TRITA-ICT/MAP AVH Report 2012:18
ISSN 1653-7610
ISRN KTH/ICT-MAP/AVH-2012:18-SE
ISBN 978-91-7501-530-9
Royal Institute of Technology
School of Information and
Communication Technology
SE-164 40 Kista
Akademisk avhandling som med tillstånd av Kungliga Tekniska högskolan framlägges till offentlig granskning för avläggande av teknologie doktorsexamen i mikroelektronik och tillämpad fysik måndagen den 19 november 2012 klockan 10:30 i sal C2, KTH-Electrum, Isafjordsgatan 26, Kista.
© Vytautas Liuolia, November 2012
Tryck: Universitetsservice AB

InGaN based blue and near-ultraviolet light emitting diodes and laser diodes have been successfully commercialized for many applications such as general lighting, display backlighting and high density optical storage devices. Despite having a comparably high defect density, these devices are known for their efficient operation, which is attributed to localization in potential fluctuations preventing carriers from reaching the centers of nonradiative recombination. Nitride research is currently headed towards improving deep ultraviolet AlGaN and green InGaN emitters with higher Al and In molar fractions. The efficiency of these devices trails behind the blue counterparts as the carrier localization does not seem to aid in supressing nonradiative losses. In addition, the operation of ternary nitride heterostructure based devices is further complicated by the presence of large built-in electric fields.
Although the problem can be ameliorated by growing structures in nonpolar or semipolar directions, the step from research to production still awaits.
In this thesis, carrier dynamics and localization effects have been studied in three different nitride ternary compounds: AlGaN epitaxial layers and quantum wells with high Al content, nonpolar m -plane InGaN/GaN quantum wells and lattice matched AlInN/GaN heterostructures. The experimental methods of this work mainly consist of spectroscopy techniques such as time-resolved photoluminescence and differential transmission pump-probe measurements as well as spatial photoluminescence mapping by means of scanning near-field microscopy.
The comparison of luminescence and differential transmission measurements has allowed estimating the localization depth in AlGaN quantum wells. Additionally, it has been demonstrated that the polarization degree of luminescence from m -
InGaN quantum wells decreases as carriers diffuse to localization centers. What is more, dual-scale localization potential has been evidenced by near-field measurements in both AlGaN and m -InGaN. Larger scale potential fluctuation have been observed directly and the depth of nanoscopic localization has been estimated theoretically from the recorded linewidth of the near-field spectra. Lastly, efficient carrier transport has been observed through AlInN layer despite large alloy inhomogeneities evidenced by broad luminescence spectra and the huge Stokes shift.
Inhomogeneous luminescence from the underlying GaN layer has been linked to the fluctuations of the built-in electric field at the AlInN/GaN interface.
Keywords: AlGaN, InGaN, AlInN, LEDs, near-field microscopy, carrier dynamics, alloy fluctuations, carrier localization, built-in electric field, nonpolar planes, polarized luminescence.
i

Abstract
Table of contents
List of papers
Acknowledgments
List of abbreviations
xi
1 Introduction
1
1.1
Aim and overview of the original work . . . . . . . . . . . . . . . . .
5
1.2
Thesis structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6
2 Material properties and growth of III-nitrides
7
2.1
Crystal structure and material properties
. . . . . . . . . . . . . . .
7
2.2
Polarization fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
Screening of the built-in electric field . . . . . . . . . . . . . . . . . .
11
Different growth planes
. . . . . . . . . . . . . . . . . . . . . . . . .
15
2.3
Growth techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
2.4
Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
3 Experimental techniques
21
3.1
Laser systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
3.2
Ultrafast spectroscopy techniques . . . . . . . . . . . . . . . . . . . .
23
Time-resolved photoluminescence . . . . . . . . . . . . . . . . . . . .
23
Pump-probe spectroscopy . . . . . . . . . . . . . . . . . . . . . . . .
26
3.3
Near-field optical microscopy . . . . . . . . . . . . . . . . . . . . . .
27
Aperture SNOM configurations . . . . . . . . . . . . . . . . . . . . .
28
Fiber probes
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
Scanning near-field microscope setup . . . . . . . . . . . . . . . . . .
32 iii i v ix iii

iv TABLE OF CONTENTS
4 Localization effects in ternary nitrides
37
4.1
AlGaN films and quantum wells . . . . . . . . . . . . . . . . . . . . .
37
Epitaxial films
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
38
Quantum wells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
4.2
InGaN m -plane quantum wells . . . . . . . . . . . . . . . . . . . . .
42
Polarized emission . . . . . . . . . . . . . . . . . . . . . . . . . . . .
42
Potential fluctuations and inhomogeneous broadening
. . . . . . . .
44
4.3
AlInN/GaN heterostructures
. . . . . . . . . . . . . . . . . . . . . .
48
5 Conclusions and future work
51
5.1
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
5.2
Suggestions for future work . . . . . . . . . . . . . . . . . . . . . . .
52
Bibliography
List of figures
55
63
Guide to the articles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67

[A] V. Liuolia , S. Marcinkevičius, A. Pinos, R. Gaska, and M. S. Shur, “Dynamics of carrier recombination and localization in AlGaN quantum wells studied by time-resolved transmission spectroscopy”, Appl. Phys. Lett.
95 , 091910 (2009).
[B] J. Mickevičius, E. Kuokštis, V. Liuolia , G. Tamulaitis, M. S. Shur, J. Yang, and R. Gaska, “Photoluminescence dynamics of AlGaN quantum wells with built-in electric fields and localized states”, Phys. Status Solidi A 207 , 423
(2010).
[C] V. Liuolia , S. Marcinkevičius, Y. D. Lin, H. Ohta, S. P. DenBaars, and S.
Nakamura, “Dynamics of polarized photoluminescence in m -plane InGaN/GaN quantum wells”, J. Appl. Phys.
108 , 023101 (2010).
[D] V. Liuolia , A. Pinos, S. Marcinkevičius, Y. D. Lin, H. Ohta, S. P. DenBaars, and S. Nakamura, “Carrier localization in m -plane InGaN/GaN quantum wells probed by scanning near field optical spectroscopy”, Appl. Phys. Lett.
97 ,
151106 (2010).
[E] A. Pinos, V. Liuolia , S. Marcinkevičius, R. Gaska, and M. S. Shur, “Localization potentials in AlGaN epitaxial films studied by scanning near-field optical spectroscopy”, J. Appl. Phys.
109 , 113516 (2011).
[F] V. Liuolia , S. Marcinkevičius, Q. Wang, J. Andersson, S.-M. Kim, and J. H.
Baek, “Near- and far-field optical characterization of InGaN photonic crystal light emitting diodes”, Phys. Status Solidi C 9 , 1664 (2012).
[G] V. Liuolia , S. Marcinkevičius, D. Billingsley, M. Shatalov, J. Yang, R. Gaska, and M. S. Shur, “Photoexcited carrier dynamics in AlInN/GaN heterostructures”, Appl. Phys. Lett.
100 , 242104 (2012).
v

vi LIST OF PAPERS
1. Q. Wang, S. Savage, S. Person, B. Noharet, S. Junique, J. Y. Andersson,
V. Liuolia , and S. Marcinkevičius, “Multiple functional UV devices based on InGaN/AlGaN and Al x
Ga
1-x
N/Al y
Ga
1-y
N quantum wells for biological warfare agents detections”, SPIE Photonics West 2009 , January 24–29, 2009,
San Jose, USA.
2.
V. Liuolia , S. Marcinkevičius, R. Gaska, J. W. Yang, and W. Sun, “Differential transmission dynamics in AlGaN quantum wells”, 15 th Semiconductor and Insulating Materials Conference , June 15–19, 2009, Vilnius, Lithuania.
3. J. Mickevičius, E. Kuokštis, V. Liuolia , G. Tamulaitis, M. S. Shur, J. Yang, and R. Gaska, “Carrier localization and built-in electric field in AlGaN/AlGaN quantum wells”, 15 th
Semiconductor and Insulating Materials Conference ,
June 15–19, 2009, Vilnius, Lithuania.
4. S. Marcinkevičius, V. Liuolia , Y.-D. Lin, H. Ohta, S. P. DenBaars, and
S. Nakamura, “Dynamics of polarized photoluminescence in m -plane InGaN quantum wells”, 37 th International Symposium on Compound Semiconductors , May 31–June 4, 2010, Takamatsu, Japan, paper TuC4-2.
5.
V. Liuolia , A. Pinos, S. Marcinkevičius, Y.-D. Lin, H. Ohta, S. P. Den-
Baars and S. Nakamura, “Carrier localization in m-plane InGaN/GaN quantum wells probed by scanning near field optical spectroscopy”, International
Workshop on Nitride Semiconductors 2010 , September 19–24, 2010, Tampa,
USA, paper C1-7.
6.
V. Liuolia , S. Marcinkevičius, Q. Wang, J. Andersson, S.-M. Kim, and J.
H. Baek, “Near- and Far-Field Optical Characterization of InGaN Photonic
Crystal LEDs”, 16 th
Semiconducting and Insulating Materials Conference ,
June 19–23, 2011, Stockholm, Sweden, paper Tu3-27.
7. A. Pinos, S. Marcinkevičius, V. Liuolia , J. Yang, R. Gaska, and M. S.
Shur, “Scanning near-field optical spectroscopy of AlGaN epitaxial layers”,
16 th Semiconducting and Insulating Materials Conference , June 19–23, 2011,
Stockholm, Sweden, paper Tu1-3.
8. S. Marcinkevičius, A. Pinos, V. Liuolia , J. Yang, R. Gaska, and M. S. Shur,
“Localization potentials in AlGaN epitaxial layers studied by scanning near field optical spectroscopy”, 9 th International Conference on Nitride Semiconductors , July 10–15, 2011, Glasgow, UK, paper E2.1.
9. S. Marcinkevičius, V. Liuolia , D. Billingsley, M. Shatalov, J. Yang, R. Gaska and M. S. Shur, “Photoexcited carrier dynamics and localization potentials in AlInN”, 31 st
International Conference on the Physics of Semiconductors ,
July 29–August 3, 2012, Zurich, Switzerland, paper 3.6.

LIST OF PAPERS vii
10. S. Marcinkevičius, V. Liuolia , D. Billingsley, M. Shatalov, J. Yang, R.
Gaska and M. S. Shur, “Carrier dynamics and localization in AlInN/GaN heterostructures”, 39 th International Symposium on Compound Semiconductors , August 27–30, Santa Barbara, USA, paper Mo-P.22.
1. L. Dong, F. Ye, A. Chughtai, V. Liuolia , S. Popov, A. T. Friberg, and M.
Muhammed, “Lasing From Water Solution of Rhodamine 6G/Gold Nanoparticles: Impact of SiO
48 , 1220 (2012).
2
Coating on Metal Surface”, IEEE J. Quantum Electron.

First and foremost, I owe sincere and earnest thankfulness to my supervisor Prof.
Saulius Marcinkevičius for offering me a PhD position at KTH and his patient guidance. He has always been available for help, be it experimental work or scientific writing.
I share the credit of my work with my fellow PhD student Andrea Pinos. I want to thank him for his dedicated effort on improving our near-field microscopy system as well as our entertaining discussions ranging from the most boring technical details to cultural and political aspects of the society.
It also gives me a great pleasure acknowledging other colleagues from Optics group for contributing to the nice atmosphere in our small group.
Thanks to
Srinivasan Iyer for being a nice friend and also an office roommate for a few last months, Lin Dong for our collaboration, particularly the SEM measurements, Per
Martinsson for his friendly and positive attitude, my co-supervisor Sergei Popov for all sorts of help including combating printing woes and guiding through KTH’s courses-and-credits jungle, Sergei’s newest rookie Gleb Lobov for passionately taking over the (in)famous holography student lab, and our group’s ambassador in
Finland Prof. Ari Friberg for visiting us whenever most needed.
Many thanks to our previous and current departments administrators, Marianne
Widing and Madeleine Printzsköld, for taking care of the administrative burden in an efficient way, including paperwork for the PhD defense.
This thesis would not have been possible without the hard work of our collaborators. I owe my deepest gratitude to Prof. Michael Shur and (back then a PhD student) Dr. Kai Liu for warmly welcoming me during my visits to Rensselaer
Polytechnic Institute. I am grateful to staff of the Sensor Electronic Technology,
Inc., researchers of the Nitrides group at the University of California, Santa Barbara, and Korea Photonics Technology Institute for growing the samples for our experiments. I am indebted to Prof. Takashi Yatsui and his group at the University of Tokyo for the nice demonstration of their experimental facilities and invitation to nomikai with awesome saké, sushi and good company. I would also like to thank my master thesis supervisor at Vilnius University, Prof. Edmundas Kuokštis, for his quantitative approach to physics, his dedication to teaching and involvement in organizing physics competitions for school students.
I would like to acknowledge all my colleagues, friends and visiting researchers at ix

x ACKNOWLEDGMENTS
KTH and Acreo. Thanks to Andrius Žukauskas for a frame of pool or a beer once in a while. Also many thanks to (in no particular order) Srinivasan Anand, Magnus
Hårdensson Berntsen, Romain Estève, Mats Götelid, Vytautas Grivickas, Karolis
Gulbinas, Aki-Kimmo Kallio, Sebastien Lourdudoss, Haniye Mahabadi, Vasileios
Manolopoulos, Maziar Manouchehry Naiini, Sergei Manuilov, Anneli Önsten, Henry
Radamson, Reza Sanatinia, Mahtab Sangghaleh, Aziza Sudirman, Qin Wang, and everyone else not listed here. Thanks to my other friends in Sweden and Lithuania.
Finally, I am grateful to my parents, grandparents and my sister for their support and encouragement. Last, but not least, special thanks to my beloved wife
Milda and our son Erikas.
Vytautas Liuolia
Stockholm, October 2012

CCD
CL
DBR
DOS
DUV
EL
FF
FWHM
Ga
Al
AlGaAs
AlGaInP
AlGaN
AlInN
AlN
BBO
GaN
HEMT
I-C
In aluminum aluminum gallium arsenide aluminum gallium indium phosphide aluminum gallium nitride aluminum indium nitride aluminum nitride barium triborate charge-coupled device cathodoluminescence distributed Bragg reflector density of states deep ultraviolet (usually defined as λ < 300 nm) electroluminescence far-field full width at half maximum gallium gallium nitride high electron mobility transistor illumination-collection indium xi

xii LIST OF ABBREVIATIONS
NF
NIR
PL
PLE
QW
RT
SEM
SNOM
TD
Ti:sapphire
TRPL
UV
VCSEL
WPE
InGaN
InN
LBO
LD
LED
MBE
MEMOCVD
MOCVD
MQW
N indium gallium nitride indium nitride lithium triborate laser diode light emitting diode molecular beam epitaxy migration-enhanced MOCVD metal-organic chemical vapor deposition multiple quantum wells nitrogen near-field near infrared photoluminescence photoluminescence excitation quantum well room temperature scanning electron microscope scanning near-field optical microscopy threading dislocation titanium-doped sapphire time-resolved photoluminescence ultraviolet vertical cavity surface emitting laser wall plug efficiency

