Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Study of quantum dots on solar
energy applications
Xiang-Jun Shang
Doctoral Thesis in Theoretical Chemistry and Biology
School of Biotechnology
Royal Institute of Technology
Stockholm, Sweden 2012
Study of quantum dots on solar energy applications
Thesis for Philosophy Doctor degree
Division of Theoretical Chemistry and Biology
School of Biotechnology
Royal Institute of Technology
Stockholm, Sweden 2012
c© Xiangjun Shang, 2012
pp i-xii, 1-60
ISBN 978-91-7501-352-7
ISSN 1654-2312
TRITA-BIO Report 2012:12
Printed by Universitetsservice US-AB
Stockholm, Sweden 2012
To my parents
Abstract
This thesis is devoted to experimental and theoretical study of p-i-n gallium
arsenide (GaAs) solar cells with self-assembled indium arsenide (InAs) quantum
dots (QDs) inserted. It gives detailed study in aspects: the QD growth and the de-
sign and fabrication of QD solar cells; the photovoltaic (PV) current-voltage (I-V)
measurement under solar simulator and the diode performance analysis; the pho-
toluminescence (PL) spectroscopy and the photocurrent (PC) spectroscopy; the
simulation of the device energy-band structure; and the discussion of the mech-
anisms of PC production from QDs, based on bias- and temperature-dependent
PC spectra.
The values of this work lie in three aspects. First, by comparing the solar
cell performance with QDs in the i-region and the n-region, we prove that for the
PC production from QDs, either the pure thermal activation or the intermediate
band (IB) cascade absorption proposed by Luque et al has lower efficiency than
the tunneling and the thermal activation-assisted tunneling, so, the IB-based QD
solar cells with QDs in the n-region make no sense to raise the PV efficiency
of a traditional single-junction solar cell to a level beyond the Shockley-Queisser
limit (31%). Second, the efficiency of the PC production from QDs (i.e. the QD
responsivity), characterized by the device PC spectra, is helpful to the design
of InAs QD-based near-infrared photodetectors. Third, the 20 nm-thick GaAs
spacers for QD layers are proper in thickness to form dislocation-free strain cou-
pling in adjacent QD layers and vertically aligned QDs, between which a strong
inter-QD-layer tunneling exists, leading to an effective PC production from QD
states, even the ground state, which has been reflected from the bias dependence
of each peak intensity in the PC spectra. A much broad-band PC production is
thus expected in closely stacked QDs for solar cell and photodetector applications.
Fourth, by analyzing the temperature-dependent PC spectra it is found that the
minority photohole thermal escape is dominant for PC production from QDs in
the n-region. The activation energy (Ea) characterizes the energy offset between a
QD hole-state and flat GaAs valence bandedge, providing a means to reconstruct
the QD energy level diagram.
Apart from the study of InAs QDs, in cooperation with co-authors, I also ex-
plore the time correlation of the light emission from two individual colloidal CdSe
QDs. The light emission from each colloidal QD is intermittent, i.e. so-called
‘QD blinking’, with time constant varying from 1 ms to 1000 s. We record the
blinking of a QD-pair sample by a 1000-frame movie in 50 ms/frame, extract the
evolution of the light intensity from each QD, and analyze the time correlation
vi
between the two QDs as their interval time τ varies. It is found that, for QD
distance of 1 µm or less, there is a bunched correlation at τ = 0, meaning that the
two QDs blink synchronously; otherwise, such correlation disappears gradually.
Although the time scale, 50 ms/frame, is too rough to conclude a strict simulta-
neous photon emission, the synchronous blinking still extends the understanding
of colloidal QD blinking that is related to the photocarrier tunneling from a QD
to the surface traps. This correlation, possibly caused by the stimulated emission
(∼ ps) between two QDs, is potential to generate coherent photon-pairs.
Preface
The work presented in the thesis was fulfilled at the Division of Theoretical Chem-
istry and Biology, School of Biotechnology, Royal Institute of Technology, Stock-
holm, Sweden and State Key Laboratory of Superlattices and Microstructures,
Institute of Semiconductors, Chinese Academy of Science, Beijing, China. The
Fourier transformed infrared spectroscopy of photocurrent was performed at Cen-
ter for Applied Mathematics and Physics, Halmstad University, Halmstad, Swe-
den. The photovoltaic efficiency measurement under a Xenon lamp solar sim-
ulator and the monochromator-based photocurrent spectroscopy in 300-1000 nm
were performed under the help from Department of Chemistry, School of Chemical
Science and Engineering, Royal Institute of Technology, Stockholm, Sweden.
List of papers included in the thesis
Paper 1. Effect of built-in electric field on photovoltaic effect of InAs quantum dots
in GaAs solar cells,
X.-J. Shang, J.-F. He, H.-L. Wang, M.-F. Li, Y. Zhu, Z.-C. Niu, and Y.
Fu,
Appl. Phys. A 2011, 103, 335.
Paper 2. Quantum-dot-induced optical transition enhancement in InAs quantum-dot-
embedded p-i-n GaAs solar cells,
X.-J. Shang, J.-F. He, M.-F. Li, F. Zhan, H.-Q. Ni, Z.-C. Niu, H. Pet-
tersson, and Y. Fu,
Appl. Phys. Lett. 2011, 99, 113514.
Paper 3. Photocarrier extraction from closely stacked InAs quantum dot layers em-
bedded in GaAs solar cells,
X.-J. Shang, J.-F. He, M.-F. Li, F. Zhan, H.-Q. Ni, Z.-C. Niu, H. Pet-
tersson, and Y. Fu,
In manuscript.
Paper 4. Observation of bunched blinkings from individual colloidal CdSe/CdS and
CdSe/ZnS quantum dots,
H.-Y. Qin, X.-J. Shang, Z.-J. Ning, T. Fu, Z.-C. Niu, H. Brismar, H.
Agren, and Y. Fu,
Submitted.
viii
Comments on my contributions to the papers included
As the first author, I have taken the major responsibilities, i.e., sample fabri-
cations, experimental measurements, and formulating Papers 1-3, corresponding
with the journals. As a co-author, I did the numerical analysis and writing of the
time correlation of QD-pair blinking, participated in the discussions and formula-
tion of paper 4.
Acknowledgments
I want to give my sincere gratitude to my supervisor Dr. Ying Fu. I am benefit a
lot from his continuous guidance on my study and assistance on my totally new life
in Sweden. His trust, encouragement and support help me to keep self-confidence
in front of each scientific problem and get out of valleys and difficulties on my
research road. His valuable suggestions often correct my naive ideas on science
and insight into the real interesting things of our research. I am thankful to him
for all these aspects.
I am grateful to my home supervisor, Prof. Zhichuan Niu, at Institute of
Semiconductors, Chinese Academy of Science. In his active research group I have
chance to study the molecular beam epitaxy of GaAs-based III-V compound semi-
conductors, especially self-assembled InAs QDs and QD-based optoelectronic de-
vices. I am also thankful to him because he gave me an opportunity to study
abroad in a beautiful, quiet, free and scientific environment.
Thank Prof. Hans Agren, head of Division of Theoretical Chemistry and
Biology, for building a friendly, free and scientific environment for study. Kindly
thanks to Prof. Yi Luo for his help in my study life here. Sincerely thanks to Dr.
Yaoquan Tu for friendly talking on the academic career. Thanks to Prof. Faris
Gel’mukhanov and Prof. Boris Minaev for setting good examples of true scientists
with continuous interest and assiduity on their work to young beginners in science
like me.
Thanks to Dr. Zhijun Ning, Dr. Qiong Zhang and Dr. Kai Fu for their kindly
help on my living in Stockholm. Thanks to my colleagues Zhihui Chen, Chunze
Yuan, Li Li, Dr. Xifeng Yang, Staffan Hellstrom, Guangjun Tian, Dr. Weijie
Hua, Dr. Qiang Fu, Dr. Xin Li, Xing Chen, Xiuneng Song, Yong Ma, Quan
Miao, Lili Lin, Yongfei Ji, Hongbao Li, Sai Duan, Lu Sun, Li Gao, Ying Wang, Ce
Song, Junfeng Li, Wei Hu, Yan Wang, Xiaoqiang Sun, Xinrui Cao, Lijun Liang,
Dr. Peng Cui and Dr. Liqin Xue, who have made my daily life in Sweden happy.
Thanks to Dr. Arul Murugan, Dr. Zilvinas Rinkevicius, Dr. Kestutis Aidas and
Bogdan Frecus for badminton exercise in weekend.
Sincerely thanks to Prof. Hakan Pettersson at Halmstad University for Fourier
transformed infrared spectroscopy of photocurrent and the valuable suggestions on
my manuscripts. Thanks to Dr. Haining Tian in Department of Chemistry, Royal
Institute of Technology for the help on the photovoltaic efficiency measurement
and the monochromator-based photocurrent spectroscopy.
x
I want to acknowledge the two-year financial support from the Chinese Schol-
arship Council.
Finally, I want to thank my parents for their selfless love, understanding and
support. As the thrifty holders of a peasant family, they sacrifice their material
and spiritual enjoyment to bring up two children, and allow their only son to
choose a path completely different from the surrounding peers.
Xiangjun Shang
2012-1-2, Stockholm
Contents
1 Introduction 1
1.1 Green energy: semiconductor solar cells . . . . . . . . . . . . . . . 1
1.2 Materials: Si or compound semiconductors? . . . . . . . . . . . . 3
1.3 High efficiency: by multi-junctions or nanostructures? . . . . . . . 4
1.4 Quantum dot solar cells . . . . . . . . . . . . . . . . . . . . . . . 5
2 Experimental techniques 8
2.1 Molecular Beam Epitaxy . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Atomic Force Microscopy . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Photoluminescence spectroscopy . . . . . . . . . . . . . . . . . . . 11
2.4 Photovoltaic I-V measurement . . . . . . . . . . . . . . . . . . . . 13
2.5 Photocurrent spectroscopies . . . . . . . . . . . . . . . . . . . . . 13
3 Self-assembled InAs quantum dots 17
3.1 Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3 Photoluminescence spectrum . . . . . . . . . . . . . . . . . . . . . 21
3.4 Closely stacked QDs . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5 Quantum description of optical properties . . . . . . . . . . . . . 23
4 InAs QD-embedded solar cells 26
4.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3 Photovoltaic I-V measurements . . . . . . . . . . . . . . . . . . . 31
4.4 Photocurrent spectra . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.5 Mechanisms of photocarrier escape from QDs . . . . . . . . . . . 37
xi
xii CONTENTS
5 Bunched correlation of two colloidal QD blinking 42
5.1 QD fluorescence imaging system . . . . . . . . . . . . . . . . . . . 43
5.2 Correlation analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Statistical result . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.4 Explanation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6 Summary of the included papers 53
References 55
Chapter 1
Introduction
1.1 Green energy: semiconductor solar cells
Energy is the engine of economic development in a modern society. Fossil fuels
contribute more than 85% of the global energy consumption. As BP statistical
review reported [1], during 2010, the world consumes 4350.4 million tons of oil (i.e.
87382 thousand barrels daily), 3169.0 billion cubic meters of nature gas (or 2852.1
million tons of oil equivalent) and 3555.8 million tons of oil equivalent of coal,
meaning a total energy consumption of 4.7×1013 kWh. The shortage of these
fossil energies and the overdischarge of greenhouse gases have been global crises
in recent years. Governments and scientists in many countries have been making
efforts to find alternative energy, especially “green energy” that is friendly to the
environment. There are several kinds of green energy, including nuclear power,
hydroelectric power, wind power, geothermal power, tidal power, and solar cell.
Nuclear power and hydroelectric power are most effective to obtain considerable
and sustainable power supply. However, nuclear power is limited by sophisticated
technology and only widely used in a few developed countries until now; moreover,
it is quite dangerous if there are some operation faults that have been observed
during Fukushima nuclear disaster (2011). Hydroelectric power is limited by the
geography (i.e. a river) and the rainfall amount; it would influence the ecology and
the climate possibly. Compared to them, solar cell that utilizes solar light to create
electric energy is easier on technology, safer, without gas pollution, unlimited by
the geography (especially suitable for deserts) and with sufficient resource (the
solar energy is 925 W/m2 at Sea level (AM1), see the spectrum in Fig. 1.1, and
the earth cross-section area for the normal incident sunlight is 1.3 × 1014 m2),
showing potential on energy solution in future.
1
2 1 Introduction
Figure 1.1 Solar light spectrum [2].
electron
Front grid contactSolar light
oc Voc
Isc
I
V0
ηmax
electron
hole
donor ion
acceptor ion
+
–p-region
built-in
Vo
c Voc
Back contactn-region
built-in
field
EC
EF
EV
Figure 1.2 Structure, I-V curve and band diagram of p-n junction solar cell.
Solar cell is based on photovoltaic (PV) effect. Semiconductor p-n junction
solar cell is most popular. Its structure, band diagram and I-V curve under light
are shown in Fig. 1.2. In equilibrium, in the junction interface, free electrons
provided by donors in the n-region and free holes provided by acceptors in the p-
region will recombine each other (i.e. ‘depletion’), leaving donor ions and acceptor
ions form the ‘built-in’ electric field. The Fermi level is flat everywhere and there
is no voltage output. Under sunlight, excess electrons and holes generated by
photon absorption will be drifted by this field into opposite regions. If there is no
external connection (i.e. the open-circuit case), these excess carriers will accumu-
late, e.g. see the holes in the p-region in the band diagram, and give an additional
potential, the open-circuit voltage (Voc), that reduces the built-in field (see the
red band diagram), making an equilibrium of photocarrier drift and diffusion and
thus no photocurrent (PC) output. If there is an external connection or load,
these accumulated excess photocarriers will output to produce PC; meanwhile,
the output voltage is reduced. In the short-circuit case when the output voltage is
zero and the Fermi level (EF ) is forced to be flat everywhere (see the black band
diagram), the PC output reaches the maximum (Isc). Between the open-circuit
1.2 Materials: Si or compound semiconductors? 3
case and the short-circuit case, the output voltage and the corresponding output
PC give the I-V curve, along which there is a point corresponding to the maxi-
mum electric power output (i.e. the rectangular area in the I-V curve) and the
maximum PV efficiency (ηmax).
1.2 Materials: Si or compound semiconductors?
In principle, because solar spectrum has a broad distribution from 250 to 2500
nm (see Fig. 1.1), all kinds of semiconductors with bandgap between 0.35 and
3.5 eV, such as Ge (0.66 eV), Si (1.12), InAs (0.36), InP (1.35), GaAs (1.42),
GaP (2.26), GaN (3.44), AlAs (2.16), PbS (0.37), CdTe (1.49), CdSe (1.70), CdS
(2.42) and ZnO (3.35), can absorb above-bandgap photons for solar cell appli-
cation. The lower the bandgap is, the larger the solar light absorption and the
photocurrent are. However, as the bandgap reduces, the Voc also reduces; more-
over, the photocarrier relaxation to the band edge loses more energy. There is an
optimal bandgap for solar cell. Based on the detailed balance principle, Shockley
and Queisser deduced the banggap-related ideal ηmax of a single-junction solar cell,
i.e., Shockley-Queisser limit for real cells [3]. The highest value, 31 %, is obtained
at bandgap of 0.95 ∼ 1.6 eV [2,3]. Si, GaAs, InP and CdTe are good candidates.
Si is abundant on earth, cheap (even for Si crystal wafer), and mature on
technologies of growth, purification and microelectronic device fabrication. Com-
mercial solar cells are mostly made in Si. But, on optoelectronic properties, Si
is not as better as compound semiconductors like GaAs. Si has indirect bandgap
and thus lower absorption coefficient than direct-bandgap GaAs, so Si solar cells
must be much thicker (e.g. 100 µm) than GaAs solar cells (4 µm) in order to ab-
sorb sunlight completely. Moreover, the electron mobility of Si is much lower than
GaAs, resulting in a slower photoelectron drift and diffusion in Si solar cells. As
Solar Cell Efficiency Tables (Version 38) [4] reported, under one sun AM1.5G, the
highest ηmax of single-junction crystalline Si solar cells has reached 25.0±0.5 %;
while that of single-junction crystalline GaAs cell has reached 28.1±0.8 %. GaAs
has another advantage over Si: GaAs crystals can be nearly lattice-matched grown
on Ge substrates [5] and allow InGaP crystals then grown on it; this material serial
is proper for triple-junction solar cell fabrication whose highest ηmax has reached
32.0±1.5 % under one sun AM1.5G. While for Si, the lattice-matched growth of
other materials (e.g. GaP [6]) on it for multi-junction solar cell fabrication is unde-
sired. Of course, considering the expensive GaAs substrate and the growth cost,
high-efficiency GaAs-based solar cells are only potential on satellites.
