LED.
In order to achieve white light emission, blue, red and green emission must be present in the appropriate pro-
portions. The amount of blue light absorbed and red light emitted by the WS2 can be tuned by controlling
the number of 2D TMD layers used.[28][29] Optical characterization in terms of quantifying the photolu-
minescent and absorptive properties, as well as lifetime and external quantum efficiency (EQE) analyses of
both single layer WS2 and artificial bilayer WS2 , will be presented in chapter 5 of this thesis. The nuance of
artificial bilayer is essential here. Namely, in natural bilayers of WS2 , or any multilayer for that matter, in-
direct transitions are energetically favorable.[27] Consequently, the PL of natural bilayers is negligible with
respect to that of a monolayer.[30] Instead, two individually obtained monolayers of WS2 will be stacked,
whilst spatially separated by a spacer, into a van der Waals (vdW) heterostructure. For the purpose of sep-
aration, commonly few layers of hexagonal boron nitride (hBN) are stacked, or aluminum oxide (Al2 O3 )
is deposited using atomic layer deposition (ALD). This way, both individual monolayers retain their direct
band gap property, and the overall absorption and PL can be increased.[28] In reality, many different factors
may influence the optical properties upon stacking layers. For instance, the choice of spacer material may
influence the different excitonic transitions due to dielectric screening of the environment.[31] Furthermore,
the interlayer interactions and thus resulting optical properties are highly dependent on the thickness of the
spacer.[28] Also, during exfoliation and transfer of the monolayers, mechanical deformation may induce the
formation of bubbles and/or wrinkles, affecting the interface properties of the 2D material.[32] Altogether,
all of these factors will need to be considered in the fabrication for tailoring the red/blue emission ratio
in the proposed device. In a later stage, by adding quantum dots (QDs) for instance, green emission can
be generated, in order to achieve the final goal of a white emission spectrum.[33] A schematic of the final
structure of the proposed device for efficient white light emission from micro-LEDs is provided in figure
1.2.
7
Figure 1.2: Schematic of future final device for efficient white light emission from micro-LEDs.
This thesis shows how the use of plasmonic NPs can significantly improve the performance of micro-
LEDs, to enable their application in high-resolution displays for emerging technologies. In order to achieve
this, we introduce aluminum (Al) NP arrays that support SLRs at the emission wavelength of the blue
InGaN/GaN LEDs, i.e. around 445 nm. The investigated arrays are arranged in a hexagonal lattice with
an intended period of 215 nm. This approach relies on the near-field coupling between the SLRs from the
Al metasurfaces and the MQWs. This near-field coupling leads to reshaping of the far-field emission and
beaming or focusing the generated light into a narrow solid angle which is essential for several applications,
as well as for improving light outcoupling. Additionally, monolayers of WS2 are introduced to the sapphire
side of the micro-LED to operate as small size, highly absorbing down-converters. These monolayers
will be stacked to create artificial bilayers in order to increase the absorption and PL. This way, the ratio
of blue to red emission can be quantified which is essential for obtaining white light emission. Then,
after encapsulation by hBN for protection against defects, QDs can be fabricated on top of the multilayer
structure in the future to achieve the final goal of efficient, directional, white LED emission with micrometer
pixel size.
8
2 Theoretical background
In this chapter, a theoretical background of several different topics will be provided to the reader. In section
2.1, the theory of diodes (2.1.1), and that of LEDs in particular, will be treated in detail. The recombination
mechanisms in LEDs (2.1.2) will be discussed, and the recombination rates will be used for defining the
carrier lifetime (2.1.3). In 2.1.4, colorimetry for white LED emission will be treated.
Then, in section 2.3, the unique optical properties of 2D materials will be discussed. The properties of
TMDs (2.3.1), and WS2 in particular (2.3.2), will be treated in more detail. Finally, the possibility of
fabricating vdW heterostructures using 2D materials will be discussed in 2.3.3.
At last, the theory of SLRs and how they can be exploited for enhanced and directional emission in LEDs
will be treated in section 2.2.
2.1 Light-emitting diodes

The definition of an LED is rather straightforward to break down: The devices is a diode, which is a special
type of junction, that generally only allows current to flow in one direction. Furthermore, the device is
known for the emission of light, under application of a forward bias. [34] In this section, the working
principle of diodes and LEDs will be discussed in detail, based on the descriptions by Kitai and Schubert,
respectively.[35] [11]
2.1.1 Diodes
Originally designed for logic circuits, diodes were used to provide current in one direction while blocking
current in the other direction. More recently, diodes have become widely implemented into semiconductor
devices like solar cells and LEDs. We will start this section with a qualitative description of the working
principle of a p-n diode. A p-n diode consists of a positively doped semiconductor brought into contact
with a negatively doped semiconductor. In the presence of a concentration gradient, charges start to flow in
a process called diffusion current. Near the p-n junction, free electrons originating from donors may diffuse
from the n-side to the p-side, while holes originating from acceptors may diffuse from the p-side to the
n-side. Although originally both the p-side and the n-side are neutrally charged, after diffusion of majority
carriers the p-side has a net negative charge while the n-side has a net positive charge. Consequently, an
electric field arises from the n-side to the p-side. If now, for example, more holes on the p-side try to diffuse
to the n-side, they will experience an opposing force, preventing them from crossing the junction. A region
close to the junction, will be depleted of free charges and is therefore known as the depletion region. The
net charge on either side is the result of ionized dopants: the donor and acceptor atoms are depleted of their
free charge carriers. This region is also known as the space charge region. The space charge region causes
a potential barrier known as the diffusion voltage, or contact potential V0 . The flow of charges as a result
of the presence of an electric field is referred to as drift current. In conclusion, current experiences a lot of
resistance from the p-side to the n-side. In order to overcome this resistance, a forward bias can be applied
to the p-n diode in an electrical circuit. A positive contact, or anode, is made to the p-side, while a negative
contact, or cathode, is made to the n-side, as shown in figure 2.1.
9
2.1. LIGHT-EMITTING DIODES
Figure 2.1: Schematic of a p-n diode (a) and the corresponding symbol and orientation for an electrical
circuit (b).
For a more quantitative analysis of the p-n diode, the band diagram of the junction should be considered.
For a p-doped semiconductor, the Fermi level, which is the energy level with a 50% chance to be occupied,
is shifted down in energy towards the top of the valence band. For an n-doped semiconductor, the Fermi
level is shifted up in energy towards the bottom of the conduction band. The magnitude of the shift depends
on the doping level. Upon contacting a p-doped and an n-doped semiconductor, the Fermi levels in the band
diagram ought to be aligned. Near the junction, the conduction band and valence band levels are connected
smoothly in the transition region. The resulting band diagram is shown in figure 2.2.
Figure 2.2: Band diagram of a p-n diode with the transition region indicated near the junction, the direction
of electric field ε and the energy barrier E0 .[35]
The built-in field ε drives the electron and hole drift currents across the junction. Upon crossing the junction,
the charge carrier gains or loses an amount of energy equal to E0 , equivalent to a change in potential
V0 = E0 /q, where q represents the charge of the carrier. The strength of this potential, induced by the
space charge region, is given by

kT NA ND
V0 = ln , (2.1)
q n2i
where k is the Boltzmann constant, T the temperature, NA and ND are the acceptor and donor concentra-
tions and ni is the intrinsic carrier concentration. As described previously in the qualitative description of
10
the p-n diode, two different mechanisms drive four types of current:
In drift : The built-in electric field drives electrons (minority carriers) from the p-side to the n-side.
Ip drift : The built-in electric field drives holes (minority carriers) from the n-side to the p-side.
In diffusion : Electrons (majority carriers) diffuse from the n-side to the p-side due to a concentration gradient.
Ip diffusion : Holes (majority carriers) diffuse from the p-side to the n-side due to a concentration gradient.
Note that since electrons carry negative charge, the direction of the current I is opposite to the direction
of the flow of the electrons. The directions of the different currents are summarized in the band diagram
below.
Figure 2.3: Band diagram of a p-n diode with the directions of the different currents. Note that the direction
of the flow of the electrons is opposite to the direction of the current by definition.[35]
Under equilibrium conditions (zero bias), the following balances apply:
Ip drift + Ip diffusion = 0
(2.2)
In drift + In diffusion = 0
If we connect the diode to an external voltage source and apply a positive voltage V>0, the diode operates
under forward bias. In forward bias, the external potential opposes the built-in potential of the junction,
reducing the electric field at the junction which drives the drift current. In the band diagram, this can be
interpreted as a decrease in the energy barrier that the majority carriers experience, as can be seen in figure
2.4. Consequently, the drift current is no longer able to cancel out the diffusion current, and a net current
starts to flow from the p-side to the n-side.
Figure 2.4: Band diagram under application of a forward bias.[35]
11
Reversely, when applying a reverse bias V<0, both the potential barrier barrier and the electric field are
increased, to the effect that the drift current virtually eliminates the diffusion current. The resulting band
diagram under reverse bias is provided in figure 2.5.
Figure 2.5: Band diagram under application of a reverse bias.[35]
However, a small amount of minority carriers that are thermally generated causes a net minority carrier drift
current supported by the electric field. This current is also known as the reverse saturation current I0 , and
is practically independent of the external voltage V . This is because the magnitude of I0 is limited by the
amount of thermally generated minority carriers, and increasing the electric field strength does not increase
the number of available carriers. The total I-V characteristic of a diode is shown in figure 2.6.
Figure 2.6: I-V characteristic of the p-n diode.[35]
Qualitatively, the I-V characteristic of the p-n junction was first described by Shockley, and is therefore also
known as the Shockley equation:

qV
I = I0 exp −1 . (2.3)
kT
The Shockley diode equation, however, does not account for the behavior of a diode at large reverse bias.
At a well-defined negative voltage known as the reverse breakdown voltage Vbd , current starts to flow
abruptly. At this semiconductor material specific voltage, the reverse bias is large enough to overcome
the ionization energy, such that electrons that are normally bound can be released in a process called field
ionization. These released electrons, in turn, contribute to releasing more normally bound electrons through
impact ionization. This process, which is known as avalanche breakdown, increases rapidly upon further
increasing the reverse bias. It is important to carefully limit the current upon avalanche breakdown because
12
of the relatively high value of Vbd , in combination with a rapidly increasing current. The power P = IVbd
for high currents could be too much for the diode to dissipate. The possibility of breaking down a diode to
a short circuit will be discussed in section 3.4 of this thesis. The I-V characteristic of a p-n diode, including
the breakdown regime, is provided in figure 2.7.
Figure 2.7: I-V characteristic of the p-n diode, including I-V characteristic upon breakdown in reverse bias.
2.1.2 Recombination
In an LED, electrons and holes are injected across the junction. Near the junction, majority carriers en-
counter each other and recombine, in a diffusion driven process. This recombination of electrons and holes
can occur either radiatively, i.e. upon emission of a photon, or non-radiatively, without emission of a photon.
For obvious reasons, radiative recombination is the preferred process in LEDs. In practice, non-radiative
recombination can never be eliminated completely. This section will discuss different recombination mech-
anisms and their respective influence on the LED’s performance.
Radiative recombination
Radiative recombination involves the annihilation of an electron-hole pair upon emission of a photon. In
absence of an external stimulus such as a current or light, the law of mass action dictates that the product of
the electron and hole concentrations is constant, according to
n0 p0 = n2i , (2.4)
where n0 and p0 are the equilibrium concentrations of the electron and the hole. By means of the injection
of an external current to or the absorption of light by the semiconductor, excess carriers can be generated.
Then, the total carrier concentration can be considered the sum of the equilibrium carrier concentration and
the excess carrier concentration:
n = n0 + ∆n
(2.5)
p = p0 + ∆p,
where, ∆n and ∆p represent the excess electron and excess hole concentration, respectively. Recombi-
nation of electrons and holes can be considered as the rate at which the electron and hole concentrations
13
reduce. This rate is proportional to the probability of an electron and a hole recombining. These prob-
abilities, in turn, are proportional to the concentration of their respective carrier. By introducing the a
proportionality constant B, the recombination rate R can be described by
dn dp
R=− =− = Bnp. (2.6)
dt dt
B is also known as the bimolecular recombination coefficient, and the equation itself is also known as the
bimolecular rate equation. This equation will be the basis for our lifetime calculations in section 2.1.3.
Non-radiative recombination
Non-radiative recombination involves the annihilation of an electron-hole pair without emission of a photon.
The energy upon recombination is manifested as a phonon, that is a lattice vibration, and thus converted to
heat. When input energy is converted to heat rather than light in an LED, this undesired output reduces the
efficiency of the device. Different mechanisms can cause non-radiative recombination. Most commonly,
defects in the crystal structure are responsible for non-radiative recombination. Unwanted foreign atoms,
interstitials, vacancies, antisite defects and dislocations all constitute defect states, with energy levels differ-
ent than that of the substitutional semiconductor atoms. Because of defects, energy levels can form within
the band gap of the semiconductor, which can serve as non-radiative recombination centers. Recombina-
tion via such trap levels is known as Shockley-Read-Hall (SRH) recombination. Although most trap-assisted
recombination is non-radiative, radiative deep-level transitions are also possible for specific crystal defects.
Without providing an exact derivation, for simplicity, the SRH recombination rate is assumed to be linearly
proportional to the carrier concentration, according to
RSRH = An, (2.7)
where A is the Shockley-Read-Hall recombination coefficient.
Alternatively, non-radiative recombination can be caused excitation of a free electron (hole) to a higher
energy level within the conduction (valence) band using the energy upon recombination. Relaxation to the
band edge is subsequently achieved by means of emission of multiple phonons. This type of recombination
is known as Auger recombination. The Auger recombination rates for electrons and holes, respectively, are
given by
RAuger = Cp np2
(2.8)
RAuger = Cn n2 p,
where Cp and Cn are the Auger coefficients. Note the proportionality to np (like in the bimolecular rate
equation) due to recombination, and the additional proportionality to either p or n (causing the square de-
pendence), due to the excitation of a third carrier. This threefold dependence on the carrier concentration
causes that Auger recombination only plays a significant role at high injection current (or excitation inten-
sity). Consequently, Auger recombination can often be neglected at low carrier concentrations. Nonethe-
less, at high injection currents, the Auger rate equations reduce to
RAuger = (Cp + Cn ) n3 = Cn3 , (2.9)
where C is the Auger coefficient. The numerical value of this coefficient can be determined by a quantum
mechanical calculation based on the band structure of the semiconductor material and typically ranges from
10−28 − 10−29 cm2 /s for III-V semiconductors.
Although the recombination mechanism is the same as in SRH recombination, surface recombination de-
serves a separate mention due to the surface of semiconductors being an efficient non-radiative recombina-
tion site. In practice, semiconductor materials can not be infinite in size and thus the crystal structure has to
14
end at some point. The periodicity of the lattice ends at the surface, causing dangling bonds. These dangling
bonds locally affect the band diagram and cause electronic states within the band gap of the semiconductor
material. Trap-assisted recombination can again take place and increase the value of the A coefficient of
SRH recombination.
In micro-LEDs specifically, surface recombination effects are more pronounced because of the high perime-
ter to area ratio. Size-dependent measurements show that smaller micro-LEDs suffer more from surface re-
combination losses. Increasing the current density can partially overcome this issue and decreases surface
recombination. Nevertheless, surface recombination, together with sidewall defects, leads to an increase
of non-radiative SRH recombination.[36] So in micro-LEDs, both size and injected current density affect
the current density at which the EQE reaches its maximum, after which it will start to decrease. This
phenomenon is known as size-dependent efficiency droop.[37]
2.1.3 Lifetime
The lifetime τ of a carrier is defined as the mean time between generation and recombination of that carrier.
For a more mathematical description, the bimolecular rate equation for radiative recombination of equation
(2.6) is revisited. If we momentarily excite the semiconductor, the excess carrier concentrations in equation
(2.5) become a function of time. Substitution of equation (2.5) into the bimolecular rate equation then yields
R = B [n0 + ∆n(t)] [p0 + ∆p(t)] (2.10)
Two approximation are then applied: One for low-level and one for high-level excitation. Upon low-
level excitation, the excess carrier concentration is small with respect to the majority carrier concentration:
∆n (n0 + p0 ).
Without providing a complete derivation here (see [11]), the rate equation is given by
dn(t) ∆n0 −t/τ