N itride semiconductors such as gallium nitride (GaN), indium nitride (InN), aluminum nitride (AlN) and their alloys are key materials for efficient optoelectronic devices in ultraviolet (UV), blue and green spectral range. In addition, large bandgap and break-down electric field of GaN and AlN make them an attractive choice for power electronics. The success story of nitrides began in 1993, when
Shuji Nakamura from Nichia Corporation demonstrated the first commercially vi-
able high-brightness blue light emitting diode (LED) [1]. The device was based on
indium gallium nitride (InGaN) multiple quantum wells (MQW). The effort, which
was later recognized by the Millenium Technology Prize [2], significantly changed
the landscape of solid state light emitters, since efficient LEDs had become available in the whole visible spectrum. Development of InGaN MQW based violet laser
diodes (LD) followed shortly after LEDs [3]. Moreover, blue LEDs can be covered with yellow or a combination of green and red phosphors [4] in order to create the
perception of white light. Commercial “white” LEDs are available with luminous
efficacy of up to 150 lm/W and proof-of-concept devices have reached 250 lm/W [5].
LEDs have thus surpassed the efficacy of traditional white light sources such as incandescent (13 lm/W) and fluorescent (up to 90 lm/W) lamps, and are increasingly used for backlighting displays, automobile headlights and general home lighting. At the time of writing this thesis, the packaged LED market is expected to reach 11.4
billion US dollars in 2012, with nitride based LEDs for general lighting accounting
for more than half of overall revenue [6].
Despite the large-scale success in solid state lighting, nitride based devices are not limited to near-UV or blue color. AlN has a band gap corresponding to deep
UV emission (DUV), whereas that of InN lies in near infrared (NIR), thus enabling nitrides to cover an extremely broad range of wavelengths with a single material system. However, the efficiency of longer wavelength InGaN devices is trailing in comparison to the now mature technology of blue LEDs. On the other hand, the demand of efficient red, orange and (partially) yellow emitters is addressed by
1

2 CHAPTER 1. INTRODUCTION aluminum gallium indium phosphide (AlGaInP) devices while aluminum gallium
arsenide AlGaAs based LEDs and LDs are very efficient in NIR [7]. Therefore the
research of InGaN optoelectronics is currently focused on the most problematic
green-yellow spectral range, colloquially also known as the “green gap” [8]. Several
challenges for producing light emitters with larger indium (In) content (that, is with longer emission wavelength) have been pointed out. Firstly, the large lattice constant mismatch between hexagonal InN (2 .
19 Å) and GaN (1 .
96 Å) [9] com-
plicates growth of high In content In x
Ga
1-x
N epitaxial layers and quantum wells
(QWs). InGaN growth is long known for kinetic phase segregation, manifesting as
inhomogeneous alloy distribution and experimentally evidenced In-rich regions [10].
Surprisingly, the resulting potential fluctuations improve efficiency of blue LEDs by localizing carriers and preventing them from reaching the centers of non-radiative
recombination [11, 12]. In this way, efficient emission is achieved even from ma-
terials with relatively poor crystalline quality (threading dislocation (TD) density at 10
8
–10
10 cm
− 2
). However, carrier localization does not seem to be beneficial for green emitters, possibly due to generation of additional TDs as In content increases,
as well as very long carrier lifetimes [13] providing sufficient diffusion length to reach the centers of non-radiative recombination [12].
Furthermore, commercial devices are most commonly grown along c -axis, re-
sulting in large spontaneous and piezoelectric polarizations [14]. Polarization dis-
continuities at heterostructure interfaces lead to built-in electric field in a QW, which reduces the carrier recombination probability due to spatial separation of electrons and holes by the electric field. Polarization fields increase with higher
In content due to larger piezoelectric strain, therefore green emitters are affected more. This has spurred attempts on eliminating built-in fields by growing heterostructures perpendicularly to c -axis. However, since InGaN growth on nonpolar
planes has proved to be challenging due to formation of stacking faults [15], the
research has also shifted to semipolar planes where the polarization is smaller or,
for certain alloy compositions, vanishes completely [16]. Several semipolar planes
on these planes indeed helps eliminating issues related to the built-in electric fields, the growth of semipolar QWs experiences other challenges. For instance, the nonpolar and semipolar nitride structures must be grown on native GaN substrates cut along the appropriate planes. This makes the substrate price high and substrate area small, of the order of a few cm 2 . In incorporation into the alloy is also very different for different substrate orientations, thus, growth conditions should be optimized for each individual case. To loosely quote one of the leading researchers in the field of nonpolar and semipolar nitrides Prof. J. Speck from UCSB, InGaN grown on different planes behaves like completely different materials.
The reduction of built-in electric field particularly benefits LDs as the increased matrix element for optical transition yields higher gain and lower threshold current in comparison to polar c
-plane devices [18]. Recently, semipolar (20¯
LDs have been reported to reach 7 % wall plug efficiency (WPE) for “true” green
(530 nm) color [19]. Although green lasers based on second harmonic generation

3
exist [20], commercialization of green LDs could pave the way for energy efficient
compact full-color laser projectors for mobile applications. Moreover, research on semipolar InGaN QWs at UCSB has resulted in the fabrication of UV vertical
cavity surface emitting laser (VCSEL) based on nonpolar m-plane QWs [21]. In
addition to advantages of eliminating or reducing the built-in electric field in the
QWs, nonpolar and semipolar QW structures allow producing sources of polarized radiation. This occurs because of the strain-induced valence band splitting and the
selection rules for the optical transitions [22]. The latter feature is advantageous
for applications requiring illumination by polarized light, such as backlighting of liquid crystal displays. Using LEDs grown on nonpolar or semipolar planes allows energy savings of nearly 50%.
While InGaN is the primary material platform for blue and white light emitters, aluminum gallium nitride (AlGaN) based devices cover the UV spectral region. Because of the large AlN band gap (6 .
2 eV), Al x
Ga
1-x
N LEDs can reach wavelengths
as short as 210 nm [23]. Deep UV LEDs are already applied in many areas including
analytical and biomedical instrumentation, bio-agent detection and identification, medicine, air and water sterilization and decontamination, radiation resistant UV
sources, UV curing and agriculture [24]. Deep UV LEDs could potentially replace
mercury lamps, thus providing UV sources that have a desired emission wavelength, are compact, have a long lifetime and are environmentally friendly (do not contain
mercury). Fig. 1.1 shows that varying Al
x
Ga
1-x
N alloy composition, UVA (320 nm to 400 nm), UVB (280 nm to 320 nm) and, partially, UVC (100 nm to 280 nm) spectral ranges may be covered. However, before wide scale applications of AlGaN-
Figure 1.1: Wavelengths corresponding to Al x
Ga
1-x
N quantum wells versus Al molar fraction at different temperatures (reproduced
with permission from Ref. [24])

4 CHAPTER 1. INTRODUCTION
State of art blue LED (Osram)
EE 87% IQE x IE 75%
1134 mW
I = 0.35 A, U = 3.24 V
986 mW
LEE 81%
740 mW 600 mW WPE = 53%
State of art UV LED (SET)
EE 60% IQE x IE 15%
145 mW
I = 20 mA, U = 7.25 V
87 mW
LEE 18%
13 mW 2.3 mW
WPE = 1.6%
Figure 1.2: Typical values of different efficiencies characterizing visible
(blue) and UV LEDs. Here EE is electric efficiency, IQE is internal quantum
efficiency and LEE is light extraction efficiency (based on Ref. [24])
based LEDs are implemented, a number of issues still have to be resolved. These, among other things, include a low Mg acceptor activation level (the Mg binding energy is about 0 .
5 eV and increases with increasing Al content) and a large (10 9 to 10 10 cm
− 2 ) concentration of TDs. Dislocations, decorated with point defects,
act as nonradiative recombination channels [25] and contribute to premature de-
vice ageing [26, 27]. Fig. 1.2 illustrates a recent (2010) comparison of wall plug
efficiencies of state of the art Osram blue LEDs and the best deep UV LEDs from
Sensor Electronic Technology Inc. (SET). While the blue LEDs show nearly ideal wall plug efficiency, deep UV LEDs still need a considerable improvement to reach comparable parameters. The figure clearly summarizes the current challenges in improving DUV LEDs. These issues might be addressed by using transparent superlattice structures for the p -contact, improving material quality (i.e. reducing the TD density), and by increasing light extraction efficiency. Recent efforts in development of material quality and improving p-doping have lead to increase of
the wall plug efficiency to 10% [28]. In the context of this thesis, it is also worth
noting that in AlGaN, contrary to InGaN, regions of the lower band potential aid, rather than prevent, the nonradiative recombination (
Paper E ).
Another important application of AlGaN is in the area of high electron mobility transistors (HEMTs). The spontaneous and piezoelectric polarization difference between AlGaN and GaN causes GaN band bending at the AlGaN and GaN in-
terface and forms the HEMT channel [9]. However, the lattice mismatch between
AlGaN and GaN induces strain, relaxation of which generates dislocations. Therefore, in 2001 Kuzmik has suggested to replace AlGaN with aluminum indium ni-
tride (AlInN) [29]. At about 17% of In, AlInN is lattice matched to GaN, which
allows avoiding the strain-related issues. The large spontaneous polarization difference causes a huge interface electric field (3 .
6 MV/cm for Al
0.83
In
0.17
N/GaN) that forms the HEMT channel and accumulates electrons with a sheet density of

1.1. AIM AND OVERVIEW OF THE ORIGINAL WORK 5
~10
13 cm
− 2
. In HEMTs based on AlInN/GaN heterostructure, however, the electron mobility is low, about 100 cm 2 V
− 1 s
− 1 . This low value has been attributed to
alloy composition fluctuation related alloy scattering [30]. To reduce the scatter-
ing, structures with a thin, about 1 nm AlN layer at the interface were introduced.
Even though such thin layers, during their growth, partially intermix with GaN
[31], their introduction allows increasing the mobility by an order of magnitude, to
the order of 1000 cm 2 V
− 1 s
− 1
[30, 32]. Apart from applications in HEMTs, AlInN
layers have been successfully used to fabricate distributed Bragg reflectors (DBR)
for blue VCSELs [33]. This became possible because of the large refractive index
contrast between AlInN and GaN (7–8% at 420 nm).
In spite of the first promising results, Al
1-x
In x
N remains the most peculiar and the least researched of ternary nitride alloy semiconductors. One of the properties that stands out in comparison with AlGaN and InGaN is the large depth of localization potentials characterized by the Stokes shift. In AlGaN, the energy difference for optical transitions between the extended and the localized states is 30 to 80 meV
(
Papers A
and
E ), in InGaN – 20 to 100 meV ( Papers C , D
and [12]). In AlInN,
for application-relevant In fraction around 17%, the Stokes shift can be as large as
1 eV (
Paper G
and [33]). The large AlInN band potential fluctuations should have
a large influence on the carrier scattering in HEMTs (even with the interlayer) and the spatial uniformity of the refractive index in DBRs.
The mentioned large potential fluctuations and resulting carrier localization effects in InGaN, AlGaN and AlInN are a distinctive feature of III-nitride semiconductors. Although potential fluctuations in semiconductor alloys are expected to be considerably larger than in their binary components, the other popular material systems for LEDs, namely III-arsenides and III-phosphides, exhibit only very moderate inhomogeneities in comparison to nitrides. For instance, if alloy homogeneity is expressed in terms of the spectral linewidth at low temperature, growth processes
can be readily optimized to produce AlGaAs [34, 35] and AlGaInP [36, 37] with
spectral linewidth of about or less than 10 meV throughout the whole composition range. Due to low alloy broadening, it is not unusual that well width fluctuations often have more influence on effective band gap variations in QWs grown from these
materials [38, 39]. This is in strong contrast to nitrides where low temperature linewidths exceeding 50 meV for AlGaN [40] and 200 meV for InGaN [41] are com-
mon for higher molar fraction of Al or In, respectively. Therefore it is particularly important to study carrier localization effects in ternary nitride alloys.
This thesis project has tried contribute to the understanding of carrier localization and related effects by studying luminescence inhomogeneities in materials behind the current frontiers in nitride device development – AlGaN with high Al content for deep UV LEDs, nonpolar InGaN for efficient green emitters, and AlInN for new applications such as HEMTs. The following issues have been investigated:

6 CHAPTER 1. INTRODUCTION
(i) In contrast to blue InGaN emitters, carrier localization does not aid in supressing non-radiative recombination in AlGaN deep UV LEDs. Carrier localization has been observed in AlGaN MQWs by means of time-resolved photoluminescence and differential transmission measurements (
Papers A
and
B ).
Paper B
also addresses the screening of built-in electric field in
MQWs. In addition, potential fluctuations in AlGaN epitaxial layers have been directly studied by near-field photoluminescence mapping in
Paper E .
(ii) Nonpolar m -plane InGaN/GaN quantum wells for efficient green-blue emitters are the subject of research in
Papers C
and
D . The properties of strongly
polarized luminescence have been studied in
Paper C , whereas the alloy
homogeneity and growth peculiarities on the m -plane have been inspected by near-field photoluminescence mapping of both InGaN active layers and GaN barriers in
Paper D .
(iii) Possible alternatives for improving WPE of LEDs involve patterning devices with photonic crystals for improved light extraction efficiency. Emission from such LEDs has been explored in both near- and far- field conditions in Pa-
per F . Moreover, the paper has also investigated the effect of photonic crystals
on spatial emission homogeneity.
(iv) In
Paper G , lattice matched AlInN/GaN heterostructures (similar to those
used for AlInN HEMTs) have been studied by time-resolved photoluminescence and near-field microscopy. Carrier transport is analyzed by comparing luminescence dynamics in AlInN and GaN layers, and the inhomogeneity of the electric field at the AlInN/GaN interface is discussed in connection to the observed spatial emission distribution in the near-field.
After this introductory chapter, the remainder of the thesis is organized as follows.
The next Chapter 2 provides an introduction to basic material properties and popu-
lar growth techniques of III-nitrides. In addition, the relation between polarization
fields in heterostructures and crystal growth planes is explained. Chapter 3 de-
scribes the experimental techniques and the respective setups used in this work.
Furthermore, a few practical implementation details and experimental challenges are presented, particularly for near-field microscopy. Carrier localization effects
are discussed in Chapter 4. Separate sections address effects in AlGaN for deep
UV LEDs, m -plane InGaN for green emitters and AlInN/GaN heterostructures for
HEMTs. Next, the conclusions and suggestions for the future work are outlined in
Chapter 5. After conclusions, the lists of bibliographic citations and figures follow.
Finally, the thesis is finished with a guide summarizing the appended articles and the corresponding reprints of the papers.