4 1 Introduction
Due to their optimal properties, compound semiconductors, especially III-V
ones, have been widely applied in optoelectronics such as light emitting diode
(LED), laser, photodetector and solar cell, and in microelectronics such as mi-
crowave and radio-frequency high-speed components in mobile devices [7], playing
a dominant role irreplaceable by the Si-based devices. Beyond these successes,
many potential ‘next steps’ have shown, for example, GaN-based blue and ultra-
violet LED, laser and photodetector [8], GaN power amplifier [9], III-Vs integration
on Si [10], and GaAs-based tera-Hertz devices [11]. Meanwhile, lots of novel nanos-
tructures fabricated on compound semiconductors, e.g. GaAs quantum rods [12]
and GaN nanowire [13], also show potentials for applications.
1.3 High efficiency: by multi-junctions or nanos-
tructures?
The Shockley-Queisser limit with highest value of 31 % forms an upper limit for
the practical single-junction solar cell. Now, the highest ηmax of single-junction
GaAs cell has reached 28.1±0.8 %, very close to this limit. To further increase
the PV efficiency, it must introduce new structures. One scheme is multi-junction
tandem solar cell. Several semiconductor p-n junctions in different bandgaps,
from the top wide-bandgap junction to the bottom narrow-bandgap junction, are
connected together with tunneling junctions, in order to absorb different-energy
photons in different junctions and fully utilize the generated photocarriers with lit-
tle energy loss during their relaxation into the bandedge. A multi-junction tandem
solar cell always works under concentrated solar light (e.g. 1000 suns). It produces
higher Voc. However, the Isc of each junction must be matched to avoid the bot-
tleneck of the maximum photocurrent output in some junctions, so the bandgap
and thickness of each junction must be proper. Moreover, limited by the lattice
match requirement for dislocation-free crystal growth, the material series of multi-
junctions is not free to choose. Now, the popular series is GaInP/Ga(In)As/Ge,
with the maximum PV efficiency beyond 40 % under 500 suns. Dilute nitride
is able to tune the InGaAs lattice constant and energy bandgap, making more
flexible on the material selection. The maximum ηmax of GaInP/GaAs/GaInNAs
triple-junction solar cells has reached 43.5 % under 400∼600 suns [14].
Apart from multi-junction tandem solar cells, in recent years, the popularly
studied nanostructures, such as quantum dots and quantum wells, provide new
ideas on improving the ηmax of a single-junction solar cell. The main idea is that
1.4 Quantum dot solar cells 5
except the above-bandgap photon absorption in the bulk host semiconductor, the
embedded nanostructures absorb and recycle sub-bandgap photons, which are
potential to raise the photocurrent as well as the PV efficiency. Another idea is
that these nanostructures are potential to well utilize the energy of ‘hot carrier’
before their relaxation into the low-energy bandedge [15]. This relaxation and its
energy loss are inevitable in bulk materials. Compared to multi-junction tandem
cells, nanostructure-embedded cells are more flexible on the material selection
and the structure design; they avoid the photocurrent match requirement and
the use of tunneling junctions. However, at present, there are still few solid
evidences demonstrating that nanostructures have improved the whole ηmax of a
single-junction cell, although sometimes there is a considerable increase of the
photocurrent Isc due to their additional absorption.
1.4 Quantum dot solar cells
As a kind of nanostructures, quantum dot is thought to have many desired qual-
ities: three dimensional quantum confinement, discrete energy levels, size and
energy-level tunability, long relaxation time due to phonon bottleneck effect [16],
and high-intensity light emission and absorption due to the confinement induced
high transition rate. Some scientists even proposed two conceptual mechanisms
in QDs that are helpful to raise the photocurrent. One is the multiple exciton
generation (MEG), meaning that a ‘hot’ electron generated in a high conduction-
band (CB) QD-level by absorbing a high-energy photon, will relax onto a low
CB QD-level, with the energy loss exciting additional electrons from valence-band
(VB) QD-levels to CB QD-levels. In other words, the absorption of a high-energy
photon creates more-than-one electron-hole pairs, similar to carrier impact ioniza-
tion and opposite to Auger relaxation. The MEG effect was reported in colloidal
QDs (e.g. colloidal CdSe QDs) [17] with high confinement barriers (i.e. the vacuum
levels), but not in epitaxial QDs like self-assembled InAs/GaAs QDs due to their
low barrier. The other mechanism is the intermediate band (IB) absorption [18]. It
means that the coupling of discrete levels in close-stacked epitaxial QDs forms one
or more IBs in the host semiconductor bandgap, enabling a cascade absorption of
sub-bandgap photons from the VB to the IBs, and from the IBs to the CB. Since
the IBs are electrically isolated from VB and CB, their introduction increases Isc
and keeps Voc unreduced.
This thesis mainly focuses on epitaxial QD-embedded p-n junction solar cells.
As for colloidal QDs-sensitized solar cells that are popularly studied recently,
6 1 Introduction
because their structure and principle (photocarrier selective extraction, similar to
dye-sensitized solar cells) are quite different, I do not want to detail them here.
In the study of p-n junction-based QD solar cells, the recent experimental
works usually compare the PV performance of single-junction solar cells with or
without QDs. Although sometimes there is a considerable increase of Isc, the re-
duced Voc is actually a common problem after embedding QDs, resulting in lower
PV efficiency. The reduced Voc is mostly attributed to strain induced disloca-
tions. The strain is always a drawback for epitaxial InAs QDs. A moderate strain
drives QD formation; while excessive strain induces dislocations. The disloca-
tions are serious for very closely stacked multiple QD layers, degrading both Voc
and Isc[19]. Introducing strain-compensated layers, e.g. GaInP, in QD-embedded
GaAs cells, is valid to reduce dislocations and PV degradation [19]. Meanwhile, the
strain-compensated layers with higher bandgaps also compensate the reduction of
bulk GaAs bandgap after embedding lower-bandgap InAs QDs. But, even for
far spaced dislocation-free InAs QDs, their embedded solar cells still have reduced
Voc. There must be some more essential reason. As for the photocurrent Isc, many
works expect the above-mentioned IB absorption would increase it. But, from PC
spectroscopy, we find that the PC from QDs is at least one order of magnitude
lower than the PC from the GaAs host. So, except the photocarrier generation in
QDs, the photocarrier extraction from QDs is even more important.
In this thesis, we design p-i-n GaAs solar cell and embed 20 nm spaced InAs
QD layers in its n-region and i-region and study their PV performance, with re-
spect to the non-QD control cell. The aim for QDs embedded in different regions
is to study their different mechanisms for photocarrier escape and check the roles
of the possible IB absorption the built-in electric field. Their different PV perfor-
mances help to understand the essential reason for the reduced Voc. The factors
of QDs and dislocations are distinguishingly shown in diode performance [20]. QDs
in the i-region enhance photocarrier radiative recombination and increase J0. It is
the essential reason for the commonly reduced Voc. While, dislocations caused by
strain in many QD layers induce Shockley-Read-Hall nonradiative recombination
(the ideality factor n close to 2) that also reduces Voc.
From the PC spectra we find that the above-GaAs-bandgap PC is enhanced
after embedding QDs in the i-region [21]. These QDs reflect the GaAs band wave-
functions, especially the valence band hole wavefunction, increasing wavefunction
overlap and interband absorption. From the bias- and temperature-dependent
sub-GaAs-bandgap PC spectra, we find that a strong built-in field is necessary
for effective photocarrier escape from QDs that is why QDs are mostly embedded
1.4 Quantum dot solar cells 7
in the i-region [19,22]. Inter-QD-layer tunneling under the field is effective between
closely stacked QDs. The IB absorption, if it exists, is very low in efficiency to
produce PC. The photocarrier escape from QDs in the n-region depends on the
minority hole thermal activation. It is poor in efficiency due to electron filling
induced potential variation.
The study of photocarrier escape from QDs is valuable for the design of
infrared photodetectors, e.g. p-i-n near-infrared photodiodes and n-i-n QD mid-
infrared photodetectors (QDIPs). Focal plane arrays of QDIP have been success-
fully produced for thermal imaging [23,24].
Chapter 2
Experimental techniques
The experimental work in this thesis covers the sample growth by molecular beam
epitaxy (MBE), morphology imaging by atomic force microscopy (AFM), photo-
luminescence (PL) spectroscopy, photovoltaic I-V measurement and photocurrent
spectroscopy, as detailed below.
2.1 Molecular Beam Epitaxy
Molecular Beam Epitaxy (MBE) is a method to grow crystals by the source
molecule epitaxy on the substrate in a high-vacuum and ultra-clean environment.
Most MBE systems use solid sources, e.g. our Vecco Gen II MBE. The schematic
of MBE system is shown in Fig. 2.1. There are three vacuum chambers, the
introducing chamber, the buffer chamber and the growth chamber, made from
stainless steel and separated by valves. The vacuum of the introducing chamber
(∼ 10−7 Torr) is maintained by the cascade pumping of a membrane pump and
a turbo molecular pump. The vacuum of the buffer chamber is kept by an ion
pump. The vacuum of the growth chamber is kept by a cryogenic pump and a
liquid nitrogen-cooled cryogenic panel inside the chamber. Sometimes, to start
the cryogenic pump operation, a prepump, i.e., a cascade turbo molecular pump
and membrane pump, is used. The cryogenic panel condenses impurities such
as H2O, N2, CO and CO2, making a high-vacuum and ultra-clean environment
around the heated substrate for sample growth. The background pressure of the
growth chamber in standby is 10−10 ∼ 10−9 Torr.
Substrates (e.g. GaAs) are fixed on substrate holders, located on the vehicle,
and transported inside the introduction chamber. In this chamber, the 180 ◦C bak-
ing is performed. Then, they are transferred into the buffer chamber to undergo
8
2.1 Molecular Beam Epitaxy 9
0
Sample moving poles
Orbit
Buffer chamber
Introducing chamber
Valve
0
Beam flux
gauge
180 C baking420 C degassing
Orbit
Vehicle
Growth
chamber
Valve
Baffle
Rheed gun
Rheed
screen
Substrate
gauge
Cryogenic
pump
chamberRotational
mechanics
Quadrupole
GaIn
AsSi
Reflection
spectrometer
Liquid nitrogen-cooled
cryogenic panel
Baffle
Heating
coil
Quadrupole
mass spectr.
spectrometercoil
Figure 2.1 Schematic of MBE system.
the 420 ◦C degassing one by one. When growing samples, a degassed substrate is
put into the growth chamber by the sample moving pole. Then, the substrate is
rotated to the growth position by the rotational mechanism and gradually heated
up to 650 ◦C for deoxygenating. The deoxygenation removes oxygen atoms from
oxides of Ga and As in the substrate surface, in an As-rich atmosphere that avoids
the As atom desorption. The baking, the degassing and the deoxygenating aim
the same, desorbing impurities from the substrate surface. After deoxygenating,
the substrate temperature is reduced to 600∼620 ◦C and a GaAs buffer layer in
thickness of 300 nm is grown, forming a flat and clean GaAs surface for sample
structure growth. After the sample growth has finished, the substrate will be re-
turned back to the buffer chamber and taken out from the introduction chamber
finally. The valves keep the vacuum of each chamber during these processes.
As for the sources, high-purity (99.9999%∼99.999999%) elementary sources
such as As, Ga, In, and Si (as dopant), are stored in separate effusion cells. These
cells are actually crucibles with the body heated up to about 1000 ◦C by a primary
filament and the mouth heated to even higher temperature by a tip filament (i.e.,
the ‘hot-lip’ mode [25]). Solid source inside each cell will evaporate as gaseous
atoms and effuse out. Only for the Al effusion cell, the tip filament is colder than
the body one (i.e., the ‘cold-lip’ mode), in order to avoid the Al spilling-caused cell
10 2 Experimental techniques
damage [25]. Moreover, the As cell has a two-zone crucible: the body contains solid
As source and the long tip cracks the evaporated As4 molecules into As2 ones. By
extensively controlling the base and tip temperatures with thermocouple probes
and electronic PID temperature-control modules, the effusion rate of source atoms
or molecules from each cell will be accurately controlled and tuned. By opening
or shuttering the baffle in front of each cell, different material or compositional
structures can be grown flexibly.
The effusion rate is measured by the beam flux gauge, i.e., a hot-filament
ionization gauge [26], after it is rotated to the growth position by the rotational me-
chanics. High-energy electrons, emitted from the filament cathode and speeded
up by a high positive grid voltage, will ionize the surrounding source atoms or
molecules, having them attracted to the collector in negative voltage and produc-
ing electric current that is proportional to the amount of source ions.
During the growth, the substrate is at 500∼650 ◦C, located at the growth
position in the center of the growth chamber. These gaseous atoms and molecules
deposited on it will migrate in the surface, react with each other, and eventually
form an epitaxial layer. The growth is a dynamic process, rather than a thermal
one. The crystalline substrate, acting as the template (or seed crystal), enables
the growth of lattice-matched crystalline materials on it. The epitaxy is a both
chemical and physical process. The deposited atoms and molecules firstly form
an amorphous material (chemical reaction); under the substrate temperature, the
amorphous material crystallizes into film (physical phase transition).
The epitaxial growth is in-situ monitored by Reflection High Energy Electron
Diffraction (RHEED). High-energy electrons, emitted from the high-voltage (14
kV) electron gun, inject on the sample surface and reflect. The reflecting electron
beam has diffraction due to the sample surface roughness in atomic monolayer
level. The diffraction pattern is recorded by the RHEED screen and displayed in
computer. A flat sample surface gives a multi-line pattern; while a rough surface
(e.g. quantum dots on it) gives a dotted pattern. The intensity of the RHEED
pattern is also oscillating during one crystalline monolayer growth, from which an
accurate monolayer epitaxy is available to control.
The rotational mechanics including several gears has two functions: one is
inverting the positions of the substrate and the beam flux gauge that enables
the transfer of a substrate to the growth position and the measurement of the
effusion rate; the other is axially rotating the substrate during the growth for
space-uniform epitaxy.
The quadrupole mass spectrometer [27] can detect gaseous impurities in the
2.2 Atomic Force Microscopy 11
growth chamber, such as H2O, CO, N2 and CO2, and oxides of As and Ga. The
reflection spectrometer uses halogen lamp, optical grating and photodetector for
in-situ measurement of sample’s reflection spectrum, which is useful to monitor the
layer thickness and optical mode during the growth of distributed Bragg reflectors.
Compared to Metal Organic Chemical Vapor Deposition (MOCVD) that
grows crystals fast by gas sources in low vacuum and suits the mass produce
of most optoelectronic devices like LED and VCSEL, MBE has its own features:
accurate control of monolayer growth, flexibility on energy band design and struc-
ture growth, high vacuum and relative low growth temperature, therefore, it is
suitable for the growth of superlattices [28], quantum cascade lasers [29] and high-
electron-mobility microwave devices.
2.2 Atomic Force Microscopy
The morphology of uncapped QDs, including their sheet density, shape and size,
is characterized by atomic force microscopy (AFM). AFM has a cantilever with a
sharp tip in µm-scale size and nm-scale curvature to scan the sample surface. For
the usually used contact mode, a van der Waals repulsive force exists between the
tip and the surface atoms it touches. The surface morphology of the sample will
vary this force, leading to a cantilever deflection that is determined by Hooke’s
law. A laser illuminates on the cantilever and reflects to a four-region detector to
show a light spot. The cantilever deflection is monitored by the shift of the spot
position, and transformed into morphological information of the sample surface.
In contact force operation, the spot shift is fed back to a piezoelectric element
behind the sample board that moves the sample slightly in vertical direction to
keep constant force and deflection. The displacement describes the local height
of the sample surface. Compared to another kind of scanning probe microscopy,
scanning tunneling microscopy, that is based on electron tunneling from atoms
on an electrically conductive sample to the probe tip and in 0.01∼0.1 nm spacial
resolutions, AFM is in lower (∼nm) spacial resolution, less limited on sample
conductivity and vacuum condition, and thus more widely used.
2.3 Photoluminescence spectroscopy
Photoluminescence (PL) is a phenomenon that a material absorbs high-energy
photons and generates photocarriers; these photocarriers relax into the low-energy
12 2 Experimental techniques
states, especially the ground state, of the material, and emit photons by the tran-
sition from these states. By spectroscopy of the luminescence, energy states of the
material are informed. For example, InAs QD samples, excited by a He-Ne laser
(λ =632.8 nm), give infrared luminescence due to the electron-hole recombination
on discrete QD states, whose spectrum with several peaks can characterize the
QD energy states.
The PL instrument is shown in Fig. 2.2. The power-tunable He-Ne laser beam
is incident from the central hole of the concave copper mirror onto the sample.
The luminescence emits from the sample, together with the reflected laser light.
It is reflected by the concave copper mirror and incident into the detector module
by the lens focus. The detection is based on the Fourier transform (FT) of the
incident light by a Michelson interferometer (for more detail, see Section 2.5). The
Ge detector is cooled by liquid nitrogen. The sample is located in a chamber at
room temperature.
He-Ne lasermA
FT spectrometer
Michelson
interferometerConcave
copper
Mirror
Power-tunable electric power for laser
0
Lens
Sample
copper
mirror
Beamsplliter
0Liquid N2-cooled
Ge detector
Sample
Figure 2.2 Setup of PL spectroscopy.