R= = e , (2.11)
dt τ
where the characteristic time constant τ is also known as the carrier lifetime
1
τ= . (2.12)
B(n0 + p0 )
Figure 2.8 summarizes the behavior of the electron and hole concentrations before, during and after low-
level optical excitation for a p-doped semiconductor.
15
Figure 2.8: Optical generation rate as a function of time (top) and the corresponding course of
the hole (middle) and electron (bottom) concentrations for low-level optical excitation on a p-doped
semiconductor.[11]
In the case of high-level excitation, the excess carrier concentration is large with respect to the majority
carrier concentration: ∆n (n0 + p0 ). The bimolecular rate equation (2.6) can then be reduced to
dn(t) −B
R= = 2 . (2.13)
dt Bt + ∆n−1
0
Using optical excitation, the recombination rate can be measured by measuring the PL as a function of
time. For these measurements, the exponential recombination rate from equation (2.11) belonging to low-
excitation levels will appear as a linear decrease on a logarithmic y-axis. Upon high-excitation levels, the
non-exponential decay can be represented by an exponential function with a time-dependent characteristic
time τ (t). As τ (t) increases over time, it will approach the constant time characteristic of the exponential
decay. The corresponding function is therefore also called a stretched exponential decay function. The
luminescence intensity and characteristic time τ as a function of time are provided in figure 2.9, both for
low and high power excitation.
16
Figure 2.9: Photoluminescence intensity as a function of time for low- and high-level excitation (a) and the
characteristic time τ as a function of time for low- and high-level excitation (b).
When we neglect Auger recombination upon low power excitation, an exponential function can be fit to the
raw data of the time dependent PL measurement, which allows us to retrieve the value of the lifetime, as
will be discussed in section 2.1.3. Recombination, however, can also occur non-radiatively, as described in
section 2.1.2.
Knowing the lifetime of both the carriers that decay radiatively and non-radiatively provides a possibil-
ity to calculate the efficiency of your LED. , the overall recombination rate is given by the sum of the
recombination rates for radiative and non-radiative recombination, according to
τ −1 = τr−1 + τnr
−1
, (2.14)
where τr and τnr represent the radiative and non-radiative lifetimes, respectively. If we define the recombi-
nation rate γ as the inverse of the lifetime τ then, for each recombination event, the probability of it being
radiative recombination is given by
γr γr
ηint = = , (2.15)
γ γr + γnr
where ηint is known as the internal quantum efficiency (IQE).
This principle is not only applicable to LEDs, but can be applied to any photoluminescent material. For
the purpose of this thesis, the lifetime will be calculated for heterostructures of 2D materials in order to
draw conclusions about the performance of these materials. The lifetime of carriers in 2D materials will be
briefly revisited in section 2.3.
2.1.4 White LEDs

White light can be used for many applications such as illuminating homes and offices, lighting the streets
or automotive lighting. However, according to the working principle of LEDs explained thus far, the emis-
sion from LEDs is (almost) monochromatic. White light is polychromatic by definition, and consequently
obtaining white light from LEDs to replace traditional white light sources such as incandescent lamps and
fluorescent lamps provides an additional challenge. [11]
17
Colorimetry
Colorimetry, or the science of color, involves the quantification of color. In order to quantify color, we dis-
tinguish between radiometric and photometric units. Radiometric units characterize the physical quantities
of light: photon energy, number of photons, optical power, etc. Although these quantities are measurable
and to some extent characterize the light, they do not account for the human perception of the light. For this
purpose, photometric units are required.
The human retina contains light-sensitive rod cells and cone cells. The cones mediate what is called photop-
tic vision, that is the human vision at high ambient levels. There are three different types of cone cells, with
peak sensitivity in the red, green, and blue part of the visible spectrum. Rod cells, on the other hand, mediate
what is scotopic vision, that is the human vision at low ambient levels. Rod cells are much more sensitive
to light than cones, which comes at the cost of losing the sense of color in these cells. At intermediate
luminance, vision is mediated by both cones and rods in what is called the mesopic vision regime.
Conversion between radiometric and photometric units can be achieved by means of the eye sensitivity
function V (λ). This function can be considered a weighted average of the sensitivities of the red, green,
and blue cone cells, normalized to the overall maximum sensitivity. The eye sensitivity function can be
seen in figure 2.10.
Figure 2.10: Eye sensitivity function V (λ) (left-hand ordinate) and luminous efficacy (right-hand ordinate)
as a function of wavelength. The maximum sensitivity is reached at a wavelength of 555 nm and corresponds
to a luminous efficacy of 683 lm/W.[11]
Now, if we define the visible wavelength range as the wavelength range for which the relative sensitivity
is at least 10−3 , from figure 2.10 we then conclude that the wavelength range 390nm ≤ λ ≤ 720nm
constitutes the visible wavelength range.
The measurement of color was standardized by the International Commission for Illumination (Commision
Internationale de l’Eclairage, CIE) by introducing the chromaticity diagram, which is provided in figure
2.11.
18
Figure 2.11: CIE 1931 (x,y) chromaticity diagram. Monochromatic colors are located on the edge of the
diagram, whereas polychromatic colors (including white) can be found in the interior.[11]
The basis for the chromaticity diagram are the color-matching functions. The idea behind the color-
matching functions is that for any monochromatic color, there exists a combination of relative intensities of
red, green, and blue light that will appear identical to the observer, and thus the colors will match. The rela-
tion between the relative intensities of red, green, and blue and the monochromatic light across the visible
spectrum is then described by the color-matching functions. We distinguish three different color-matching
functions: x(λ), y(λ) and z(λ). Color-matching function y(λ) is chosen to equal the eye sensitivity func-
tion V (λ). The color-matching functions are plotted in the figure below.
Figure 2.12: CIE 1931 standard observer color matching functions x(λ), y(λ) and z(λ).[38]
Next, we define the tristimulus values as the degree of stimulation, for a given power-spectral density P (λ),
19
required to match the color of P (λ), as:
Z
X= x(λ)P (λ)dλ
λ
Z
Y = y(λ)P (λ)dλ (2.16)
Zλ
Z= z(λ)P (λ)dλ.
λ
The chromaticity coordinates of the chromaticity diagram can be calculated from the tristimulus values as
the relative weight of each component to the entire stimulus, according to
X
x=
X +Y +Z
Y
y= (2.17)
X +Y +Z
Z
z= .
X +Y +Z
As can be seen in figure 2.11, monochromatic colors are located on the perimeter of the chromaticity
diagram. Although LEDs may appear monochromatic to the human eye, their finite spectral linewidth
renders them not monochromatic. Consequently, the emission spectrum of an LED can be expected to be
located close to the edge, rather than on the edge. The broader the emission range of a source, the more the
coordinate on the chromaticity diagram moves towards the center.
The z-coordinate, however, is redundant in the sense that it can be described in terms of the x- and y-
coordinate (z = 1 − x − y). Hence, the z-coordinate does not provide any new information and the full
description of the color can be expressed as a coordinate (x, y) on the chromaticity diagram.
Now we are interested in the coordinates (xmix , ymix ) that result from mixing two colors (x1 , y1 ) and
(x2 , y2 ). The tristimulus values for different colors are additive, and in case of each emission band being
much narrower than any of the color-matching functions can be approximated according to
Z Z
X= x(λ)P1 (λ)dλ + x(λ)P2 (λ)dλ ≈ x(λ1 )P1 + x(λ2 )P2
Zλ Zλ
Y = y(λ)P1 (λ)dλ + y(λ)P2 (λ)dλ ≈ y(λ1 )P1 + y(λ2 )P2 (2.18)
Zλ Zλ
Z= z(λ)P1 (λ)dλ + z(λ)P2 (λ)dλ ≈ z(λ1 )P1 + z(λ2 )P2 .
λ λ
where P1 and P2 are the powers emitted by the two sources. If we define
L1 = x(λ1 )P1 + y(λ1 )P1 + z(λ1 )P1

(2.19)
L2 = x(λ2 )P2 + y(λ2 )P2 + z(λ3 )P3 ,
then substitution of equation (2.18) into (2.17) yields
x1 L1 + x2 L2
xmix =
L1 + L2
(2.20)
y1 L1 + y2 L2
ymix = .
L1 + L2
20
2.2. SURFACE LATTICE RESONANCES
These equations can be generalized to a mix of n colors according to
n
P
xi Li
xmix = i=1
Pn
Li
i=1
n
(2.21)
P
yi Li
ymix = i=1
Pn .
Li
i=1
In this thesis, we strive for obtaining white light emission by means of combining three colors. Upon
drawing a line between the three sets of chromaticity coordinates in the chromaticity diagram, a triangle is
formed, also known as the color gamut. By adjusting the relative luminances of the three individual colors,
any color within the color gamut can be obtained. The color gamut for our final device will be treated in
section 4.10.
2.2 Surface lattice resonances

SLRs exploit the possibility of achieving a higher quality factor optical response and field enhancement by
means of an array of plasmonic NPs when compared to that of the LSP response of its constituents.[39] An
LSP involves the coupling of the oscillation of the free electrons in a NP to the electromagnetic field. The
resonance frequency of the LSP depends on several factors: size, shape, optical environment and material.
Typically, the resonance occurs in the visible to near-infrared of the spectrum in NPs made of noble metals
or Al.[40] The LSP resonance leads to local electric field enhancements in sub-wavelength volumes around
the NP. If we bring the NPs into close proximity of each other, coupling of the near-fields of the NPs
can occur. When organized into arrays where the period of the lattice is in the order of the wavelength
of the electromagnetic radiation, even higher field enhancement and higher quality factor resonances can
be generated.[41] Long-range radiative coupling of LSPs to NPs can be mediated by diffraction modes in
the plane of the array. This requires a uniform dielectric environment around the NPs to generate SLRs.
Contrary to LSPs, SLRs are able to produce field enhancements that extend over large volumes.
The existence of SLRs was postulated many years before they were discovered experimentally. Commonly,
the coupled dipole approximation (CDA) is used to explain the physics behind the coupling NPs in arrays.
In this approximation, we treat each individual NP in a 1D chain at position ri as an electric dipole with
polarizability αi . Then, the strength of the local electric field Eloc,i induces a dipole Pi = αi Eloc,i in each
NP. If the total array consists of N particles, the local field at each NP upon plane wave excitation is the
sum of the incident electric field Einc,i and the dipole field induced by the other N − 1 particles:
X
Eloc,i = Einc,i + Edipole ,i = E0 exp (ik · ri ) − Aij · Pj (i = 1, 2, . . . , N ), (2.22)
j=1
j6=i
In this equation, E0 is the amplitude, and k = |k| = 2π/λ is the wavenumber of the incident plane wave,
and Aij is the dipole interaction matrix. This matrix can be expressed as
2
2 rij × (rij × Pj ) rij Pj − 3rij (rij Pj )
Aij ·Pj = k exp (ik · rij ) 3 +exp (ik · rij ) (1 − ik · rij ) 5 (2.23)
rij rij
where rij is a vector pointing from dipole i to dipole j (i = 1, 2, . . . , N ; j = 1, 2, . . . , N ; i 6= j).

By assuming an infinite array of identical particles with polarizability α, equation (2.23) can be solved
analytically such that the polarization of each particle is found to be
1
P = a∗ E0 = E0 , (2.24)
1/α − S
21
2.3. 2D MATERIALS
where a∗ is the effective polarizability and S is the dipole sum, given by
"
X (1 − ik · rij ) 3 cos2 θij − 1 exp (ik · rij ) k 2 sin2 θij exp (ik · rij )
S= 3 + , (2.25)
rij rij
j6=i
where θij is the angle between the direction of the polarization induced by the incident electric field and
rij .
From the effective polarizability a∗ , then the absorption, scattering, and extinction cross sections can be
calculated according
Cabs = 4πk=(a∗ )
8
Cscat = πk 4 |a∗ |2 (2.26)
3
Cext = Cabs + Cscat .
The criterion for excitation of an SLR is that the magnitude of the polarization P of a NP in the array is
large. From equation (2.24), we conclude this condition is met when < (1/α) = < (S) and the difference
between = (1/α) and = (S) is small. The values of the real and imaginary parts of S and 1/α as a function
of energy are plotted in figure 2.13 (right). Dashed vertical lines indicate the energy values where <(S)
and <(1/α) cross. At the smaller energy value, the difference between =(S) and =(1/α) is small, whereas
this is not the case at the larger energy value. We conclude that the left vertical dashed line represents the
energy where the condition for an SLR is met. Figure 2.13 (left) shows the extinction cross section as a
function of energy for a single particle and a particle in an array. Indeed, the peak extinction cross sections
occurs at the same energy as the left vertical dashed line. Furthermore, we can see that the extinction cross
section of the SLR is enhanced with respect to the single particle resonance. Moreover, the quality factor
of the particle in the array is significantly higher, given its much narrower bandwidth.
(a) (b)
Figure 2.13: Calculated spectra of the extinction cross section of a single particle (green), and a lattice
of particles (blue) as a function of energy (left) and the real and imaginary components of S and 1/α
(right).[42]
The conditions for an SLR illustrate how the coupling of the individual LSP and the diffraction modes of
the lattice as a whole can lead to electric field enhancement. The polarizability α can be calculated using
electromagnetic theory, and then the term 1/α describes the properties of a single particle. Parameters such
as size, shape and composition can be varied to tailor the value of this term. The S term, on the other hand,
describes the geometry of the entire structure, think of lattice configuration and period. This provides an
additional degree of freedom with respect to tuning the resonance when compared to LSPs.
22
2.3. 2D MATERIALS
2.3 2D materials
When compared to 3D materials, 2D materials often display unique properties. Most notably, quantum
confinement in the out-of-plane direction causes optical properties different from their bulk counterparts.
Furthermore, integration of 2D materials into heterostructures is possible without lattice mismatch due to
the surface of 2D materials being naturally passivated, without any dangling bonds.[25] This section will
treat the properties of specific 2D materials in more detail.
2.3.1 Transition metal dichalcogenides

TMDs are materials with chemical formula MX2 , where M is a (group VI) transition metal and X a chalco-
genide. Typically, M = W, Mo; X = S, Se. Bulk crystals of TMDs consist of vdW bonded monolayers, most
commonly stacked in the ABAB configuration, also known as the 2H polytype. Bulk TMDs are semicon-
ductors with an indirect band gap, of which the conduction band minimum and the valence band maximum
are located at the Q point and the Γ point respectively. [26] The unit cell of TMDs both in real space and in
reciprocal space is provided in figure 2.14.
(a) (b) (c)
Figure 2.14: Trigonal prismatic structure of monolayer TMDs (a), top view of the honeycomb lattice struc-
ture (b) and the first Brillouin zone with the high-symmetry points (c).[26]
Upon decreasing the number of layers, the quantum confinement in the out-of-plane direction increases.
As a result, the indirect gap size increases, whereas the direct gap at the K/K’ points remains largely un-
changed. In the monolayer limit, the gap at the K/K’ points becomes energetically favorable, rendering
TMD monolayers direct band gap semiconductors. [26] Consequently, TMD monolayers have drawn con-
siderable scientific attention.
2.3.2 WS2
One specific TMD material, WS2 , will be of specific interest for the purpose of this thesis. WS2 , in
its monolayer form, is particularly interesting because of its high PL quantum yield due to low defect
density when compared to other TMD monolayers and its band gap in the red part (∼ 2 eV) of the visible
spectrum.[27] The influence of the number of layers on the band structure of WS2 is provided in figure
2.15.
Figure 2.15: The band structure of W S2 (a) monolayer, (b) bilayer, (c) trilayer and (d) quadlayer. [27]
23
2.3. 2D MATERIALS
2.3.3 Van der Waals heterostructures