T his chapter gives a brief overview of nitride semiconductor material properties, focusing on wurtzite structure ternary alloys studied in this work. Different growth planes and their effect on piezoelectric and spontaneous polarizations in heterostructures are explained together with the presentation of carrier dynamics modeling for an AlGaN quantum well with the built-in electric field. Additionally, common nitride growth techniques and related defect types are discussed.
Like most semiconductors, the atom arrangement in the nitride semiconductors is tetrahedrally coordinated; therefore, each atomic site has the four nearest neighbors occupying the vertices of a tetrahedron. Nitride semiconductors grow in cubic zinc blende and hexagonal wurtzite structures. The latter one is thermodynamically
most stable [42] and is used in the absolute majority of nitride-based devices. All
the work presented in the thesis has been performed on wurtzite symmetry mate-
rials and, therefore, only their properties will be briefly reviewed below. Fig. 2.1
shows the crystal structure of wurtzite GaN. The wurtzite group-III nitrides lack an inversion symmetry plane perpendicular to the polar c ([0001]) axis. As a consequence, two possible stacking orders can be distinguished. If the bonds along the c -axis in the direction of growth go from a metal atom to a nitrogen (N) atom, the layer has a (Ga,In,Al)-polarity, otherwise it has a N-polarity. The polarity of nitride epitaxial layers determines the sign of the polarization charge in heterostructures.
In ternary nitrides, the ideal wurtzite structure is distorted because of the different Al-N (1 .
89 Å), Ga-N (1 .
95 Å) and In-N (2 .
15 Å) bond lengths (Fig. 2.2).
This is one of the reasons leading to atom clustering and local alloy composition
7

8 CHAPTER 2. MATERIAL PROPERTIES OF III-NITRIDES
Figure 2.1: Schematic drawing of the crystal structure of wurtzite Ga-
face and N-face GaN (reproduced with permission from Ref. [43])
nonuniformities of the ternary materials. Different cation arrangement around an N atom affects the local band structure. For instance, in the uniform In x
Al
1-x
N alloy, there is almost no change in bond lengths going from the binaries to In
0.25
Al
0.75
N, whereas in the clustered alloy the In-N bonds are shorter than those in pure InN
(Fig. 2.3). Shorter In-N bonds reflect that the interaction between indium and
nitrogen atoms is stronger than in pure InN, which leads to an increased hybridization between In( p , d ) and N( p ) states. As a result, the N( p ) states, which primarily
Figure 2.2: Schematic arrangement of atoms for InN, AlN, and GaN.
Nitrogen is in the middle and its four nearest-neighbor cations are shown.
The bond lengths in Å are indicated (reproduced with permission from
Ref. [44])

2.1. CRYSTAL STRUCTURE AND MATERIAL PROPERTIES 9
Figure 2.3: Schematic arrangement of atoms for In
0.25
Al
0.75
N in the uniform and clustered cases. Nitrogen atom is in the middle, and only the configuration around nitrogen atoms with largest number of In atoms as
nearest neighbor is shown (reproduced with permission from Ref. [44])
form the valence band, are pushed up, changing the band edge shape and its width
[44].
AlN, GaN, InN and their alloys have direct band gaps. The maximum of the valence band and the minimum of the conduction band are situated at the Γ point of the Brillouin zone. The alloy band gaps can be tuned from 0.7 eV for InN,
through 3.42 eV for GaN, up to 6.2 eV for AlN, as shown in Fig. 2.4. In practice,
however, InGaN and AlInN alloys with a high In content are yet hardly used because
Figure 2.4: Energy band gap as a function of lattice parameter a for
III-nitride alloys (adapted from Ref. [45])

10 CHAPTER 2. MATERIAL PROPERTIES OF III-NITRIDES the large bond length mismatch prevents from growing layers of an acceptable structural quality. In addition to broad band gap range, nitride-based devices are well suited to operate in high-temperature environments due to the strong bonds in nitride materials. Besides, the III-nitrides exhibit good thermal conductivity, which allows efficient heat dissipation from devices operating in high-current conditions
as high brightness LEDs [46]. Unfortunately, these advantageous properties are
often counterbalanced by the poor thermal conductivity of the normally employed
sapphire substrates [47] and the high density of native extended defects is among the
limiting factors. The extended defects reduce the lifetime of AlGaN-base devices
at high current/temperature operating conditions [27, 48].
Strong polarization fields in III-nitride heterostructures have important consequences on the optical and electrical properties of nitride-based devices. The existence of a macroscopic polarization in III-nitride layers was first predicted from ab initio
calculations by Bernardini and Fiorentini [49]. It was experimentally confirmed by photoluminescence (PL) measurements [50].
N
p
0 p
TOT
N p
0
Ga p
0 N p
0
N
+
Figure 2.5: Bond distribution and dipole moments around Ga in the
wurtzite, Ga-face GaN (reproduced with permission from Ref. [51])
Fig. 2.5 shows the bond arrangement around a Ga atom in GaN crystal. N
is more electronegative than a group III atom. As a result, a fractional negative charge accumulates close to the N atom in every covalent bond of the crystal. Hence, every bond in the crystal is associated with a permanent dipole moment, which is schematically shown by a green arrow. The sum of the four dipole moments around every group-III atom would add to zero in the ideal wurtzite structure, where the bonds are arranged as perfect tetrahedron. However, the electrostatic interaction between non-bonded anion-cation pairs aligned along the c -axis compresses the
ideal tetrahedral structure [43]. Therefore the dipoles along the
c -axis are not

2.2. POLARIZATION FIELDS 11 compensated by those off the axis and every lattice c-plane exhibit a net charge.
Such charge separation causes the spontaneous polarization. Additionally, strain contributes to the total deformation of the crystal from the ideal wurtzite structure and alters the net charge at every lattice c -planes. The strain-induced contribution to the total polarization is called piezoelectric polarization.
The magnitude of piezoelectric polarization along the c -axis (here labeled as z component) can be
expressed as [43]:
P pz,z
= 2 a − a
0 a
0 e
31
− e
33
C
13
C
33
, (2.1) with a and a
0 being strained and unstrained lattice constants, e
31 electric coefficients, whereas C
13 and C
33 and e
33
– piezoare elastic constants. The total polarization vector P is the sum of the piezoelectric ( P sp
) and spontaneous polarization
( P sp
):
P = P sp
+ P pz
.
(2.2)
The polarization charge within every c -plane cancel the contribution from the adjacent planes within the material. However, at the external boundaries or at the interface between layers with different compositions a net surface charge remains uncompensated. In the latter case, it is the different average value of polarization between the adjacent layers that produces an incomplete charge cancellation. The total polarization is related to the surface bound charge density, σ b
, as:
P · ˆ = σ b
.
(2.3) where ˆ is the surface normal vector. Provided P b and P w are total polarization
(Eq. 2.2) components parallel to the growth direction in the barrier and the well, the
resulting built-in electric field in the wells of an MQW structure can be expressed
as [52]:
F w
= l b
( P b l b
ε w
−
+ l
P w w
ε b
)
, (2.4) where l b and l w are widths of the barrier and the well, ε b and ε w are the respective dielectric constants. In nitride semiconductors, the polarization typically attains large values, creating built-in electric fields at heterostructure interfaces or in the quantum wells of the order of 1 MV/cm. The spontaneous polarization is dominant in AlGaN, while InGaN mainly exhibits the piezoelectric polarization due to the large strain originating from mismatch of lattice constants between InN and GaN
[52].
The presence of the built-in electric field shifts electrons and holes to the opposite sides of the well. In addition to reduced wavefunction overlap and consequently decreased recombination probability, the energy levels of a QW are also affected.
The electric field bends the potential profile so that the quantum well becomes more

12 CHAPTER 2. MATERIAL PROPERTIES OF III-NITRIDES triangular-like, resulting in shifts of both hole and electron levels towards lower energies. However, if a large amount of carriers is photoexcited or injected electrically, electric field arises from unevenly distributed electrons and holes in the opposite direction to the built-in field, effectively screening it. In case of the screened built-in field, the potential profile resembles that of a rectangular quantum well, effectively manifesting as a blueshift of the effective energy band-gap in comparison to the originally shifted energy levels in a distorted potential. For strong built-in fields, the spectral shift can be relatively large, for instance, the peak wavelength of emission from a c -plane green LED shifts from 530–535 nm at low operating current to less than 500 nm at 10 kA/cm 2
[53].
The effect of screening the electric field in AlGaN/AlGaN QWs with photogenerated carriers has been modeled in
Paper B
by coupling Schrödinger and Poisson equations. In this model, energy levels of an electron or a hole in a quantum well are calculated by the stationary Schrödinger equation in the approximation of the
envelope wavefunction [54]:
where V ( z
−
~
2
2 m ∗ d
2 χ ( d z 2 z )
+ V ( z ) χ ( z ) = Eχ ( z )
) is the potential profile of the quantum well, χ ( z ) and m
∗
(2.5) are the wave-
function and the effective mass of a quasiparticle. The Eq. 2.5 is solved numerically
by discretizing the wavefunction in the real space along the z axis to become an
N -dimensional vector:
 χ
1

X =




χ
2
.
..




.
(2.6)
χ
N
The potential is also discretized into the same number of nodes, and the Hamiltonian ˆ = − ~
2
2 m ∗ d
2 d z 2
+ V ( z ) is expressed by a tridiagonal matrix:
ˆ
=
 V
1
+




 m ∗
~
2
∆ z 2
− ~
2 m ∗
2
∆ z
0
2
V
− ~
2
2 m ∗ ∆ z 2
2
+
. .
m ∗
~
2
∆ z 2
.
−
. .
. .
~
.
.
2
2 m
∗
∆ z 2
0 


V
−
N
~
2
2 m ∗ ∆ z 2
2
+ ~ m
∗
∆ z 2



, (2.7) where ∆ z
is the spatial discretization step. Eq. 2.5 can then be approximated by a
typical matrix eigenvalue problem:
ˆ
= EX, (2.8) which is solved numerically for both holes and electrons to find the energy levels and the corresponding wavefunctions. Once the energy levels are known, the twodimensional density of states (DOS) in a QW can be expressed as g c,v
( E ) = m
∗ c,v
π
~
2
X
Θ ( E − E i c,v
) , i =1
(2.9)

2.2. POLARIZATION FIELDS 13 where m ∗ c are i th and m ∗ v are conduction and valence band effective masses, E i c and E v i energy levels of the conduction and valence bands, respectively, and Θ is the
Heaviside step function. Significant screening of the built-in electric field usually involves a considerably large carrier density, therefore different Fermi energies for electrons and holes must be taken into account when computing the thermal distribution of the carriers. These Fermi levels for conduction and valence bands, E c
F and E v
F
, are related to the photogenerated electron-hole plasma sheet density N eh
by the Fermi-Dirac integral [55]:
N eh
=
Z
+ ∞ g c,v
( E ) exp
0
E − E kT c,v
F
+ 1
− 1 d E.
(2.10)
Electron and hole energies here, and further in this thesis, are measured upwards from the bottom of the conduction band, and downwards from the top of the valence
band, respectively. Since no general analytic solution exists for Eq. 2.10, the electron and hole Fermi levels are obtained from Eq. 2.10 by numerical methods. After
resolving the thermal carrier distribution, electron and hole densities corresponding to each energy level ( N i e and N i h
) can be integrated in a similar fashion:
N i e,h
=
Z
E c,v i +1
E c,v i g c,v
( E ) exp
E − E kT c,v
F
+ 1
− 1 d E.
(2.11)
The average spatial distribution of electron and holes corresponding to these levels can be found from the probability for a particle to be found at a given location, defined by the wavefunction. The total densities of electrons and holes are then summed over all the energy levels: n p
(
( z z
)
)
=
=
P i
P i
N i e
N i h
| χ e i
( z ) |
2
2
χ h i
( z )
,
;
(2.12) where χ e i
( z ) and χ h i
( z ) are electron and hole wavefunctions corresponding to conduction and valence band energy levels E i c and E i v . The presence of electric field shifts electron and hole wavefunctions to the opposite sides of the well, therefore creating asymmetric charge distribution in the well. Electrostatic potential created by space charge is expressed by the Poisson equation:
∇ 2 ϕ = e ( p ( z ) − n ( z ))
,
ε
(2.13) with ε being permittivity of the nitride material, and e – elementary charge. The potential originating from the space charge distribution is found by integrating
Eq. 2.13:
V ch
= extra potential: eϕ
. The Schrödinger equation in Eq. 2.5 is adjusted to include this
−
~
2
2 m ∗ d
2
χ ( z ) d z 2
+ ( V ( z ) + V ch
( z )) χ ( z ) = Eχ ( z ) .
(2.14)

14 CHAPTER 2. MATERIAL PROPERTIES OF III-NITRIDES
Eqs. 2.14 and 2.13 form a non-trivial equation system, since
V ch is dependent on the charge distribution which is dependent on wavefunctions and energy levels, which in turn are solutions of this equation. The system could be solved in an iterative
fashion, i.e. finding the wavefunctions and the energy levels from Eq. 2.14, then
finding the new potential V ch
from Eq. 2.13, and repeatedly performing the same
procedure until the self-consistent solution is found. Unfortunately, such equations are known for stiff convergence (especially at higher carrier densities), and various
methods are suggested [56]. In this work, the problem is solved by means of the
iterative relaxation method. The full charge induced potential is not substituted
directly into Eq. 2.14, albeit the current potential is modified by a small part of it,
controlled by the relaxation parameter ω :
V n +1 ch
= (1 − ω ) V n ch
+ ωV
0 ch
, (2.15) where V
0 ch is the newly calculated charge induced potential from the Poisson’s
Eq. 2.13. A smaller value of
ω allows avoiding divergence at a cost of slower computation. In this work, a suitable relaxation parameter is estimated dynamically by comparing the initial charge density induced potential V 1 ch
(i.e. the one evaluated
after solving Eq. 2.5) to the built-in electric field potential
eEd :
ω =
αeEd
,
V 1 ch
(2.16) where α is an extra coefficient to further limit the chance of divergence, E is the built-in electric field in a QW, and d is the QW width. In this work, a value of
α =
1
4 has been used.
In
Paper B , this model has been used to analyze the spectral peak shift due to
the built-in electric field screening by free carriers in Al
0.35
Ga
0.65
N/Al
0.49
Ga
0.51
N
MQWs. Firstly, the sheet carrier density in the experiment has been estimated from the measured PL dependence on the excitation intensity. The PL band peak shift is calculated as the change of the difference between the first conduction and valence band levels, E c
1
− E v
1
, once the self-consistent solution is obtained from Eqs. 2.14
and 2.13. This shift is computed for the range of sheet densities corresponding
to the values attained in the experiment. The comparison between the calculated
dependencies and the experimental data points is depicted in Fig. 2.6, illustrating
the theoretical curves corresponding to the best fit value of the built-in electric field at 300 kV/cm. This value is considerably lower than the theoretical estimate for these MQWs (1 .
2 MV/cm), calculated from Eq. 2.4 using the established sponta-
neous polarization and material constants for AlGaN [57]. Apart from the possibly
inaccurate material constants, the reasons for this discrepancy are unclear. The blueshift due to band filling could only further increase experimental shift instead of making it smaller. In addition, the diminished blueshift could be attributed to band gap renormalization, which has been also neglected in these calculations.
However, in this case the spectral redshift due to the band gap renormalization should be observable in the narrowest MQWs which are not significantly affected