There are some more advanced PL spectroscopies, such as excitation power-
varied PL, time-resolved PL, PL excitation (PLE) and selectively excited PL.
Excitation power-varied PL spectra are measured under varied exciting laser in-
tensity. It is benefit to characterize high-energy states due to their considerable
occupation under high-intensity excitation. Time-resolved PL uses a pulsed ex-
citing laser (e.g. a mode-locked Ti: sapphire laser in femtosecond-scale pulses)
and a pump-probe optical setup to measure the decayed PL intensity at a given
delay time, which is useful to define the lifetime of an energy state. PLE fixes the
detection energy by a monochromator instead of the FT spectrometer, and scans
the exciting photon energy by monochromizing a broad-spectrum light source,
e.g. a low-intensity tungsten lamp or a high-intensity continuous-wave Ti: sap-
2.4 Photovoltaic I-V measurement 13
phire laser. Selectively excited PL is the opposite, i.e., fixing the exciting photon
energy while scanning the detection energy by a monochromator. Both are helpful
to study the photocarrier relaxation on a low-energy state. On the other side, the
dependence of PL on temperature and bias voltage is also studied, extending the
understanding of photocarrier occupation, relaxation and escape from an energy
state.
2.4 Photovoltaic I-V measurement
For p-n junction solar cells, as said in Chapter 1, the maximum photovoltaic
(PV) efficiency is calculated by measuring the I-V relation under a solar light
simulator. The I-V relation is measured by a multimeter, e.g. a Keithley one, by
applying a scanning forward bias on the measured sample from 0 V to Voc when the
photocurrent becomes zero. The solar light simulator is a Xenon lamp with several
filters to get rid of several sharp emission peaks in Xenon spectrum. The lamp
power is properly tuned to 1 sun AM 1.5G (i.e., 100 mW/cm2) with a standard
Si cell calibration. The measurement is performed closely in a chamber to avoid
background light. The position of samples is fixed for an identical light intensity.
Along the resulted I-V relation, there is a point corresponding to the maximum
electric power output. The maximum PV efficiency is obtained by dividing the
maximum power by 100 mW/cm2. Sometimes, the concentrated PV efficiency is
measured. The light intensity of solar simulator is tuned to 10 ∼ 104 of the above
1 sun intensity; the sample temperature is well controlled by an additional air
condition; and the maximum efficiency is calculated after considering the increased
light power. Besides, the I-V relation in darkness is also useful to explore the diode
performance of these p-n junction solar cells.
2.5 Photocurrent spectroscopies
From the photovoltaic I-V measurement, we know the total photocurrent (PC)
of a solar cell. For QD solar cells, we want to study the contribution from QD
absorption; hence, the PC spectrum as function of photon energy is required. The
PC spectroscopy is performed by either monochromator or Fourier transform (FT)
infrared spectrometer.
The PC spectroscopy in 300∼1000 nm is performed under a Xenon lamp with
a monochromator. A beamsplitter is used to equally divide the monochromatic
14 2 Experimental techniques
light into two beams: the transmission one is incident on the sample; while the
reflection one is incident on a Si photodiode as reference. The sample is located
at a fixed position behind a hole for identical intensity and spot of light. As the
monochromatic light wavelength (λ) scans, the PC signals of both the sample
and the Si photodiode, Isample and ISi, are measured by a Keithley multimeter
simultaneously. The ISi(λ) is useful to eliminate the influence of the peakful Xenon
spectrum on Isample(λ), by considering Isample(λ)/ISi(λ). In other words, since the
external quantum efficiency (EQE) of the sample is obtained by EQEsample(λ) =
EQESi(λ) × Isample(λ)/ISi(λ) [30] and EQESi(λ) is almost saturated and constant
in 500∼900 nm, Isample(λ)/ISi(λ) becomes a good measure of EQEsample(λ) in this
wavelength region. The spectroscopy is in resolution of ∆λ = 5 nm/point, defined
by the automatic controlled monochromator.
The PC spectroscopy in the near-infrared region (i.e., QD contributions) is
measured in detail by a Bruker Vertex 80v FT infrared spectrometer equipped
with a quartz-halogen tungsten (QHT) lamp. The spectral range is 1200∼30000
cm−1 at most. The resolution is ∆k = 4 cm−1/point. FT infrared spectrometer is
often used for the absorption spectroscopy of materials. It can also characterize
PL spectrum as said before and PC spectrum, by modifying the setup accordingly.
The structure of FT spectrometer and the spectra of the QHT lamp and the CaF2
beamsplitter efficiency are presented in Fig. 2.3.
The QHT lamp emits light whose spot size is trimmed by the aperture (APT).
After being reflected by a concave mirror, the parallel light is incident into the
interferometer. The CaF2 UV/VIS/NIR beamsplitter (BMS) divides the incident
light into two parts: the reflection beam is then reflected back by the static planar
mirror and transmitted through the BMS; while the transmission beam is then
reflected back by the moving planar mirror and reflected by the BMS. Both beams
experience the same optical length in the BMS, leaving their optical length dif-
ference (∆L) only tuned by the moving mirror whose position is monitored by a
He-Ne laser inside. The two beams form an interferential light that is focusing
on the sample located in the dark chamber. For absorption spectroscopy, the
transmission light through the sample is collected by the installed detector (D1)
to produce PC signal; while for PC spectroscopy, the detector is unused and the
sample’s PC signal is directly acquired and amplified by an external current am-
plifier. Unlike monochromator, the spectral scan here is based on the ∆L-varied
interference and a Fourier transform of the resulted PC signal as a function of
2.5 Photocurrent spectroscopies 15
∆L. The principle is indicated as
PC(∆L) =
∫R(k)I(k) [1 + cos(2πk∆L)] dk
=
∫PC(k) [1 + cos(2πk∆L)] dk
(2.1)
Here, k is wave number [cm−1]; I(k) is light intensity [W]; PC(k) and PC(∆L)
are photocurrent [A]; R(k) = PC(k)/I(k) is responsivity [A/W] of the detector
or sample. Making Fourier transform on PC(∆L),
FTPC(∆L) =
∫PC(∆L)cos(2πk∆L)d∆L
=
∫ ∫PC(k′)× [1 + cos(2πk∆L)]dk′cos(2πk∆L)d∆L
=
∫PC(k′)×
∫[1 + cos(2πk′∆L)]cos(2πk∆L)d∆Ldk′
=
∫PC(k′)δ(k − k′)dk′ = PC(k)
(2.2)
The modified setup to measure PC spectrum is described in Fig. 2.4. The
sample is fixed in a cryostat. Its electrodes extend to the cryostat top plate. By
using a PCB connection, its PC signal is introduced to an external current ampli-
fier (Keithley 428), which amplifies the current signal and converts it into a voltage
signal, under tunable amplification gain and bias voltage. The voltage signal out-
puts to a multimeter (Keithley 2602) for voltage monitor; meanwhile, it feedbacks
through the serial-to-parallel converter to the FT spectrometer for Fourier trans-
form. The final data is sent to a computer for spectrum display. Sometimes,
to clarify the contributions from QDs, a GaAs filter is used at the incident light
aperture, only allowing the transmission of sub-GaAs-bandgap photons.
In the low-temperature PC spectroscopy, a liquid nitrogen-cooled cryostat
(Oxford OptistatDN ) and a temperature controller (Oxford ITC503 ) are needed.
On the top plate of the cryostat there is a 10-pin seal electrical port for temper-
ature control, behind which there is a thermometer and a heater. The electronic
communication between the multi-pin port and the temperature controller en-
ables an accurate control of the sample temperature from 77 K to 300 K. Before
cooling down the cryostat, gas evacuation and helium gas inflation are performed,
in order to avoid problems related to water vapor freeze. For room-temperature
PC spectroscopy, the temperature controller is not used and the cryostat is not
cooled.
16 2 Experimental techniques
Figure 2.3 FT spectrometer [31] and QHT lamp and BMS transmission spectra [32].
Serial->parallel
converter
Electronics (FT)
Computer
Multimeter
(V-monitor) FT spectrometer
+_
gainbias
IN
OUT
I->V amplifier cryostat
converter
GaAs filter
Temperature controllerelectric access for temperature control
Temperature setting I/O
sample
Figure 2.4 Setup for PC spectroscopy. Red arrows: the signal transport.
Chapter 3
Self-assembled InAs quantum
dots
Quite different from bulk materials, nanostructures show novel optical features,
such as wavelength-tunable emission and high emission intensity due to their nm-
scale size and quantum confinement. In recent two decades, epitaxial InAs QDs,
self-assembled on GaAs or InP substrates, have been extensively studied and
widely used in optoelectronic devices for infrared light emission and detection,
such as QD lasers and QD infrared photodetectors, because they are able to be
freely integrated into III-V bulk semiconductors. In this chapter, I will focus on
their growth, morphology imaging, quantum theory and photoluminescence (PL)
spectroscopy, in order to provide an overall knowledge of this kind of QDs.
3.1 Growth
The self-assembled growth of InAs QDs on GaAs (or InP) substrate is driven by
strain. InAs lattice constant (6.058 A) is larger than GaAs (5.653 A) and InP
(5.869 A), so the InAs epitaxy on GaAs (or InP) surface will accumulate strain.
There are two energies competing mutually during the dynamic growth process
and determining the growth mode, the strain energy that is proportional to the
volume of the strained material and the surface energy that is determined by the
surface area [33,34]. Fig. 3.1 schematizes the growth process. At first the strain
energy is zero while the surface energy is not. Indium and arsenic atoms are
effused on the substrate surface. They will migrate in the surface and bond each
other. The InAs deposition is dominated by two-dimensional (2D) film growth,
forming InGaAs wetting layer (WL) due to In-Ga intermixing. During the WL
17
18 3 Self-assembled InAs quantum dots
growth, the surface energy keeps constant while the strain energy accumulates as
the WL thickness increases. When the WL reaches a critical thickness, the strain
energy becomes dominant. Many nm-scale three-dimensional (3D) InAs islands
form suddenly. The strain is released by increasing the total surface area by these
3D islands. The growth is called Stranski-Krastanov (SK) growth, a mode of
crystal epitaxy [34]. Continuing the InAs epitaxy, these InAs islands will become
bigger and the strain will accumulate further. Finally, big clusters filled with
dislocations will form, reducing both the surface energy and the strain energy.
So, the InAs deposition amount must be proper. These islands become QDs after
cladding a GaAs barrier layer. Usually, an additional 5 nm In0.15Ga0.85As layer
is clad before cladding the GaAs layer. It flattens the surface strain, uniforms
the QD size and improves the QD optical properties by forming the Dot-in-Well
structure. Of course, the strain still exists in these QDs.
In As
InGaAs WL
InAs island InAs QD
GaAs
Strain
EnergyS-K growth mode
2D WL 3D island
Cluster and
dislocation
Surface
Strain
=> 2D growth
2D WL 3D island dislocation
InAs amount
Figure 3.1 InAs QDs growth in SK mode.
The InAs island nucleation is also related to the indium deposition, migra-
tion, evaporation and bonding on the substrate, and the indium segregation from
the WL after its content saturates [35]. It is affected by the growth temperature,
rate and interruption, the arsenic flux, and the surface lattice condition. In gen-
eral, high temperature enhances indium migration and evaporation; slow indium
effusion rate and interruption promote indium migration and island nucleation to
form big QDs; very high arsenic flux suppresses indium migration and forms big
clusters while very low arsenic flux affects indium adsorption; on a strain-relaxed
surface big QDs are formed with less strain; introducing little Sb elements on the
surface formes small QDs in an ultrahigh density ∼ 1000/µm2.
There are at least four examples to show the flexibility of the self-assemble
of InAs QDs on GaAs substrates. First, the normal QD growth is at 500 ◦C,
3.2 Morphology 19
0.1 ML/s in rate and 10 s interruption after each 1 s deposition, with the total
amount of 2.3 monolayer (ML). The formed QDs have density of 300/µm2 and
light emission around 1200 nm. Second, if increasing the growth temperature to
550 ◦C and effusing indium at a rate ∼ 0.005 ML/s with a total amount ∼ 1.9 ML,
critical for island formation, the formed InAs QDs will have a low density about
10/µm2 and show discrete emission lines due to single transitions in individual
QDs. Such QDs has been used to fabricate single photon sources [36]. Third, if
growing a thick metamorphic InGaAs buffer layer on GaAs substrate with the
indium composition increasing gradually, the InAs island nucleation on it will be
strain-relaxed, showing longer-wavelength emission and potentials for the 1.55 µm
telecom application [37,38]. Fourth, if growing closely stacked QD layers with GaAs
spacer ∼ 10 nm, the strain coupling between QD layers will lead to a vertical
aligned QD formation [39].
3.2 Morphology
AFM and PL spectrum are used to characterize the QD morphology and light
emission property and optimize the QD growth conditions. The test samples for
AFM and PL spectrum have three layers of InAs QDs with 5 nm In0.15Ga0.85As
capping layer and 50 nm GaAs spacer and an uncapped InAs island layer in the
end. The AFM image characterizes the surface morphology of the uncapped InAs
islands, such as shape, surface density and size distribution.
With different indium deposition amounts and growth interruptions, the nor-
mally grown QDs show AFM images in Fig. 3.2. The growth temperature is 500◦C. The arsenic partial pressure is 8.7× 10−7 Torr. QDs are lens-shaped, in small
height and large base diameter. Their density is ∼ 102/µm2. Their PL spectra
are shown in the left of Fig. 3.4. We find that the 2.3 ML indium amount and the
5 s interruption, i.e. sample (a), are optimal for QD light emission. The QD size
distribution in sample (a) is also analyzed in Fig. 3.2. The QD density is about
270/µm2. Most QDs are 7 ∼ 9 nm in height and 28 ∼ 32 nm in base diameter.
Sample (a) has no big clusters while the other samples have.
Fig. 3.3 gives the AFM image of QDs grown on a metamorphic In0.08Ga0.92As
buffer. The QDs are in a lower density about 120/µm2, mostly in height of 7 ∼ 10
nm and base diameter of 35 ∼ 50 nm, larger than the QDs in Fig. 3.2. Besides,
there are some 100 nm-sized big clusters.
20 3 Self-assembled InAs quantum dots
2 3 4 5 6 7 8 9 10 110
20
40
60
80
QD height [nm]10 15 20 25 30 35 40 450
10
20
30
40
QD base diameter [nm]
QD total number
269 /µm2
Figure 3.2 AFM images of normal QDs grown with different indium amounts andinterruptions: (a) 2.3 ML, 5 s; (b) 2.4 ML, 5 s; (c) 2.3 ML, 3 s; (d) 2.4 ML, 3 s. Thegrowth temperature is 500 ◦C and the arsenic partial pressure is 8.7 × 10−7 Torr, forall. The QD size distribution of sample (a) is analyzed.
2 3 4 5 6 7 8 9 10 11 120
10
20
30
40
QD height [nm]
25 30 35 40 45 50 55 60 65 70 75 800
5
10
15
20
25
30
35
40
Metamorphic
InGaAs buffer
QD
Num
ber
QD base diameter [nm]
QD total number
126 /µm2
Figure 3.3 AFM of InAs QDs formed on a metamorphic InGaAs buffer layer.
3.3 Photoluminescence spectrum 21
3.3 Photoluminescence spectrum
PL spectrum was used to characterize the light emission property of QDs. Fig. 3.4
shows the PL spectra of the above mentioned QD test samples. In the left, the
four samples are the ones in Fig. 3.2. Their ground PL peaks are 1250 ∼ 1280
nm. A 5 s growth interruption after each 0.1 ML indium deposition is proper for
indium migration to form islands. The 2.4 ML QDs (L-b) have longer emission
wavelength than the 2.3 ML QDs (L-a).
800 1000 1200 1400 1600800 1000 1200 1400 1600
0
50
100
150
200
250
R-c
R-b
Wavelength [nm]
R-a
PL intensity [arb. uni.]
L-a 2.3ML, 5 s
L-b 2.4ML, 5 s
L-c 2.3ML, 3 s
L-d 2.4ML, 3 s
Figure 3.4 PL spectra of InAs QD samples. Left: normal QDs on GaAs in Fig. 3.2.Right: (R-a) 2.3 ML QDs on GaAs; (R-b) 2.8 ML QDs on metamorphic In0.08Ga0.92Asbuffer; (R-c) 2.9 ML QDs on metamorphic In0.15Ga0.85As buffer.