Because of the atom thin nature of the absorption efficiency and PL intensity of monolayers is rather low.
In order to increase the absorption and PL from WS2 , one could opt for stacking several monolayers into a
vdW heterostructure. This is not straightforward though, because of the transition to energetically favorable
indirect band gap transitions in mulitlayers. Two layers of WS2 should be spatially separated in order
to retain the direct band gap property. Stacking thin layers may have significant effects on the optical
properties, depending on the dielectric environment (e.g. material type and thickness). The presence of
field lines of the Coulomb interaction in every layer is ’sensed’ by the adjacent material, in a process called
dielectric screening.[31] The effect of a change in the dielectric environment on the electric field is provided
schematically in figure 2.16.
Figure 2.16: Effect of a change in the dielectric environment on the electric field between ideal (a) and
realistic interface (b) and its influence on the observed band gap energy (c).[43]
It can be readily concluded that interlayer interactions are highly dependent on the specific sample char-
acteristics. Nevertheless, there are some general notions, that are worthy to mention for this thesis. For
instance, the choice of spacer material may influence the different excitonic transitions due to dielectric
screening of the environment.[31] Furthermore, the interlayer interactions and thus resulting optical prop-
erties are highly dependent on the thickness of the spacer.[28] Generally, interlayer excitons, exclusively
present in vdW heterostructures, have longer lifetime than intralayer excitons, which dominate in isolated
monolayers.[28] This information is useful when analyzing the carrier lifetime of fabricated vdW het-
erostructures in section 4.9. Additionally, during exfoliation and transfer of the 2D materials, mechanical
deformation may induce the formation of bubbles and/or wrinkles, affecting the interface properties of the
multilayer structure.[32]
In this thesis, a particular focus will be on using hBN and Al2 O3 as spacer materials between layers of WS2 .
Generally, quantum confinement effects in vdW heterostructures increase the observed band gap energy. For
hBN/WS2 stacks, however, a decrease in band gap energy is reported.[30] This contradiction once again
exemplifies how interlayer interactions are highly dependent on the specific sample characteristics.
24
3 Methodology and setups
This chapter treats the used methods for fabrication of the samples under investigation as well as the used
setups for performing the measurements. In section 3.1, the fabrication of the InGaN/GaN LED (3.1.1) and
the fabrication of the NP arrays (3.1.2) on top of this LED will be discussed.
In section 3.6, the principle of mechanical exfoliation for obtaining atomatically thin layers and the fabri-
cation of vdW heterostructures (3.6.1) will be treated. In this section, we will also define the relevant areas
on our heterostructures that will be investigated in terms of PL, absorption, lifetime and efficiency.
In section 3.2, the InGaN/Gan LED is modelled with and without NPs and the emission enhancement is
simulated based on the finite element method in a software environment called COMSOL Multiphyisics.
Next, the used setups for our measurements will be described. The confocal microscope is described in
section 3.7, and is used for PL (3.7.1), transmission (3.7.2), and lifetime (3.7.3) measurements. The Fourier
plane imaging setup, which is used for BFP imaging and laser excited PL measurements, is described in
3.3. The newly built microprobe setup that can take EL spectra and BFP images is described in section 3.5.
This same setup can also be used for diode burning (3.4). Lastly, the working principle of the integrating
sphere, used for EQE measurements, will be treated in section 3.8.
3.1 Sample fabrication

This section will treat the fabrication of the InGaN/GaN LED that will serves as a starting point for this
thesis (3.1.1), as well as the fabrication of NP arrays on top of the InGaN/GaN LED (3.1.2).
3.1.1 InGaN/GaN LED

The wafer used for the measurements in this thesis was fabricated by Lumileds Holding B.V. Using a
metal-organic chemical vapor deposition (MOCVD) reaction, the InGaN/GaN was grown on a flat sapphire
substrate. In the next step, a 6 µm GaN layer, two effective InGaN/GaN quantum wells with a total thickness
of 10 nm, and a p-GaN layer of 90 nm were grown, by means of epitaxial growth. Then, the substrate was
thinned down to 240 µm and the backside was polished. Finally, the wafer was covered by a 20 nm layer
of indium tin oxide (ITO). This transparent conductive oxide allows a homogeneous current distribution to
form over the device.
The wafer as described above was used as a substrate for depositing the 2D materials later. For electrical
excitation, however, additional metal contacts were required. On the wafers intended for electrical charac-
terization, 200µm × 200µm Al square contacts were deposited on the ITO side and the ITO all around the
contacts is etched. In section 3.4 we describe how we excited the LED electrically with both the anode and
the cathode contact on the same side of the wafer. Although the Al contacts are reflective, we were still able
to excite this LED optically as long as we excited from the sapphire side. An image of the LED structure
with Al contacts as well as the emission spectra upon optical excitation (with a 405 nm laser) and electrical
excitation are provided in figure 3.1.
25
3.1. SAMPLE FABRICATION
(a) (b)
Figure 3.1: Schematic cross section of the InGaN/GaN LED wafer with Al contacts (a) and the emission
spectra of this LED upon optical excitation and electrical excitation (b).
We concluded that the peak wavelength upon optical excitation lies at 440 nm, and the full width half
maximum (FWHM) was 14 nm. Upon electrical excitation, the peak wavelength was red-shifted to 452
nm, and the spectrum was broader with a FWHM of 18 nm.
3.1.2 Nanoparticle fabrication

Arrays of Al metasurfaces were fabricated by drs. Mohamed S. Abdelkhalik using electron beam lithog-
raphy (EBL) and a lift-off process. A poly(methyl methacrylate) (PMMA) positive resist was coated onto
the ITO with a thickness of 240 nm on the LED wafer and prebaked for three minutes at 135 °C, followed
by four minutes at 185 °C. The positive resist was exposed using an EBL system (Raith EBPG 5150,100
kV) to nanopattern Al nanodisks in hexagonal arrays with dimensions of 200µm × 200µm and with an
intended period of 215 nm. The resist was developed using MIBK/IPA 1:3 developer for 80 seconds. Then,
32 nm of Al was evaporated using electron beam (e-beam) evaporation and using acetone to remove the
resist covered by Al, a lift-off process was performed. At last, 108 nm of TiO2 was deposited using ALD,
performed by Lumileds Germany GmbH. To characterize the fabricated sample, a focused ion beam (FIB)
image of the final device was taken. Both a schematic of the sample as well as the resulting FIB image of
the final sample are provided in figure 3.2.
(a) (b)
Figure 3.2: Schematic of the intended InGaN/GaN LED wafer with NP arrays (a) and focused ion beam
image of the actual structure made by drs. Mohamed S. Abdelkhalik (b).
The FIB image shows that the actual period is considerably smaller than intended (125 nm instead of 215
26
3.2. FINITE ELEMENT METHOD SIMULATIONS
nm). This was caused by a mistake with the layout design provided to the EBL. The significant difference in
period will have an influence on the extent to which the NP array supports SLRs, and thus on the emission
enhancement and reshaping of the far-field emission. This notion is crucial to take into account when
analyzing the results of the measurements. Moreover, a conformal rather than a flat layer of TiO2 formed
around the NPs. Since the conformal layer formed around the NPs, a flat rather than a conformal layer was
still present in our reference samples without NPs. The wafer structure with NPs and conformal layer as
well as without NPs and without conformal layer will be modelled with COMSOL Multiphysics, as shown
in section 3.2, to simulate the influence on the performance of the LED.
3.2 Finite element method simulations

The InGaN/GaN LED with conformal structure as shown in figure 3.2 was modelled by drs. Mohamed S.
Abdelkhalik and compared to a model without the NPs and conformal layer. The emission enhancement
was simulated based on the finite element method in a software environment called COMSOL Multiphysics.
The x-y and x-z plane of the modelled unit cell with labels for the structure with NPs and conformal layer
are provided in figure 3.3.
(a) (b)
Figure 3.3: COMSOL snapshot of the x-y plane (a) and the x-z plane (b) of the simulated geometry unit
cell.
In the model, the LED was excited using a plane wave for both 440 nm and 452 nm to simulate the enhance-
ment upon optical excitation and electrical excitation, respectively (see figure 3.1). Furthermore, the ratio
of the in-plane to out-of-plane emission of the LED was set to 90:10, based on the average dipole emission
as stated by the wafer manufacturer. The simulation was based on the reciprocity principle. This principle
dictates that the far-field emission from the metasurface in a certain direction with integrated incoherent
emitters is reciprocal to the intensity of the near-field excited in the metasurface by the incident external
light in the same direction, averaged over the positions of the emitters.[44] A graphical interpretation of the
reciprocity principle is provided in figure 3.4.
27
3.3. PHOTOLUMINESCENCE AND FOURIER PLANE IMAGING SETUP
Figure 3.4: Schematic image displaying the reciprocity between the far-field emission from the metasurface
and the intensity of the near-field averaged over the positions of the emitters.[44]
The electric field distributions of both the structure with and without NP arrays will be compared in section
4.1, and from this the emission enhancement will be calculated. Furthermore, a simulated BFP image of
the intended structure with period of 215 nm is provided to see what the reshaping of the far-field would
have looked like in case the fabrication would have been performed according to plan.
3.3 Photoluminescence and Fourier plane imaging setup

Fourier plane imaging, also known as BFP imaging, allows for the retrieval of the angular distribution of the
emission of a sample. The technique is based on the ability of an objective lens to create a Fourier transform
of an object at its BFP. Any plane wave impinging on the objective lens with an angle θ is mapped onto
a single point in the BFP. Typically, the BFP of an objective lens lies within the objective body, requiring
additional external optics to image the BFP. For this purpose, a two-lens configuration is used. A first lens,
also known as a tube lens, is placed at its focal distance f1 away from the BFP of the objective lens. This
lens performs a Fourier transform, yielding an intermediate image plane at a distance f1 from the tube lens.
A second lens, commonly referred to as a Bertrand lens, is placed at a distance f1 + f2 from the tube
lens, with f2 the focal distance of the Bertrand lens. This lens performs a second Fourier transform, which
recovers the BFP of the objective at a distance f2 away from the Bertrand lens. A schematic of the above
described system in the used setup can be found in figure 3.5.
Figure 3.5: Schematic of a Fourier microscope setup.
As can be seen in figure 3.5, two light rays emitted at the same angle but at different positions will end up
at the same position in the BFP. This principle can be extended to a 3D case, where a plane wave is emitted
from a position (x,y) on the sample under an angle (θx , θy ). By placing a CCD camera at exactly one focal
length of the Bertrand lens, where the BFP is recovered, a circular image is obtained, where each pixel
corresponds to a unique emission direction (θx , θy ). The radius of this circle is limited by the maximum
emission angle as collected by the objective lens, that is the numerical aperture (NA). The image provides
information on how many photons are detected for each emission angle. This way, the BFP image provides
us with information on the directionality of the emission. The principle of angular decomposition in BFP
imaging is summarized in figure 3.6.
28
3.4. DIODE BURNING
Figure 3.6: Schematic of the working principle of angular decomposition in back-focal (Fourier) plane
imaging. Courtesy of Dr. Gabriel W. Castellanos.
In our case, the LED wafer was placed in a sample holder with the contacts down, with an objective lens
(Nikon TU Plan Fluor 100X/0.90) placed at the working distance (1.0 mm) away from the sample. Because
of the reflection of the Al contacts of the LED, the setup must be operated in reflection mode, meaning the
objective lens functions both as a focusing objective and as a collection objective. The LED was excited
by a 405 nm laser (NKT Photonics SuperK SELECT) through an objective lens, and both the emission and
the reflected laser light were collected by the same objective lens. Furthermore, a flip mirror allowed us to
switch between retrieving the BFP image and measuring the PL from the sample. A 405 nm notch filter
was used to filter out the laser emission that is collected upon reflection.
3.4 Diode burning

In order to electrically excite an LED, current must be injected at the anode side of the diode, and extracted at
the cathode side of the diode. In the proposed device, both the positive and negative contact were fabricated
on the upper surface. After the current has flown through the diode in the forward direction once, the current
needs to flow back through the diode in the reverse direction in order to reach the negative contact. The
resistance in this latter direction, however, is very high. From a circuit modelling point of view, two diodes
are connected in series, in the anti-parallel orientation, to a voltage supply. A schematic of the described
current path together with a model of the equivalent electrical circuit is shown in figure 3.7.
(a) (b)
Figure 3.7: Schematic of the considered current path (a) and the model of the equivalent electrical circuit
upon electrical excitation (b).
To overcome this problem of high resistance of the reverse diode, commonly, an etching process is applied
to the p-side of the junction and the MQWs. Then, the negative contact can be placed on top of the n-side
29
3.5. ELECTROLUMINESCENCE AND FOURIER PLANE IMAGING SETUP
of the junction. This etching process proves to be difficult and time-consuming, though. For this reason,
alternatives to etching were investigated. The proposed solution to the problem is called diode burning,
where a short, high-voltage pulse is applied in order to reach the breakdown region of the diode in reverse
bias. After burning the one diode, the LED in forward bias will operate at low voltages as expected from
when etching would be applied. This process was performed in the same setup as for measuring EL, by
replacing the continuous voltage supply with a high-voltage pulse generator. For this purpose, a pulse
generator (Agilent 33250A 80 MHz Function/Arbitrary Waveform Generator) was used in combination
with a home-made power amplifier. To ensure precise control over the output signal of the power amplifier,
the signal was monitored using an oscilloscope (Rigol DS1074B). This process will be characterized in
terms of the used pulse amplitude and pulse width. The results of this characterization of the burning
process will be discussed in section 4.3.
3.5 Electroluminescence and Fourier plane imaging setup

Although for many purposes LEDs can be characterized by optical excitation, a working device will always
need to be electrically excited. Additionally, part of the emission enhancement in PL measurements of LEDs
with NP arrays can be attributed to the enhancement of the pump. In other words, upon optical excitation,
part of the incident laser light is scattered by the NPs, causing improved absorption by the InGaN/GaN
MQWs. This can be explained by the occurrence of constructive interference between the incident and
reflected waves at the position of the MQWs, causing the MQWs to be exposed to a higher intensity of
light. In turn, this results in enhanced emission when compared to electrical excitation of the same sample,
since pump enhancement is not expected to occur in EL measurements. Altogether, the comparison of
optical measurements and electrical measurement may provide interesting insights when characterizing
LEDs.
Electrical excitation of micro-LEDs, however, is not as straightforward as it is for conventional LEDs. The
small size of the LED itself automatically results in the contacts being small as well. A new setup was
specifically built for the purpose of electrically exciting the micro-LEDs and the possibility to measure the
EL, as well as to observe the real and Fourier image of the emission simultaneously. A schematic of the
setup can be found in figure 3.8.
Figure 3.8: Top view of the microprobe setup used for electrical excitation of the LED. A Nikon TU Plan
Fluor 100X/0.90 objective and a 45° mirror are located underneath the sample holder to collect the emitted
light and direct it parallel to the optical table.
A sample holder was specifically designed and 3D-printed for accommodating a one inch LED sample. The
contact to the LED was made from the top, using microprobes with a tip diameter of 25 µm. Using a camera,
the position of the microprobe tips with respect to the Al contacts was observed. Two three-axis linear
translation stages allowed for micrometer accurate control over the x-, y- and z-position of the microprobes.
These probes were connected to the positive and negative terminal of a continuous voltage supply. In power
30
3.6. MECHANICAL EXFOLIATION OF WS2
dependent measurements, a multimeter was connected parallel to the terminals of this voltage supply for
a more accurate control over the voltage. Furthermore, a second multimeter was connected in series to
measure the current before returning to the ground of the voltage supply, and thus, the current flowing
through the LED device.
The emitted light is collected by the same objective as in the setup for optical excitation, from the bottom
of the sample holder. A mirror is placed under 45° to direct the collected light parallel to the table, towards
a lens with a focal distance of 15 cm. Next, the light encounters a flip mirror, where in the upright position
the light will be sent to a fiber coupling unit which is connected to a spectrometer to measure the emission
spectrum. In case the flip mirror is in the downward position, the light will propagate through a beamsplitter.
The part of the light that is reflected to the right will end up at a CCD camera, where the real plane image
is measured. The part of the light that continues straight at the beamsplitter, will be focused once more by
another 15 cm focal distance lens, after which the Fourier image can be measured on a second CCD camera.
The laser, together with the optical density (OD) filter, fixed mirror, flip mirror, and two pinholes, are solely
present for alignment purposes. Additionally, alignment of the z-position of the sample with respect to
the objective upon emission from the sample is difficult because it would also require the microprobes to
move equal distances in the z-direction. Instead, using the laser and a grating, the focus can be found by
observing the diffraction orders of the grating in the Fourier image, without the requirement of moving the
microprobes.
3.6 Mechanical exfoliation of WS2