2.2. POLARIZATION FIELDS 15
Figure 2.6: Comparison of experimental PL peak position dependences
(symbols) on sheet carrier densities and calculated curves in AlGaN/AlGaN
MQWs at 8 K (a) and 300 K (b). Well thicknesses are indicated by the built-in field. Another possible mechanism for the reduced built-in electric field involves carrier localization in small scale composition fluctuations and the resulting in-plane spreading of the electric field lines outside the localization area. The evidence of potential fluctuations and carrier localization in these AlGaN
MQWs is presented in Chapter 4. For instance, a rough estimate based on results
reported for grained nitride materials [58] shows that the field reduction by a factor
of two might occur due to this effect.
Nitride structures are typically grown along the c axis utilizing sapphire or, lately,
Si (111) as substrates. However, different spontaneous and piezoelectric polarizations in different constituents of a heterostructure create huge electric fields at the

16 CHAPTER 2. MATERIAL PROPERTIES OF III-NITRIDES interfaces or QWs. While in some cases the field is being utilized, e.g. to form two-dimensional (2D) channels in electronic devices (HEMTs), in photonic devices the fields are usually detrimental. As discussed in the previous section, the built-in electric field increases the radiative lifetime due to reduced electron-hole wave function overlap and causes the pronounced redshift of emission at low carrier density.
Because of that, the QW widths in nitride light emitting devices are limited to about 3 nm. In narrow QWs, the energy levels are shifted high up from the material band gap reducing confinement and increasing the leakage current. One way to overcome this drawback is to grow QWs on nonpolar planes. For such structures, usually grown on (11¯ a 100) ( m
) planes (depicted in Fig. 2.7), the growth
direction is perpendicular to the c axis.
Figure 2.7: Polar c -plane and non-polar a - and m -planes
of a wurtzite structure (based on Ref. [51])
In the work presented in this thesis,
Papers C
and
D
explore properties of
InGaN/GaN QWs grown on non-polar m plane. In such QW structures, the well layer experiences biaxial compressive strain, which modifies alignment of the valence bands. In unstrained nitride semiconductors with a wurtzite structure, three uppermost valence bands at the Brillouin zone center have symmetry Γ
9
, Γ upper
7 and
Γ lower
7
[59]. Usually, they are labeled heavy hole (HH), light hole (LH) and split-off
hole bands. The HH and LH bands are a mixture of p x and p y atomic orbitals with wave functions of the character | X ∓ iY i , and the split-off band is formed of p z orbitals with the | Z i -like wave function. Here wurtzite c axis defines z direction.
When an InGaN layer is grown on c -plane GaN, the layer experiences a biaxial compressive strain, and symmetry and polarization properties of the valence bands
are essentially unchanged [22]. If an InGaN layer is grown on a nonpolar plane,
e.g. the m -plane, the x and y directions are no longer equivalent, and the original
valence band states are strongly modified (see Fig. 2.8). The original
| X ∓ iY i
HH and LH states become | X i and | Y i -like. The in-plane compression along the x axis induces expansion along the y -axis; as a result, the | Y i -like band is shifted in energy below the | Z i -like band. Optical transitions between the conduction and

2.2. POLARIZATION FIELDS
(a) compressive in the c plane (b) compressive in the conduction band edge m plane
17
Figure 2.8: Energy level diagram at the Γ point of compressively strained
InGaN on (a) the c plane and (b) the m -plane. In the c -plane InGaN case, biaxial strain (in the x – y plane) is isotropic.
| HH i and | LH i maintain their character of an equal mixture of | X i and | Y i under strain. In the m -plane
InGaN, biaxial stress (in the z – x plane) induces tensile strain along y . As a result, | X i and | Y i have different energies. Downward arrows indicate
electronic transitions with their polarizations (adapted from Refs. [22, 60])
the three valence bands become polarized; for an InGaN layer grown on m -plane
GaN, the uppermost transition is predominantly x polarized while the second one is z - polarized. The splitting between the two highest hole bands increases with
strain [59], thus, increasing InN molar fraction in an
m -plane InGaN layer would result in a larger valence band separation. Then, assuming thermal hole distribution between the valence bands, optical transitions at elevated temperatures would be more | X i
-like band related and have a larger degree of linear polarization [22].
Quantum confinement further modifies the valence band states, however, the two upper-most levels are considered to originate from the | X i -like and | Z i -like bands, respectively.
In spite of a considerable progress in development of the LEDs and, recently, lasers based on m-plane InGaN QWs, In incorporation into the alloy during the material growth is poor. This complicates achieving the goal of developing green light emitters based on this material platform. Therefore, during the recent years, the interest shifted towards QWs grown on semipolar (20¯ 21) planes (see
Fig. 2.9). Degree of polarization for light emitted from such QWs is lower than
that for the m -plane, but the semi-polar planes seem to be superior in terms of In incorporation and growth of InGaN QWs with a high In content emitting in the
green [61].

18 CHAPTER 2. MATERIAL PROPERTIES OF III-NITRIDES
Figure 2.9: The schematic view of different crystal planes
21), (20¯ 1) and m -plane in wurtzite crystal structure
(reproduced with permission from Ref. [61])
Nitride epitaxial layers are grown using several versions of molecular beam epitaxy
(MBE) and metal-organic chemical vapor deposition (MOCVD) [42]. Due to the
higher growth rates, nitride-based devices are primarily grown using MOCVD. After the substrate preparation by its exposure to ammonia at 600
◦
C, called nitridation,
N and group III atoms are brought close to the growing surface via precursor compounds: ammonia for N, trimethylgallium for Ga, trimethylaluminum for Al and trimethylindium for In. Molecular hydrogen and nitrogen are used as carrier gases. For the growth of doped layers, trimethylmagnesium ( p -type) and silane ( n type) are also flown into the chamber. The deposition occurs via pyrolysis of the precursors on the heated surface. InGaN QWs studied in
Papers C
and
D
have been grown by MOCVD. A modification of MOCVD, called migration-enhanced
MOCVD (MEMOCVD) [62] has been used for AlGaN epitaxial layers and QWs,
as well as AlInN/GaN heterostructures studied in the thesis. In this technique, precursor pulse durations and overlaps have been modified to significantly reduce the density of growth defects and deposit high quality, very thin (with QW less than
2 nm) heterostructures with sharp heterointerfaces and large composition variations
[63]. Nevertheless, the heterostructure interfaces are not as abrupt as in MBE [61].
The growth of ternary nitrides by MOCVD is often complicated because of different optimal temperatures for the binary constituent growth. In the extreme case of AlInN, the binary growth temperatures differ by as much as 600
◦
C (1200
◦
C for AlN and 600
◦
C for InN [64]), which makes it difficult to grow uniform and
defect-free layers. Growth temperatures and III/V ratios should also be individually
optimized for material growth on each of the non-polar or semipolar planes [61].

2.4. DEFECTS 19
Extended defects, primarily dislocations, occupy the central role in the development of nitride material and device technology. Due to the lack of cheap GaN substrates, heteroepitaxy on c-plane sapphire or, to a lesser extent, hexagonal SiC or Si (111) is still the technology of choice for the nitride devices. The heteroepitaxy of nitrides, however, suffers from a large lattice and thermal expansion coefficient mismatch. The growth of µm thick structures introduces dislocations and strain and even strain-induced cracking. A high dislocation density is introduced during the nucleation of the first nitride layer and they tend to thread upwards. The high dislocation density is detrimental to the optical and electrical performance of devices. To reduce the dislocation density in the active region of a device, interme-
diate layers are used, for instance, AlN/AlGaN superlattice [63] or epitaxial lateral overgrowth [65] are being employed. Nevertheless, even in a good quality ternary
nitride layer the dislocation density is about 10
8 to 10
9 cm
− 2
.
The main species of extended defects in nitride semiconductors are different
types of threading dislocations (Fig. 2.10) and stacking faults (Fig. 2.11). The
detrimental effect of the extended defects to the device performance is manifold.
Firstly, they can accumulate point defects, such as anion or N vacancies and serve
as efficient nonradiative recombination centers [25]. Secondly, they may serve as
high conductivity channels in, e.g., LEDs transporting injected carriers through
the structure without recombination [27, 48]. Surprisingly, in some cases, namely,
InGaN QWs, dislocations seem to be surrounded by potential barriers, which pre-
vent carrier capture and the subsequent nonradiative recombination [68]. On the
Figure 2.10: Cross-section microstructure of a 5 µm thick
MOCVD-grown GaN film on a sapphire substrate demonstrating threading dislocations propagating through the film to the surface
(reproduced with permission from Ref. [66])

20 CHAPTER 2. MATERIAL PROPERTIES OF III-NITRIDES
Figure 2.11: Wurtzite structure (a) and a stacking fault (b) projected along the h 1¯ i direction. The broken line indicates the stack-
ing fault plane (reproduced with permission from Ref. [67])
contrary, in AlInN, strong strain-induced alloy composition fluctuations have been
found, which should lead to an efficient nonradiative recombination [69].
In spite of the extensive research, the influence of dislocations to the carrier recombination in nitrides is still under debate. InGaN QW LEDs achieve very high internal quantum efficiency even for dislocation densities of the order of 10 9 cm
− 2 .
Meanwhile, GaAs or InP-based devices cease to emit light at dislocation densities
4 to 5 orders smaller than that. In the research papers included in this thesis, the influence of dislocations on the carrier localization and the nonradiative recombination is also discussed.

T his chapter presents the optical spectroscopy methods and the corresponding experimental setups that have been used to study ternary nitride alloys AlGaN, In-
GaN and AlInN. In this work, the temporal dynamics of photogenerated carriers has been investigated by differential pump-probe spectroscopy and time-resolved photoluminescence (TRPL) measurements. In addition, scanning near-field microscopy
(SNOM) has been used to characterize non-homogeneous PL and EL emission in the near-field (NF) with the subwavelength spatial resolution.
All ultrafast spectroscopy and SNOM (except electroluminescence collection from
LEDs) experiments at KTH use the second and the third harmonic pulses of a mode-locked titanium-doped sapphire (Ti:sapphire) laser as the excitation source.
Ti:sapphire lasers operate at near infrared (NIR) frequencies, thus typically producing the second harmonic in the blue spectral region, and the third in UV. In this work, the second harmonic is used to excite InGaN/GaN QWs, whereas wide bandgap AlGaN and AlInN materials require UV excitation by the third harmonic.
Two Ti:sapphire systems are in use at KTH. The first one is based on a Coherent
Mira 900 femtosecond laser. The repetition rate is 76 MHz. The wavelength is tunable in the range of 700–980 nm, and the typical pulse length is about 150 fs.
The second and third harmonics are generated in a two stage setup [70]. The
principal scheme of the setup is depicted in Fig. 3.1. Firstly, the second harmonic
is generated by focusing the input beam onto a lithium triborate (LBO) nonlinear crystal. The second harmonic is then split from the fundamental wavelength by a dichroic mirror which is highly reflective for blue light (B-DM in the figure), but transmits NIR. In order to maximize the third harmonic generation, the two beams
21

22
2ω
CHAPTER 3. EXPERIMENTAL TECHNIQUES
ω
Ti:sapphire laser
SHG (LBO)
B-DM
ω
λ/2 plate
ω
2ω
B-DM delay stage
THG (BBO)
3ω
UV-DM
ω
2ω
Figure 3.1: Two stage generation of the third harmonic must be matched in terms of space, time and polarization. The polarization of the second harmonic is rotated by a half-wave plate so that it becomes parallel to the polarization of the fundamental. Furthermore, the temporal delay of the fundamental is adjusted to counterbalance the delay introduced by dispersion in the LBO crystal. Finally, the two beams are combined again and focused tightly to coincide on a barium triborate (BBO) nonlinear crystal, where the third harmonic generation takes place. The resulting third harmonic beam is separated by a UV dichroic mirror. Considering the power of laser emission is about 1 .
3 W in NIR
(when aligned to 800 nm wavelength), 25–30 mW of UV can be typically produced by the setup.
The second system features an improved successor of the first one, a Coherent
Chameleon II Ti:sapphire laser. The laser emission is tunable in the wider spectral range from 680 to 1080 nm, allowing to span the contiguous range of wavelengths

3.2. ULTRAFAST SPECTROSCOPY TECHNIQUES 23 between 540 and 230 nm with the second and the third harmonics respectively.
The repetition rate is 80 MHz. In this system, an automated harmonics generation setup (Harmonixx by Angewandte Physik & Elektronik) is used. In this setup, the fundamental and the second harmonics always travel in the same optical path until the third harmonic is generated. Instead of separating the beams for temporal delay correction, they cross the delay compensator which is consists of two retractable prisms. The material of the prisms has slightly different speeds of light for the fundamental and the second harmonic, therefore providing an adjustable delay between those beams. This allows for a less complicated setup that is easier to align, automate and is less sensitive to external disturbances. The Ti:sapphire laser in this system can attain the output power in excess of 3 .
5 W (when tuned to
800 nm ), and the third harmonic of 0.4–0 .
5 W can be readily generated.
Ultrafast spectroscopy provides a non-destructive way to investigate time-resolved carrier dynamics in semiconductors. In these techniques, a short laser pulse is used to excite electron and hole pairs in the sample material. Subsequently, the temporal evolution of various properties such as luminescence spectrum, transmittance, reflectance and Raman scattering is studied as a function of time after the excitation. Two ultrafast spectroscopy techniques, time-resolved photoluminescence and pump-probe, are presented in this section.
After the photoexcitation, the generated carriers recombine either in a radiative or a non-radiative way. Radiative recombination yields photoluminescence from the sample, which can be recorded and analyzed. In the time-resolved photoluminescence technique, the instantaneous intensity (or the whole spectrum) of photoluminescence emission is measured as a function of time after the excitation. Assuming bimolecular recombination, the total energy-integrated emission intensity I ( t ) is proportional to the instantaneous total electron and hole densities n ( t ) and p ( t ):
I ( t ) ∝ n ( t ) p ( t ) , (3.1)
If electron and hole pairs form excitons, then the photoluminescence intensity is proportional to the density of excitons n ex
( t
) [71]:
I ( t ) ∝ n ex
( t ) .
(3.2)
In this work, TRPL experiments have been performed by means of a streak camera (model C5680 by Hamamatsu Photonics). The streak camera is an ultrafast device which delivers a two-dimensional image depicting evolution of optical input
channels over time [72]. In our setup, PL emitted from the sample is spectrally
resolved by a spectrograph and directed towards the input slit so that the input

24 trigger signal
CHAPTER 3. EXPERIMENTAL TECHNIQUES sweep circuit sweep electrodes micro-channel plate entrance slit lenses accelerating electrode incident light streak image time ħω ħω time photocatode
(photons → electrons) phosphor screen
(electrons → photons)
Figure 3.2:
Operating principle of the streak camera (adapted from Ref. [72])
channels correspond to a range of different photon energies. The operating principle
of the streak tube is summarized in Fig. 3.2. Firstly, lenses focus the spectrally
resolved photoluminescence from the sample onto a photocatode which converts the incident photons to electrons in a proportional fashion. The electrons are then accelerated through a tube by an accelerating electrode. Furthermore, two sweep electrodes activated by the trigger unit span the electron path in the chamber. The electrodes create a linear voltage across the traversing electrons so that electrons arriving at different times end up at different vertical positions on the image plane.
Thus a conversion from arrival time to space on screen is obtained. Afterwards, the photoelectrons hit the micro-channel plate (MCP), where their number is multiplied many times in order to enable detection of weak signals. Finally, the electrons from the MCP are converted back to light by inducing fluorescence on a phosphor screen, which is imaged by a charge-coupled device (CCD) camera.
Weak TRPL signals can be resolved by means of a long integration time in synchroscan mode. In this mode, the trigger unit detects a portion of the laser pulse from the mode-locked laser and generates the synchroscan signal at the same frequency as the laser repetition rate. Hence each shot is temporally aligned to the arrival of the excitation pulse. Therefore many shots representing the same

3.2. ULTRAFAST SPECTROSCOPY TECHNIQUES 25
PL decay can be digitally accumulated. The acquired two-dimensional image represents the photoluminescence intensity as a function of photon energy and time after the excitation. An example of photoluminescence decay in InGaN QW is
shown in Fig. 3.3. Photoluminescence intensity transients for a desired wavelength
range, and vice versa, PL spectra at a desired time can be obtained by integrating
this two-dimensional image. The process is illustrated in Fig. 3.3, where the total
integrated PL spectrum and PL intensity transient are obtained. The temporal resphoton energy PL intensity ħω
ħω
1
2
I ( ħω,t ) d( ħω )
Figure 3.3: Example of an acquired streak image olution of the streak camera working in synchroscan mode depends on many factors such as fluctuations of the laser power, the noise of the accelerating voltage and the
dispersion of electron initial velocity [73]. In general, synchronization fluctuations,
also known as jitter, can be partially corrected by realigning an image of each single shot according to the center of gravity of the captured part of the excitation beam or another reference. In this work, the resolution of about 3 ps has been achieved in, for instance, TRPL measurements in
Paper G .