The right of Fig. 3.4 compares the PL spectra of InAs QDs formed on meta-
morphic InGaAs buffer and directly on GaAs. R-a is the best sample in Fig. 3.2,
i.e. L-a in the left of Fig. 3.4, with 2.3 ML InAs on GaAs. Its ground PL peak is
at 1270 nm and the 1st-excited peak is at 1180 nm, spaced by 75 meV. R-b is the
sample in Fig. 3.3, with 2.8 ML InAs QDs grown on a metamorphic In0.08Ga0.92As
buffer, capped by 5 nm In0.25Ga0.75As and spaced by 50 nm In0.08Ga0.92As. The
growth temperature is 510 ◦C. Its ground peak is at 1370 nm and the 1st-excited
peak is at 1290 nm, spaced by 56 meV. R-c is another QD sample, with 2.9 ML
InAs deposition, formed on a metamorphic InxGa1−xAs buffer (x step graded from
0.02 to 0.18 and then set back to 0.15), capped by 3.5 nm-thick In0.33Ga0.67As and
spaced by 50 nm-thick In0.15Ga0.85As. The QD growth temperature is 510 ◦C. Its
ground peak is at 1440 nm and its 1st-excited peak is at 1370 nm, spaced by 44
meV. The high indium composition in the metamorphic InGaAs buffer relaxes the
strain and realizes longer-wavelength emission in QDs. The reduced PL peak sep-
aration is related to the strain relaxation and the reduced barrier for QDs. From
22 3 Self-assembled InAs quantum dots
R-a to R-c, the PL intensity also decreases since lots of dislocations have been
introduced in the metamorphic buffer and the QDs growth on it is not optimal.
3.4 Closely stacked QDs
Based on the optimal QD growth conditions, in this thesis, the InAs QDs in QD
solar cells were deposited at 500 ◦C, in rate of 0.1 ML/s and interrupted 10 s after
each 0.1 ML, in arsenic partial pressure of 8.7 × 10−7 Torr, on a rotating GaAs
substrate. The total InAs deposition amount is 2.3 ML. InAs starts to island after
1.8 ML; the rest 0.5 ML makes these islands growing bigger. These InAs islands
are directly clad by 20 nm-thick GaAs layer in a rate of 1 ML/s, instead of the 5
nm InGaAs capping layer and the 50 nm GaAs spacer. After the 20 nm GaAs,
the next QD layer would be grown. Five such QD layers were grown as a QD
matrix.
As seen in the cross-sectional transmission electron microscopy [33,39], verti-
cally aligned QDs have formed in closely stacked QD layers, due to the inter-QD-
layer strain coupling. The thicknesses of the GaAs spacer is critical for such QD
alignment [33,40–42]. We use closely stacked QDs for QD solar cells, because these
closely spaced and aligned QDs could allow an interlayer state coupling [43,44] and
form intermediate bands (IBs), for free carrier transport and cascaded absorption
of sub-bandgap photons [45].
0.9 1.0 1.1 1.2 1.3800 1000 1200 1400 1600
0
50
100
150
200
Sample A
Energy [eV]
Sample B(b) Coupled
QDs
PL intensity [arb. uni.]
Wavelength [nm]
(a) Uncoupled
QDs
Figure 3.5 PL spectra of coupled QDs used for solar cells.
Fig. 3.5 shows the PL spectra of closely stacked QDs in our QD solar cells.
There is a broad profile including many closely spaced peaks. These peaks come
from the energy state coupling and the QD size distribution. Compared to the
best uncoupled QD sample mentioned above, the PL intensity of coupled QDs is
much reduced because of the strain induced dislocations.
3.5 Quantum description of optical properties 23
3.5 Quantum description of optical properties
The essence of the novel optoelectronic properties in nanostructures is the quantum
confinement in their nm-scale size. Carriers (e.g. electrons) are confined in a
region comparable to their de Broglie wavelength, and the carrier wave property
therefore becomes considerable. Quantum wells (QWs) confine carriers in one
dimension, leaving the other two dimensions (2D) free for carrier motion (e.g.
2D electron gas); quantum wires (QWires) confine carriers in 2D, allowing carrier
free motion along the wire. Quantum dots (QDs), or namely, nanocrystals, confine
carriers in three dimensions (3D), providing all-space discrete energy states and δ
function-like density of states. The energy of these confined states is tunable by
changing the QD size. The smaller the QD is, the higher the confinement and the
state energy are, as demonstrated in the multi-color visible light emission from
colloidal II-VI QDs with the same material [46]. Here, I will introduce some basic
concepts to describe QDs.
The carrier wave property in a QD can be depicted by the 3D steady-state
Schrodinger equation
[−~2
2m0
∇( ∇
me(r)
)+ V (r)− En
]Ψ(r) = 0, V (r) =
EGaAs, r ∈ GaAs
EInAs, r ∈ QD(3.1)
where V (r) is the electric potential; En and Ψ(r) are the QD state energy and
wavefunction; me is the carrier effective mass; EGaAs (EInAs) are conduction band
(CB) edge for electron and valence band (VB) edge for hole in GaAs barrier (InAs
QD). The QD band diagram is shown Fig. 3.6. Both CB electron and VB hole are
confined in QDs. The solutions of this equation are the bound states s, p and so on,
with even or odd parities. This equation is the same as the Schrodinger equation
to describe electronic states in an atom, so QDs are also called “artificial atoms”.
For spherical colloidal QDs, V (r) has spherical symmetry, so this equation can be
analytically solved by setting the wavefunctions as Ψ(r) = Rnl(r)Ylm(θ, ϕ). The
radial component Rnl has quantum numbers (n, l) and the spherical harmonic
function Ylm(θ, ϕ) has quantum numbers (l, m), analog to atomic states. For
lens-shaped QDs, there is only axial symmetry, the QD states can be numerically
rather than analytically solved. The 3D confinement can be separated into the
confinements in the normal direction and the lateral plane. Since the QD height
is 7 ∼ 9 nm and base diameter is 28 ∼ 32 nm, the normal confinement is so large
that only one confined state is allowed. Bayer [47] thus proposed that the lateral
24 3 Self-assembled InAs quantum dots
confinement is responsible for the discrete QD states in separation of 50 ∼ 100
meV. Their wavefunctions also show parity features in the lateral plane, such as s,
p and so on. The s-state can accommodate two carriers with opposite spins; the
p-state can accommodate four carriers with either different spatial distributions
or opposite spins; while the d-state can accommodate six carriers. The confined
carriers tend to relax and distribute on low-energy states, e.g. the ground s-state.
Since the QD material is still periodic crystal and in each QD (∼ 103 nm3)
there are ∼ 104 atoms, the potential V (r) and the wavefunction Ψ(r) must be able
to describe the lattice periodicity, i.e., the periodic nuclear potential on the lattice
with distance (i.e. lattice constant) of 0.56∼0.6 nm. So, similar to bulk crystals,
the Bloch theorem is still valid for QDs, i.e., Ψ(r) = ϕ(r)u(r), where ϕ(r) is the
slowly-varied envelope around QD and the Bloch function u(r) is periodic, i.e.,
u(r + r) = u(r), where r is a lattice vector.CB
Intraband
p-state
WL state
)()()( rurrrvr ϕψ =
Interband
transition
Intraband
transition
s-state
QD
)(rvϕ
)(rur
p-state
s-state
QD
Lattice constantNucleus
VB
WL state
Figure 3.6 Schematics of QD band diagram, wavefunction and transition.
There are two kinds of optical transition between QD states, one is the in-
terband transition, the other is the intraband transition, as seen in Fig. 3.6. The
transition rate Ti→f from an initial state i to a final state f depends on the tran-
sition matrix element, as the Fermi golden rule Ti→f ∝ (2π/~) |<Ψf |E · r|Ψi >|2,where the assumption of electrical dipole E ·r transition is used. Using the Bloch
wavefunction, the transition matrix element is expressed as
< ϕecuc|E · r|ϕh
vuv >=< ϕec|ϕh
v >< uc|E · r|uv >, (interband)
< ϕec1uc|E · r|ϕe
c2uc >=< ϕec1|E · r|ϕe
c2 >< uc|uc >, (intraband)(3.2)
where ϕec1,2 are CB electron states and ϕh
v is VB hole state; uc and uv are CB
and VB Bloch functions. For the interband transition, the rapidly varied Bloch
functions uc and uv give a constant dipole moment while the overlap of the slowly
3.5 Quantum description of optical properties 25
varied envelops ϕec and ϕh
v determines the size of the matrix element. The maxi-
mum overlap is obtained between the same states (e.g. s-state) for electron and
hole. This transition describes the interband absorption of near-infrared photons
in QDs. For the intraband transition between two CB electron states, since the
two states have the same rapidly varied Bloch function, the matrix element is
determined by the dipole moment of the envelope functions, <ϕec1|E · r|ϕe
c2 >. It
requires opposite parities of the wavefunction envelops (e.g. s-state and p-state).
Moreover, the external electric field must be parallel to the confined dimension
that shows opposite parities. It is why in QW infrared photodetectors only the
electric field component in the normal direction is effective to create the intraband
transition by absorbing mid-infrared photons.
The optical transition also satisfies the angular momentum selection rule.
For example, s-state has two degenerations, | ± 1/2〉 for electron, | ± 3/2〉 for
heavy holes; the transition only occurs between |1/2〉 and |− 3/2〉 (emit left-hand
circularly polarized light), or between |−1/2〉 and |3/2〉 (emit right-hand circularly
polarized light). The exciton state in a single InAs QD thus has two configurations
for polarized light emission, based on which the polarization-entangled photon-pair
emission has been realized [48].
The optical transition is also sensitive on electric field. Under an electric
field, the wavefunctions of CB electron and VB hole in a QD separate, reducing
their overlap (transition rate) and transition energy. It is so called quantum Stark
effect. Fig. 3.7 shows the bias-dependent PL spectra of our QD solar cell sample
A, with QDs in the built-in field region. The lowest QD transition (QD6*) is most
reduced as the electric field enhances.
0.9 1.0 1.1 1.40 1.45 -1.0 -0.8 -0.6 -0.4 -0.2 0.0
Photo
lum
inescence [arb
. units]
Photon energy [eV]
(a)0 ~-1 V (-0.1 V/step)
(b)
Bias [V]
QD4
QD5
QD6
QD6*
Relative PL
intensity
Figure 3.7 Bias-dependent QD PL spectra, vertically shifted by ×10 for clarity.
Chapter 4
InAs QD-embedded solar cells
In this chapter, I will describe the design, fabrication and photovoltaic (PV) ef-
ficiency measurement of our InAs QD solar cells. The QD performance on pho-
tocurrent (PC) production is analyzed from the bias- and temperature-dependent
PC spectra. The reduced Voc of both QD solar cells is because of: 1) the Shockley-
Read-Hall (SRH) recombination in QDs and dislocations; 2) the reduced photo-
carrier lifetime when QDs are embedded in the depletion region. The interlayer
tunneling in closely stacked QD layers helps the PC production from QDs in the
depletion region.
4.1 Design
p-n junction solar cells are based on the light absorption in semiconductors and
the photocarrier transport and extraction under the built-in electric field before
recombination. The light absorption satisfies the Beer law ln(It/I0) = −αL where
α is absorption coefficient; I0 is the initial light intensity; It is the light intensity
in depth of L. The photocarrier generation is thus exponentially dependent on
the depth. The α is dependent on wavelength. For GaAs, it is 2 × 103 cm−1 at
870 nm (i.e. the GaAs bandedge), 3 × 104 cm−1 at 600 nm and even higher for
shorter wavelengths. Most ultraviolet photons are absorbed in the front surface
and recombine before drift. Solar cell design mainly focuses on visible and near-
infrared photons absorbed in ∼ µm depth. The built-in electric field, determined
by the Poisson equation ∇2ψ = ∇·E = −(e/ε)(p−n+N+D−N−
A ), distributes in a
width of W (i.e. ‘depletion region’), with the total potential Vbi and the maximal
26
4.1 Design 27
field Em expressed by
Vbi =kBT
qln(
NAND
n2i
),W =
√2e
ε(Vbi − V )(
NA + ND
NAND
), Em =2(Vbi − V )
W(4.1)
where NA and ND are acceptor and donor densities; ψ and E are electric potential
and field; n2i (=n0p0) is equilibrium carrier product; V is bias voltage; ε is dielectric
constant; kBT is 25.8 meV at 300 K.
The function of a p-n diode can be described by the drift-diffusion model [49,50]
in Eqs. 4.2. n and p represent electron and hole; µn,p are carrier mobilities; Dn,p
(=µn,pkBT/e) are diffusion coefficients; Jn and Jp are current density. There
are two kinds of currents: the drift current (e.g. enµnE) proportional to the
electric field and the diffusion current (e.g. eDn∇n) related to the carrier density
gradient. In the steady state, the local carrier generation G and recombination U
keep conservation with their output in form of current. U depends on the extra
carriers, np− n0p0. For electrons (i.e. the minority carriers) in the p region (np),
U ≈ (np − np0)/τn, where τn is their lifetime; Dn∇2np ≈ (np − np0)/τn gives an
effective diffusion length before recombination, Ln =√
Dnτn. It is similar for the
minority holes in the n region (pn), i.e. Lp =√
Dpτp.
Jn = enµnE + eDn∇n, ∇ · Jn/(−e) = G−R
Jp = epµpE − eDp∇p, ∇ · Jp/e = G−R (4.2)
Fig. 4.1 shows the principle of a p-n junction solar cell. In darkness (G = 0), in
zero bias (flat Fermi level), the carrier diffusion, e.g. electrons from the n region to
the p region, keep equilibrium with the carrier drift in opposite directions. Under
a forward bias V , the built-in potential becomes lower; the diffusion is enhanced
while the drift is reduced, contributing to a total current Jtotal,
Jtotal = Jn + Jp = J0(eqV/kBT − 1)
J0 = e(np0
√Dn
τn
+ pn0
√Dp
τp
) (4.3)
Under light, there are extra minority photocarriers in both p and n regions, Gτn
and Gτp. They will diffuse to the depletion region and drift to the other sides.
Their diffusion lengths Ln and Lp, plus the width of the depletion region W ,
are the total area for effective photocarrier extraction, producing a photocurrent
JL ∼ qG(Lp + W + Ln). The total current is J = J0(eqV/kBT − 1)− JL. For direct
band semiconductors like GaAs, the dominant recombination in the depletion
28 4 InAs QD-embedded solar cells
region leads to a total current J = J0(eqV/nkBT − 1)− JL where the ideality factor
n > 1.
EC
EF V
free electron diffusion
free hole
Photocarrier
drift Photocarrier
diffusion
EV W
Depletion
region
npn0 p0
free hole
diffusion
pn0 p0
W
p =n 2/nnp0=ni
2/p0
Lp
Ln
ln(n),
ln(p)
p =n 2/nnp0=ni
2/p0
Lp
Ln
Gτp
Gτn
pn0=ni2/n0
Jtotal
pn0=ni2/n0
Jp Jn
0
Jp Jn
JL
0
In darkness
Under light
In darkness
Figure 4.1 Band structure, carrier and current densities in darkness or light.
The lifetime τ and the mobility µ of the minority carriers are two essential
parameters to determine the performance of a p-n junction. µ is degraded by
dopant ion scattering; τ is inversely dependent on the majority carrier density.
Low doping levels will keep high mobility and long lifetime. Since µp is much
lower than µn, most single junction solar cells use a thin n layer as the top emitter
for hole diffusion and a thick p layer as the base for electron diffusion. The emitter
is relatively heavily doped to keep a strong built-in field. In my solar cells, as seen
in Fig. 4.2, low doped p-i-n structure is applied: the p region in 2×1017 cm−3; the
n region in 2× 1016 cm−3; the i region intrinsic. The µn and µp are 5000 and 300
cm2/(Vs) respectively; estimating τ ∼ ns, the diffusion lengths Ln and Lp will be
3.6 and 0.9 µm. But, limited by the n+ GaAs substrate, our solar cells use p-GaAs
as the emitter and n-GaAs as the base. The three layers are in thickness of 500
nm (p region), 140 nm (i region) and 1860 nm (n region). The built-in field is ∼
4.1 Design 29
45000 V/cm in the i region; the depletion region is in width of 400 nm, extending
into the n region by 220 nm and the p region by 40 nm.
Outside the p-i-n structure, there is a 500 nm-thick n+ : 1×1018 cm−3-doped
GaAs layer and p+ Al0.85Ga0.15As layer as the top window and n+ Al0.2Ga0.8As
layer as the back-surface field. The AlGaAs layers, both in thickness of 100 nm,
promote the electron extraction from the back electrode and the hole extraction
from the top electrode while prevent the opposite carrier extraction, and thus
reduce the surface recombination. In the top, a 20 nm p+ GaAs doped at 2×1018
cm−3 is used as the top electrode.
Al0.85Ga0.15As 100nm (Be:1E18)
GaAs 20nm (Be: 1E18)
Au-alloy electrode
GaAs 140nm (undoping) GaAs 80nm (undoping)
light
GaAs 0.5µm (Be:2E17)
GaAs 1.86µm (Si: 2E16)
GaAs 140nm (undoping)
GaAs 360nm (Si: 2E16)
GaAs 488nm (Si: 2E16)
GaAs 140nm (undoping) GaAs 80nm (undoping)
5××××InAs QD (undoped) spaced by GaAs 20nm
Top 2 layers are undoped, GaAs 1.86µm (Si: 2E16)
GaAs 0.5µm (Si:1E18)
Al0.2Ga0.8As 100nm (Si:1E18)
GaAs 488nm (Si: 2E16)GaAs 1.78µm (Si: 2E16)
5××××InAs QD 2.3ML (undoped) spaced by GaAs 20nm (doped)
Top 2 layers are undoped,
i.e. 80+20××××2=140nm
Al0.2Ga0.8As 100nm (Si:1E18)
GaAs buffer 300nm (Si:1E18)
GaAs substrate (Si:1.6E18)
GaAs 100nm (Si: 2E16)
BC
A
Figure 4.2 Solar cell structures with or without QDs embedded. Taken from Paper1. Reprinted with permission of Springer.