In order to obtain atomically thin layers of WS2 , the principle of mechanical exfoliation was applied to
a bulk crystal. The method is rather straightforward and identical to how the groundbreaking discovery
of graphene was achieved.[45] Using adhesive tape, single atomic layers of WS2 was extracted from the
bulk crystal by means of a peeling force.[46] For the purpose of optimizing this process, another piece of
adhesive tape was fixed upside down on a sheet of aluminum foil, which in turn was fixed to a table. A
small piece of bulk crystal was then plucked from a larger crystal (HQ Graphene) and put on the adhesive
tape on the table. Then, a second piece of tape was brought into contact with the piece on the table, while
the small piece of bulk crystal was in between. After applying a small amount of lateral force for ensuring
sufficient contact, the top piece of tape was peeled off, leaving material behind on both pieces of tape. This
principle is shown schematically is figure 3.9.
Figure 3.9: Schematic of the principle of mechanical exfoliation, showing a bulk piece of W S2 crystal
sandwiched between two pieces of tape. A peeling force overcomes the van der Waals force and cleaves
thin layers from the bulk crystal.
This process was then repeated over and over, until a sufficient amount of material was present on the top
piece of tape. Then, a soft transfer was performed by placing the piece of tape containing the potential
monolayer on polydimethylsiloxane (PDMS), which in turn was deposited on glass. After applying a small
amount of force to both ensure the contact and remove possible air bubbles, the tape was peeled off and part
of the material was transferred to the PDMS substrate. Optical microscopy was then performed to very that
a sizeable monolayer flake was successfully obtained.
31
3.6.1 Stacking of atomically thin layers

In order to achieve sufficient absorption of blue light and intense emission of the red light, a single mono-
layer WS2 will not suffice. 2D atomic crystals can be stacked on top of each other to create what is called
a vdW heterostructure.[46] However, monolayers of WS2 can not be stacked directly on top of each other,
because of the direct band gap property only applying to monolayers. A layer of hBN can be used to spa-
tially isolate the layers. In this thesis, heterostructures were fabricated onto LED wafers. Hence, first the
process of stamping the bottom monolayer onto the LED wafer will be addressed.
Stacking of the different layers is commonly achieved by preparing monolayers onto a transparent thin
film. In our case, the PDMS substrate, which resides on a glass slide, was cut to a size as small as possible
around the target monolayer. The glass side was fixed to a three-axis translational stage using screws with
the monolayer facing down. The target substrate, in our case an InGaN/GaN LED wafer, was fixed to a
second three-axis translational stage using double sided adhesive tape. Upon bringing the two samples into
contact, a contact line can be observed using an optical microscope. Although the exact position of the
monolayer with respect to the substrate was irrelevant in this step, the optical microscope was still used
for observing the contact line. This is to ensure not to apply more force than required, which may inflict
damage. After allowing the contact some time, the stamping stage was lifted up again and the monolayer
was transferred to the target. This conceptually straightforward process proves to be difficult in practice and
requires clean-room level cleaning, and micrometer accuracy for the alignment of different layers.[46] Even
then, applying excessive force may cause breaking of the monolayer flakes. Also, isolating the monolayer
requires cutting of the PDMS. In case of cutting close to the monolayer, damage may be inflicted. On
the other side, if the PDMS is not cut closely to the monolayer, surrounding bulk flakes may complicate
a successful transfer. A last complication arises from the lack of angular control over the samples upon
stamping in the current setup. The schematic of the setup used for this stamping process can be found in
figure 3.10.
Figure 3.10: Schematic of the stamping setup used for fabricating van der Waals heterostructures, consist-
ing of two micrometer translational stages and an optical microscope.
In the second step of fabrication, a thin (several atomic layers) layer of hBN was stamped onto the bottom
monolayer. The principle here is similar to that of the initial step, except for that now alignment of the
position of the hBN flake with respect to the monolayer is crucial. For this purpose, IC Measure image
acquisition and measurement software was used. First, the focus of the microscope objective was set to
the hBN flake by adjusting the z-position of the stamping stage (figure 3.10). Then, using the IC Measure
software, the perimeter of the flake was traced and saved on screen. Next, the stamping stage was moved
32
up and the substrate stage with the monolayer was brought into focus. The monolayer appeared on screen
in the IC Measure software and the x- and y-position of the substrate stage were adjusted to align to the
hBN flake. Next, the stamping stage was moved down in the z-direction until the contact line was observed
under the microscope.
Because of the high risk of introducing defects associated with manually stacking atomically thin layers
on top of each other, an alternative method for isolating the monolayers was proposed for comparison. A
second sample was fabricated where ALD of 2 nm of Al2 O3 was used for spatial isolation of the WS2
monolayers.
On both samples, we then defined three areas that will be investigated separately in the results section of this
thesis: Area 1 corresponds to the overlapping bilayer area, area 2 corresponds to the only the top monolayer,
and area 3 corresponds to only the bottom monolayer. The different areas of investigation are marked by
blue circles with corresponding numbers for both samples, in the blue lamp optical microscope image as
provided in figure 3.11.
(a) (b)
Figure 3.11: Final device with partial overlap of the monolayers and the areas under investigation indi-
cated, for Al2 O3 (a) and hBN (b) as a spacer.
In the sample where hBN was used as a spacer, the bilayer was also encapsulated by hBN, using the same
stamping principle. This encapsulation will be relevant when we apply QDs onto the WS2 monolayers
in the future, without inflicting damage. In case we choose hBN as a spacer, our proposed white LED,
including NP arrays, TMDs and QDs (to be fabricated in the future) will have the structure as provided in
figure 3.12.
33
3.7. CONFOCAL MICROSCOPY SETUP
Figure 3.12: Schematic of future final device for efficient white light emission from micro-LEDs.
3.7 Confocal microscopy setup

A confocal microscope (Nikon Ti-E Eclipse C2) was used for performing PL, transmission and lifetime
measurements. The microscope is an inverted microscope, meaning the objective is located underneath the
sample, and it operates in reflection mode. Two different turret stages allow for the insertion of different
types of beamsplitters and filters. Using different optical paths, different types of measurements can be
performed, as will be discussed in the remainder of this section. A schematic image of a typical confocal
microscope can be found in figure 3.13.
Figure 3.13: Schematic image of a typical confocal microscope, in many aspects similar to the one used
in our measurements. The main difference lies in the fact that not all detectors connected to our confocal
microscope are displayed in the image.
In traditional wide-field fluorescence miscropy, the entire sample is illuminated evenly and all the fluores-
cent light, including out-of-focus background light, is collected. Confocal microscopy aims to enhance the
optical resolution and contrast by means of a spatial pinhole to block unfocused background light. A laser
(Spectra-Physics Mai Tai Ti:Sapph oscillator in combination with Radiantis Inspire Femtosecond OPO) ex-
cites the sample very locally and also only the emission from a small area is collected to increase resolution.
A scanning head then moves across the sample to retrieve spatial information of the emission in order to
reconstruct a 2D image.
34
3.8. INTEGRATING SPHERE
3.7.1 Photoluminescence
The standard illumination source of the confocal microscope is a mercury arc lamp coupled to an optical
fiber. By means of a movable filter, either attenuated white light or blue light centered around 420 nm will
be transmitted. The latter was used for PL measurements. The light is directed towards the objective lens
through a longpass filter cube that is also a beamsplitter in the upper turret. Upon reflection from the sample,
the blue light of the excitation source is filtered out and the PL can be directed either towards the eye piece,
a CCD camera, a spectrometer or a confocal scanning head. For performing PL measurements, the sample
is loaded onto a motorized three-axis translational stage. By adjusting the z-position of the sample stage,
the focus with respect to the objective lens can be found. Then, by using the blue lamp filter and choosing
the correct optical path in the microscope settings, the PL of the sample can be directed to the spectrometer
(Shamrock - SR303i). By means of the Andor Solis 64 software, first a background is taken without any
PL directed to the spectrometer. When taking the measurement, the software automatically corrects for the
dark counts.
Alternatively, the PL measurements can be performed with a free space laser (Spectra-Physics Mai Tai
Ti:Sapph oscillator in combination with Radiantis Inspire Femtosecond OPO) as the excitation source. In
accordance with the emission wavelength of the InGaN/GaN LED, a 450 nm laser was used for the excita-
tion. A beamsplitter cube was placed in the lower turret for directing the laser to the objective lens, onto the
sample. The working principle is the same as for the blue light excitation, with the microscope operating
in reflection mode, and the reflected light can be directed towards the output of preference. However, this
time not a longpass filter cube was used, but a 50/50 beamsplitter in combination with an additional notch
filter to filter out the 450 nm laser excitation.
3.7.2 Transmission
Apart from the photoluminescent properties of the samples under investigation, we are also interested in
their absorptive properties. For this purpose, a measurement in transmission mode is required. An external
white light source (Thorlabs SLS201/M) is fiber coupled and focused onto the sample from the top using
a lens in free space. In order to acquire a transmission spectrum of the WS2 , this transmission needs to be
corrected for the presence of the substrate. Besides taking the background and the signal like for the PL
measurements, we now also take a reference measurement, that is the number of counts as detected after
transmitting through the substrate (e.g. glass + PDMS). Upon taking a signal on the WS2 , the Andor Solis
64 software then automatically corrects for the contribution of the substrate to the overall transmission.
Conservation of energy dictates that T + R + A = 1, where T is transmission, R is reflection and A is
absorption. By (boldly) neglecting the reflection, we conclude that A = 1 − T , or equivalently A[%] =
100% − T [%] when expressed in percentages. The latter is the preferred method for showing the results in
section 4.7.
3.7.3 Fluorescence-lifetime imaging microscopy

A fluorescence-lifetime imaging microscopy analysis may provide interesting information on recombina-
tion dynamics and the efficiency in both monolayers and multilayers of 2D materials. For measuring the
lifetime of the WS2 , the confocal scanning microscope in combination with a time correlated single photon
counter (TCSPC) diode (Picoquant TimeHarp 300) is used. A 450 nm laser is fiber coupled to the scanning
head, which scans the excitation source over an area of interest. The scanning head sends laser pulses to the
selected area and over a period of time, re-emitted photons are being collected of which both the spatial and
temporal information are stored. Using the x- and y-coordinate and arrival time of each photon, a micro-PL
image of the sample can be reconstructed with the SymPhoTime 64 software. The micro-PL image can
represent the events, that is the number of detected photons (PL), for each position (x,y) using a color bar.
Alternatively, based on the arrival time of each photon, the mean lifetime can be displayed using a color bar
for each position (x,y). Altogether, these images may provide interesting information on the efficiency for
monolayers versus artificial bilayers, and the influence of edges or cracks on the recombination of excitons.
Additionally, within the same software, an area of interest can be selected for which an exponential fit can
be made to the raw data of events versus time. Increasing the order of the exponential fit may reduce the
residuals of the fit, however, also adds complexity to interpreting the results. Every extra order corresponds
to an extra lifetime and thus an extra recombination mechanism. In the case of a biexponential fit, the rel-
ative amplitude of the respective lifetimes will determine which of the two is the dominant recombination
mechanism.
35
3.8 Integrating sphere

When attempting to measure the quantum efficiency of photoluminescent materials, all emitted light must
be collected. This may be a problem in high refractive index materials due to the angular distribution of the
emission.[47] An integrating sphere is a hollow sphere of which the inner surface is coated with a diffu-
sively reflecting material. When a photoluminescent sample is placed within the sphere, all emitted light is
redistributed isotropically over the surface of the sphere, independent of the direction of the emission. An
illustration of the integrating sphere can be seen in figure 3.14 (a). The integrating sphere has a small open-
ing on the side, through which the laser used for excitation can enter. An optical fiber collects the emission
and sends it to a spectrometer, which measures the spectrum of both the laser and the PL. Furthermore, the
integrating sphere contains a sample holder, in which the photoluminescent material can be placed.
The EQE of a photoluminescent sample is defined as the number of photons emitted over the number of
photons absorbed. Three measurements with the integrating sphere are required for the calculation of the
EQE. In the first measurement, the sphere is left empty and the measured spectrum is solely based on the
output of the laser. In the second measurement, the photoluminescent sample is placed inside of the sphere,
but is oriented such a way that the laser is not directly incident onto it. In the third measurement, the laser
is directly incident onto the sample. A summary of the three measurements is given by figure 3.14.
Figure 3.14: The integrating sphere and the three different configurations used in the experiments. In
experiment (a), no sample is placed in the sphere. In experiment (b), the sample is placed in the sphere but
is only excited with indirect illumination. For experiment (c), the laser is directly incident upon the sample.
The area under the laser profile, which we shall call L, is proportional to the amount of unabsorbed light.
In experiment (a), we expect to see only this narrow laser profile (La ) around 450 nm. In experiment (b)
and (c), part of the laser light will be absorbed, decreasing the peak at 450 nm (Lb and Lc ), and will be
re-emitted at longer wavelengths. These areas of PL will be defined as Pb and Pc for experiments (b) and
(c) respectively. An example of an expected spectrum can be observed in figure 3.15.
36
Figure 3.15: Example of an expected spectrum for experiment (a) in black, experiment (b) in red and
experiment (c) in blue.
Next, we define µ as the fraction of scattered laser light that is absorbed by the sample in experiment (b).
This allows us to define the relation between the laser profiles in experiment (a) and (b) as
Lb = La (1 − µ). (3.1)
Then, we define A as the fraction of the incident laser light that will be absorbed directly upon hitting the
sample in experiment (c). This means that a fraction (1 − A) is either reflected or transmitted, and will then
scatter on the interior of the integrating sphere. From experiment (b), we know that a fraction µ of this light
will be absorbed by the sample. Altogether, we conclude that
Lc = La (1 − A)(1 − µ). (3.2)
Now, substitution of Lb from equation (3.1) into equation (3.2) yields the following expression for the
absorption coefficient A:

Lc
A= 1− (3.3)
Lb
The final spectrum Lc + Pc as measured in experiment (c) originates from two contributions: The contri-
bution of the scattered light is equal to (1 − A)(Lb + Pb ), while the light emitted due to the absorption of
the laser light is equal to ηLa A, where η is the EQE. Adding this up, leaves the energy balance
Lc + Pc = (1 − A)(Lb + Pb ) + ηLa A. (3.4)
Finally then, by means of elimination of Lb and Lc using equation (3.3) and rearranging equation (3.4), we
find
Pc − (1 − A)Pb
η= . (3.5)
La A
37
In this way, the EQE of arbitrary photoluminescent samples can be calculated based on the results of the
three aforementioned experiments. The only requirement then is that the photoluminescent material is
larger in size than the laser spot and that its PL is sizeable when compared to the laser peak.
For the case of WS2 monolayers, unfortunately, this criterion is not met. With a typical size of the order
of tens of micrometers, measuring the EQE in the integrating sphere is very difficult. Therefore, a relative
EQE measurement with respect to a reference sample was performed. A dye with strong absorption around
450 nm and PL at longer wavelengths was chosen in the form of Lumigen F Green and was spin coated
onto a microstructured substrate, for more efficient outcoupling. The EQE of this reference sample was
measured in the integrating sphere using the process described above. Then, the PL of this sample with
known EQE was measured using the same laser excitation in the confocal microscope setup. Next, the
PL of the WS2 monolayer (or artificial bilayer) was measured under the very same conditions. A simple
relative calculation then teaches us the EQE of the WS2 sample according to
P LW S2
ηW S2 = · ηref . (3.6)
P Lref
38
4 Results and discussion
In this chapter, the results of the performed measurements will be provided and analyzed. First, the results of
the finite element method simulations will be treated in section 4.1. These simulated results will then be used
for comparison to the experimental results. Section 4.2 will treat the power dependent PL measurements
and the observed emission enhancement. After the principle of diode burning is characterized in section
4.3, also the power dependent EL measurement and the respective emission enhancement can be analyzed
in section 4.4.
The focus will then shift to the 2D materials. The PL and absorption measurements of different samples
will be discussed in sections 4.6 and 4.7, respectively. Then, the EQE of these samples will be treated in
section 4.8, and the results will be compared to the lifetime measurements in section 4.9. Lastly, we will
provide the color gamut for our proposed device based on the measured spectra in section 4.10.
4.1 COMSOL simulations