26 CHAPTER 3. EXPERIMENTAL TECHNIQUES
In this form of ultrafast spectroscopy, the sample is excited by one pulse (pump), while the changes it induces are investigated by the second pulse (probe), which is delayed with reference to the pump. One of the main advantages of this technique is the time resolution, which is principally limited only by the temporal width of the laser pulse, or the jitter between the lasers if two different lasers are used. In this thesis, the changes in transmittance of AlGaN MQW has been studied. The differential transmission method assumes that the absorption coefficient is modified by the presence of free photogenerated carriers. Neglecting the time dependence for many body effects and changes in the matrix elements for interband transitions, the change in the absorption coefficient at photon energy hν can be expressed as
[74]:
∆ α ( hν ) = α
0
( hν ) (1 − f e
( E e
) − f h
( E h
)) , (3.3) where α
0
( hν ) is the absorption coefficient of the unexcited semiconductor at hν , f e
( E e
) and f h
( E h
) are the electron and hole distribution functions, whereas E e and
E h represent electron and hole energies, corresponding to the photon energy hν .
laser
BS pump
λ/2 plate probe chopper sample iris lock-in amplifier detector polarizer delay stage
Figure 3.4: Differential transmission pump-probe setup

3.3. NEAR-FIELD OPTICAL MICROSCOPY 27
The scheme of our differential transmission pump-probe setup is illustrated in
(Fig. 3.4). In this work, a degenerate pump-probe configuration is employed, that
is, both the pump and the probe originate from the same third harmonic beam, which is divided by a beam splitter. Both beams are then focused on the same point of interest on the sample, and the intensity of the transmitted probe is measured by a detector. The probe is delayed by passing through a mechanical delay stage. In order to improve experiment sensitivity, the pump beam is mechanically chopped.
Thus even small changes in the probe intensity are detected by means of lock-in amplification provided their periodicity matches the actual modulation of the pump by a chopper.
Depending on the pump scattering from the sample, a fraction of the pump intensity may reach the detector, thus deteriorating the signal-tonoise ratio. In this case, the problem is mitigated by rotating the polarization of the probe pulse by means of a half-wave plate. Afterwards, the undesired pump contribution is rejected by a polarizer which is aligned orthogonally with respect to the pump polarization and is placed before the detector. Furthermore, all optical noise originating from both the pump and the environment is partially blocked by passing the probe through an iris diaphragm. Ultimately, the probe pulse can be coupled into an optical fiber for delivery to the detector. The fiber core aperture acts as a tiny diaphragm which is sensitive to any coupling misalignment. Such an arrangement has proved to be unnecessarily complex for regular measurements since the coupling needs to be adjusted each time the laser wavelength is tuned.
However, coupling to a fiber can be used to verify the parallelness of the delay stage as well as to check whether the probe beam is not significantly spreading or focusing as a result of the introduced delay.
Conventional spectroscopy techniques can be combined with microscopy to study spatial distribution of various properties of a semiconductor, for instance, by focusing the excitation beam into a tight spot and repeatedly scanning over a chosen line or area. Spatially resolved spectroscopy is particularly relevant in research of nitride ternary alloys, since they are long known to exhibit inhomogeneous composi-
tion [10]. Nevertheless, classical microscopy is subject to the well known diffraction
limit as even a perfect optical lens can not focus light onto a spot that is smaller than so called Airy disk pattern for the given optical configuration. The minimum size of an object that can be resolved by a diffraction-limited system is expressed
by the limit of Abbe [75]:
d min
≈
λ
2 n sin θ
, (3.4) where λ is the wavelength of the light and n sin θ is the numerical aperture used for imaging. For imaging in air ( n ≈
1), equation 3.4 effectively limits the res-
olution at λ/ 2. However, the remaining optical information is contained in the non-propagating near-field (NF) modes, also known as the evanescent field. The

28 CHAPTER 3. EXPERIMENTAL TECHNIQUES amplitude of the evanescent field decays exponentially on a subwavelength scale
[76]. The basic idea behind NF microscopy systems is to use a sharp probe to
perturb the evanescent field so that a part of it is scattered to a propagating wave, which can be later detected by conventional methods. The first proposal to increase optical resolution by harvesting the evanescent field dates back to as far as
1928, when the Irish physicist E. H. Synge proposed the idea of using a subwavelength pinhole in a metal plate or a metal-coated quartz cone with the open tip
[77]. Although he also envisioned the use of piezoelectric positioning for scanning
microscopy, such a system unfortunately could not be fabricated at the time as the required technology took more than 50 years to mature. Accelerated by the progress in atomic force microscopy and scanning tunneling microscopy, the first
experimental tests of a complete SNOM system were performed in 1984 [78, 79].
In general, there are many possible configurations of scattering the NF light, however only so called aperture SNOM has been employed in the experimental work of this thesis. In the aperture SNOM, the evanescent field interacts with a subwavelength aperture, for example a tapered optical fiber, and the scattered light is accordingly transmitted through the same aperture. Alternatively, the sample can be excited in the NF by the light propagating from the aperture. The most popular aperture SNOM configurations for investigating PL and EL from semiconductors are briefly reviewed in the next section.
Illumination mode
In the illumination mode (Fig. 3.5 (a)), the sample is excited in the NF through
a probe aperture, and the emitted luminescence is collected by conventional optics in the far-field (FF). Although the size of the excitation spot is defined by the probe aperture size, photogenerated carriers can diffuse, effectively reducing the resolution to the diffusion length. For instance, carrier diffusion lengths compara-
ble to emission wavelength are experimentally evidenced in InGaN QW [80], thus
negating the resolution advantage of illumination-mode SNOM versus conventional microscopy techniques in such cases. On the other hand, if the illumination mode is complemented by other SNOM modes which are unaffected by diffusion, this effect
can be exploited to study carrier transport properties [12, 81]. Illumination mode
has not been used in this thesis.
Collection mode
In terms of light propagation, the collection mode (Fig. 3.5 (b)) is nearly identical
to the illumination mode, albeit illumination and collection are reversed. However, the collection mode does not suffer from the aforementioned resolution drawback.
In addition, the sample might be excited by total internal reflection in the so-
called dark-field mode [82]. In the collection mode, a larger area is illuminated,
thus a considerable amount of FF light can be collected through the sides of the

3.3. NEAR-FIELD OPTICAL MICROSCOPY
(a)
fiber tip
(b)
fiber tip
(c)
sample fiber tip sample sample
collection optics
Figure 3.5: Aperture SNOM configurations: (a) illumination mode, (b) collection mode, and (c) illumination-collection mode
29 fiber probe, possibly even masking the near-field contribution or deteriorating the resolution. Therefore metal-coated fiber tips have been used for collection mode experiments included in this thesis.
Illumination-collection mode
The illumination-collection (I-C) mode (Fig. 3.5 (c)), sometimes also called reflec-
tion or internal reflection mode, combines the resolution advantages of both the collection and the illumination modes. Light passes through the same probe aperture twice, thus significantly enhancing the spatial resolution. It has been predicted
[83, 84] and demonstrated experimentally [83, 85] that a subwavelength resolution
can be obtained by I-C mode even with uncoated fiber probes. However, the superior resolution of I-C mode comes at a cost of reduced signal level. Firstly, only a tiny fraction of the excitation beam exits at the aperture and is coupled to the sample in the NF. What is more, not all photogenerated carriers result in radiative recombination, particularly if experiments are carried out at RT. Lastly, not all the emitted photons are collected back through the same probe aperture. Aside from possibly low signal intensity, optical signal in I-C mode SNOM is also particularly vulnerable to be obscured by the luminescence from the fiber itself. Since the coupled excitation intensity potentially exceeds the PL signal by many orders of magnitude, even relatively weak emission emission from the fiber can completely overwhelm the signal. In this work, the relatively lowest self-luminescence in I-C mode has been observed with Optran UVNS multimode (50 µm core diameter) pure

30 CHAPTER 3. EXPERIMENTAL TECHNIQUES
1 . 0
0 . 5
4 0 0 w a v e l e n g t h ( n m )
3 5 0 3 0 0 f i b e r s e l f - l u m i n e s c e n c e l a s e r r e p l i c a l a s e r
( f i l t e r e d )
0 . 0
3 . 0 3 . 5 4 . 0 p h o t o n e n e r g y ( e V )
4 . 5
Figure 3.6: Self luminescence from a pure silica core fiber silica core fiber from CeramOptec. In addition, this fiber type has not shown any significant degradation during measurements even under prolonged excitation with
UVC (wavelengths up to 250 nm). However, even pure silica is known to emit a
broad photoluminescence spectrum consisting of several emission bands [86]. These
bands span the near-UV and visible range (with the highest energy bands situated around 4 .
2 eV) and are attributed to oxygen vacancies and other defects in SiO
2
.
An example of such emission is shown in Fig 3.6. Additionally, a Raman-like replica
of the exciting laser pulse is seen. The effect of fiber self-luminescence has proved to be most challenging in the slope regions of the silica emission bands, where it can distort the shape of acquired NF spectra if not correctly eliminated by background subtraction. For instance, NF measurements of Al
0.86
In
0.14
N (emitting at
310–340 nm) and GaN (around 365 nm) would be affected most.
Uncoated fiber probes for visible and UV range have been produced from the afore-
mentioned pure silica core fiber by the tube-etching method [87]. This method
requires a fiber to be covered with a coating that is resistant to hydrofluoric acid
(HF). Acrylate jacket has been preferred in this work as it is not only suitable for etching but also makes fibers flexible and mechanically robust. Firstly, an opening in the jacket is scratched in order to be able to remove it after the etching procedure. The fiber is then submerged in a Teflon vessel with 50 % HF solution and left to etch. HF is covered with 0.5–1 centimeter layer of isooctane in order to prevent
HF from evaporating and etching the fiber in an unintended way. Pure silica core multimode fibers etch much slower than telecommunication fibers described in Ref.
[87], therefore the etching time has been experimentally adjusted to approximately
3 .
5 h for a 50% HF solution to obtain the desired aperture size. The etching pro-
cedure is outlined in Fig 3.7. In the beginning, HF and reaction products diffuse

3.3. NEAR-FIELD OPTICAL MICROSCOPY
(a) (b) (c)
31
(d) (e) fresh HF products
Figure 3.7: Tip manufacturing using tube etching method. Timeline of the etching process: (a) initial etching, (b) thinning, and (c) tip formation. Mechanisms of the etching process: (d) initial diffusion-controlled etching and (e) convection-controlled formation of the tip. The flow of
HF and reaction products is indicated (adapted from Ref. [87])
around end of the fiber, however the outer regions of the of the fiber are believed to etch slightly faster. Once a preliminary tapered shape is created, the tip formation
is controlled by convection of HF to the upper region as depicted in Fig. 3.7 (e).
Finally, the fiber is removed from the vessel and the jacket is gently pulled away.
The newly produced tip is washed in a methanol (or isopropanol) bath to remove the remaining HF. New tips are inspected with an optical microscope. Although the diffraction limited microscope can not resolve the apex of a tip, the obvious failures of the process such as accidentally broken or under-etched fibers are easily discriminated. The reproducible quality of this method has been also verified
by a scanning electron microscope (SEM) (see Fig. 3.8). The large scale image
demonstrates different etching speeds for pure silica and fluorine-doped cladding.
Typical aperture sizes vary between 75–150 nm. It can be shown that in the simple dipole-dipole interaction approximation, the SNOM resolution depends only on the
aperture of the probe and the sample-probe separation distance [76]. In this sense,
tips would be desired to be as sharp as possible. However, it has been demonstrated that throughput of a subwavelength aperture is inversely proportional to the fourth

32
(a) core (50 μm)
CHAPTER 3. EXPERIMENTAL TECHNIQUES
(b)
Figure 3.8: SEM images of fiber probes manufactured by tubeetching from UV pure silica core fiber: (a) large scale image of the etched fiber and (b) close-up of the apex region
power of the aperture size [88]. Hence the resolution improvement would be insep-
arable from the greatly diminished NF contribution, and tips of about 100 nm have been found to provide the optimal compromise between the resolution and the throughput for materials studied in this thesis. The resolution is also dependent on a low cone angle of the probe, which introduces huge propagation losses in a long waveguide which can not fit a single mode of propagation. This problem can be addressed by multiple-tapered probes where a sharp-angled probe is attached to a
large-angled mesa [89].
Tapered probes can later be coated with metal, aluminum being a common
choice [90, 91]. Afterwards, the aperture is opened by focused ion beam milling
[92] or just by applying a definite amount of pressure onto the tip by approaching the surface [93]. In this work, only uncoated probes have been manufactured since,
as discussed in the previous section, they provide adequate resolution in I-C mode.
Furthermore, in-house fabricated uncoated fiber tips are also very cost effective for experimentation, considering that SNOM setup often requires a great deal of trial and error. Ready-made metal coated probes have been instead purchased from a commercial company, Lovalite. They have been used to analyze electroluminescence emission from InGaN LEDs (
Paper F ).
The SNOM system has been designed and built by the Max-Born institute (the
description of such an apparatus can be found in Ref. [94]). Since then, it has been
customized for easier operation, for instance, a more convenient magnetic probe holder design has been implemented. The principal scheme of the SNOM setup
working in I-C mode is depicted in Fig. 3.9. The sample-tip distance is controlled by