To study the QD performance on PV conversion and the built-in field role on
PC production from QDs, we use five 20 nm-GaAs spaced InAs QD layers as a QD
matrix (mentioned in Chapter 3) and design three samples: sample A with one
QD matrix in the built-in field; sample B with three separated QD matrices in the
n region, away from the built-in field; sample C has no QD as the control sample,
as shown in Fig. 4.2. Each region is well designed to keep the same thickness
between samples. For example, in sample A, the top two QD layers (QD height
∼ 8 nm) are spaced by 20 nm undoped GaAs, so the total thickness of the i
region is 80 + (20 + 8) × 2 ≈ 140 nm; in sample B, the 360 nm n-GaAs keeps
QD layers away from the built-in field and the total thickness of the n region is
30 4 InAs QD-embedded solar cells
360 + 488× 2 + 100 + (20 + 8)× 15 ≈ 1860 nm. Sample B is also used to study
the QD intermediate band (IB) absorption where IBs must be flat and partially
filled [51].
Samples’ one dimensional (1D) band structures are presented in Fig. 4.3. In
sample A, QDs are empty of free carriers because the built-in field avoids carrier
occupation in QDs. In sample B, electrons from the surrounding n-GaAs host fill
in QDs, inducing a potential variation around each QD matrix. The QD states
are not completely filled. The band structures are still valid under solar light (100
mW/cm2, 3.6× 1017 photons/(s cm2)), since the photocarrier density Gτ ∼ 1012
cm−3 is much less than the free carrier density (e.g. 2× 1016 cm−3).
-2
-1
0
1
2
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5-2
-1
0
1
2
-2
-1
0
1
2
(a) sample A
VB
CB
5 InAs QD layers
n+-AlGaAs
p+-AlGaAs
Fermi level
z axis [µm]
(c) sample C
Electron energy [eV]
(b) sample B
Figure 4.3 1D energy band structures. Each QD layer is equivalent to 0.7 nmInAs quantum well, with δ function-like densities of states featuring the 3D discrete-ness of QD levels. The 1D Schrodinger equation and Poisson equation are solvedself-consistently [52,53]. Taken from Paper 1. Reprinted with permission ofSpringer.
4.2 Fabrication
The three samples were grown on n+ GaAs substrates by solid-source MBE. Before
growing the solar cell structures, a 300 nm n+ GaAs was grown as buffer. Except
the growth of QD matrices at 500 ◦C, the other structures were grown at 600
4.3 Photovoltaic I-V measurements 31
◦C. The QD growth has been mentioned in Chapter 3. The donor dopant is Si
and the acceptor dopant is Be. After the sample growth, the top contact was
fabricated by evaporating Au and Cr and metalizing them into alloy. The circular
window is optical lithographed and etched for the light incidence. Its diameter,
defined by the lithography pattern, is 2 mm for sample B and C and 6 mm for A.
Then, the samples were thinned down from the back side to 130 µm to fabricate
the AuGeNi-alloy back contact. Both Ohmic contacts tunnel the corresponding
carriers from the semiconductor to the contacts, as seen in Fig. 4.4(a). The back
electrode was mounted with indium on an Al-alloy heat sink. The heat sink has
an insulated base for the top electrode bonding with gold wires and extension with
metal wires. The device is schematized in Fig. 4.4(b). There is no antireflection
coating layers or the top grid electrode in our samples.
Ф = 5.1 eVФ = 4.8
vacuum (a)
Ф = 5.1 eVФ = 4.8
(4.5~5.1)χ = 4.07
Eg = 1.42CB
VB
ψBn=1.03
ψBp= 0.69
AuGeNi
n-contact
CrAu
p-contactn+ GaAs
substraten- GaAs p GaAs
VB
Sample (b)
Gold wire bonding
Al-alloy heat sink
(back contact)
Indium
mounting Metal wire
Sample
Circular light
window
(b)
mounting Metal wire
(top contact)
Figure 4.4 (a) Band structure of Ohmic contacts; (b) device schematic.
4.3 Photovoltaic I-V measurements
The results of photovoltaic (PV) I-V measurements are shown in Fig. 4.5. First
of all, the efficiencies (ηmax) of all three samples, 3∼5 %, are much lower than
the record of single junction GaAs solar cells (28 %). It is due to the un-optimal
designs, such as no top grid electrode, no antireflection coating layer, no effective
surface passivation, defects in material and the structure limit (i.e., p-GaAs as the
32 4 InAs QD-embedded solar cells
0.0 0.2 0.4 0.6 0.80.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
0.0 0.2 0.4 0.6 0.80
2
4
6
8
10
12
14
Current [m
A]
External bias [V]
D
A: D=0.6 cm
C: D=0.2 cm
B: D=0.2 cm
Current density [mA/cm
2]
A: η=4.3%FF=0.63
B: η=3.1%FF=0.69
C: η=5.3%FF=0.67
Figure 4.5 PV measurements under Xenon-lamp solar simulator whose intensity hasbeen adjusted to 100 mW/cm2. D: diameter of the light window. Taken from Paper1. Reprinted with permission of Springer.
top emitter and n-GaAs as the base). We do not care these low efficiencies since
our interest is to study QD role in PV conversion.
Compared to the control sample C (Voc = 0.74 V, Jsc = 10.1 mA/cm2 and ηmax
= 5.3 %), QD samples A and B behave differently. The Voc of sample A reduces
to 0.50 V while its Jsc increases to 13.5 mA/cm2; the Voc and Jsc of sample B
reduce to 0.62 V and 7.4 mA/cm2. The ηmax of sample B is the lowest.
The decrease of Voc is common for most QD solar cells [19,22,51,54,55] and at-
tributed to dislocations in multiple closely stacked QD layers [19,51]. Our sample
A has 5 layers of QDs while sample B has 15 layers. If the dislocations are the
reason, more dislocations and larger decrease of Voc will be expected in sample B.
But, the reduced Voc of sample B is still larger than that of sample A. So, apart
from dislocations, there are more essential reasons for the degradation of Voc.
The 1/3 increase of Jsc in QD sample A is interesting here. The increase of Jsc
in QD solar cells has been reported before [55,56] and attributed to the improved
additional sub-bandgap absorption in QDs. While, some others [19,54] reported
degraded Jsc as well as Voc in QD solar cells. On the theoretical aspect, both
multiple exciton generation effect and intermediate band effect are expected to
enhance Jsc. These conflicting results confuse the understanding of QD roles on
PC production.
A drawback here is that the comparison cannot avoid the factor of disloca-
tions. The strain-induced InAs QD growth is flexible. If the growth is not optimal,
many dislocations will be introduced; then, both Jsc and Voc will degrade, for ex-
ample, QD solar cells with 10 nm-GaAs spaced multiple QD layers [19,51]. If the
4.3 Photovoltaic I-V measurements 33
GaAs spacer is proper in thickness (20∼50 nm) and the QD layer number is few
(e.g. 5), there will be not so many dislocations.
Apart from dislocations, the comparison of Jsc makes sense only when the
solar cell structure especially the depletion region keeps the same. However, most
previous works just embedded multiple QD layers in a p-i-n structure and did not
keep the same width of each region. For example, Martı et al [51] put 10, 20 and 50
layers of QDs with 10 nm GaAs spacers in the i region respectively; Zhou et al [54]
embedded the same 5 layers of QDs in the p, i and n regions respectively, without
considering the width increase in the region and its influence on the other regions;
Sablon et al [56] applied a Si:δ-doping layer with different doping levels in each 50
nm-thick GaAs spacer of the 20 layers of QDs, without considering the built-in
field change by these Si dopant ions. My samples here keep the same width of each
region. The comparison between them will indicate the QD performance. In next
section, I will present the photocurrent spectra and determine which wavelength
of light absorption is responsible for the enhanced Jsc.
Let us return to the reduced Voc. It can be analyzed in the viewpoint of
the p-n diode performance. The diode performance is extracted from the PV I-V
curves by plotting Jsc − J ∼ V , as shown in Fig. 4.6. In the semi-log plot, the
linear region near Voc will give the J0 and the ideality factor n, characterizing the
drastic reduce of photocurrent near Voc.
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0.01
0.1
1
10
100
J0 (estimated in the line-fitted region):
A: 6.60; B: 0.353; C: 0.572 [µΑ/cm2]
nC = 3.02
nB = 2.34
Sample A
Sample B
Sample C
Current density [mA/cm2]
Voltage [V]
J = J0(eqV/nkT
-1)
nA = 2.53
Figure 4.6 Diode performances of three solar cells. Taken from Paper 1.Reprinted with permission of Springer.
The ideality factor n is 3.02 for sample C, 2.53 for A and 2.34 for B. The
reduced n of QD samples A and B is because of a dominant SRH recombination,
induced by QDs and/or dislocations. The n is near 2 when this recombination is
34 4 InAs QD-embedded solar cells
dominant. Sample C has no QDs and few dislocations; the direct recombination
is dominant, giving a high n.
The main difference lies in J0. It is 6.60 µA/cm2 for sample A, 0.353 µA/cm2
for B and 0.572 µA/cm2 for C. The one-order of magnitude increase of J0 in sample
A, according to Eq. 4.3, is due to a large decrease of photocarrier lifetime. The
QD confinement increases the chance that electrons and holes meet and recombine
with each other, enhancing the recombination current in the depletion region. The
slightly reduced J0 in sample B is due to the potential variations around QDs (see
Fig. 4.3). QDs with electron filling, served as Coulomb scattering centers, will
reduce the carrier mobility in the n region, as well as J0 and Jsc. The same scale
of J0 in samples B and C indicates the similar depletion region and the same scale
of photocarrier lifetime.
The Voc can be deduced when J = 0, i.e.,
Voc =nkBT
qln(
Jsc
J0
+ 1) (4.4)
The Jsc of all samples are in the same scale, so the reduced Voc of sample A
is related to the large increase of J0, i.e. the reduced photocarrier lifetime. In
sample B, both J0 and Jsc are reduced and their ratio Jsc/J0 is little changed, so
the reduced Voc of sample B is because its n is most reduced, related to the most
enhanced SRH recombination.
4.4 Photocurrent spectra
The photocurrent (PC) spectra are presented here, from which we can find the
contribution in different wavelengths and the reason for the increased Jsc. The
300∼1000 nm PC spectra are measured by a monochromator and a Si reference
photodiode under a Xenon lamp, as shown in Fig. 4.7. The sub-GaAs-bandgap
PC spectra are measured by a Fourier Transform (FT) spectrometer, as presented
in Fig. 4.9.
In Fig. 4.7, the PCs of samples (Isample(λ)) and the reference Si diode (ISi(λ))
are recorded at the same time. The illumination area of samples is fixed to
be the same by an aperture with the area of 0.32 cm2. The relative intensity
Isample(λ)/ISi(λ) spectra, as shown in the inset, reflect the external quantum ef-
ficiency (EQE) of samples, EQEsample(λ), since the EQE of the Si diode is little
changed in 500∼900 nm [1]. For sample A, the >900 nm spectrum is clearly
observed, due to QD absorption.
4.4 Photocurrent spectra 35
300 400 500 600 700 800 900 1000
0
1
2
3
4
5
400 500 600 700 800 900 10000
1
2
3
4
5
Photocurrent [10-5 A]
Wavelength [nm]
A
BC
Si
Relative intensity A C
B
Figure 4.7 300 ∼ 1000 nm PC spectra. Inset: relative intensity Isample(λ)/ISi(λ).Taken from Paper 2. Reprinted with permission of American Institute ofPhysics.
The PC spectral integrations in the dominant 300∼1000 nm region agree with
the 1/3 raise (decrease) of Jsc in sample A (sample B) compared to sample C in the
PV measurement, so the Isample(λ) of each sample can be compared directly. In
sample B, the lowest Isample(λ) is because the potential variation around each QD
matrix reduces the carrier mobility in the n region. Besides, the most dislocations
and SRH recombination are also the reasons.
In sample A, the enhanced above-GaAs-bandgap PC spectrum is responsible
for the increased Jsc. There are several possible reasons for it. One is the multiple
exciton generation (MEG) in QDs. Since sample A has a 80 nm-thick undoped
GaAs layer above the QD matrix, the photoelectrons in the p region will drift
through this layer and accelerate without dopant ion scattering; when they reach
QDs, these electrons with high kinetic energy will impact ionize electron-hole pairs
and produce more photocarriers. But, the built-in field in our samples cannot
increase the photoelectron kinetic energy to twice the QD ground-state transition
energy (∼1 eV) that is believed to be the threshold for the MEG [57].
We think the above-bandgap photon absorption is enhanced after embed-
ding QDs. In the depletion region, most photocarriers move at the drift velocity.
QDs will reflect electron and hole wavefunctions and increase their overlap and
transition (i.e. absorption) rate.
We make a simulation in Fig. 4.8. In the depletion region in width of W ,
the five QD layers are modeled as five 1D quantum wells with thickness of 0.7
nm; the electron and hole wavefunctions are modeled as 1D free particles, e±ikz,
incident into the QD matrix from opposite directions [58]. The k characterizes
36 4 InAs QD-embedded solar cells
0
1
2
0
1
2
VB
CB
QD QD QD QD QD
transmission
amplitude of CB electron
incidencereflection
reflection
transmission
amplitude of VB hole
incidence
Figure 4.8 Wavefunctions of CB electron and VB hole in the depletion region withQDs embedded (solid) and without QDs (dotted). ~ω = 1.593 eV is chosen to showa complete reflection of the hole wavefunction. Taken from Paper 2. Reprintedwith permission of American Institute of Physics.
the drift velocity. QDs will reflect and transmit the carrier wavefunctions, giving
components re±ikz and te±ikz. The hole wavefunction is reflected serious due to
its large effective mass. It is completely reflected when the transition energy is
1.593 eV. If there is no QD, the wavefunction overlap is
WC = 〈eikez|e−ikhz〉 =
∫
W
ei(−ke−kh)zdz (4.5)
If there are QDs, assuming a complete hole reflection and a complete electron
transmission, the wavefunction overlap is
WA = 〈eikez|e−ikhz + eikhz〉 =
∫
W
ei(−ke−kh)zdz +
∫
W
ei(−ke+kh)zdz (4.6)
Since the optical absorption in GaAs is mainly direct transition, i.e., ke = kh, the
WC oscillates while the second term of WA gives a large contribution to PC.
This enhancement is too large as compared with the experiments, since the 1D
quantum wells exaggerate the wavefunction reflection in 3D QDs; not all-energy
holes are completely reflected. Besides, the direct absorption in GaAs creates
electron and hole moving in the same direction and one of them will re-orientate
under the built-in field in ∼10 ps (~k/eE), contributing to the PC as well.
The FT-based PC spectra are shown in Fig. 4.9. Compared to sample C,
both QD samples A and B show sub-GaAs-bandgap spectral features, including a
clear wetting-layer peak near 920 nm and QD features extended in 950∼1300 nm.
To clarify the sub-bandgap PC spectra, Fig. 4.9(b) uses a GaAs-filter to filter the
4.5 Mechanisms of photocarrier escape from QDs 37
incident QHT light. There are multiple QD peaks, due to the size distribution
and/or the energy state coupling in closely spaced QD layers.
The sub-bandgap PC spectrum of sample B is about 30 times lower than that
of sample A, even for the wetting-layer peak. It indicates that the photocarrier
extraction from QDs in the depletion region is effective while that from QDs in
the n region is weak. The QDs in the n region are partially filled with electrons,
so the QD interband absorption is suppressed. Besides, the potential variation
around each QD matrix reduces the carrier mobilities (see the spectra backgrounds
nB < nC) and increases the thermal activation energy for the minority photohole
extraction from QDs. In addition, if the intermediate band absorption exists in the
partially filled and coupled QDs, its PC production via the minority photocarrier
diffusion (Lp ∼ 0.9µm) is still quite lower in efficiency than the built-in field
extraction and drift.
900 1000 1100 1200 1300
0.00
0.05
0.10
0.15
0.30
0.35
0.40
0.45
400 600 800 1000 120010
-5
10-4
10-3
10-2
10-1
100
WL
Wavelength [nm]
Photocurrent intensity [arb unit]
1
(b)
AB(x30)
4
6
5
3
7
2
A
B
C
(a) WL
nA
nC
nB
Figure 4.9 PC spectra measured by a FT infrared spectrometer under a quartz halo-gen tungsten lamp (a) and GaAs-filtered light (b). Taken from Paper 2. Reprintedwith permission of American Institute of Physics.