As discussed in section 3.2, the enhancement was calculated for an incident wavelength of 440 nm, corre-
sponding to the PL peak wavelength, and 452 nm, corresponding to the EL peak wavelength.
The simulated electric field distribution throughout the modelled structure of the fabricated sample is shown
in figure 4.1.
Figure 4.1: Simulations of the spatial distribution of the electric field in the y-z plane of the model of the
fabricated structure for a wavelength of 450 nm.
We observe a high field enhancement in the vicinity of the NP, around the conformal layer. Standing waves
occur due to constructive interference between the incident plane waves and the reflected waves from the
NPs.
39
4.2. POWER DEPENDENT PHOTOLUMINESCENCE AND ENHANCEMENT
From this spatial distribution of the electric field, we can estimate the emission enhancement in the far-field.
In a volume at a distance from 60 to 70 nm away from the ITO, the electric field strength of the structure
with NP arrays and conformal layer was divided by that of the reference to obtain an EL enhancement factor
of 2.1, and a PL enhancement factor of 3.08. These values will be used for comparison to the results of the
experimental PL and EL measurements that will be treated in the following sections.
4.2 Power dependent photoluminescence and enhancement

On the LED structure as described in section 3.1, power dependent PL measurements were performed in
the Fourier setup described in section 3.3. The PL spectrum was obtained both for a reference sample and
a sample with NP arrays, while the incident power of the laser (λ = 405nm) was measured using a power
meter and varied from 0.25 mW to 18 mW. For each individual spectrum, the counts under the PL peak
were integrated and the accumulated PL and the corresponding power were plotted as a single data point in
figure 4.2. For optical excitation, the incident power is proportional to the number of photons and thus to
the number of excited carriers. Consequently, as long as Auger recombination can be neglected, we assume
a linear relation between the incident power on the x-axis and the PL on the y-axis. The raw data of the
power dependent PL measurements for both the reference and the NP array sample are provided in figure
4.2.
Figure 4.2: Integrated PL intensity for the reference sample and sample with NP arrays (on log scale) as a
function of incident laser power.
Clearly, enhanced PL is observed for the sample with NP arrays when compared to the reference. Measuring
the total extraction efficiency, however, is practically impossible since collecting all the emitted light would
require an objective lens with NA=1. We introduce a collection efficiency to quantify the amount of light
as collected by the NA = 0.9 objective used in the experiment. The overall enhancement then should be
distinguished into two components: Firstly, there is an expected increase in outcoupling due to the presence
of the NP arrays, which will be referred to as the collection enhancement. Additionally, part of the overall
enhancement can be assigned to the enhancement of the pump, as discussed in section 3.5.
The PL was measured for different powers in order to quantify the contributions of the collection and pump
enhancement, following the recipe of Abdelkhalik.[48] To better understand this, we analyze the power
dependent carrier dynamics of the InGaN/GaN MQWs and the influence of the NP arrays. In the stationary
state, carrier generation and recombination balance each other upon optical excitation, according to
aα(N )W = AN + BN 2 + CN 3 , (4.1)
40
4.2. POWER DEPENDENT PHOTOLUMINESCENCE AND ENHANCEMENT
where A, B and C represent the coefficients of Shockley-Read-Hall recombination, radiative recombination

and Auger recombination, respectively. Furthermore, α(N ) is the single-photon absorption coefficient of
the laser beam by the InGaN/GaN MQWs, W is the power of the excitation laser beam and N is the photo-
excited carrier density. The introduction of the additional coefficient a is required to correct for the influence
of the photonic environment on the absorption of the laser, that is the pump enhancement due to scattering
from the NP arrays.
The number of PL counts measured in our experiment is proportional to the radiative recombination as
described by
P L(aW ) ∝ BN (aW )b, (4.2)
where b is the collection efficiency of the objective. We assume the value of b to be 1 by definition in the case
of the reference sample, which makes the value of the collection efficiency b in the case of the NP arrays
automatically equal to the collection enhancement. Based on how we expect the reference PL to change
with changing power compared to how we expect the NP enhanced PL to change with changing power we
can shift the reference PL vs power curve to fit the measurement of the NP PL. Generally speaking, NP
arrays may also influence the value of the B coefficient related to radiative recombination by means of the
Purcell effect.[49] However, we will neglect the influence of the Purcell effect on the value of B based on
the assumption that lack of near-field coupling of the MQWs to the NP arrays will cause its effect to be
insignificant with respect to pump and collection enhancement.[50] [51] Then, based on equation (4.2), the
x-axis as depicted in figure 4.2 is scaled linearly by the pump enhancement a. The collection enhancement
b, on its turn, linearly scales the measured PL counts under the logarithm in the y-axis of the same figure.
Knowing this, we could scale the reference measurement to fit the NP array measurement, and retrieve
the values of a and b. A MATLAB script was written for this purpose and can be found in section B.1 of
appendix B. After making a linear fit to our measured reference data, we perform these scaling operations
on the reference fit by sweeping over a range of values for a and b, with step size of 0.01. For every unique
combination (a, b), the mean square error (MSE) of the resulting curve is calculated with respect to a linear
fit to the NP array measurement. The result of this MSE calculation is shown by a color scale for every
(a, b) in figure 4.3.
Figure 4.3: Mean square error of the fitted curve as a function of the pump and collection enhancement.
A local maximum in the logarithm of the inverse of the MSE (equivalent to a minimum in the MSE) is
observed at a pump enhancement of 1.48 and a collection enhancement of 3.38. Based on this we conclude
that scaling the x-axis by 1.48 ± 0.005 and the logarithm of the y-axis by 3.38 ± 0.005 would yield the best
fit of the shifted reference curve to the NP array measurement. The quality of the fit is verified in figure 4.4.
41
4.3. DIODE BURNING
Figure 4.4: Shifted reference curve compared to non-shifted nanoparticle curve for optimal values of a and
b.
Additionally, we conclude that the collection enhancement has a value of 3.38. Since pump enhancement
is only present in optical measurements, this enhancement value of 3.38 will be used for comparison to EL
measurements on the same NP array with the same objective. Furthermore, this value compares well to the
value of 3.08 from our simulation of the PL enhancement from section 4.1.
4.3 Diode burning

As described in section 3.4, high-voltage pulses can create a permanent current path through the p-GaN
of the LED. Initially, the current induced by the high-voltage pulse is still relatively low due to the high
resistance of the reverse diode. As soon as this diode breaks down to a short circuit though, sending another
high-voltage pulse induces a very high current through the forward diode, potentially breaking down this
diode as well. It is therefore desirable to monitor the current through the device at all times. The best
way to monitor the current through a system is by placing a resistor with known resistance in series, and by
measuring the potential difference across this resistor. In our setup, an oscilloscope was used for monitoring
both the voltage of the source as well as over the resistor at the same time. It should however be noted that
introducing this additional resistor in series increases the overall resistance of the circuit, hence requiring
higher voltage pulses to induce the same amount of current. Alternatively, the resistor can be omitted as long
as the light emission from the contacts is monitored carefully. As soon as a bright flash is observed upon
sending a pulse through the device, the high-voltage pulsing should immediately stop. This flash serves
as an indicator for the current having increased, that is the resistance having decreased, thus the burning
being successful. Similarly, as soon as a sudden increase in the measured current is observed in presence
of a resistor, indicating successful burning, the pulsing should be stopped. Afterwards, blue light emission
can be observed after switching from the high-voltage pulse generator to a continuous low voltage supply.
Light was indeed observed from the LED at voltages as low as 2.5 V. An image of the emitting micro-LED
(without being etched) after contacting a continuous voltage supply using microprobes is provided in figure
4.5.
42
4.4. POWER DEPENDENT ELECTROLUMINESCENCE AND ENHANCEMENT
Figure 4.5: Light emission from the micro-LED at low continuous voltage after diode burning was per-
formed successfully.
Although the diode burning turned out to be inconsistent at times, an attempt was made to characterize the
process in terms of used amplitude and pulse width. It should be noted that diode burning is a permanent
effect that cannot be repeated for characterization purposes at the same position of the sample. Nevertheless,
in case of no additional resistor, burning was visually observed at voltages in the range of 25 V, and pulse
widths in the range of 1−10 ms. The frequency of pulses used was 0.4 Hz, such that there was sufficient time
to manually stop the pulsing in case successful burning was observed. Usually, an amount of approximately
10 pulses was required before burning was observed.
4.4 Power dependent electroluminescence and enhancement

In comparison to the power dependent measurements of the PL in section 4.2, the EL was measured on
the very same reference LED and LED with NP arrays, with the same objective lens. Contrary to the PL
measurement, however, in the EL measurement it is not the applied power but the current that linearly scales
the number of excited carriers. The measured spectra at different current densities for both the reference
and the NP array sample can be found in figure A.2 of appendix A. After taking the FWHM of all spectra,
the result of the raw data of the EL measurement versus current density was plotted and can be found in
figure 4.6.
43
Figure 4.6: Reference EL and nanoparticle EL (on log scale) as a function of current density.
The measurement was performed by controlling the voltage over the junction, and therefore the x-coordinates
of the two measurements in figure 4.6, which represent the current density, do not correspond exactly. The
same EL measurement plotted as a function of voltage can be found in figure A.3 of appendix A. At first
sight, the observed trend in figure 4.6 seems to be comparable to that of the optical measurements in figure
4.2 However, like mentioned before, pump enhancement does not play a role upon electrical excitation.
Therefore, we compare the total enhancement of the EL measurement to the collection enhancement of the
PL measurement. Preferably, we would calculate the enhancement at the current density where the perfor-
mance of the LED with NP arrays is best. In order to obtain a qualitative measure of the EQE, we normalize
the output EL counts to the input current density. At the current density where the output over input value
reaches its maximum, the EQE is at its highest. Then by also normalizing the y-axis to have a maximum
value of 1, we obtain the relative EQE as shown in figure 4.7.
Figure 4.7: The relative EQE as a function of the current density for both the reference LED and the LED
with nanoparticle arrays.
Within the measured current density range, the relative EQE appears to increase to its maximum relatively
fast to then approach a constant value. From literature, however, we expect a efficiency droop to occur at
44
high current density. When we compare our figure to similar ones in literature, though, the main difference
lies in the measured range of current densities. Generally, when the efficiency droop is evaluated, the
EQE is plotted for current densities up to at least 100 A/cm2 or even 1000 A/cm2 . [36] [37] Then, the
position of the peak can be observed more clearly when the data is plotted with the current density on a
logarithmic scale. It is therefore plausible that a decrease in EQE would be observed upon increasing the
current density further. From literature, we also conclude that for micro-LEDs the maximum EQE is often
observed for current densities between 0 and 20 A/cm2 .[36] Therefore, for now we carefully conclude that
the maximum EQE lies somewhere between 2.0 A/cm2 (relative EQE = 97.3%) and 8.2 A/cm2 (relative
EQE = 100%), based on the data points of figure 4.7. However, we also conclude that the EQE is reasonably
constant within the investigated current density range, and that it is therefore difficult to draw conclusions
about the exact optimal current density value. Nevertheless, it appears that the optimal EQE in the case of
the LED with NP arrays is reached at a lower current density when compared to the reference LED. A peak
in the EQE at lower current density corresponds to a lower operating voltage, which would be beneficial for
applications. This argument is supported by figure A.3 in appendix A.
Another crucial difference with our measurement when compared to literature is that of the investigated
voltage range.[37] [36] Where comparable experiments require 5 Volts to reach 100s to 1000s of A/cm2 ,
we required 12.5 Volts to even reach 18 A/cm2 . For a smaller voltage range starting below the emission
threshold, the current density is measured again as a function of voltage to characterize the (diode) relation,
of which the result is plotted in figure 4.8.
Figure 4.8: The measured current density as a function of the input voltage over the junction for the LED
with nanoparticle arrays.
Although the observed current density is much lower at given voltages than expected from literature, the re-
lationship between the current density and input voltage does appear to adhere the Shockley diode equation.
From this we conclude that the expected (single) diode characteristic is present, however, the resistance in
the operating voltage regime in absolute terms is higher than expected after burning.
In order to calculate the EL enhancement, we assume a linear relation between the current density and the
EL (figure 4.6) in the range of 2.0 to 8.2 A/cm2 . Then, by evaluating the ratio of the slopes, we draw a
conclusion about the overall enhancement. The resulting plot is given in figure 4.9.
45
4.5. BACK-FOCAL PLANE IMAGING
Figure 4.9: A linear fit to the EL versus current density measurement, within the range of optimal EQE.
From the ratio of the slopes we conclude that there is a factor 2.42 ± 0.08 enhancement in the EL measure-
ment when comparing the LED with NP arrays to the reference sample. This does not completely account
for the collection enhancement factor of 3.38 obtained from the PL measurements. This, however, was
expected from the difference in peak wavelength from our simulations. The enhancement at 452 nm was
calculated to be 2.1 from our simulations, which compares well to the factor 2.42 calculated experimentally.
4.5 Back-focal plane imaging