3.3. NEAR-FIELD OPTICAL MICROSCOPY preamplifier
OUT
IN lock-in amplifier
IN REF
OUT function generator
OUT proportionalintegral controller y-piezo driver
OUT sample tuning fork dither piezo fiber y-piezo
Nanocube x, y, z excitation beam spectrograph interference filter collimator entrance slit f / D ≈ 6.5
collection illumination
Figure 3.9: Scheme of the SNOM setup in I-C mode. Illumination and collection parts of the optical setup are highlighted. Some tiny parts like the tuning fork are intentionally enlarged out of scale to be visible
33

34 CHAPTER 3. EXPERIMENTAL TECHNIQUES
means of shear-force feedback [95]. In this method, the tip of a fiber probe is glued to
one arm of a commercial quartz tuning fork. Tuning forks have a factory resonance frequency at 32 .
768 kHz, however the resonance shifts (usually to lower frequencies such as 30–33 kHz) as a tip is attached. Forks are soldered to a stripboard based holder as solder is more stable than glue and ensures a higher quality factor of a system. In general, resonance quality factors between 10 3 and 2 × 10 3 have been observed to give the best results in terms of stable tracking of the surface and tip wearing. Tested holders with forks can be reused as the fiber is unglued by washing the holder in the acetone bath. The fork is vibrated mechanically by a dither piezo, which is driven by a sinusoidal signal from a function generator. The same signal is used as a reference in lock-in amplification of the signal from the tuning fork. The dither piezo voltage is set not to exceed 500 mV peak-to-peak, which ensures that the tip does not vibrate more than a few nanometers. Even at such small vibration, the fork is able to generate the signal of a few hundred µV at resonance frequency, which is amplified by the preamplifier and fed to the lock-in. The signal from the lock-in is analyzed in time by a proportional-integral controller, which continues to extend y -piezo until the signal from the fork decreases to the predefined level.
As the tip approaches the sample, shear forces cause friction and the vibration amplitude is damped. The proportional-integral controller then enters a feedback loop maintaining the constant tip-sample distance by extending or retracting y piezo to keep the fork vibration amplitude at a constant level, usually set at about
90–95 % of the resonance value. Since the maximum extension range of y -piezo is less than 10 µm, a combination of a y -motor and y -piezo is used for approach procedure in practice. The computer program tries to reach contact by extending y -piezo to the maximum, and retracts again in case of failure. The y -motor is then rotated a few steps and the procedure is repeated until the contact is reached.
Once the tip is verified to be close to the surface, the control is surrendered to the proportional-integral electronics for the scanning experiment. During the scan, the tip is controlled in x and z directions by Nanocube, a commercial component which can precisely move in all directions within a 100 µm range.
In this work, the optical signal from the probe is recorded by a spectrograph
(SPEC-10 system by Princeton Instruments) equipped with a liquid nitrogen cooled
CCD. In collection mode (see “collection” part in Fig. 3.9), the light from the fiber
is first collimated to a parallel beam by an objective lens. The collimated beam is then guided to the spectrograph and focused onto a slit by another lens. This lens should have an f-number that is equal or close to the f-number of the spectrograph
(6.5 for this system) so that all the incident light is analyzed by the spectrograph without sacrificing the spectral resolution. Furthermore, it is usually desirable to filter out the excitation beam, therefore an interference filter is placed in the path of the collimated beam. When working in I-C mode (see “illumination” part in
Fig. 3.9), the excitation beam must be delivered through the same fiber. In this
case, the same interference filter also acts as a dichroic mirror. By using external mirrors, the laser beam is guided to reflect from the filter in such a way that it propagates exactly in the same path as the collimated beam from the fiber, albeit

3.3. NEAR-FIELD OPTICAL MICROSCOPY 35 in the opposite direction. Hence the same objective lens is employed to couple the excitation beam into the fiber. In practice, the excitation alignment is carried out by observing the signal level of laser light emanating from the other end (that is, the “tip” side) of the fiber. All the alignment is normally performed without a tip as the fiber probe is spliced by using an arc fusion splicer and subsequently glued to the tuning fork only as a last step.
During the measurements, it is important to distinguish NF light within the collected signal. Therefore the probe is first retracted 1–2 µm from the sample and the
FF background is acquired. Moreover, the background spectrum also accounts for fiber self-luminescence and other system noises, if any are present. The tip is then approached to reach the NF contact and the desired area is scanned. Afterwards, the background data is subtracted from the acquired spectra so that only NF data
4 0 0 0
4 0 0 3 8 0 w a v e l e n g t h ( n m )
3 6 0 3 4 0 3 2 0
3 0 0 0
2 0 0 0
1 0 0 0
G a N P L
3 0 0
A l I n N P L
x 1 / 3 b a c k g r o u n d ( f i b e r a n d F F P L )
N F P L a f t e r b a c k g r o u n d s u b t r a c t i o n
N F P L a f t e r b a c k g r o u n d s u b t r a c t i o n
a n d i n t e n s i t y c o r r e c t i o n
G a u s s i a n f i t o f G a N P L l a s e r r e p l i c a
3 . 0 3 . 2 3 . 4 3 . 6 3 . 8 p h o t o n e n e r g y ( e V )
4 . 0 4 . 2
Figure 3.10: Analysis of an acquired near-field spectrum of Al
0.86
In
0.14
N/GaN heterostructure. In this example, the background emission (scaled down 3 times for visualization purposes), consisting of the fiber self-luminescence and some FF PL, is more intense than the collected NF spectrum. The final NF spectrum (orange line) is found by subtracting the background, which is normalized according to the laser replica intensity. The topmost part of the GaN peak is fitted with a Gaussian bell in order to precisely obtain the peak wavelength and intensity

36 CHAPTER 3. EXPERIMENTAL TECHNIQUES is left. This method should be used with caution as the FF spectrum can also vary over the sample surface. Moreover, if the laser peak, the replica or fiber-self luminescence bands are not known to overlap with the expected PL spectra, they can be used to estimate the current level of laser intensity, and the background intensity can be accordingly normalized before subtraction. Since the time of a single scan is limited in order to cover the whole area of interest in a reasonable time, the resulting data is usually relatively noisy. Therefore a smoothing procedure is applied. Fourier smoothing has been used most often. Alternatively, if spectra do no vary much in shape and resemble Gaussian, the top part of a spectra can be fitted with a Gaussian bell. After smoothing, the desired parameters such as peak wavelength, peak intensity or full width at half maximum (FWHM) are extracted.
An example of such analysis of a single spectrum is illustrated in Fig. 3.10. The
same procedure is repeated for each spectrum, resulting in two-dimensional maps for chosen parameters. Examples of such maps can be found in the next chapter as well as in the included papers.

E ven though AlGaN, InGaN and AlInN have the same crystal structure and, in many ways, similar physical properties, carrier localization effects, determined by band potential fluctuations, are largely different. Therefore, in this chapter, these effects will be discussed for each of the ternary nitrides separately.
In this work, carrier localization potentials and carrier dynamics in the extended and localized states have been studied in both, bulk-like, several µm thick epitaxial layers, and QWs. Potential fluctuations in QWs, in addition to those occurring in thin films, are also affected by fluctuations of the QW width as well. In photonic devices, such as LEDs or lasers, the active region is usually based on multiple QWs.
For the best device operation parameters, e.g. the narrowest LED line or the lowest laser threshold, the QW levels in different QWs should be close in energy. Thus, both, alloy composition and defect-based potential fluctuations and well width variations have a critical influence on the light emitting device performance. In bulk
Al x
Ga
1-x
N, several kinds of compositional inhomogeneities are commonly observed: random alloy disorder, alloy fluctuations, alloy segregation and ordering. Randomly distributed alloy systems always present a certain degree of disorder at the scale
of the lattice cell as a result of the statistical occupancy of the lattice sites [34].
Alloy composition variations on a larger scale than the unit cell are usually defined as composition fluctuations. Alloy segregation is the extreme situation for which one sort of atoms or binary constituent condensates within a region of the crystal.
Alloy fluctuations and segregation usually occur within irregular volumes that have limited lateral size and do not present internal structure. On the other hand, or-
37

38 CHAPTER 4. LOCALIZATION EFFECTS IN TERNARY NITRIDES dering refers to the formation of domains containing a superlattice of spontaneous
QWs along the growth direction within small areas of the sample [96] or even extending over the whole sample surface [97]. Besides, other phase inclusions, such as zinc-blende in a wurtzite lattice, may occur [98]. Ordering is likely to occur
as there are several ordered configuration of atoms with formation energy close to
each other and to the perfectly random configuration [99]. In AlGaN, Al segrega-
tion has been observed around threading dislocations in MOCVD grown samples by
TEM measurements [100]. The Al surplus comes from Al-depleted regions which
have been observed within a few nanometers of the dislocation lines. A similar redistribution of Al atoms has been revealed in an AlGaN QW LED under current
crowding conditions [26].
Localization potentials and related carrier dynamics are best visualized via light emission spectroscopy. The spatial resolution of the far-field PL, the most common technique, is typically limited to the emission wavelength by the diffraction limit.
Confocal PL microscopy provides a somewhat better resolution [101], however, does
not allow relating the PL spatial distribution to the surface morphology. The luminescence spatial distribution with a high spatial resolution is usually obtained by cathodoluminescence (CL) spectroscopy. The resolution of the CL measurement is limited by the excited volume and by the carrier diffusion within the sample. Moreover, the measurement requires conductive samples to limit the charging effect. The preparation of doped samples for CL measurements may create artifacts. In fact, doping of GaN related materials has a significant influence on their structure qual-
ity [99]. In this work, light emission (and related compositional) inhomogeneities
have been studied in AlGaN with the help of scanning near-field optical microscopy
(SNOM). So far, SNOM has found a limited application to the study of inhomogene-
ity in AlGaN alloys [102]. This can be explained with the difficulties of using this
technique in the UV range (see Chapter 3). On the other hand, NF measurements
have some advantage with respect to CL measurements. In illumination-collection or collection mode, near-field measurements are less affected by the carrier diffusion in the sample and do not require any specific sample preparation.
In
Paper E , SNOM measurements have revealed distinctive differences among Al-
GaN layers with different alloy compositions. In the SNOM measurements, the presence of localization on a spatial scale larger than the experimental resolution
(in our case, ~100 nm) can be detected by means of the PL peak shift.
Inhomogeneities occurring on a scale smaller than the instrument resolution may be evaluated indirectly from the spectral broadening.
The SNOM measurements show clearly defined microscopic PL energy variations in AlGaN epitaxial layers with 30, 42 and 50% Al content. In the maps of PL
intensity and peak energy (depicted for 42% Al in Fig. 4.1 (a) and (b)), domain-like
structures with clearly defined boundaries can be observed. The correlation with
the PL intensity map (Fig. 4.1 (e)) clearly shows that the regions emitting at longer

4.1. ALGAN FILMS AND QUANTUM WELLS 39
Figure 4.1: Near-field PL maps of the peak intensity (a), peak energy (b), and FWHM (d) for the Al
0.42
Ga
0.58
N layer, along with the near- and far-field spectra (c) and the correlation graphs (e, f) wavelengths (and having a lower AlN molar fraction) experience a shorter carrier lifetime, i.e. domain boundaries have a more pronounced nonradiative recombination. The inverse correlation between the peak energy and the full width at half
maximum (FWHM) of the PL peak (Fig. 4.1 (f)) indicate a larger inhomogeneous
broadening in the boundary areas.
The experiments described in
Paper E
show that PL spectra in the major part of the measured area for the 30% Al layer experience only the homogeneous broadening. However, in the domain boundaries, and, especially, at layers with a larger
Al content the spectra become broader suggesting an inhomogeneous contribution.
This is especially well pronounced in the Al
0.5
Ga
0.5
N layer. The inhomogeneous broadening indicates that band potential fluctuations with a characteristic distance below the experimental resolution, i.e. ~100 nm occur. These potential fluctuations

40 CHAPTER 4. LOCALIZATION EFFECTS IN TERNARY NITRIDES have been evaluated using a similar model to the one employed for m -plane In-
GaN/GaN QWs, just with the density of states adapted to reflect the 3D bulk
material. The model is described later in Section 4.2.
The observed microscopic and nanoscopic potential fluctuations provide an insight into growth of AlGaN layers with a different content. The red shift of the PL emission at the domain boundaries is consistent with previously observed Ga-rich
regions at grain boundaries of an AlGaN layer (Fig. 4.2) [103, 104]. During the
Figure 4.2: Growth schematics of an AlGaN structure showing domains, their boundaries, and preferential dislocation sites initial growth stage, adjacent small islands, from which the growth starts, coalesce
into larger grains [97]. As the islands enlarge, Ga adatoms, having a larger lateral
mobility than Al adatoms, reach the island boundaries more rapidly. Hence, the
Ga concentration in the coalescence regions is higher than in the center of the is-
lands [105]. The composition pattern, which is formed during the coalescence, is
maintained as the growth proceeds vertically. As a result of the coalescence, the domain boundaries usually contain extended defects that form to accommodate the relative difference in crystal orientation among the islands. Decorated with point
defects [25] these extended defects function as nonradiative recombination centers
reducing the time-integrated PL intensity.
Localization potentials in AlGaN QWs have been studied by the transient transmission technique. To that end, five Al
0.35
Ga
0.65
N/Al
0.49
Ga
0.51
N MQW structures of different QW width have been used (
Paper A ). In wavelength-dependent trans-
mission transients, three different regions may be distinguished (depicted in Fig
4.3). At intermediate photon energies (red curves), carriers are excited in the QW

4.1. ALGAN FILMS AND QUANTUM WELLS 41
265 nm
270 nm
272 nm
276 nm
282 nm
0 50 100 150 200
Figure 4.3: Normalized differential transmission transients of 3 .
3 nm
MQW sample for different pump and probe pulse wavelengths extended states above the band gap and the induced transmission decays with a characteristic time of 80–150 ps. At these energies, corresponding to the extended
QW states, the carriers are mobile, can find nonradiative recombination centers and, therefore, the transmission decay is determined by the nonradiative recombination. At lower energies, the transients have a step-like shape indicating that the carrier lifetimes are much longer than the experimental time scale. The long lifetime shows that at these energies carriers are excited into localized states and cannot freely move in the QWs. At higher photon energies close to the barrier band gap
(green curve), the transmission transients change sign from induced transmission to induced absorption. For excitation in the wells and in the barriers compared to excitation just in the wells, a larger number of carriers are excited in the active region. Therefore, the field in the wells is screened shortly after the excitation and the overlap of electron and hole wave functions increases, increasing the, the absorption of the probe pulse. The changes of transmission transient shapes from the step-like to the decaying one has allowed estimating localization potential depths for wells of
different widths. Fig. 4.4 summarizes the results, putting the localization potential
depth at about 80 meV.
This is similar but somewhat larger than the sum of microscopic and nanoscopic localization potentials (~50 meV) obtained by the SNOM measurements on AlGaN epilayers of the same composition as the QWs. The deeper localization potentials