4.5 Mechanisms of photocarrier escape from QDs
The photocarrier extraction from QDs is essential for QD applications in pho-
todiodes, photodetectors and solar cells. Two mechanisms have been proposed,
tunneling and thermal activation [59,60]. In our QD sample A, the strong built-in
field enables photocarrier tunneling out from QDs to GaAs. The thermal activa-
tion is also enhanced due to the reduced barrier height under the field, i.e. the
Poole-Frenkel effect [61]. In our QD sample B, the photocarrier extraction only
38 4 InAs QD-embedded solar cells
relies on thermal activation. By analyzing their bias- and temperature-dependent
PC spectra, I find that sample A has an inter-QD-layer tunneling and sample B
is dominated by the hole thermal escape.
I first position the energy of each PC peak, by combining with the Gaussian-
line fitted PL spectra in Fig. 4.10. There are seven QD peaks and two wetting-layer
(WL) peaks (dashed Lorentzian lines). These densely (50 meV) packed QD peaks,
unlike the usual ∼70 meV separation for uncoupled QDs [59,60], are caused by the
QD size distribution and the energy state coupling between adjacent QD layers.
The non-uniform QD height, discrete by a step of atomic monolayer (ML), will
give a stepwise change of the QD state energy. It is profound on small QDs when
the QD height is as low as 3 nm [62,63]. The QD1 to QD3 in our samples may
be from the non-uniform small QDs. This size distribution in large QDs only
induces a continuous modification of QD state energy, corresponding to the QD4
to QD6 in our samples. QD6* in sample A, separated from QD6 by 36 meV in
PL spectrum and 43 meV in PC spectrum, is possible due to a coupling between
the QD6 exciton and the LO phonon [64]. The two WL peaks, separated by ∼20
meV, are from the non-uniform wetting layer thickness [65].
0.9 1.0 1.1 1.2 1.3 1.4
0.9 1.0 1.1 1.2 1.3
0.9 1.0 1.1 1.2 1.3
WL1
WL2
QD1QD2
QD3QD4
QD5QD6
PC
(a) Sample A
QD6*
no-filter
WL1
WL2
QD1QD2
QD3
QD4QD5PC
(b) Sample B
Photon energy [eV]
QD6
filter
QD2
QD3
QD4
QD5PL
QD6
QD4
QD5
QD6PL
QD6*
Figure 4.10 Room temperature PC and PL spectra. Each peak position is marked.
4.5 Mechanisms of photocarrier escape from QDs 39
The bias-dependent PC spectra of QD samples A and B and the bias depen-
dence of each peak are presented in Fig. 4.11. As the reverse bias increases, the
PC intensity enhances. WL1 becomes increasingly important as compared with
WL2, because it is closer to the GaAs bandedge. A ∼2 meV red-shift of WL2
is observed in sample A when the bias scans from 0.5 V to -1.0 V due to the
quantum Stark effect [66]. QD peaks show no shift when the bias changes.
For sample A, as the external bias changes from 0.5 to -1 V, the PCs from QDs
and WLs first increase rapidly, then increase slowly and finally become saturated,
similar to the behavior at the GaAs bandedge. It implies that the photocarrier
extraction from WLs and QDs is effective. For sample B, the PC contributions
from WLs and QDs are linearly dependent on the external bias from 0.2 V to
-0.4 V, due to the extension of the built-in field and the change of the minority
hole diffusion. The linear bias dependence is very different from that at the GaAs
bandedge, implying that the photocarrier extraction from WLs and QDs in sample
B is very poor.
In detail, the bias dependence of each PC peak shows fine features. For sample
A, the bias dependences of GaAs, WL1, WL2, QD1 to QD3 are similar while QD4
to QD6 and QD6* show stronger bias dependence. The lower the energy is, the
stronger the bias dependence is, indicating a tunneling mechanism for photocarrier
escape from QDs [67,68]. For uncoupled QDs, e.g. single QD layer or multiple QD
layers with GaAs spacers thicker than 40 nm, the photocarrier tunneling escape
from deep QD levels is believed to be exponentially dependent on the external
bias [68]; however, the deep QD peaks here such as QD6 show bias dependence in
similar trend with the PC from GaAs, suggesting a significant interlayer tunneling
between the 20 nm GaAs spaced QD layers. Closely stacked QD layers will form
vertically aligned QDs and vertical state coupling. Photocarriers in deep QD
states in one QD layer can tunnel into adjacent QD layers and finally escape, as
schematized in Fig. 4.12. For sample B, all QD peaks show an identical linear
bias dependence, weaker than the linear bias dependence of WL peaks, because
the photocarrier thermal activation from WL is larger than that from QD states
and photocarriers cannot escape from QDs via inter-QD-layer tunneling.
Then, I measured the temperature-dependent PC spectra of sample B, which
are presented in Fig. 4.13(a). Below 200 K, there is no PC spectrum (see the 200 K
grey line). From 250 K to 295 K, the PC intensity increases. The position of each
QD peak is almost unchanged; while WL peaks are red-shifted. WL2 is shifted by
20 meV and WL1 is shifted from the filter region of the room-temperature GaAs
filter to 1.368 eV at 295 K.
40 4 InAs QD-embedded solar cells
0.000
0.002
0.004
0.006
0.008
0.010
0.012
0.9 1.0 1.1 1.2 1.3 1.40.00
0.02
0.04
0.06
0.08
0.10
0.01
0.02
0.03
0.04
0.12
0.16
0.20
1.0
1.2
1.4
1.6
1.8
2.0
-0.8 -0.4 0.0 0.4
1.0
1.2
1.4
1.6
y-shift:
0.0002 / line
QD1
QD2
QD3
QD4
QD5
QD6
(a) Sample A
Room temperature,
104 V/A
Photocurrent
QD6*
-1.0, -0.8, -0.6, -0.4, -0.2,
0.0, 0.2, 0.4, 0.5V
-0.4, -0.2, 0.0, 0.2V
y-shift:
0.003 / line
WL1
WL2
QD1
QD2QD3
QD4QD5
QD6
Photon energy [eV]
(b) Sample B
Room temperature,
107 V/A
y-shift:
0.0015 / line
WL2WL1
Relative PC intensity
GaAs
WL1
WL2
QD1
QD2
QD3
QD4
QD5
QD6
QD6*
(c) sample A
(d) sample B
WL1
WL2
QD1
QD2
QD3
QD4
Bias [V]
Figure 4.11 Bias-dependent PC spectra at room temperature. (a) Sample A underQHT light. External current amplifier gain of 104 V/A, from -1.0 V to 0.5 V with 0V in red and 0.5 V in blue for clarity. The QD range (0.9∼1.3 eV) is shown in theleft, vertically shifted by 0.0002/line; the WL and GaAs bandedge range (1.3∼1.44 eV)is shown in the right, in a large vertical scale and vertically shifted by 0.0015/line.(b) Sample B under GaAs-filtered QHT light. Gain 107 V/A, from -0.4 V to 0.2 V,vertically shifted by 0.003/line. (c) and (d) compare the relative bias dependence ofeach peak, with the intensity accumulated in its width and scaled to the intensity at0.5 V for A and 0.2 V for B.
thermal activation
interlayer tunneling
QD-stateWL-state
WL-state
QD-state
(a) thermal activation
tunneling
tunneling
thermal activation
(b)
VB
QD state
QD state
QD state
QD state
CB
Figure 4.12 Interlayer tunneling from closely stacked QDs (a) compared to the ordi-nary tunneling from far spaced QDs (b).
4.5 Mechanisms of photocarrier escape from QDs 41
The temperature dependence of each peak intensity, as shown in Fig. 4.13(b),
is exponentially dependent on 1/kBT , indicating the thermal activation e−Ea/kBT
(Ea: activation energy). The deduced Ea of each peak is larger than 210 meV,
even for WL2, much larger than the result of a single layer of QDs in the depletion
region [60]. The large Ea suggests that the thermal activation from QDs is dom-
inated by the minority photoholes, because the potential variation around each
QD matrix increases the barrier for photohole escape from QDs. Therefore, the
Ea characterizes the offset between the QD hole-state and the flat GaAs valence
bandedge. Based on Ea and the energy (Eph) of each PC peak, the QD matrix’s
energy-level diagram is reconstructed, as seen in Fig. 4.13(c).
1.0 1.1 1.2 1.3 1.4
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
1/3001/290
1/2801/270
1/2601/250
1/240
10-4
10-3
10-2
10-1
1.0 1.1 1.2 1.3
0.0
0.4
0.8
1.2
1.6
2.0
1200 1300 1400 1500
0.0
0.2
0.4
1.0
1.2
1.4
1.6
295 K
290 K
285 K
280 K
275 K
260 K
250 K
200 K
Photo
current
Photon energy [eV]
WL2
(a) (b)
236.5
236.5
253.7
QD1-3
QD5, 296.7
QD4, 275.2
QD6, 322.5
1/T [K -1]
WL2
Ea=210.7 meV
QD4
QD5
x 100
QD6
Ea
Eph
QD6
QD6
WL2 (c)
Energ
y [eV]
Position [nm]
WL2QD1-QD3
Figure 4.13 (a) Temperature-dependent PC spectra of sample B (unbiased), with QDcontributions zoomed in the inset; (b) Temperature dependence and activation energyEa of each peak; (c) Reconstructed energy level diagram of the QD matrix.
The photohole extraction also dominates the PC production from QDs in the
depletion region [59,60]. That is why in closely stacked QDs when the temperature
varies from 300 K to 20 K the PC spectral intensity shows negligible temperature
dependence; while in uncoupled QDs the PC intensity of each QD peak, especially
the ground QD peak, is very sensitive on the temperature [59].
Chapter 5
Bunched correlation of two
colloidal QD blinking
In this chapter I study another subject, colloidal QD blinking. Colloidal QDs
(e.g. CdSe QDs) have big potentials on applications of photon emission such as
bioimaging [69] and LED [70], and light absorption such as QD solar cells with QDs
replacing the dye molecules in dye-sensitized solar cells [71]. For light emission,
a colloidal QD has an intrinsic drawback: its emission is intermittent and often
switches between the on and off states. The fluorescence blinking is related to
the surface traps, i.e., dangling bonds in the QD surface. Under light excitation,
electron-hole pairs are created in a QD. They will relax to the ground QD states
and recombine radiatively to emit photons (i.e. the on state). Sometimes, a pho-
tocarrier will escape from QD and occupy the surface state. The left photocarrier
in the QD, as a charge, will suppresses the bright exciton formation if there is
another electron-hole pair excited subsequently. Now, electrons and holes will
recombine non-radiatively (i.e. the off state) until the trapped carrier returns to
the QD. Fig. 5.1 gives an example of the light intensity evolution of colloidal QD
blinking. The blinking seems random. But, the previous studies have shown some
law. By introducing a threshold above which the intensity is defined to be 1 (the
on state) while below which the intensity is defined to be 0 (the off state), the
statistical probability distribution of the on/off state shows an inverse power law
on the on/off time [72,73]. This law is valid for the time from 1 ms to several 100
s, implying a multi-time constant process. From Fig. 5.1 we also find that the
multi-shell QD light intensity is higher and more steady than the single-shell QD
light intensity, because the multi-shell is thick for photocarrier escape and blinking
occurring. In this chapter, I found that the light intensity from two individual
42
5.1 QD fluorescence imaging system 43
blinkingfull QDs with distance less than 1 µm tend to have a time correlation, i.e.
bunched photon emission, due to their mutually stimulated emission.
0
500
1000
1500
2000
0 100 200 300 400 500 600 700 800 900 10000
2500
5000
7500
10000
12500
Light inensity [counts/frame]
CdSe/ZnS core/shell water-soluble QD
CdSe/CdS/Zn0.5Cd
0.5S/ZnS core/multi-shell oil-soluble QD
Time frames [50 ms/frame]
Figure 5.1 Colloidal QD blinking, 1000 frames, 50 ms/frame. Top: a single-shell QD;Bottom: a multi-shell QD.
5.1 QD fluorescence imaging system
The colloidal CdSe QD solution is diluted into a concentration less than 10−9
mol/L and deposited on a clean glass. The samples are dried in air and imaged
under a microscope under a mercury lamp (λ = 436nm) excitation. The imaging
system includes a 100× oil-immersed objective (NA=1.45) and a camera. The
camera can record 128×128 pixel movies in 50 ms per frame and 64×64 pixel
movies in 5 ms per frame. The camera scene covers single, double or several
QDs, and records their blinking for a long time (1000 frames). A typical blinking
movie is shown in Fig. 5.2. Each QD has a light spot whose maximal intensity is
extracted as the QD blinking intensity, I(t), as a function of the time frame.
5.2 Correlation analysis
The time correlation function of two QD blinking is defined to be
gAB(τ) =<IA(t)IB(t + τ)>t
<IA(t)>t · <IB(t + τ)>t
(5.1)
where < . . . >t is the average on the time; τ is the delay between the two QD
signals. The unit of τ is a frame, i.e. 50 ms. The following correlation analysis
is valid in this time resolution. If the two QD blinking affect each other and tend
to emit photons synchronously (or bunched), the correlation function will give a
44 5 Bunched correlation of two colloidal QD blinking
0.72
Figure 5.2 A frame of CdSe QD blinking movie that includes 1000 frames in 50ms/frame. QDs A and B are 0.72 µm distant. Their light spot diameters are 7 pixelsat most.
peak at τ = 0. If their blinking affect each other and emit photons mutually
exclusively (antibunched), gAB(τ) will give a dip at τ = 0. If their blinking have
no correlation, gAB(τ) will be flat around τ = 0. The time correlation function is
often used to decide if there is a single photon emission [74,75,80] or a two-photon
emission [76,77] or a two-photon absorption [78], with the delay time smaller than
the emission time (∼ ns) and absorption time (∼ ps) and to characterize the
shape of an ultra-short (∼ fs) light pulse with an optically nonlinear crystal [79].
In principle, photon emission is a trigger process in time resolution smaller than
the emission lifetime. Here, we use correlation analysis to study the QD blinking
emission since it is a multi-time constant process and its dominant time constant
is possible comparable with 50 ms. For most multi-shell QDs with steady light
emission and few blinking observed (see Fig. 5.1), in order to record the real-time
blinking emission, the frame time must be much reduced. Furthermore, if the
above correlation is caused by the blinking itself, the steady emission with few
blinking will give no correlation.
In fact, our statistical correlation analysis is performed on 1000 frames rather
than infinite time; each frame records the total light intensity in 50 ms rather
than a real-time photon-number counting in resolution smaller than the intrinsic
photon emission lifetime (∼ ns), so gAB(τ) has a random fluctuation. In a large τ
range (e.g. ±200 frames = ±10 s), the fluctuation is serious, providing no valuable
information. We also deny the existence of a long-time correlation between two
QD blinking. In a local region around τ = 0 (e.g. ±5 frames = ±0.25 s), the fluc-
tuation is small. We use a scheme to compare gAB(0) and the surrounding gAB(τ)
5.2 Correlation analysis 45
and decide if there is a correlation: analyzing the average g and the standard
derivation σ of the surrounding 10 points of gAB(τ) where τ=±1, ±2, ±3, ±4 and
±5); if gAB(0) > g + σ, we believe gAB(0) is a peak (bunching); if gAB(0) < g−σ,
gAB(0) is a dip (anti-bunching); if gAB(0) is in the g±σ range, we believe it is flat
(no correlation). The decision will be more reliable if the tolerance range increases
to ±2σ, i.e., the bunched correlation must satisfy gAB(0) > g + 2σ.
Another problem is that for closely (e.g. 0.5 µm) spaced two QDs, their light
spots overlap and affect the independent acquiring of light intensity from each
QD. In this case, the above correlation will contain some contribution from the
autocorrelation of each QD that is defined to be
gA(τ) =< IA(t)IA(t + τ) >t
< IA(t) >t · < IA(t + τ) >t
(5.2)
For example, if the recorded IA(t) = a(t)IA(t)+b(t)IB(t) where IA(t) and IB(t) are
the real light intensity from each QD and a(t) and b(t) are contribution factors in
[0, 1], gAB(τ) will have a term from < IB(t)IB(t + τ) >t, i.e. gB(τ). The autocor-
relation function is symmetric with τ and has a peak at τ = 0 (see Fig. 5.3), even
for the background signal in the movie. It cannot conclude a bunched correlation.
-20 0 20 40 60 80 100
1.0
1.1
1.2
1.3
0 200 400 600 800 1000
100
200
300
400
500
600
700
800
0 200 400 600 800 1000
100
120
140
-20 0 20 40 60 80 100
1.000
1.002
1.004
Oil, multi-shell (steady-state)
Water, single-shell (trigger-like)
Autocorrelation g
A(τ)
Light intensity [counts/frame] Trigger-like (QD-A blinking in S32)
Time [x50ms]
Background signal
Background singal
Delay τ [x 50ms]
Figure 5.3 Light intensity and autocorrelation of a trigger-like QD blinking (water-soluble single-shell QD-A in S32) and the background, compared with the oil-solublemulti-shell QD in Fig. 5.1.