Due to the error in the fabrication, the period of the actual lattice did not meet the criterion for SLRs to occur.
For this reason, we do not expect to see any reshaping in the BFP images. This is confirmed by the BFP
images that were taken in the newly built electroluminescence setup, which can be found in figure A.4 of
appendix A. From this, we conclude that SLRs are not responsible for the observed emission enhancement.
Rather, the emission enhancement is attributed to the scattering from the individual NPs combined with the
conformal layer of TiO2 . This conclusion is in line with the observed enhancement in our simulations of
section 4.1.
For completeness sake, the BFP images of an older sample, where the NP arrays were fabricated correctly
as intended, are shown in figure 4.10. Although the structure of the LED was slightly different, it can be
assumed the obtained BFP images are similar to that of the expected BFP images of the newer sample if the
fabrication would be performed correctly.
46
4.6. PHOTOLUMINESCENCE
(a) (b) (c)
(d) (e)
Figure 4.10: Back-focal plane images for a reference sample (a) and periods of 220 nm (b), 230 nm (c),
240 nm (d) and 250 nm (e).
In these images, reshaping of the far-field emission is observed for the given periods. In the future, these
images are intended to be reproduced in the EL setup with the new LED structure and correct fabrication.
Additionally, the BFP images were simulated for how the sample was intended to be fabricated. The results
of this simulation can be found in figure A.1 of appendix A.
4.6 Photoluminescence
The PL of a WS2 monolayer on PDMS on glass was investigated using blue light excitation (λ = 420 nm)
as described in section 3.7.1. The resulting spectrum can be found in figure 4.11.
Figure 4.11: PL of a W S2 monolayer on PDMS on glass and its peak wavelength at 617 nm.
47
A sharp (FWHM = 9nm) PL peak is observed at a peak wavelength of 617 nm, corresponding to the
recombination of the A-exciton of WS2 .
Next, we wanted to investigate the influence of stamping the monolayer on the LED wafer on the PL
spectrum. A comparison of the PL spectrum before and after stamping is provided in image 4.12.
Figure 4.12: PL of a W S2 monolayer on PDMS on glass compared to the PL of the same monolayer after
stamping it onto an LED wafer.
After stamping the monolayer onto the LED substrate, we observe approximately a factor 3 decrease in the
PL intensity in figure 4.12. Moreover, the FWHM of the peak increased from 9 nm to 14 nm, and a red-shift
of the peak wavelength of 0.9 nm is observed. All of these observations are expected and can be explained
by the fact that the introduction of defects upon stamping from a soft substrate to a hard substrate reduces
the quantum yield.[52] Furthermore, the different dielectric environment of the LED wafer when compared
to the PDMS could also play a role. A last difference can be observed around 695 nm, where a small
additional peak seems to emerge. The origin of this peak remains unknown, however, this peak was never
observed upon stamping onto glass substrates. Upon optically exciting the LED without WS2 monolayer
with a laser (section 4.2) the peak is also observed. On the other hand, upon electrically exciting the LED
without WS2 monolayer (section 4.4) the peak is not present. From these observations we conclude that the
peak must originate from some optical interaction between the excitation light source and the LED wafer.
After the observed decrease in PL upon stamping, the question was raised whether there are ways to recover
some part of this loss. Therefore, the sample was annealed for 10 minutes at 140 °, and the resulting spectra
of before and after annealing are provided in figure 4.13.
48
Figure 4.13: PL of a W S2 monolayer stamped onto the LED wafer before and after annealing for 10
minutes at 140 °C.
The annealing of the WS2 monolayer led to an 11% increase of the PL intensity, as shown in figure 4.13.
11%. The FWHM did not change upon annealing. However, the peak wavelength underwent a slight blue-
shift (0.3 nm), partially reversing the observed red-shift after transfer in figure 4.12. This can be explained
by the annealing releasing some of the strain at the interface of the monolayer and the substrate.[53] [54] We
conclude that although we can increase the PL reasonably upon annealing, we still lose a substantial amount
of the PL intensity upon transferring the WS2 monolayer from PDMS to the LED wafer by introducing
defects.
Next, we investigate whether we can enhance the PL emission intensity by creating a bilayer structure as
described in section 3.6.1. For the Al2 O3 sample, the same flake as for the previous measurements was
used as the bottom layer. For the hBN sample, new monolayers for both the bottom and the top flake had
to be exfoliated. The PL spectra for both the samples with Al2 O3 and hBN spacers, for the three different
areas as defined in figure 3.11, are provided in figure 4.14.
(a) (b)
Figure 4.14: PL spectrum of the Al2 O3 sample (a) and hBN (b) for the three areas under investigation.
Although the PL is clearly highest in the bilayer region in the case of Al2 O3 , when we compare the absolute
number of counts per second to that of the monolayer directly after stamping, the value decreased. It can be
seen in figure A.5 of appendix A that the PL of the bottom layer had already decreased after the ALD was
performed, before stamping the top layer. Furthermore, we know from figure 4.12, that the introduction of
49
4.7. ABSORPTION
defects upon stamping leads to a decrease in PL counts. Strangely enough, it appears that in the hBN sample,
the bilayer shows less PL than the top monolayer. However, when we integrate the total number of counts
under both curves, the bilayer slightly exceeds the top monolayer, due to a broader FWHM. In this case, it
seems that stamping additional layers introduces additional defects, decreasing the PL contribution of the
bottom monolayer, broadening the overall spectrum, and rendering the top monolayer the dominant emitter.
Nonetheless, for the hBN sample, the bilayer does appear to lead to increased PL when compared to the
monolayer before stamping. This leads us to conclude that for achieving enhanced PL in artificial bilayers,
hBN is a more suitable candidate than Al2 O3 . This may have to do with more pronounced dielectric
screening in Al2 O3 when compared to hBN. The PL of the artificial bilayer in the hBN sample is dominated
by the PL of the top flake, though.
4.7 Absorption
The absorption of a WS2 monolayer on PDMS on glass was investigated as described in section 3.7.2. The
resulting absorption spectrum is provided in figure 4.15.
Figure 4.15: Absorption spectrum of a W S2 monolayer on PDMS on glass, with the peak of the C-exciton
located at 435 nm and a corresponding absorption of 16.7%
The highest absorption of 17.9% is observed at 615 nm, corresponding to the absorption of the A-exciton.
We are, however, more interested in the absorption of the C-exciton, since its peak is closer to the emission
spectrum of the InGaN/GaN LED. Here an absorption of 16.7% is reached at a wavelength of 435 nm.
However, a steep drop in absorption can be observed towards higher wavelengths. The slight mismatch
between the peak wavelength of the LED upon electrical excitation (452 nm) and the peak wavelength of
the absorption of the C-exciton (435 nm) may lead to a considerable performance issue. Namely, at 452
nm the absorption has dropped to 10.3%. The EL emission intensity at 435 nm drops down to 13% of its
maximum value. The slight misalignment of the two peaks is visualized in figure 4.16.
50
4.7. ABSORPTION
Figure 4.16: Comparison of the absorption spectrum of a W S2 monolayer on PDMS on glass (left-hand
ordinate) and the emission spectrum of the LED upon electrical excitation (right-hand ordinate), displaying
a slight mismatch in peak wavelengths.
Next, we analyze the absorption of the two bilayer samples, with hBN and Al2 O3 as a spacer. A comparison
of the absorption spectra of again the same three areas (defined in figure 3.11) for both samples is provided
in figure 4.17.
(a) (b)
Figure 4.17: Absorption spectrum for the three different areas on the Al2 O3 sample (a) and the hBN sample
(b).
A first obvious observation is the wavy behavior of the absorption curves, which was not observed on PDMS
on glass before. This behavior is due to the layered structure of the LED, causing Fabry-Pérot resonances
between the interfaces. Furthermore, the curves of area 3 of both samples were shifted down to match
the zero-crossing of the other areas. The original raw data and justification can be found in section A.5 of
appendix A. The maximum absorption of the C-exciton of WS2 is 18.4% at 439 nm for the hBN sample and
14.8% at 436 nm for the Al2 O3 sample. We conclude that the absorption of the bilayer is slightly enhanced
for the hBN sample, though slightly decreased for the Al2 O3 sample. In the case of the hBN sample, the
absorption of the single layer (bottom) approaches the value of the monolayer on PDMS on glass, at least in
the wavelength regime of interest. The absorption of the A-exciton, however, is significantly reduced with
respect to the monolayer on PDMS on glass. The absorption of only the top layer is less than that of only
the bottom layer. This can be explained by the fact that two additional stamping steps have been performed
(hBN spacer + top monolayer), with additional defects introduced every step. Apparently, the amount of
51
4.8. EXTERNAL QUANTUM EFFICIENCY
defects introduced is highest at the position where most layers overlap, since the bilayer absorption is less
than the sum of the individual absorptions. This holds for both samples. The Al2 O3 sample shows two
main differences compared to the hBN sample, though. The individual layers are decreased more when
compared to the monolayer on PDMS on glass, indicating the Al2 O3 worsens the performance of the single
layers more than the hBN does. However, because the Al2 O3 was applied using ALD, less defects are
introduced upon stacking, and the enhancement in the bilayer area is more pronounced. Overall, absorption
enhancement was achieved and we conclude that hBN has more potential for absorption enhancement, since
it hardly effects the absorption of the single layers, as long as we can limit the amount of defects introduced
in the stamping process.
4.8 External quantum efficiency

Since the EQE of a sample may depend on the excitation power, the integrating sphere measurement ought
to be performed under the same conditions as the illumination of our WS2 flakes. Initially, a microstructured
LED wafer was used as a reference and measured with a 405 nm excitation laser in the integrating sphere.
The EQE was measured as a function of power, and the resulting plot can be found in A.7 of appendix A.
The EQE indeed turned out to be power dependent. There were two problems with this reference, though.
For starters, the reference could not be optically excited with the 450 nm laser, which was the wavelength
at which the PL of the monolayers was going to be measured. Additionally, the required power to excite
the LED was too much to excite the monolayers with. Since the WS2 may bleach upon high excitation
powers, we limited the power of the laser to 100 µW. A Lumigen F Green dye was now chosen, because
of its absorption peak around 450 nm, in line with the emission wavelength of the LED. To get sufficient
PL at 100 µW incident power, the dye was spincoated onto a microstructured glass substrate (1 minute at
1000 RPM), for more efficient outcoupling. The result of the integrating sphere measurement is provided
in figure 4.18.
Figure 4.18
Although a slight decrease in the laser peak from measurement a to b to c, indicating absorption, can be
observed, no clear PL was observed from the dye in the integrating sphere measurement. This can partially
be explained by the fact that the core of the used fiber is rather small. One could argue to increase the power
of the laser in order to obtain more PL, however, at the same time we must ensure the laser does not saturate
the spectrometer. This is crucial, because the area under the laser curve is also required for the calculation
of the EQE. In addition, using higher power may damage the sample. This makes it impossible to calculate
the EQE of the dye using equation (3.5). At the moment of writing this thesis, reference samples that are
measurable in the integrating sphere that also absorb at 450 nm low (100 µW) incident power are being
investigated. In conclusion, no suitable reference sample was fabricated yet and the EQE from here on will
be assumed to be equal to ηdye .
52
4.9. LIFETIME MEASUREMENTS
Then, the PL was measured under the aforementioned conditions. The obtained PL spectrum of the Lumi-
gen F Green sample is given in figure 4.19.
Figure 4.19: Measured PL spectrum of the Lumigen F Green dye.
We can now compare the PL spectrum of the WS2 to that of the dye to obtain a relative value for the EQE.
Since the PL spectra of figure 4.14 were obtained under the exact same conditions, we can use these results
for this purpose. By means of equation (3.6), we calculate the EQE on the three different areas (as defined
in 3.11), for both samples. The results are summarized in the table below.
Table 4.1: Summary of the calculated EQEs for the three different areas of investigation for both the hBN
and the Al2 O3 sample.
Area 1 Area 2 Area 3

Al2 O3 η = 6.3ηdye η = 5.2ηdye η = 1.5ηdye
hBN η = 29.4ηdye η = 28.5ηdye η = 8.7ηdye
The fact that the efficiency of the bilayer is highest, arises from the notion that in total more of the incident
power is converted to light. The PL intensity of the bilayer, however, has not doubled with respect to that
of the monolayer, such that the EQE of the individual layers in the multilayer structure is less than that of
the monolayers.
4.9 Lifetime measurements

A micro-PL image was obtained for both samples. Both the number of collected photons (events) and the
lifetime were measured as a function of position, and the resulting color maps can be found in figure 4.20.
53
4.9. LIFETIME MEASUREMENTS
(a) (b)
(c) (d)
Figure 4.20: Micro-PL image of the events of the final Al2 O3 device (a) and of the final hBN device (b),
and of the lifetime of the final Al2 O3 device (c) and the final hBN device (d).
In the left top image, clearly that the most counts are detected from the bilayer overlap area, whereas in the
right top image the total number of counts appears to be dominated by the top layer. This observation is in
line with the conclusions from our PL measurements of figure 4.14.
From the left bottom image, we observe that the lifetime in the overlap region is predominately green, which
is relatively fast based on the color scheme. Both the top and bottom monolayer seem to display a mixture
of green and red, representing a combination of relatively fast and slow decay.
In the right bottom image, we observe that lifetime of the entire top monolayer is predominately green.
This is again expected from the fact that the PL is dominated by the top monolayer and equation (2.15).
The bottom monolayer in this sample, shows a similar mixture of fast and slow (green and red) decay,
when compared to the monolayers of the Al2 O3 sample. The red areas in the right bottom correspond to
areas where no emission is detected in the top image, because the bottom monolayer is covered with bulk
material here. From experience we learned that mechanical exfoliation often yields monolayers that are still
connected to multilayered or bulk flakes, rather than isolated flakes.
Furthermore, the image shows even shorter lifetimes (blue areas) at the edges of the overlap between the top
and bottom monolayer. This observation seems contradictory to the fact that the same edge in the top image
shows less events. The mean lifetime, however, is a combination of the radiative and non-radiative lifetime,
and thus we conclude this must be the consequence of a shorter non-radiative lifetime. This conclusion in
accordance with the edges serving as effective non-radiative recombination sites.
Furthermore, a micro-PL image of the bottom monolayer covered in Al2 O3 , before putting the top mono-
layer, is provided in figure A.8 of appendix A. This image shows that some of the cracks that can be seen
in the bilayer were already there in the monolayer, meaning these defects do not originate from stamping.
54
4.10. COLOR GAMUT
Next, for each of the areas of interest (as defined in section 3.6.1), an exponential fit was made to the raw
data to obtain the values of the lifetime. The plots for the bilayer areas of both samples are provided in
figure 4.21. The other plots and residuals can be found in figure A.9 and A.10 of appendix A.
(a) (b)
Figure 4.21: Biexponential fit to the intensity as a function of time for area 1 of the final Al2 O3 device (a)
and the final hBN device (b).
From the parameters of these fits we obtain the values of the lifetime(s) and the corresponding amplitudes.
For all areas, a biexponential fit was used, with exception of area 2 of the hBN device (single exponential),
for optimal fit. The resulting parameters are summarized in the tables below.
Table 4.2: Summary of the lifetimes and respective amplitudes for the three different areas of investigation
of the Al2 O3 device.

A1 0.4806 ± 0.0027 0.1734 ± 0.0093 0.1537 ± 0.0052
τ1 0.4852 ± 0.0049 ns 0.389 ± 0.021 ns 0.44 ± 0.02 ns
A2 0.0249 ± 0.0025 0.0475 ± 0.0093 0.0161 ± 0.0039
τ2 2.1 ± 0.4 ns 1.040 ± 0.089 ns 1.60 ± 0.18 ns
Table 4.3: Summary of the lifetimes and respective amplitudes for the three different areas of investigation
of the hBN device.

A1 0.674 ± 0.012 0.8806 ± 0.0018 0.1218 ± 0.0043
τ1 0.960 ± 0.013 ns 1.2230 ± 0.0043 ns 0.3095 ± 0.0072 ns
A2 0.148 ± 0.012 N.A. 0.1041 ± 0.0052
τ2 2.970 ± 0.083 ns N.A. 1.110 ± 0.055 ns
The observed values of the lifetime can be related to the radiative and non-radiative lifetime according to
equation (2.14). The observed lifetime alone, provides insufficient information to draw conclusions about
the radiative and non-radiative lifetimes individually. The radiative and non-radiative lifetimes, however,
are also related by the IQE, according to equation (2.15). If we then approximate the IQE to be equal to
the EQE for the monolayer flakes, we can draw conclusions about the radiative and non-radiative lifetimes
individually, based on our obtained values for the EQE from section 4.8.
Without definitive value for the EQE, however, it is difficult to distinguish radiative and non-radiative life-
time from the observed lifetime. Drawing conclusions at this point would therefore be premature, and the
future steps will be discussed in the outlook.
55
4.10. COLOR GAMUT
4.10 Color gamut