42 CHAPTER 4. LOCALIZATION EFFECTS IN TERNARY NITRIDES
4 . 8
4 . 7
4 . 6
4 . 5
4 . 4
4 . 3
2 6 0
2 6 5
2 7 0
2 7 5
2 8 0
2 8 5
2 9 0 e l l w i d t h ( n m )
Figure 4.4: Energy dependence of the time-resolved transmission shape for
AlGaN MQW samples of different widths. Solid circles ( ) mark PL peak position for a low excitation power, solid triangles ( H ) – estimated band gap of the barrier alloy. DT traces are indicated by the hollow shapes: upward triangles ( 4 ) indicate decaying transients, rectangles ( ) – step-like transients, and downward triangles ( 5 ) – increased absorption in the multiple QW structure suggests that part of the localization potential may be attributed to the well width fluctuations.
m
In most of the previous studies of carrier localization in InGaN layers and QWs structures grown along the c axis have been studied. In general, the localization potentials in InGaN are deeper that in AlInN, preventing localized carriers from reaching sites of the nonradiative recombination. This efficient localization is probably the main reason of the high internal quantum efficiency of InGaN QW-based
LEDs.
In this work, another modification of InGaN QWs has been studied, namely, QWs grown on the nonpolar m
-plane. As mentioned in Chapter 2,
m -plane QWs do not experience built-in electric field, and the electron and hole function overlap is improved, compared to the polar plane, leading to the shorter radiative lifetime

4.2. INGAN M -PLANE QUANTUM WELLS 43 and more efficient recombination. Besides, because of the strain-modified valence band structure, the interband light emission in m -plane QWs is highly polarized.
Firstly, localization potentials have been evaluated from the time-resolved PL measurements, namely, from the dynamics of the degree of polarization of the emitted light. The degree of polarization is defined as
P =
I
⊥
− I k
,
I
⊥
+ I k
(4.1) where I
⊥ and I k are PL intensities perpendicular and parallel to the c axis. The idea between distinguishing between the extended and the localized states through the PL polarization degree is the following: in the extended states, the polarization degree is determined solely by the arrangement of the valence bands (see Chap-
ter 2), and their separation with respect to the carrier distribution function. The
localized, quantum dot-like states might have a different symmetry, different band arrangement compared to the QWs, and, consequently, different polarization prop-
erties [106–108]. Moreover, by performing a time-resolved experiment (see Fig. 4.5)
and applying a diffusion model, it has been possible to evaluate not only the energy
1 . 0 0
0 . 9 5
0 . 9 0
0 . 8 5
8 0 K , 1 0 7 p s
1 6 0 K , 1 0 2 p s
2 4 0 K , 5 2 p s
1 . 0
0 . 8
0 . 6
0 . 4
A
B
0 . 2
0 . 0
- 1 0 0 - 5 0 5 0 1 0 0
A n a l y z e r a n g l e ( d e g )
0 . 8 0
1 0 0 2 0 0
D e l a y t i m e ( p s )
3 0 0 4 0 0
Figure 4.5: Temporal change of the polarization degree for sample A at three different temperatures.
The inset shows normalized room temperature PL intensities for samples A (2 .
5 nm thick
In
0.17
Ga
0.83
N/GaN QW) and B (2 .
5 nm In
0.2
Ga
0.8
N/GaN QW) as a function of the analyzer angle

44 CHAPTER 4. LOCALIZATION EFFECTS IN TERNARY NITRIDES depth of the localization potentials, but also their density. These investigations, in detail described in
Paper C , indeed confirmed that PL from the localized states
have a reduced degree of polarization, and allowed evaluating an average distance between localization sites (200–500 nm) in these technologically important InGaN structures.
Localization potentials in the same m -plane InGaN single QWs have been examined by SNOM as well. These investigations are described in
Paper D . The near-field
maps of different parameters of the PL spectra are presented in Fig. 4.6. Here,
4
3
2
1
(a)
Peak intensity (a. u.)
5
[0001]
0
0 1 2 3 4
µm
(b)
Peak wavelength (nm)
5
A
4
3
D
2 C
1
B
0
0 1 2 3 4
µm
(c)
FWHM (nm)
1 .
0
0 .
9
0 .
8
34
32
30
0 .
7 28
495
494
493
492
491
490
489
0 .
6
0 1 2 3 4
µm
(d) Wavelength (nm)
540 520 500 480 460
26
D
C
B
A
2 .
3 2 .
4 2 .
5 2 .
6 2 .
7
Photon energy (eV)
Figure 4.6: Near-field peak PL intensity (a), peak wavelength (b), and FWHM
(c) maps for the m -plane In
0.2
Ga
0.8
N QW together with the near-field spectra
(d) measured at selected positions indicated in the peak wavelength map

4.2. INGAN M -PLANE QUANTUM WELLS
(a) (b) conduction band
1.5
σ
L
σ
L
σ
L
= 20 meV
= 50 meV
= 100 meV
σ
C 1.0
E g
0.5
σ
V
0.0
valence band space coordinate
0.0
0.2
0.4
0.6
ħω - E g
(eV)
Figure 4.7: Gaussian density of states model: potential fluctuations (a) and the joint DOS computed for different localization depths (b)
45 similarly to the near-field spectra of AlGaN epitaxial layers, a dual localization potential has been observed. The larger scale potential minima are directly revealed by the SNOM, while the smaller, sub-aperture scale, fluctuations have been evaluated from the inhomogeneous broadening of the near field spectra, using the
Gaussian density of states (DOS) model similar to the one suggested by Eliseev et
al. [109].
In this model (illustrated in Fig. 4.7), the effective band gap fluctuations are
assumed to approximately obey a Gaussian distribution . The joint density of states of a homogeneous QW can be expressed in a similar fashion as electron and hole
DOS in Eq. 2.9:
g
0
( E ) =
π
µ
~
2
X i =1
Θ ( E − E g
− E c i
− E i v
) , (4.2) where µ = 1 m
∗ c
+ 1 m
∗ v
− 1 is the reduced effective mass of the electron-hole pair with effective conduction and valence band effective masses m
∗ c
, respectively, E g
, E i v and m
∗ v is the band gap of the QW material, and E i c i th are conduction and valence band
energy levels as in Eq. 2.9. Optical transitions are only allowed between the
same quantum numbers of conduction and valence band energy levels due to the orthogonality of the QW wavefunctions. Considering Gaussian fluctuations of the conduction and valence bands with standard deviations σ c and σ v
(see Fig. 4.7 (a)),

46 CHAPTER 4. LOCALIZATION EFFECTS IN TERNARY NITRIDES the fluctuation of the band gap corresponds to the localization depth σ
L
[109]:
σ
L
= p
σ 2 c
+ σ 2 v
.
(4.3)
The joint Gaussian DOS reflects the superposition of various localization sites according to Gaussian distribution and can be integrated as: g
G
( E ) =
1
√
2 πσ
L
Z
+ ∞
−∞ exp −
E 2
1
2 σ 2
L g
0
( E − E
1
) d E
1
, (4.4) where g
0
( E
) is the initial two-dimensional joint DOS defined in Eq. 4.2. Examples
of the broadened DOS (calculated using theoretically estimated m -InGaN/GaN
band offsets [110]) for
m -plane In
0.2
Ga
0.8
N/GaN heterostructure are depicted for various σ
L
values in Fig. 4.7 (b). Furthermore, the SNOM measurements in I-
C mode can be assumed to correspond to low-excitation conditions, therefore the carrier distribution can be approximated with Boltzmann statistics. Hence the PL intensity would be proportional to:
E
I
P L
( E ) ∝ g
G
( E ) exp − k
B
T
, (4.5) with g
G
( E
) being the Gaussian joint density of states from Eq. 4.4, and
k
B
–
Boltzmann’s constant. Conveniently, the same approximation is also valid for excitons, as the excitonic density of states would be broadened by the same Gaussian tail, and the Boltzmann distribution would represent the low population case of the Bose-Einstein distribution. In addition, the thermal broadening should not be neglected for room temperature experiments. In this work, the model of exciton interaction with phonons has been used in order to estimate the total linewidth Γ as a function of temperature T
[111]:
Γ( T ) = Γ
0
+ βT + exp
γ
~
ω
LO k
B
T
− 1
(4.6) where β represents the exciton–acoustic phonon coupling strength, γ is the exciton–
LO phonon coupling strength, and
~
ω
LO is the LO phonon energy. Γ
0 is calculated
numerically as the linewidth of a PL spectrum from Eq. 4.5.
LO phonon energy
~
ω
LO
= 88 meV is linearly interpolated from values for GaN [112] and InN
[113]. The values of
β = 0 .
035 meV/K and γ = 0 .
75 eV are obtained in a simi-
lar manner to Ref. [114] by fitting Eq. 4.6 to the previous temperature-resolved
far-field PL measurements on these samples, presented in
Paper C . Numerically
calculating the linewidth for room temperature according to Eq. 4.6 reveals almost
linear dependence upon the localization depth σ
L
Γ
RT
≈ 2 .
26 σ
L
+ 47 .
for all but very small σ
L values:
3 (meV). Therefore the FWHM variations of the NF spectra
from 25 nm to 34 nm (portrayed in Fig. 4.6 (c)) correspond to nanoscopic localiza-
tion depths of 36–55 meV, with the average σ
L is 2 .
9 meV).
= 46 meV (the standard deviation

4.2. INGAN M -PLANE QUANTUM WELLS 47
As can be seen from the InGaN PL NF maps (Fig. 4.6 (b)), the microscopic
areas of the lower potential align along the [0001] direction. In the maps of the
near-field PL measured from the top GaN layer (shown in Fig. 4.8), the areas of the
3
2
(a)
Peak intensity (a. u.)
5
4
1
0
0 1 2 3 4
µm
1 .
0
(b)
Peak wavelength (nm)
0 .
9
0 .
8
0 .
7
[0001]
0 .
6
0 .
5
0 1 2 3 4
µm
365 .
2
365 .
0
364 .
8
364 .
6
364 .
4
Figure 4.8: Near-field peak PL intensity (a) and wavelength (b) maps of the top GaN layer in m-plane InGaN/GaN QW structure the SNOM maps of these two layers of the structure shed light into the origin of the localization minima in the QWs. The monolayer steps in m-plane GaN primarily align along [0001] direction, while dislocations tend to align along both, [0001] and as the efficient nonradiative recombination sites, align. This allows assign the lower potential sites in the InGaN QW to the monolayer steps.
The possibility of comparison of the SNOM maps originating from different layers of a heterostructure once again emphasizes the great potential of the technique as compared to the pure morphology or structural microscopy experimental methods.
As revealed by the inhomogeneous broadening analysis above, in m -plane InGaN
QWs the localization potentials are relatively shallow, of the order of 40–50 meV.
This is smaller than typical values of about 100 meV for c -plane QWs with compa-
rable indium content [12]. Taking into account the above-mentioned polarized light
emission properties of the extended and the localized states, the shallow localization potentials are advantageous for applications in polarized light emitters.

48 CHAPTER 4. LOCALIZATION EFFECTS IN TERNARY NITRIDES
The SNOM maps of the different layers of a heterostructure may also be used to correlate nonradiative recombination centers in different layers. This can be illustrated by an example of an Al
0.86
In
0.14
N/GaN heterostructure (
Paper G ). In this
structure, the AlInN layer is 95 nm thick (as compared to the absorption length of
~100 nm), thus, PL is excited in both AlInN and GaN layers. The maps are pre-
sented in Fig. 4.9. The Pearson product-moment correlation coefficient coefficient
(the definition of the Pearson coefficient can be found in, for instance, Ref. [116])
between AlInN and GaN PL intensity maps is equal to r = 0 .
26. Though it is far from the ideal correlation with r = 1, the analysis indicates that it is statistically significant, therefore hinting that some of the nonradiative recombination centers
5
4
3
2
(a)
AlInN PL intensity (a. u.)
6 1 .
0
0
0
.
.
9
8
(b)
GaN PL intensity (a. u.)
1
0
0
.
.
.
0
8
6
1
0
0 1 2 3 4 5
µm
(c)
0 .
AlInN-GaN correlation
7 0 .
4
0 1 2 3 4 5
µm
(d)
GaN peak wavelength (nm)
6 361 .
7
1 .
0
5
0 .
8 4
361 .
6
0 .
6
3
2
0 .
4
1
0 .
2
0 .
7 0 .
8 0 .
9
AlInN PL intensity (a. u.)
0
0 1 2 3 4 5
µm
361
361
.
.
5
4
Figure 4.9: SNOM maps of the AlInN PL intensity (a), the GaN PL intensity (b), correlation between the GaN and AlInN PL intensities,
(c) and the GaN peak wavelength (d)

4.3. ALINN/GAN HETEROSTRUCTURES 49 are at common locations in both layers. These are probably extended defects, i.e.
point-defect decorated dislocations.
As it can be seen from Fig. 4.9, the map of the spectrally integrated AlInN
PL intensity is very different from those of AlGaN and InGaN. It consists of small features with sizes close to the aperture diameter and modest intensity variations
(Fig. 4.9 (a)). No micrometer-size islands are observed. The coefficient of the inten-
sity variation (the standard deviation divided by the mean value) is only 0.05. The near-field PL spectra are very broad (0.4 eV), indicating a very large inhomogeneous broadening.
In general, the properties of the localization potentials in AlInN are very different from those in AlGaN and InGaN. These properties are studied by means of PL dynamics and SNOM in
Paper G . For instance, the huge Stokes shift of
0 .
9 eV (Fig. 4.10) between the absorption edge and the emitting states is uncom-
mon for crystalline semiconductors. The Stokes shift indicates that the localized
1 0
3 6 0 3 4 0 a v e l e n g t h ( n m )
3 2 0
3 0 0 K
3 0 0
1 0
1 0
4 . 0 4 . 5
E n e r g y ( e V )
5 . 0
1 0
1 0
1 0
3 . 4
6 5 0 W / c m 2
3 . 6 3 . 8
E n e r g y ( e V )
4 . 0
2 8 0 K
2 2 0 K
1 6 0 K
8 0 K
2 0 K
4 . 2
Figure 4.10: PL spectra of the AlInN/GaN heterostructure at different temperatures. The inset shows AlInN PLE spectrum at 300 K states, from which the emission occurs, lie deep below the band edge(s). Interestingly, no PL has been observed at the band gap (defined by the photoluminescence excitation (PLE) spectrum) energy, suggesting that the carriers, excited into the extended band edge states, are immediately captured into the localized states. To

50 CHAPTER 4. LOCALIZATION EFFECTS IN TERNARY NITRIDES
- 2 1 0
D e l a y t i m e ( p s )
Figure 4.11: Transient photoreflection of the AlInN/GaN heterostructure measured at the photon energy of 4 .
75 eV, corresponding to the extended state band gap of AlInN evaluate the time of carrier trapping from the extended to the localized states, pump-probe experiment in reflection geometry at the extended state band gap en-
ergy has been applied (Fig. 4.11). As the transient shows, the carriers are removed
from the band edge to the localized states in about a picosecond, thus, no diffusion to find localization minima is involved.
Another peculiar observation is the large (0 .
4 eV) spectral width of the AlInN PL peak. The FWHM of the PL peak does not change with temperature (illustrated in
Fig. 4.10), excitation power density or time (in the time-resolved PL experiments).
Moreover, it is the same in the far- and the near-field spectra. The spectral width, the ultrafast trapping and the SNOM data indicate that the luminescing localized states are small and closely spaced. This allows assigning the origin of these states
to the In clusters, discussed in Chapter 2. The cluster states, depending on the
different cation atom arrangements, should have different energies ensuring the large width of their distribution. On the other hand, the fast (~10 ps) AlInN PL decay and the prolonged (~8 ps) GaN PL rise, as well as the large, compared to
AlInN, GaN PL intensity suggest that carriers, captured to the sub-band edge states of AlInN are not strongly localized, but can participate in an efficient transport towards the GaN layer. Such transport through the sub-band edge states most
probably proceeds via dislocation-related high conductivity channels [69, 117].