The third problem is that the recorded movie includes a noise caused by the
46 5 Bunched correlation of two colloidal QD blinking
dark counting in the camera CCD, I(t) = I(t) + N(t). The background noise
N(t) is random on both time and position. If the real QD light intensity is as
low as N(t), the noise will give a considerable influence on gA(τ) and a random
local-region fluctuation on gAB(τ). In this case, gAB(τ) cannot reflect clearly the
real correlation information of two QDs.
A proper QD blinking sample for correlation analysis must have many obvious
blinking events, i.e., the recorded light intensity must be high trigger-like rather
than steady state-like. Fig. 5.3 shows a trigger-like QD blinking. Its short-time
autocorrelation, i.e., gA(1) is much lower than gA(0), is compared with the auto-
correlation of a steady state-like QD blinking that gradually decays from τ = 0.
The background signal and its autocorrelation are also presented, showing ‘white
noise’ features.
5.3 Statistical result
We made and measured 43 samples of dilute water-soluble CdSe/ZnS single-shell
QDs. Their blinking intensity is shown in Fig. 5.4. The samples grouped together
are actually the same sample measured several times. The other once-measured
samples are also listed together as the QD distance increases from 0.52 to 2.11
µm. Their correlation analysis is presented in Fig. 5.5 and 5.6. In Fig. 5.7,
we compare gAB(0) and the surrounding gAB(τ) values by the above-mentioned
scheme and determine which has bunched correlation. Fig. 5.8 makes a statistics
on the correlated samples as a function of QD distance. We find that for closely
spaced two QDs in distance of 0.7∼ 1.1 µm, gAB(0) is a peak, indicating a bunched
correlation; as the QD distance increases, this phenomenon disappears gradually;
for QDs distant from each other by 0.52 ∼ 0.55 µm, the two QD light spots
overlap obviously, although there is a gAB(0) peak, it cannot conclude a bunched
correlation. I will discuss them in detail below.
For QD distance between 1.33 and 5.78 µm, all the nine samples (N5, N6, N9,
N10, N12, N14, N15, S10 and S15) have no gAB(0) peak definitely. Their gAB(τ)
fluctuate randomly; the same sample measured in different times shows different
profiles, e.g. N5 and N6, N10 and N12, N14 and N15. When the QD distance
is between 1.23 and 1.29 µm, in the six samples (S2, S6, S9, S22, S23 and S24),
there are only two samples, S2 (1.23 µm) and S6 (1.24), satisfying gAB(0) > g+σ,
and only S2 satisfying gAB(0) > g + 2σ.
5.3 Statistical result 47
0
500
1000
1500
2000
2500
3000
0 200 400 600 800 10000
300
600
900
1200
1500
0 200 400 600 800 10000
500
1000
1500
0
500
1000
1500
2000
0
200
400
600
800
1000
0 200 400 600 800 10000
500
1000
1500
0 200 400 600 800 10000
500
1000
1500
2000
0
500
1000
1500
2000
2500
0
300
600
900
1200
1500
0
300
600
900
1200
0 200 400 600 800 10000
100
200
300
400
500
0
200
400
600
0 200 400 600 800 10000
200
400
600
800
1000
0
100
200
300
400
500
0 200 400 600 800 10000
100
200
300
400
500
0 200 400 600 800 10000
100
200
300
0
300
600
900
1200
0
200
400
600
0
200
400
600
0
100
200
300
400
500
0
100
200
300
400
500
0
200
400
600
0 200 400 600 800 10000
100
200
300
400
0
200
400
600
800
0 200 400 600 800 10000
200
400
600
800
0
200
400
600
800
0
200
400
600
0 200 400 600 800 10000
200
400
600
0
200
400
600
800
1000
0
200
400
600
800
1000
0
200
400
600
800
1000
0
200
400
600
800
0 200 400 600 800 10000
200
400
600
800
0
200
400
600
800
0
200
400
600
0 200 400 600 800 10000
200
400
600
800
0
200
400
600
800
0 200 400 600 800 10000
100
200
300
400
500
0
200
400
600
800
0
200
400
600
800
0
200
400
600
800
0
200
400
600
800
0
200
400
600
0 200 400 600 800 10000
200
400
600
N5 ~ N6 Light intensity I(t)
N5 (5.76)
N6 (5.73)
N15 (1.39)
N14 ~ N15N14 (1.33)
N10, N12
N10 (3.98)
N12 (3.98)
N9 (2.11)
N16 (1.09)
Rest: N9, N13, N16, S1,
S2, S5, S8, S11, S15N13 (0.52)
S3 ~ S4
S3 (0.77)
S4 (0.84)
S6 ~ S7
S6 (1.24)
S7 (1.14)
S9 ~ S10
S9 (1.29)
S10 (1.36)
Background of N13
S8 (0.55)
S1 (0.78)
S5 (0.96)
S2 (1.23)
S19 ~ S21
S19 (0.52)
S20 (0.52)
S21 (0.52)
S12 ~ S13S12 (1.14)
S13 (1.14)
S16 ~ S18
S16 (1.16)
S17 (1.16)
Time t [x 50ms]
S18 (1.16)
S22 ~ S24S22 (1.23)
S23 (1.24)
S11 (0.78)
S15 (1.45)
S24 (1.23)
S25 ~ S27
S25 (0.80)
S26 (0.94)
S27 (0.94)
S28 ~ S29
S28 (0.72)
S29 (0.72)
S30 ~ S35S30 (0.82)
S31 (0.82)
S32 (0.72)
S33 (0.72)
S34 (0.72)
S35 (0.72)
Figure 5.4 Two QD blinking intensity as function of time. Samples as a group arethe same sample measured several times. Sample name is, e.g. S5, followed with theQD distance [µm].
48 5 Bunched correlation of two colloidal QD blinking
-20 -10 0 10 200.96
0.98
1.00
1.02
1.04
-20 -10 0 10 20
0.99
1.00
1.01
1.02-20 -10 0 10 20
1.00
1.01
1.02
1.03
1.04
-20 -10 0 10 20
0.96
0.98
1.00
-20 -10 0 10 20
1.00
1.02
1.04
-20 -10 0 10 20
1.04
1.06
1.08
-20 -10 0 10 20
0.98
1.00
1.02
-20 -10 0 10 200.995
1.000
1.005
-20 -10 0 10 20
1.00
1.02
1.04
-20 -10 0 10 200.97
0.98
0.99
1.00
-20 -10 0 10 200.992
0.996
1.000
-20 -10 0 10 20
1.000
1.004
-20 -10 0 10 201.06
1.08
1.10
1.12
-20 -10 0 10 20
1.00
1.05
1.10
1.15
1.20
-20 -10 0 10 201.00
1.01
1.02
1.03
1.04
-20 -10 0 10 20
1.03
1.04
1.05
1.06
-20 -10 0 10 20
1.00
1.02
-20 -10 0 10 20
0.98
1.00
1.02
1.04
-20 -10 0 10 200.9996
1.0000
1.0004
-20 -10 0 10 20
1.03
1.04
-20 -10 0 10 200.99
1.00
-20 -10 0 10 20
1.01
1.02
1.03
N5 (5.76)
Corr
ela
tion g
AB
(τ)
N6 (5.73)
N10 (3.98)
N12 (3.98)
N14 (1.33)
N15 (1.39)
S3 (0.77)
S4 (0.84)
S6 (1.24)
S7 (1.14)
S9 (1.29)
Delay τ [ X 50ms]
S10 (1.36)
N13 (0.52)
S8 (0.55)
S1 (0.78)
S11 (0.78)
S5 (0.96)
N16 (1.09)
Background cross-correlation
S2 (1.23)
S15 (1.45)
N9 (2.11)
Figure 5.5 Correlation functions of samples (the left two columns in Fig. 5.4).
5.3 Statistical result 49
-20 -10 0 10 200.99
1.00
1.01
1.02
-20 -10 0 10 201.00
1.01
-20 -10 0 10 20
1.02
1.04
1.06
-20 -10 0 10 200.99
1.00
-20 -10 0 10 201.00
1.01
-20 -10 0 10 20
1.00
1.01
-20 -10 0 10 201.01
1.02
1.03
1.04
-20 -10 0 10 201.00
1.01
-20 -10 0 10 20
0.94
0.96
0.98
1.00
-20 -10 0 10 200.98
0.99
1.00
-20 -10 0 10 20
1.00
1.01
1.02
-20 -10 0 10 20
1.03
1.04
1.05
-20 -10 0 10 20
1.00
1.01
1.02
-20 -10 0 10 20
0.99
1.00
1.01
-20 -10 0 10 200.99
1.00
1.01
-20 -10 0 10 200.98
0.99
1.00
1.01
-20 -10 0 10 20
1.00
1.01
1.02
-20 -10 0 10 20
0.98
1.00
1.02
-20 -10 0 10 20
1.00
1.01
1.02
-20 -10 0 10 200.98
1.00
1.02
-20 -10 0 10 20
0.99
1.00
1.01
1.02
-20 -10 0 10 20
1.00
1.01
1.02
S12 (1.14)
Correlation g
AB(τ)
S13 (1.14)
S16 (1.16)
S17 (1.16)
S18 (1.16)
S19 (0.52)
S20 (0.52)
S21 (0.52)
S22 (1.23)
S23 (1.24)
S24 (1.23)
S25 (0.80)
S26 (0.94)
S27 (0.94)
S28 (0.72)
S29 (0.72)
S30 (0.82)
S31 (0.82)
S32 (0.72)
S33 (0.72)
S34 (0.72)
S35 (0.72)
Delay τ [ X 50ms]
Figure 5.6 Correlation functions of samples (the right two columns in Fig. 5.4).
50 5 Bunched correlation of two colloidal QD blinking
When the QD distance is in 1.14 ∼ 1.16 µm, all the six samples (S7, S12,
S13, S16, S17 and S18) satisfy the requirement gAB(0) > g +σ and only S7 (1.14)
and S17 (1.16) do not satisfy gAB(0) > g + 2σ. S12, S13, S17 and S18 actually
have a bright QD and a dark QD. The dark QD often stays in the dark state
and has very few blinking triggers that are well resolved in the time line, so the
correlation function is less influenced by the fluctuation mentioned above. The
underline physics here is that the equivalent photon emission lifetime of the dark
QD (i.e., blinking time) is longer than the frame time 50 ms, so the camera CCD
can count the emitted photons in real time, leading to a good correlation analysis
and a clear peak at τ = 0 if there is correlation.
0.98
1.00
1.02
1.12
1.13
Noise
S21 (0.52)
S20 (0.52)
N13 (0.52)
S19 (0.52)
S11 (0.78)
S1 (0.78)
S8 (0.55)
S35 (0.72)
S34 (0.72)
S33 (0.72)
S32 (0.72)
S29 (0.72)
S28 (0.72)
S3 (0.77)
S25 (0.80)
S31 (0.82)
S30 (0.82)
S4 (0.84)
S27 (0.94)
S26 (0.94)
S5 (0.96)
N16 (1.09)
S13 (1.14)
S12 (1.14)
S7 (1.14)
S18 (1.16)
S17 (1.16)
S16 (1.16)
S24 (1.23)
S23 (1.24)
S22 (1.23)
S2 (1.23)
S6 (1.24)
S9 (1.29)
S10 (1.36)
N14 (1.33)
N15 (1.39)
S15 (1.45)
N12 (3.98)
N9 (2.11)
N6 (5.73)
N10 (3.98)
g(2)(0)/<g
(2)(τ≠0)>
1+σ/<g(2)(τ≠0)> 1-σ/<g(2)(τ≠0)>
Samples
N5 (5.76)
Figure 5.7 Correlation decision. The background noise has no correlation.
0
1
2
3
4
5
6
7
8
9
10
5.73-5.78
3.98
2.11
1.33-1.45
Distance [µm]
1.23-1.29
1.14-1.16
0.94-1.09
0.80-0.84
0.72-0.78
0.52-0.55
Number of QD pairs
Distance between paired QDs
QD pairs studied
with bunching, g(0) > <g(τ≠0)>+σ with bunching, g(0) > <g(τ≠0)>+2σ
Figure 5.8 Statistics on QD distance.
5.3 Statistical result 51
When the QD distance is between 0.72 and 1.09 µm, all the 17 samples show
the gAB(0) peak under the requirement gAB(0) > g + σ. Only 4 samples, S5, S29,
S31 and S32, cannot satisfy gAB(0) > g + 2σ. In S32 there is a side peak around
τ = 2 and 3; in S31 the gAB(0) peak is not clear. By comparing the different
measurements of the same sample (S30 to S35, 0.72 ∼ 0.82), we conclude that
the gAB(0) peak always exists although sometimes the background fluctuation is
serious. The same samples (S25, S26 and S27, 0.80 ∼ 0.94) and (S3 and S4, 0.77
∼ 0.84) show the gAB(0) peak definitely. The same sample S28 and S29 sometimes
shows a clear gAB(0) peak (S28) while sometimes not (S29). In all, for QD distance
of 0.7 ∼ 1.1 µm, most samples show a clear gAB(0) peak in most cases.
When the QD distance is between 0.52 and 0.55 µm, the QD light spots over-
lap obviously, e.g. N13 and S8 (see their images in Fig. 5.5). The two QD signals
are not independent; the autocorrelation gA(0) peak of each QD will contribute
to the cross-correlation gAB(0) peak. For N13, the QD light intensity is high and
steady state-like, i.e., most photons are emitted without correlation, so its gAB(0)
peak is small. S8 has trigger-like blinking from each QD and thus a huge gAB(0)
peak. Compared to N13 and S8, the same sample S19, S20 and S21 (0.52) has
lower light intensity and smaller overlap. Their gAB(0) peaks are similar to the
cases of 0.72 ∼ 1.09 µm.
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
0 2000 4000 6000 8000 100000
200
400
600
800
1000
-20 -15 -10 -5 0 5 10 15 20
0.99
1.00
1.01
1.02
1.03
Blinking intensity
Time frame [50 ms/frame] QD-A
QD-B
Time frame [5 ms/frame]
A
B
gAB(τ)
movie in 5 ms/frame
Time delay [x50 ms]
movie in 50 ms/frame
Figure 5.9 Steady-state QD blinking in 50 ms/frame or 5 ms/frame and their cross-correlation. QD distance is 0.61 µm.
Later, we tried closely spaced CdSe/CdZnS/ZnS multi-shell QDs. These QDs
have high and steady state-like light intensity. We extended the frame number
52 5 Bunched correlation of two colloidal QD blinking
to 10000 and reduced the frame time to 5 ms. But, the blinking intensity is still
steady state-like, as shown in Fig. 5.9. There is no correlation between two QDs in
distance of 0.61 µm. We think the multi-shell QD light emission is continuous and
independent because of its thick and high-energy shells for photocarrier tunneling
out to the surface traps. There is seldom blinking and no correlation.
5.4 Explanation
The correlation between two closely-spaced single-shell colloidal QDs lies in the
trigger-like blinking itself. The light emission in one QD will stimulate the light
emission from the other QD if their distance is small. The former QD will be dark
if there is a carrier trapped in its surface state; until the trapped carrier returns
to it the QD cannot emit light to stimulate the light emission from the other QD.
The latter QD has more traps and often stays in the off state. Without the former
QD light stimulation, it tends to keep dark. The time correlation of their light
emission reflects their mutually stimulated emission. The stimulation rate is very
fast, typically ∼ ps−1. It is potential to generate coherent photon-pairs from the
blinking of the two QDs.
The Fourier transform of the autocorrelation function gives the power spec-
trum of each QD blinking (i.e., Wiener-Kintchine theorem). It satisfies the 1/f law
that is well known in random telegraph process, reflecting the random tunneling
and trapping process. The trigger-like QD blinking has a rapid reduce from gA(0)
to gA(1) and a thin gA(τ) profile. It indicates that the dominant time constant
for blinking is ∼ 50 ms and there are many triggers of light emission in the 1000
frames. For a dark QD, the difference of gA(0) and gA(1) is very small, close to
that of the background noise, implying that there is seldom trigger of light emis-
sion; for a steady-state QD blinking, the gradually reduced gA(τ) profile implies
that the dominant time constant for blinking is much larger than 50 ms, i.e., the
light emission is more continuous.
Chapter 6
Summary of the included papers
In Paper I, I studied the photovoltaic performance of InAs quantum dots (QDs)
in p-i-n GaAs cells. I find that the QDs in the depletion region will increase the
diode J0 and reduce the carrier lifetime. The sample with many layers of QDs in
the n-region, away from the built-in electric field, gives an open-circuit voltage Voc
higher than the sample with QDs in the depletion region. It indicates that the
reduced Voc in most QD solar cells has a more intrinsic reason except the strain-
induced dislocations. By analyzing the diode performance, we find that both QD
samples have enhanced SRH recombination due to dislocations and QDs; while
the QDs in the n-region do not reduce the carrier lifetime but reduce the electron
mobility due to the electron filling induced potential variation around QDs. The
photocurrent is also enhanced in the sample with QDs in the depletion region and
reduced in the sample with QDs in the n-region.