Based on the theory discussed in section 2.1.4, the chromaticity coordinates for the EL emission spectrum
(figure 3.1), WS2 monolayer (figure 4.11) and a hypothetical emission spectrum of green QDs (figure
A.11 of appendix A) were calculated. A MATLAB script was written for this purpose, that calculates the
chromaticity coordinates of a spectrum based on the input of a text file spectrum. The script can be found
in section B.2 of appendix B. For the emission spectrum of the QDs a Gaussian distribution around 530
nm with a FWHM of 20 nm was chosen based on the specifications of high-end commercial QDs.[55]
The calculated coordinates were plotted in the chromaticity diagram and the resulting color gamut of the
proposed final device is provided in figure 4.22.
Figure 4.22: The color gamut for the proposed final device, with the indicated chromaticity coordinates for
the EL emission, W S2 monolayer, and hypothetical spectrum of green QDs.
56
5 Conclusion and outlook
In conclusion, the emergence of new technologies such as smart wearable devices, augmented reality (AR),
and virtual reality (VR) displays, coincides with an increasing demand from industry for high-resolution
displays that are small in size. Micro-LEDs have emerged as a promising technology that could overcome
the limitations of directly emissive LED displays. The efficiency of these micro-LEDs, however, is limited
by the LEE, due to the large difference in refractive index at the interface of the semiconductor and air.
In large area devices, a profiled sapphire substrate is commonly used for increasing the outcoupling. This
method, however, can not be applied into small sized micro-LEDs. Etching through the sidewalls to create
mesa sidewalls can increase the LEE, but increases dangling bonds and surface recombination.
In this thesis, the possibility of exploiting SLRs for enhancing the LEE in micro-LEDs was investigated. For
this purpose, plasmonic NPs were fabricated in periodic arrays onto a InGaN/GaN LED. By means of ALD,
a conformal layer of TiO2 was deposited onto the NPs to create a homogeneous dielectric environment.
Due to a design error during fabrication, however, the fabricated period was smaller than the intended
period, which affects the extent to which the fabricated NP arrays support SLRs.
The fabricated sample was modelled into COMSOL Multiphysics software and we performed a finite ele-
ment method analysis using to calculate the collection and pump enhancement. The simulation predicted
a factor 2.1 enhancement for EL and factor 3.08 enhancement for PL measurements when compared to a
reference sample without NP arrays and conformal layer. Experimentally, the enhancements for EL and
PL were measured to be 2.42 ± 0.08 and 3.38 ± 0.005, respectively, which is in good agreement with the
simulations.
BFP images showed no reshaping of the far-field emission though, leading to the conclusion that the NP
arrays did not support SLRs. Nevertheless, emission enhancement was observed, both in the simulations as
well as in the experiments. The observed emission enhancement is thus attributed to the scattering of the
NPs in combination with the presence of the conformal layer.
To perform the EL measurements, etching through the p-GaN to make the negative contact was avoided
by applying a novel alternative in the form of diode burning. Several 25 V pulses with pulse width of 10
ms were applied to the unetched device to burn a permanent current path through the p-GaN. Successful
burning is concluded by observing the light intensity emitted upon electrical pulsing. For future reference,
the burning process should be more carefully characterized by using a resistor to monitor the current flow.
Furthermore, the possibility of integrating 2D materials into micro-LEDs for white light generation was
investigated. Monolayers of WS2 were fabricated by means of mechanical exfoliation. This TMD displays
a direct band gap in the red part of the visible spectrum only in its monolayer form. Together with an
absorption peak in the blue part of the visible spectrum, this material is a potential candidate for future
down-converting micro-LEDs. To increase the absorption and emission intensity, mulilayers of 2D mono-
layers need to be made. Two different types of multilayers were fabricated, with hBN and Al2 O3 as a
spacer material between the WS2 monolayers. It was concluded that the PL and absorption of the formed
bilayer can be enhanced for the case of hBN. For the Al2 O3 sample, however, both the PL and absorption
were quenched in the bilayer with respect to the monolayer due to more pronounced dielectric screening
and the additional defects introduced upon sample fabrication. A layer of hBN is also stamped on top of
the multilayer structure for encapsulation, to allow for fabrication of green QDs on top for efficient white
LED emission from micro-LEDs in the future. To limit the number of defects introduced upon stamping the
layers, better control over the position and angle of the sample is required. This will require an improved
setup, which is already a work in progress at the time of writing this thesis.
57
An effort was made to analyze the EQE and lifetime of the fabricated multilayer structures. Because of
difficulties with the reference sample measurement in the integrating sphere, however, no definitive con-
clusions can be drawn at this point. At the moment of writing this thesis, a suitable reference sample is
sought after and possible issues with the integrating sphere setup are being evaluated. When an appropriate
reference sample measurement is performed, the measurements on the multilayer structures of this thesis
can still be used for the analysis of the EQE and lifetime.
Altogether, when the burning of the p-GaN of the InGaN/GaN LED can be properly controlled and the
stamping process of the WS2 monolayers is optimized, the next step of this research will be to stamp WS2
monolayers on the back of LEDs for EL measurements. This will provide us with the exact ratio of the blue
and red emission, to know where the emission spectrum is in the color gamut. This is difficult to evaluate
with optical measurements, since exciting the LED with a laser will cause additional absorption of the laser
beam by the WS2 . After that, we can encapsulate the WS2 with hBN, and deposit QDs on top for green
emission. This will be the final step towards achieving our final goal of efficient white light emission from
micro-LEDs for future applications.
58
Bibliography
[1] Yifan Wu, Jianshe Ma, Ping Su, Lijun Zhang, and Bizhong Xia. Full-color realization of micro-led
displays. Nanomaterials, 10(12), 2020.
[2] Tao Zhan, Kun Yin, Jianghao Xiong, Ziqian He, and Shin-Tson Wu. Augmented reality and virtual
reality displays: Perspectives and challenges. iScience, 23(8):101397, 2020.
[3] Vincent W. Lee, Nancy Twu, and Ioannis Kymissis. Micro-led technologies and applications. Infor-
mation Display, 32(6):16–23, 2016.
[4] What microled means for pixel density, Oct 2022.
[5] A.I. Zhmakin. Enhancement of light extraction from light emitting diodes. Physics Reports,
498(4):189–241, 2011.
[6] Suihu Dang, Chunxia Li, Tianbao Li, Wei Jia, Peide Han, and Bingshe Xu. Light-extraction en-
hancement and directional emission control of gan-based led with transmission grating. Optik,
125(14):3623–3627, 2014.
[7] Hee Kwan Lee, Yeong Hwan Ko, Ganji Seeta Rama Raju, and Jae Su Yu. Light-extraction enhance-
ment and directional emission control of gan-based leds by self-assembled monolayer of silica spheres.
Opt. Express, 20(22):25058–25063, Oct 2012.
[8] Kirill A. Bulashevich, Sergey S. Konoplev, and Sergey Yu. Karpov. Effect of die shape and size on
performance of iii-nitride micro-leds: A modeling study. Photonics, 5(4), 2018.
[9] H. W. Choi, C. W. Jeon, M. D. Dawson, P. R. Edwards, R. W. Martin, and S. Tripathy. Mechanism
of enhanced light output efficiency in ingan-based microlight emitting diodes. Journal of Applied
Physics, 93(10):5978–5982, 2003.
[10] Francois Olivier, Anis Daami, Christophe Licitra, and Francois Templier. Shockley-read-hall and
auger non-radiative recombination in gan based leds: A size effect study. Applied Physics Letters,
111(2):022104, 2017.
[11] E. Fred Schubert. Light-Emitting Diodes. Cambridge University Press, 2 edition, 2006.
[12] Prasad Iyer, Ryan DeCrescent, Yahya Mohtashami, Guillaume Lheureux, Nikita Butakov, Abdullah
Alhassan, C. Weisbuch, Shuji Nakamura, Steven Denbaars, and Jon Schuller. Unidirectional lumines-
cence from ingan/gan quantum-well metasurfaces. Nature Photonics, 14:1–6, 09 2020.
[13] K. McGroddy, A. David, E. Matioli, M. Iza, S. Nakamura, S. DenBaars, J. S. Speck, C. Weisbuch,
and E. L. Hu. Directional emission control and increased light extraction in gan photonic crystal light
emitting diodes. Applied Physics Letters, 93(10):103502, 2008.
[14] Khanh T.Q., Peter Bodrogi, Vinh Trinh, and Winkler H. LED Lighting: Technology and Perception.
12 2014.
[15] Jun Tatebayashi, Tomoya Yamada, Tomohiro Inaba, Shuhei Ichikawa, and Yasufumi Fujiwara. En-
hanced luminescence efficiency of gan:eu-based light-emitting diodes by localized surface plasmons
utilizing gold nanoparticles. Japanese Journal of Applied Physics, 58(SC):SCCC09, may 2019.
[16] Gabriel Lozano, Davy Louwers, Said R. K. Rodriguez, Shunsuke Murai, Olaf Janssen, Marc Ver-
schuuren, and Jaime Rivas. Plasmonics for solid-state lighting: Enhanced excitation and directional
emission of highly efficient light sources. Light: Science Applications, 05 2013.
59
BIBLIOGRAPHY
[17] V. G. Kravets, A. V. Kabashin, W. L. Barnes, and A. N. Grigorenko. Plasmonic surface lattice res-
onances: A review of properties and applications. Chemical Reviews, 118(12):5912–5951, 2018.
PMID: 29863344.
[18] Xuefeng Gu, Teng Qiu, Wenjun Zhang, and Paul K Chu. Light-emitting diodes enhanced by localized
surface plasmon resonance. Nanoscale research letters, 6:1–12, 2011.
[19] G. Lozano, G. Grzela, M. A. Verschuuren, M. Ramezani, and J. Gómez Rivas. Tailor-made directional
emission in nanoimprinted plasmonic-based light-emitting devices. Nanoscale, 6:9223–9229, 2014.
[20] Sherry Bhalla, David T. Melnekoff, Adolfo Aleman, Violetta Leshchenko, Paula Restrepo, Jonathan
Keats, Kenan Onel, Jeffrey R. Sawyer, Deepu Madduri, Joshua Richter, Shambavi Richard, Ajai Chari,
Hearn Jay Cho, Joel T. Dudley, Sundar Jagannath, Alessandro Laganà, and Samir Parekh. Patient
similarity network of newly diagnosed multiple myeloma identifies patient subgroups with distinct
genetic features and clinical implications. Science Advances, 7(47):eabg9551, 2021.
[21] Research challenges to ultra-efficient inorganic solid-state lighting. Laser Photonics Reviews,
1(4):307–333, 2007.
[22] Peter Schlotter, Rolf Schmidt, and Juergen Schneider. Luminescence conversion of blue light emitting
diodes. Applied Physics A, 64:417–418, 1997.
[23] Gabriel Lozano, Said R. K. Rodriguez, Marc Verschuuren, and Jaime Rivas. Metallic nanostructures
for efficient led lighting. Light: Science Applications, 5:e16080, 01 2016.
[24] Yuxuan Lin, Xi Ling, Lili yu, Shengxi Huang, Allen Hsu, Yi-Hsien Lee, Jing Kong, Mildred Dres-
selhaus, and Tomas Palacios. Dielectric screening of excitons and trions in single-layer mos2. Nano
letters, 14, 09 2014.
[25] Fengnian Xia, Han Wang, Di Xiao, Madan Dubey, and Ashwin Ramasubramaniam. Two-dimensional
material nanophotonics. Nature Photonics, 8(12):899–907, nov 2014.
[26] Kin Mak and Jie Shan. Photonics and optoelectronics of 2d semiconductor transition metal dichalco-
genides. Nature Photonics, 10:216–226, 03 2016.
[27] C. Cong, Jingzhi Shang, Yanlong Wang, and Ting Yu. Optical properties of 2d semiconductor ws2.
Advanced Optical Materials, 6:1700767, 11 2017.
[28] Wenshuo Xu, Daichi Kozawa, Yingqiu Zhou, Yizhi Wang, Yuewen Sheng, Tian Jiang, Michael S.
Strano, and Jamie H. Warner. Controlling photoluminescence enhancement and energy transfer in
ws2:hbn:ws2 vertical stacks by precise interlayer distances. Small, 16(3):1905985, 2020.
[29] Pawan Kumar, Jason Lynch, Baokun Song, Haonan Ling, Francisco Barrera, Kim Kisslinger, Huiqin
Zhang, Surendra Anantharaman, Jagrit Digani, Haoyue Zhu, Tanushree Holme Choudhury, Clifford
McAleese, Xiaochen Wang, Ben Conran, Oliver Whear, Michael Motala, Michael Snure, Chris Mu-
ratore, Joan Redwing, and Deep Jariwala. Light–matter coupling in large-area van der waals superlat-
tices. Nature Nanotechnology, 17, 02 2022.
[30] Myoung-Jae Lee, David H. Seo, Sung Min Kwon, Dohun Kim, Youngwook Kim, Won Seok Yun,
Jung-Hwa Cha, Hyeon-Kyo Song, Shinbuhm Lee, MinKyung Jung, Hyeon-Jun Lee, June-Seo Kim,
Jae-Sang Heo, Sunae Seo, and Sung Kyu Park. Measurement of exciton and trion energies in mul-
tistacked hbn/ws2 coupled quantum wells for resonant tunneling diodes. ACS Nano, 14(11):16114–
16121, 2020. PMID: 33140970.
[31] Matthias Florian, Malte Hartmann, Alexander Steinhoff, Julian Klein, Alexander W. Holleitner,
Jonathan J. Finley, Tim O. Wehling, Michael Kaniber, and Christopher Gies. The dielectric impact of
layer distances on exciton and trion binding energies in van der waals heterostructures. Nano Letters,
18(4):2725–2732, 2018. PMID: 29558797.
[32] Zhihui Cheng, Huairuo Zhang, Son T. Le, Hattan Abuzaid, Guoqing Li, Linyou Cao, Albert V. Davy-
dov, Aaron D. Franklin, and Curt A. Richter. Are 2d interfaces really flat? ACS Nano, 16(4):5316–
5324, 2022. PMID: 35290014.
[33] Yu-Ming Huang, Konthoujam James Singh, An-Chen Liu, Chien-Chung Lin, Zhong Chen, Kai Wang,
Yue Lin, Zhaojun Liu, Tingzhu Wu, and Hao-Chung Kuo. Advances in quantum-dot-based displays.
Nanomaterials, 10(7), 2020.
60
BIBLIOGRAPHY
[34] C. Michael Bourget. An introduction to light-emitting diodes. HortScience horts, 43(7):1944 – 1946,
2008.
[35] The PN junction Diode, chapter 2, pages 69–122. John Wiley Sons, Ltd, 2011.
[36] Matthew Wong, Shuji Nakamura, and Steven Denbaars. Review—progress in high performance iii-
nitride micro-light-emitting diodes. ECS Journal of Solid State Science and Technology, 9:015012, 01
2020.
[37] Pengfei Tian, Jonathan J. D. McKendry, Zheng Gong, Benoit Guilhabert, Ian M. Watson, Erdan Gu,
Zhizhong Chen, Guoyi Zhang, and Martin D. Dawson. Size-dependent efficiency and efficiency droop
of blue ingan micro-light emitting diodes. Applied Physics Letters, 101(23):231110, 2012.
[38] M. Amara, Fabien MANDORLO, Romain Couderc, Félix Gérenton, and Mustapha Lemiti. Temper-
ature and color management of silicon solar cells for building integrated photovoltaic. EPJ Photo-
voltaics, 9:1, 01 2018.
[39] Anton D. Utyushev, Vadim I. Zakomirnyi, and Ilia L. Rasskazov. Collective lattice resonances: Plas-
monics and beyond. Reviews in Physics, 6:100051, 2021.
[40] V. G. Kravets, A. V. Kabashin, W. L. Barnes, and A. N. Grigorenko. Plasmonic surface lattice res-
onances: A review of properties and applications. Chemical Reviews, 118(12):5912–5951, 2018.
PMID: 29863344.
[41] The rich photonic world of plasmonic nanoparticle arrays. Materials Today, 21(3):303–314, 2018.
[42] Mohammad Ramezani, Matthijs Berghuis, and Jaime Gómez Rivas. Strong light–matter
coupling and exciton-polariton condensation in lattices of plasmonic nanoparticles
invited
. J. Opt. Soc. Am. B, 36(7):E88–E103, Jul 2019.