In Paper II, I further explore the photocurrent (PC) contribution in different
wavelength regions. The above-GaAs-bandgap absorption is dominant for the
enhanced PC of QD solar cell with QDs in the depletion region and the reduced
PC of QD solar cell with QDs in the n-region. The enhanced PC is understood
by the enhanced absorption due to the hole wavefunction reflection; while the
reduced PC is caused by the reduced carrier mobility due to the up-bent potential
variation around QDs. The sub-GaAs-bandgap PC spectra show the wetting
layer and QD contributions clearly. The contributions are considerable for QDs
in the depletion region since photocarriers can tunnel out via the built-in electric
field; the contributions for QDs in the n-region, away from the built-in field, are
much lower, since only the thermal activation works for photocarrier escape from
QDs. It demonstrates that although there are many QD layers and good photon
absorption as expected, the photocarrier extraction is more important for PC
53
54 6 Summary of the included papers
production. The cascade photon absorption via intermediate bands formed by
closely spaced QD layers, even if it exists, will have a problem on photocarrier
extraction through the depletion region.
In Paper III (in manuscript), I study the bias- and temperature-dependent
photocurrent spectra of the above QD samples. From the bias and temperature
dependence of each photocurrent peak that corresponds to QD state or wetting-
layer state absorption, I find different photocarrier escape mechanisms in the two
samples. The interlayer tunneling is responsible for the similar trend of bias
dependence in the sample with closely spaced QDs in the i-region; while the
minority hole thermal escape is responsible for the high activation energy in the
sample with QDs in the n-region, due to the up-bent potential variations around
QD matrix.
In Paper IV (submitted), I study the time correlation between two colloidal
QDs. CdSe/ZnS single-shell QDs have trigger-like blinking emission that can be
real-time recorded by the camera CCD in frame of 50 ms. If the two QD distance
is as small as 0.7∼1.1 µm, their blinking emission shows a clear correlation peak
at the zero delay, suggesting bunched emission. For CdSe/CdxZn1−xS/ZnS multi-
shell QDs, the light intensity is high and steady state-like, with no correlation.
It indicates that there are few blinking events and the light emission from each
QD is independent. We think the mutually stimulated light emission between
two closely spaced QDs is possible the reason for this bunched correlation. The
blinking of one QD affects its stimulation on the light emission from the other
QD, as a result, the two QDs show bunched blinking.
References
[1] BP statistical review of world energy 2011, www.bp.com/statisticalreview.
[2] http://en.wikipedia.org/wiki/File:Solar Spectrum.png.
[3] W. Shockley, and H. J. Queisser, J. Appl. Phys., 32, 510, 1961.
[4] M. A. Green, K. Emery, Y. Hishikawa, W. Warta, and E. D. Dunlop, Prog.
Photovolt: Res. Appl., 19, 565, 2011.
[5] J.-F. He, Z.-C. Niu, X.-Y. Chang, H.-Q. Ni, Y. Zhu, M.-F. Li, and X.-J.
Shang, Chin. Phys. B, 20, 018102, 2011.
[6] S. R. Huang, X. Lu, A. Barnett, and R. L. Opila, ”GaP films grown on
Si by liquid phase epitaxy”, 34th IEEE Photovoltaic Specialists Conference
(PVSC), 241-243, 2009.
[7] K.-S. Lee, Y.-S. Kim, Y.-K. Hong, and Y.-H. Jeong, IEEE Electron Device
Lett., 28, 672, 2007.
[8] Y. Zhang, S.-C. Shen, H. J. Kim, S. Choi, J.-H. Ryou, R. D. Dupuis, and B.
Narayan, Appl. Phys. Lett., 94, 221109, 2009.
[9] http://www2.electronicproducts.com/Using GaN power amplifiers-article-
farr nitronex nov2007-html.aspx
[10] Yu. B. Bolkhovityanov, and O. P. Pchelyakov, The Open Nanoscience Jour-
nal, 3, 20, 2009.
[11] M. Feiginov, C. Sydlo, O. Cojocari, and P. Meissner, Appl. Phys. Lett., 99,
233506, 2011.
[12] M. DeJarld, J. C. Shin, W. Chern, D. Chanda, K. Balasundaram, J. A.
Rogers, and X. Li, Nano Lett., 11(12), 5259, 2011.
55
56 REFERENCES
[13] http://www.compoundsemiconductor.net/csc/news-
details.php?cat=news&id=19734300.
[14] M. Wiemer, V. Sabnis, H. Yuen, Proc. of SPIE, 8108, 810804, 2011.
[15] M. A. Green, G. Conibeer, D. Konig, S. Shrestha, S. Huang, P. Aliberti,
L. Treiber, R. Patterson, B. P. Veettil, A. Hsieh, Y. Feng, A. Luque, A.
Marti, P. G. Unares, E. Canovas, E. Antolln, D. F. Marron, C. Tablero, E.
Hernandez, J-F. Guillemoles, L. Huang, A. L. Bris, T. Schmidt, R. Cladl,
and M. Tayebjee, ”Hot carrier solar cells: Challenges and recent progress”,
35th IEEE Photovoltaic Specialists Conference (PVSC), 57-60, 2010.
[16] K. Ikeda, H. Sekiguchi, F. Minami, J. Yoshino, Y. Mitsumori, H. Amanai, S.
Nagao, and S. Sakaki, J. Lumin., 108, 237, 2004.
[17] A. J. Nozik, Physica E, 14, 115-120, 2002.
[18] A. Luque and A. Marti, Prog. Photovolt: Res. Appl., 9, 73, 2001.
[19] S. M. Hubbard, C. D. Cress, C. G. Bailey, R. P. Raffaelle, S. G. Bailey, and
D. M. Wilt, Appl. Phys. Lett., 92, 123512, 2008.
[20] X. J. Shang, J. F. He, H. L. Wang, M. F. Li, Y. Zhu, Z. C. Niu, and Y. Fu,
Appl. Phys. A, 103, 335, 2011.
[21] X.-J. Shang, J.-F. He, M.-F. Li, F. Zhan, H.-Q. Ni, Z.-C. Niu, H. Pettersson,
and Y. Fu, Appl. Phys. Lett. 99, 113514, 2011.
[22] D. Guimard, R. Morihara, D. Bordel, K. Tanabe, Appl. Phys. Lett., 96,
203507, 2010.
[23] S. Krishna, S. D. Gunapala, S. V. Bandara, C. Hill, and D. Z. Ting, Proc. of
the IEEE, 95, 1838, 2007.
[24] S. D. Gunapala, S. V. Bandara, C. J. Hill, D. Z. Ting, J. K. Liu, S. B. Rafol,
E. R. Blazejewski, J. M. Mumolo, S. A. Keo, S. Krishna, Y. C. Chang, and
C. A. Shott, Proc. of SPIE, 6206, 62060J, 2006.
[25] http://www.veeco.com/markets/materials-science/effusion-cells,-valved-
crackers,-and-gas-injectors.aspx.
[26] http://en.wikipedia.org/wiki/Hot-filament ionization gauge.
REFERENCES 57
[27] http://www.pfeiffer-vacuum.com/know-how/mass-spectrometers-and-
residual-gas-analysis/introduction-operating-principle/quadrupole-mass-
spectrometers-qms/technology.action?chapter=tec4.1.2.
[28] J. B. Rodriguez, P. Christol, L. Cerutti, F. Chevrier, A. Joullie, J. Crystal
Growth, 274, 6, 2005.
[29] C. Manza, Q. Yanga, K. Kohlera, M. Maiera, L. Kirstea, J. Wagnera, W.
Sendb, D. Gerthsen, J. Crystal Growth, 280, 75, 2005.
[30] C.-C. Chen, H.-W. Chung, C.-H. Chen, H.-P. Lu, C.-M. Lan, S.-F. Chen, L.
Luo, C.-S. Hung, and E. W.-G. Diau, J. Phys. Chem. C 112, 19151, 2008.
[31] http://www.laballiance.com.my/brochure/spectroscopy/FTIR/VERTEX80v Flyer EN.pdf
[32] http://www.brukeroptics.com/downloads.html, Guide for Infrared Spec-
troscopy.
[33] M. Sugawara (volume editor), R. K. Willardson and E. R. Weber, Semicon-
ductors and Semimetals Volume 60, Self-Assembled InGaAs/GaAs Quantum
Dots, Academic Press, New York, 1999.
[34] D. J. Eaglesham, and M. Cerullo, Phys. Rev. Lett., 64, 1943, 1990.
[35] J. M. Moison, C. Guille, F. Houzay, F. Barthe, and M. van Rompay, Phys.
Rev. B, 40, 6149, 1989.
[36] C. Santori, M. Pelton, G. Solomon, Y. Dale, and Y. Yamamoto, Phys. Rev.
Lett., 86, 1502, 2001.
[37] J.-F. He, H.-L. Wang, X.-J. Shang, M.-F. Li, Y. Zhu, L.-J. Wang, Y. Yu,
H.-Q. Ni, Y.-Q. Xu, and Z.-C. Niu, J. Phys. D: Appl. Phys. 44, 335102, 2011.
[38] Z. Mi, P. Bhattacharya, and J. Yang, Appl. Phys. Lett., 89, 153109, 2006.
[39] J. S. Wang, S. H. Yu, Y. R. Lin, H. H. Lin, C. S. Yang, T. T. Chen, Y.
F. Chen, G. W. Shu, J. L. Shen, R. S. Hsiao, J. F. Chen, and J. Y. Chi,
Nanotechnology, 18, 015401, 2007.
[40] G. S. Solomon, J. A. Trezza, A. F. Marshall, and J. S. Harris, Phys. Rev.
Lett., 76, 952, 1996.
[41] A. Hospodkova, V. Krapek, K. Kuldova, J. Humlicek, E. Hulicius, J. Oswald,
J. Pangrac, and J. Zeman, Physica E, 36, 106, 2007.
58 REFERENCES
[42] Q. D. Zhuang, S. F. Yoon, H. X. Li, J.M. Li, Y. P. Zeng, M. Y. Kong, and
L. Y. Lin, Physica E, 11, 384, 2001.
[43] C. H. Chan, H. S. Chen, C. W. Kao, H. P. Hsu, and Y. S. Huang, Appl.
Phys. Lett., 89, 022114, 2006.
[44] S. Chakrabarti, N. Halder, S. Sengupta, S. Ghosh, T. D. Mishima, and C. R.
Stanley, Nanotechnology, 19, 505704, 2008.
[45] A. Marti, E. Antolin, C.R. Stanley, C.D. Farmer, N. Lopez, P. Diaz, E.
Canovas, P.G. Linares, A. Luque, Phys. Rev. Lett., 97, 247701, 2006.
[46] P. O. Anikeeva, J. E. Halpert, M. G. Bawendi, and V. Bulovic, Nano Lett.,
9, 2532, 2009.
[47] P. Michler, Single Quantum Dots Fundamentals, Applications, and New Con-
cepts, Springer, 2003.
[48] A. Mohan, M. Felici, P. Gallo, B. Dwir, A. Rudra, J. Faist, and E. Kapon,
Nature Photonics, 4, 302, 2010.
[49] S. M. Sze, Semiconductor Devices Physics and Technology, John Willey and
Sons, 1985.
[50] M. A. Green, Solar cells: Operating principles, Technology, and System Ap-
plications, Prentice-Hall, 1982.
[51] A. Marti, N. Lopez, E. Antolin, E. Canovas, A. Luque, C. R. Stanley, C. D.
Farmer, P. Diaz, Appl. Phys. Lett., 90, 233510, 2007.
[52] Y. Fu, M. Willander, J. Appl. Phys. 71, 3877, 1992.
[53] Y. Fu, M. Willander, J. Appl. Phys. 76, 4225, 1994.
[54] D. Zhou, P. E. Vullum, G. Sharma, S. F. Thomassen, R. Holmestad, T. W.
Reenaas, and B. O. Fimland, Appl. Phys. Lett., 96, 083108, 2010.
[55] T. Sugaya, Y. Kamikawa, S. Furue, T. Amano, M. Mori, and S. Niki, Solar
Energy Material & Solar cells, 95, 163, 2011.
[56] K. A. Sablon, J. W. Little, V. Mitin, A. Sergeev, N. Vagidov, and K. Rein-
hardt, Nano Lett., 11, 2311, 2011.
REFERENCES 59
[57] R. J. Ellingson, M. C. Beard, J. C. Johnson, P. Yu, O. I. Micic, A. J. Nozik,
A. Shabaev, and A. L. Efros, Nano Lett., 5, 865, 2005.
[58] Y. Fu, Superlattices Microstruct., 30, 69, 2001.
[59] E. S. Shatalina, S. A. Blokhin, A. M. Nadtochy, A. S. Payusov, A. V. Savelyev,
M. V. Maximov, A. E. Zhukov, N. N. Ledentsov, A. R. Kovsh, S. S. Mikhrin,
and V. M. Ustinov, Semiconductors, 44, 1308, 2010.
[60] W.-H. Chang, T. M. Hsu, C. C. Huang, S. L. Hsu, C. Y. Lai, N. T. Yeh, T.
E. Nee, and J.-I. Chyi, Phys. Rev. B, 62, 6959, 2000.
[61] J. L. Hartke, J. Appl. Phys., 39, 4871, 1968.
[62] S. Rodt, R. Seguin, A. Schliwa, F. Guffarth, K. Ptschke, U.W. Pohl, D.
Bimberg, J. Lumin., 122-123, 735, 2007.
[63] R. Heitz, F. Guffarth, K. Ptschke, A. Schliwa, and D. Bimberg, Phys. Rev.
B, 71, 045325, 2005.
[64] M. Gong, G. Chen, L. He, C.-F. Li, J.-S. Tang, F.-W. Sun, Z.-C. Niu, S-S
Huang, Y.-H. Xiong, H.-Q. Ni, and G.-C. Guo, epl, 90, 37004, 2010.
[65] V. Turck, F. Heinrichsdorff, M. Veit, R. Heitz, M. Grundmann, A. Krost,
and D. Bimberg, Appl. Surface Science, 123-124, 352, 1998.
[66] J.-H. Ser, Y.-H. Lee, J.-W. Kim, and J.-E. Oh, Jpn. J. Appl. Phys., 39, L518,
2000.
[67] G. Vincent, A. Chantre, and D. Bois, J. Appl. Phys. 50, 5484, 1979.
[68] Y. Fu, O. Engstrm, and Y. Luo, J. Appl. Phys. 96, 6477, 2004.
[69] I. L. Medintz, H. T. Uyeda, E. R. Goldman, and H. Mattoussi, Nat. Mater.,
4, 435. 2005.
[70] Q. Sun, Y. A. Wang, L. S. Li, D. Wang, T. Zhu, J. Xu, C. Yang, and Y. Li,
Nature Photonics, 1, 717, 2007.
[71] Z. Ning, H. Tian, H. Qin, Q. Zhang, H. Agren, L. Sun, and Y. Fu, J. Phys.
Chem. C, 110, 389, 2010.
[72] M. Kuno, D. P. Fromm, H. F. Hamann, A. Gallagher, and D. J. Nesbitt, J.
Chem. Phys., 112, 3117, 2000.
60 REFERENCES
[73] K. T. Shimizu, R. G. Neuhauser, C. A. Leatherdale, S. A. Empedocles, W.
K. Woo, and M. G. Bawendi, Phys. Rev. A, 63, 205316, 2001.
[74] Z. Yuan, B. E. Kardynal, R. M. Stevenson, A. J. Shields, C. J. Lobo, K.
Cooper, N. S. Beattie, D. A. Ritchie, and M. Pepper, Science, 295, 102, 2002.
[75] V. Zwiller, H. Blom, P. Jonsson, N. Panev, S. Jeppesen, T. Tsegaye, E.
Goobar, M.-E. Pistol, L. Samuelson, G. Bjok, Appl. Phys. Lett., 78, 2476,
2001.
[76] A. J. Shields, R. M. Stevenson, R. M. Thompson, M. B. Ward, Z. Yuan, B.
E. Kardynal, P. See, I. Farrer, C. Lobo, K. Cooper, and D. A. Ritchie, phys.
stat. sol. (b), 238, 353, 2003.
[77] A. Dousse, J. Suffczynski, A. Beveratos, O. Krebs, A. Lemaıtre, I. Sagnes, J.
Bloch, P. Voisin, and P. Senellart, Nature, 466, 217, 2010.
[78] P. Aivaliotis, E. A. Zibik, L. R. Wilson, J. W. Cockburn, M. Hopkinson, and
N. Q. Vinh, Appl. Phys. Lett., 92, 023501, 2008.
[79] http://www.rp-photonics.com/autocorrelators.html
[80] H. Paul, Rev. Mod. Phys., 54, 1061, 1982.