[43] Matthias Florian, Malte Hartmann, Alexander Steinhoff, Julian Klein, Alexander W. Holleitner,
Jonathan J. Finley, Tim O. Wehling, Michael Kaniber, and Christopher Gies. The dielectric impact of
layer distances on exciton and trion binding energies in van der waals heterostructures. Nano Letters,
18(4):2725–2732, 2018. PMID: 29558797.
[44] Aleksandr Vaskin, Radoslaw Kolkowski, A. Femius Koenderink, and Isabelle Staude. Light-emitting
metasurfaces. Nanophotonics, 8(7):1151–1198, 2019.
[45] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and
A. A. Firsov. Electric field effect in atomically thin carbon films. Science, 306(5696):666–669, 2004.
[46] Enlai Gao, Shao-Zhen Lin, Zhao Qin, Markus J. Buehler, Xi-Qiao Feng, and Zhiping Xu. Mechanical
exfoliation of two-dimensional materials. Journal of the Mechanics and Physics of Solids, 115:248–
262, 2018.
[47] John C de Mello, H Felix Wittmann, and Richard H Friend. An improved experimental determination
of external photoluminescence quantum efficiency. Advanced materials, 9(3):230–232, 1997.
[48] Mohamed S Abdelkhalik, Aleksandr Vaskin, Toni López, Aimi Abass, and Jaime Gómez Rivas. Un-
veiling light collection and pump enhancement from quantum wells with plasmonic metasurfaces
using power dependent measurements. Journal of Physics: Photonics, 5(2):025001, apr 2023.
[49] E. M. Purcell. Spontaneous Emission Probabilities at Radio Frequencies. Physical Review, 69:681,
1946.
[50] K.H. Drexhage. Influence of a dielectric interface on fluorescence decay time. Journal of Lumines-
cence, 1-2:693–701, 1970.
[51] Lukas Novotny and Bert Hecht. Principles of Nano-Optics. Cambridge University Press, 2 edition,
2012.
[52] Mengfan Zhou, Wenhui Wang, Junpeng Lu, and Zhenhua Ni. How defects influence the photolumi-
nescence of TMDCs. Nano Research, 14(1):29–39, January 2021.
61
BIBLIOGRAPHY
[53] Maryam Bizhani, Ruhi Thorat, William Poston, Thushan Wickramasinghe, Shrouq H. Aleithan, and
Eric Stinaff. Thermal annealing effects on naturally contacted monolayer mos2. physica status solidi
(b), 258(9):2000426, 2021.
[54] Fang Wang, Suhao Li, Mark A Bissett, Ian A Kinloch, Zheling Li, and Robert J Young. Strain
engineering in monolayer ws2 and ws2 nanocomposites. 2D Materials, 7(4):045022, aug 2020.
[55] Avantama AG. The full width half maximum of quantum dots. url: https://avantama.com/full-width-
half-maximum-quantum-dots/, Oct 2018. (visited on 31-03-2023).
62
Appendices
63
A Appendix
A.1 COMSOL simulation

Although no directionality was observed in the fabricated sample, a simulation of the BFP of the intended
sample did dislplay reshaping of the far-field emission, as can be seen in figure A.1.
Figure A.1: The simulated back-focal plane image of the intended design with a period of 215 nm, showing
reshaping of the far-field emission.
A.2 Power dependent EL and enhancement

To calculate the number of counts at a given current density, the FWHM of every spectrum was taken. The
raw data of the measurements of the EL spectra for different current densities are provided in figure A.2.
64
APPENDIX
(a) (b)
Figure A.2: EL spectrum for a range of input current densities for (a) the reference and (b) the sample with
nanoparticles.
In the thesis, the EL intensity was plotted as a function of the current density, since this was expected to
be proportional to the number of excited carriers. The measurement, however, was performed by changing
the voltage. When plotting the EL as a function of voltage, the data points are at the same position with
respect to the x-axis, which is not the case for the current density measurement. When doing this, the graph
of figure A.3 is obtained.
Figure A.3: Reference EL and nanoparticle EL (on log scale) as a function of voltage.
It appears that the exponential behavior for the nanoparticle array LED becomes evident at lower operating
voltages when compared to the reference LED, where the relation could be reasonably approximated by a
linear function within the given range.
A.3 Back-focal plane imaging

Back-focal plane images were obtained in the newly built EL setup. The investigated samples, however, did
not show any directionality. The BFP images of both the reference sample and the sample with NP arrays
APPENDIX
are provided in figure A.4.
(a) (b)
Figure A.4: Back focal plane image of (a) the reference and (b) the LED with nanoparticles upon electro-
luminescence.
A.4 Photoluminescence
The PL reduction upon putting the Al2 O3 layer can be readily seen under the confocal microscope by
sending the light to the camera. The images of figure A.5 were obtained under the same conditions (blue
lamp excitation), showing a PL reduction after ALD.
(a) (b)
Figure A.5: Bottom monolayer before (a) and after (b) ALD.
A.5 Absorption
Comparing the absorption of three different areas with different heights to one and the same reference with
a possibly different height may be difficult. In the thesis, we chose to keep the reference the same for all
APPENDIX
three areas. If there is a height difference, this might cause troubles with the automatic correction to the
reference by the software. To obtain realistic values for the absorption peaks, we checked the overlap of
the zero-crossings at the lower boundary of the spectra. It appeared that in both cases area 3 seemed to be
shifted up. We corrected for this by introducing an offset of 4 percentage points and 2 percentage points for
the hBN sample and the Al2 O3 sample, respectively. The original measurements, without correction, are
provided in figure A.6.
(a) (b)
Figure A.6: Original absorption spectrum for the three different areas on the Al2 O3 sample (a) (Area 3:
+2 percentage points) and the hBN sample (b) (Area 3: +4 percentage points).
A.6 External quantum efficiency

For measuring the EQE, intially a different reference sample was used. The EQE of a microstructured blue
LED was measured in the integrating sphere, at different powers, with a 405 nm laser. The resulting power
versus EQE graph can be found in figure A.7.
Figure A.7: External quantum efficiency of a microstructured blue LED as a function of incident laser
power
APPENDIX
In the end, this sample turned out not to be suited for reference to the WS2 sample, because it had to be
excited with a 405 nm laser, rather than 450nm, and because at these magnitudes of power the WS2 would
bleach.
A.7 Lifetime measurements

Before putting the top monolayer, a micro-PL image of the bottom monolayer covered in Al2 O3 was taken.
The resulting measurement can be found in figure A.8.
Figure A.8: Micro-PL image of the bottom monolayer.
The image shows that cracks were already present in the sample prior to fabricating the multilayer structure.
Furthermore, the exponential fits of the remaining (non-bilayer) areas are provided in figure A.9, where the
right bottom image corresponds to a single exponential fit, while the others are double exponential.
APPENDIX
(a) (b)
(c) (d)
Figure A.9: The exponential fits for the area 2 (a) and 3 (b) for the Al2 O3 sample and area 2 (c) and 3 (d)
for thethe hBN sample.
In addition, figure A.10 provides the residuals of the exponential fits (both from the fits in the thesis itself
and the ones in the appendix).
APPENDIX
(a) (b)
(c) (d)
(e) (f)
Figure A.10: The residuals of the exponential fits to area 1 (a), 2 (b) and 3 (c) of the Al2 O3 bilayer and
area 1 (d), 2 (e) and 3 (f) of the hBN bilayer.
A.8 Color gamut

For obtaining the color gamut, a green light spectrum had to be analyzed. In the future, we intend to use
green QDs for this purpose. Because we have no access to measurement data of such QDs, a Gaussian
distribution with peak wavelength at 530 nm and a FWHM of 20 nm was assumed. This was based on the
specifications of commercially available green QDs, from Avantama. The assumed spectrum can be found
in figure A.11.
APPENDIX
Figure A.11: Hypothetical emission spectrum of green quantum dots with a peak wavelength at 530 nm and
a FWHM of 20 nm, based on Avantama specifications.[55]
B MATLAB scripts
B.1 Power dependent PL and enhancement
clear all
% Loading the folder
myFolder = 'C:\AP MSc 1\Graduation Project\Nanoparticles enhancement\
Power dependent PL\23-01-2023\With NPs';
data = [];
% The powers that I measured with the power meter during measurement
power = [0.25, 0.50, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
16, 17];
% Get a list of all files in the folder with the desired file name
pattern.
filePattern = fullfile(myFolder, '*.txt'); % Change to whatever pattern
you need.
% filePattern = fullfile(myFolder, '**/*.exe'); % To look in subfolders
too, add **/.
theFiles = dir(filePattern);
%Loop through the files to read all the data

for k = 1 : length(theFiles)
baseFileName = theFiles(k).name;
fullFileName = fullfile(theFiles(k).folder, baseFileName);
fileID = fopen(fullFileName);
%find the integration time
C = textscan(fileID,'%s');
t = str2double(C{1,1}(22,1));
% Discard the acquisition data and read the wavelength and counts
data{k}=dlmread(fullFileName,'',14,0);
counts = data{1,k}(:,2)/t;
wavelength = data{1,k}(:,1);
maxcounts1(1,k) = max(counts);
acc_counts1(1,k) = sum(counts(wavelength>420&wavelength<500));
end
%Repeat for without nanoparticles

myFolder = 'C:\AP MSc 1\Graduation Project\Nanoparticles enhancement\
Power dependent PL\23-01-2023\Without NPs';
data = [];
% Get a list of all files in the folder with the desired file name
pattern.
filePattern = fullfile(myFolder, '*.txt'); % Change to whatever pattern
you need.
% filePattern = fullfile(myFolder, '**/*.exe'); % To look in subfolders
too, add **/.
theFiles = dir(filePattern);
for k = 1 : length(theFiles)
72
MATLAB SCRIPTS
baseFileName = theFiles(k).name;
fullFileName = fullfile(theFiles(k).folder, baseFileName);
fileID = fopen(fullFileName);
C = textscan(fileID,'%s');
t = str2double(C{1,1}(22,1));
% Now do whatever you want with this file name,
data{k}=dlmread(fullFileName,'',14,0);
%time = dlmread(fullFileName,'',[6 0 6 0]);
% such as running the program with system() or whatever.
counts = data{1,k}(:,2)/t;
wavelength = data{1,k}(:,1);
maxcounts2(1,k) = max(counts);
acc_counts2(1,k) = sum(counts(wavelength>420&wavelength<500));
end
%Neglect first couple data points for fitting

power(1:3)=[];
acc_counts1(1:3)=[];
acc_counts2(1:3)=[];
%Maxima for the plotting range

amax=400;
bmax=400;
norm = 4/amax;
% Create a matrix where every line represents power times 'a' (pump
% enhancement). So a = 0.1*i = 0.1:0.1:10.
P=zeros(amax,length(power));
for ia = 1:amax
P(ia,:)=(norm*ia)^-1*power;
end
% Create a matrix where every line represents PL times 'b' (collection

% enhancement). So b = 0.1*i = 0.1:0.1:10.
PL = zeros(bmax,length(acc_counts2));
for ib = 1:bmax
PL(ib,:)=(norm*ib)^-1*acc_counts2;
end
%Empty matrix that is going to represent the mean square difference
msd=zeros(amax,bmax);
y={};
%Make fit for the NP measurement
g = polyfit(power,acc_counts1,1);
xnp=power;
ynp=g(1)*xnp+g(2);
%For all combinations (a,b), make the first order linear fit, and then
%compare y(x) to the measured NP PL, square the difference
%and average it to get msd.
for ja = 1:amax
for jb = 1:bmax
f=polyfit(P(ja,:),PL(jb,:),1);
x=power;
y{ja,jb}=f(1)*x+f(2);
msd(ja,jb)=sum((log10(ynp(y{ja,jb}>0))-log10(y{ja,jb}(y{ja,jb}>0)))
.^2)/length(y{ja,jb});
semilogy(x,y{ja,jb},'r')
hold on
MATLAB SCRIPTS
end
f=polyfit(P(ja,:),PL(jb,:),1);
x=power;
y{ja,jb}=f(1)*x+f(2);
msd(ja,jb)=sum((log10(ynp(y{ja,jb}>0))-log10(y{ja,jb}(y{ja,jb}>0)))
.^2)/length(y{ja,jb});
semilogy(x,y{ja,jb},'r')
hold on
end
hold off
%Plot 1/msd in log scale

imagesc([1,norm*amax],[1,norm*bmax],log10(1./msd(1/norm:amax,1/norm:
bmax)))
set(gca,'YDir','normal')
xlabel('Pump enhancement')
ylabel('Collection enhancement')
hcb=colorbar;
hcb.Title.String = "log(1/MSE)";
%find values of a and b

[row,col]=ind2sub([amax, bmax],find(msd==min(min(msd(1/norm:amax,1/norm
:bmax)))));
a=(norm*col);
b=(norm*row);
%%Plot the measured reference versus the shifted NP fit

semilogy(power,y{row,col},"-")
hold on
semilogy(xnp,acc_counts1,"o")
semilogy(xnp,ynp,"-")
xlim([0,18])
ylim([10^6, 10^10])
hold off
xlabel('Power (mW)')
ylabel('PL (counts/s)')
legend('Reference (shifted)','With nanoparticles')
B.2 Color gamut
clear all
%import the color matching functions
cmf = readmatrix('C:\AP MSc 1\Graduation Project\Color Gamut\
color_matching_functions.txt');
wavelength_cmf = cmf(:,1); x_cmf = cmf(:,2); y_cmf = cmf(:,3); z_cmf =
cmf(:,4);
%import the measured spectrum

measurement = readmatrix('C:\AP MSc 1\Graduation Project\Color Gamut\
PL_monolayer.txt');
wavelength = measurement(:,1); counts = measurement(:,2);
%calculate the wavelengths (and corresponding number of counts) of the

measurement
%that are closest to the wavelengths of the color matching functions
matrix = repmat(wavelength,[1,length(wavelength_cmf)]);
MATLAB SCRIPTS
for i = 1:length(wavelength_cmf)
diff(:,i) = matrix(:,i) - wavelength_cmf(i);
end
abs_diff = abs(diff);
matching_wavelength = matrix(find(abs_diff==min(abs_diff)));
%find the corresponding number of counts

for i = 1:length(wavelength_cmf)
PL(i) = counts(find(wavelength==matching_wavelength(i)));
end
%calculate the tristimulus values

X = sum(x_cmf'.*PL);
Y = sum(y_cmf'.*PL);
Z = sum(z_cmf'.*PL);
%calculate the chromaticity coordinate

x = X/(X+Y+Z);
y = Y/(X+Y+Z);

Raaij T.T.L. Van - MSC Thesis Thesis - MAP

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Raaij T.T.L. Van - MSC Thesis Thesis - MAP

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Raaij T.T.L. Van - MSC Thesis Thesis - MAP

Uploaded by

Copyright:

Available Formats

Eindhoven University of Technology

Plasmonics and 2D Materials for Micro-LED Applications

van Raaij, Thomas-Jan T.L.

Plasmonics and 2D Materials for Micro-LED

T.T.L. van Raaij

Assessment committee Graduation

ALD Atomic layer deposition

BFP Back-focal plane

EQE External quantum efficiency

FWHM Full width at half maximum

hBN Hexagonal boron nitride

IQE Internal quantum efficiency

LED Light-emitting diode

LEE Light extraction efficiency

LSP Localized surface plasmon

MQW Multiple quantum well

SLR Surface lattice resonance

TMD Transition metal dichalcogenide

vdW van der Waals

3 Methodology and setups 25

4 Results and discussion 39

5 Conclusion and outlook 57

2.1 Light-emitting diodes

Under equilibrium conditions (zero bias), the following balances apply:

Figure 2.4: Band diagram under application of a forward bias.[35]

Figure 2.5: Band diagram under application of a reverse bias.[35]

Figure 2.6: I-V characteristic of the p-n diode.[35]

RSRH = An, (2.7)

where A is the Shockley-Read-Hall recombination coefficient.

RAuger = (Cp + Cn ) n3 = Cn3 , (2.9)

R = B [n0 + ∆n(t)] [p0 + ∆p(t)] (2.10)

dn(t) ∆n0 −t/τ

where ηint is known as the internal quantum efficiency (IQE).

2.1.4 White LEDs

required to match the color of P (λ), as:

L1 = x(λ1 )P1 + y(λ1 )P1 + z(λ1 )P1

then substitution of equation (2.18) into (2.17) yields

These equations can be generalized to a mix of n colors according to

2.2 Surface lattice resonances

where rij is a vector pointing from dipole i to dipole j (i = 1, 2, . . . , N ; j = 1, 2, . . . , N ; i 6= j).

where a∗ is the effective polarizability and S is the dipole sum, given by

2.3.1 Transition metal dichalcogenides

(a) (b) (c)

2.3.3 Van der Waals heterostructures

3.1 Sample fabrication

3.1.1 InGaN/GaN LED

3.1.2 Nanoparticle fabrication

3.2 Finite element method simulations

3.3 Photoluminescence and Fourier plane imaging setup

Figure 3.5: Schematic of a Fourier microscope setup.

3.4 Diode burning

3.5 Electroluminescence and Fourier plane imaging setup

3.6 Mechanical exfoliation of WS2

3.6.1 Stacking of atomically thin layers

3.7 Confocal microscopy setup

3.7.3 Fluorescence-lifetime imaging microscopy

3.8 Integrating sphere