T his thesis has been primarily devoted to investigation of localization effects in nitride ternary compounds. In contrast to most nitride near-field spectroscopy studies focusing on c -plane InGaN, SNOM has been employed to analyze potential fluctuations in AlGaN with large Al content, m -plane InGaN and AlInN (to author’s knowledge, it is the first published NF PL assessment of AlInN). Also complemented with conventional ultrafast spectroscopy measurements, this work has helped to improve the current understanding of carrier localization phenomena in these alloys.
The main results are categorized in the conclusions section below.
(i) Carrier dynamics in AlGaN QWs has been analyzed by degenerate pumpprobe (
Paper A ) and PL spectroscopy ( Paper B ). The slow decay of trans-
mission transients for sub-band-gap excitation has suggested that the depth of localization potential is about 80 meV for QWs. Furthermore, carrier localization potentials have been investigated in AlGaN epilayers by SNOM
(
Paper E ). Double scale localization has been observed and explained in con-
nection to AlGaN growth kinetics. The small scale localization has been evaluated from NF spectral width according to inhomogeneous broadening model
(also applied to InGaN in
Paper D ). The localization depth increases with
Al molar fraction and has been estimated at up to 50 meV for Al
0.5
Ga
0.5
N.
(ii) Strongly polarized emission from m -plane InGaN/GaN QWs has been investigated by means of TRPL at different temperatures and polarizations in
Paper C . The degree of polarization has been found to decrease in time after
the photoexcitation. The effect has been explained by carrier transfer to localization sites at the potential minima and supported by the corresponding temporal redshift of the PL peak wavelength. Moreover, potential fluctuations
51

52 CHAPTER 5. CONCLUSIONS AND FUTURE WORK in m -plane InGaN have been further studied by NF spectroscopy in
Paper D ,
and dual localization profile has been evidenced experimentally. The average nanoscopic localization depth of 46 meV has been calculated from the inhomogeneous broadening of NF spectra.
(iii) The light extraction from photonic crystal patterned InGaN/GaN QW LEDs has been studied by NF spectroscopy in
Paper F , and a complex NF mode
mixture has been observed at the device surface. It has been demonstrated that photonic crystals make EL emission more homogeneous in comparison to otherwise identical conventional LEDs.
(iv) Carrier dynamics in AlInN/GaN heterostructures have been examined by
TRPL and NF spectroscopy in
Paper G . PL from AlInN layer exhibits
huge Stokes shift (approx. 0 .
9 eV) and very broad spectra (FWHM up to
0 .
4 eV), suggesting large potential fluctuations. However, in contrast to inhomogeneous InGaN, efficient sub-band-gap carrier transport through AlInN has been evidenced (even at low temperature) by rise of PL from the GaN layer underneath AlInN. Also contrary to AlGaN and InGaN, no pronounced variation in NF spectral peak or width has been observed in AlInN, hinting at densely packed localization centers. In addition, pronounced PL variations of the underlying GaN layer have been explained by fluctuations of built-in electric field at the AlInN/GaN interface.
In terms of possible future directions for research of localization effects and related device efficiency issues in ternary nitrides, this thesis work has barely scratched the surface in this rapidly developing area. New prototype devices and structures often appear without a priori theoretical foundation and spur related research for optimization. That being said, provided the framework of this thesis is to be strictly followed, most relevant (in the author’s opinion) potential improvements and ideas for future work are:
(i) SNOM measurements in illumination-collection mode, particularly those of
AlInN/GaN and InGaN (relatively less so due to better quantum efficiency of
InGaN QWs) suffer from the fiber self-luminescence overlapping with PL in the same spectral range. Although this has been addressed by proper background subtraction, it still contributes to decreased signal-to-noise ratio and prevents from using very low excitation conditions. Under lower excitation power density, localization centers are less likely to be saturated, on that account providing more accurate information on the potential profile. Therefore it is desirable to get rid of the fiber self-luminescence altogether, for instance by upgrading the SNOM setup to use hollow core fibers for I-C mode.

5.2. SUGGESTIONS FOR FUTURE WORK 53
(ii) It is preferable to study carrier localization at cryogenic temperatures as carriers do not have the thermal energy to overcome potential fluctuations, revealing a more pronounced picture of them. In addition, spectral linewidths can serve more straightly as a measure for band gap fluctuations since thermal broadening can be neglected. However, despite the SNOM apparatus being in principle capable of low temperature operation, only RT measurements have been successfully carried out within the timeframe of this thesis. The future work would thus involve repeating currently described NF experiments at low temperature, probably with the exception of AlInN PL, which has not shown significant temperature dependency.
(iii) Lower intensity of NF PL can not be always unambiguously associated with higher rate of non-radiative recombination. Firstly, considering I-C mode, this might rather suggest stronger carrier diffusion away from the tip. Additionally, light collection in any SNOM configuration can be affected by surface topography. Although topography is recorded together with the optical signal, the interpretation of collected intensity from areas with significant surface features is less straightforward. Therefore the SNOM system should be extended by time-resolved PL technique, for instance, by means of time-correlated single photon counting. Analyzing dynamics of spatially resolved PL transients could then aid in distinguishing between the influence of non-radiative recombination, carrier transport and just collection efficiency.
(iv) The reduction of PL polarization degree at localization centers in m -plane In-
GaN QWs should be confirmed at the nanoscale, i.e. by means of polarizationsensitive SNOM mapping. However, the multimode fibers currently used for
I-C mode SNOM do not preserve the polarization well. Therefore a combination of polarization-preserving tips and fibers would need to be explored.

[1] S. Nakamura, T. Mukai, and M. Senoh, Appl. Phys. Lett.
64, 1687 (1994).
[2] 2006 Millenium Technology Prize awarded to UCSB’s Shuji Nakamura , http://www.ia.ucsb.edu/pa/display.aspx?pkey=1475.
[3] S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita,
H. Kiyoku, and Y. Sugimoto, Jpn. J. Appl. Phys.
35, L74 (1996).
[4] U. Kaufmann, M. Kunzer, K. Köhler, H. Obloh, W. Pletschen, P. Schlotter,
J. Wagner, A. Ellens, W. Rossner, and M. Kobusch, Phys. Status Solidi A
192, 246 (2002).
[5] Y. Narukawa, M. Ichikawa, D. Sanga, M. Sano, and T. Mukai, J. Phys. D:
Appl. Phys.
43, 354002 (2010).
[6] Packaged LED market to grow from $11.4bn in 2012 to $17.1bn in 2018 , http://www.semiconductor-today.com/news_items/2012/AUG/LED_090812.html.
[7] T. Gessmann and E. F. Schubert, J. Appl. Phys.
95, 2203 (2004).
[8] M. Crawford, IEEE J. Sel. Topics Quantum Electron.
15, 1028 (2009).
[9] O. Ambacher, J. Majewski, C. Miskys, A. Link, M. Hermann, M. Eickhoff,
M. Stutzmann, F. Bernardini, V. Fiorentini, V. Tilak, B. Schaff, and L. F.
Eastman, J. Phys.: Condens. Matter 14, 3399 (2002).
[10] S. Chichibu, K. Wada, and S. Nakamura, Appl. Phys. Lett.
71, 2346 (1997).
[11] S. F. Chichibu, H. Marchand, M. S. Minsky, S. Keller, P. T. Fini, J. P.
Ibbetson, S. B. Fleischer, J. S. Speck, J. E. Bowers, E. Hu, U. K. Mishra,
S. P. DenBaars, T. Deguchi, T. Sota, and S. Nakamura, Appl. Phys. Lett.
74, 1460 (1999).
[12] A. Kaneta, M. Funato, and Y. Kawakami, Phys. Rev. B 78, 125317 (2008).
[13] M. Funato and Y. Kawakami, J. Appl. Phys.
103, 093501 (2008).
[14] F. Bernardini and V. Fiorentini, Phys. Status Solidi B 216, 391 (1999).
55

56 BIBLIOGRAPHY
[15] F. Wu, Y.-D. Lin, A. Chakraborty, H. Ohta, S. P. DenBaars, S. Nakamura, and J. S. Speck, Appl. Phys. Lett.
96, 231912 (2010).
[16] A. E. Romanov, T. J. Baker, S. Nakamura, J. S. Speck, and E. U. Group, J.
Appl. Phys.
100, 023522 (2006).
[17] R. M. Farrell, E. C. Young, F. Wu, S. P. DenBaars, and J. S. Speck, Semicond.
Sci. Technol.
27, 024001 (2012).
[18] W. G. Scheibenzuber, U. T. Schwarz, R. G. Veprek, B. Witzigmann, and A.
Hangleiter, Phys. Rev. B 80, 115320 (2009).
[19] S. Takagi, Y. Enya, T. Kyono, M. Adachi, Y. Yoshizumi, T. Sumitomo, Y.
Yamanaka, T. Kumano, S. Tokuyama, K. Sumiyoshi, N. Saga, M. Ueno, K.
Katayama, T. Ikegami, T. Nakamura, K. Yanashima, H. Nakajima, K. Tasai,
K. Naganuma, N. Fuutagawa, Y. Takiguchi, T. Hamaguchi, and M. Ikeda,
Appl. Phys. Express 5, 082102 (2012).
[20] M. T. Hardy, D. F. Feezell, S. P. DenBaars, and S. Nakamura, Materials
Today 14, 408 (2011).
[21] UCSB researchers achieve world’s first violet nonpolar vertical-cavity laser technology , http://engineering.ucsb.edu/news/643.
[22] H. Masui, H. Yamada, K. Iso, S. Nakamura, and S. P. DenBaars, J. Phys. D:
Appl. Phys.
41, 225104 (2008).
[23] Y.Taniyasu, M. Kasu, and T. Makimoto, Nature 441, 325 (2006).
[24] M. Shur, M. Shatalov, A. Dobrinsky, and R. Gaska, “Deep Ultraviolet Light-
Emitting Diodes,” in GaN and ZnO-based Materials and Devices , Vol. 156 of Springer Series in Materials Science , S. Pearton, ed., (Springer Berlin
Heidelberg, 2012).
[25] A. Pinos, S. Marcinkevičius, M. Usman, and A. Hallén, Appl. Phys. Lett.
95,
112108 (2009).
[26] A. Pinos, S. Marcinkevičius, J. Yang, Y. Bilenko, M. Shatalov, R. Gaska, and
M. S. Shur, Appl. Phys. Lett.
95, 181914 (2009).
[27] A. Pinos, S. Marcinkevičius, and M. S. Shur, J. Appl. Phys.
109, 103108
(2011).
[28] Sensor Electronic Technology, Inc. reaches milestone UVC LED efficiencies of over 10% , http://www.s-et.com/news/2012-04-23.pdf.
[29] J. Kuzmik, IEEE Electron Device Lett.
22, 510 (2001).

BIBLIOGRAPHY 57
[30] A. Teke, S. Gökden, R. Tülek, J. H. Leach, Q. Fan, J. Xie, U. Özgür, H.
Morkoç, S. B. Lisesivdin, and E. Özbay, New J. Phys.
11, 063031 (2009).
[31] L. Zhou, D. A. Cullen, M. R. McCartney, J. H. Leach, Q. Fan, H. Morkoç, and D. J. Smith, Phys. Status Solidi C 7, 2436 (2010).
[32] D. Billingsley, J. Yang, R. Gaska, and M. Shur, J. Cryst. Growth 327, 98
(2011).
[33] R. Butté, J.-F. Carlin, E. Feltin, M. Gonschorek, S. Nicolay, G. Christmann,
D. Simeonov, A. Castiglia, J. Dorsaz, H. J. Buehlmann, S. Christopoulos, G.
Baldassarri Höger von Högersthal, A. J. D. Grundy, M. Mosca, C. Pinquier,
M. A. Py, F. Demangeot, J. Frandon, P. G. Lagoudakis, J. J. Baumberg, and
N. Grandjean, J. Phys. D: Appl. Phys.
40, 6328 (2007).
[34] E. F. Schubert, E. O. Göbel, Y. Horikoshi, K. Ploog, and H. J. Queisser,
Phys. Rev. B 30, 813 (1984).
[35] D. C. Reynolds, K. K. Bajaj, C. W. Litton, P. W. Yu, J. Klem, C. K. Peng,
H. Morkoç, and J. Singh, Appl. Phys. Lett.
48, 727 (1986).
[36] M. Ikeda, K. Nakano, Y. Mori, K. Kaneko, and N. Watanabe, J. Cryst.
Growth 77, 380 (1986).
[37] J. R. P. Schneider, E. D. Jones, J. A. Lott, and R. P. Bryan, J. Appl. Phys.
72, 5397 (1992).
[38] Y. Yayon, A. Esser, M. Rappaport, V. Umansky, H. Shtrikman, and I. Bar-
Joseph, Phys. Rev. Lett.
89, 157402 (2002).
[39] S. Hong and J. Singh, Appl. Phys. Lett.
49, 331 (1986).
[40] K. B. Nam, J. Li, M. L. Nakarmi, J. Y. Lin, and H. X. Jiang, Appl. Phys.
Lett.
84, 5264 (2004).
[41] K. P. O’Donnell, T. Breitkopf, H. Kalt, W. V. der Stricht, I. Moerman, P.
Demeester, and P. G. Middleton, Appl. Phys. Lett.
70, 1843 (1997).
[42] H. Morkoç, Handbook of Nitride Semiconductors and Devices, Materials Properties, Physics and Growth , Vol. 1 of Handbook of Nitride Semiconductors and
Devices (Wiley, Weinheim, 2008).
[43] O. Ambacher, J. Smart, J. R. Shealy, N. G. Weimann, K. Chu, M. Murphy,
W. J. Schaff, L. F. Eastman, R. Dimitrov, L. Wittmer, M. Stutzmann, W.
Rieger, and J. Hilsenbeck, J. Appl. Phys.
85, 3222 (1999).
[44] I. Gorczyca, S. P. Łepkowski, T. Suski, N. E. Christensen, and A. Svane,
Phys. Rev. B 80, 075202 (2009).

58 BIBLIOGRAPHY
[45] D. Yoo, Ph.D. thesis, Georgia Institute of Technology, 2007.
[46] S. C. Jain, M. Willander, J. Narayan, and R. V. Overstraeten, J. Appl. Phys.
87, 965 (2000).
[47] A. Khan, K. Balakrishnan, and T. Katona, Nat. Photonics 2, 77 (2008).
[48] A. Pinos, S. Marcinkevičius, J. Yang, R. Gaska, M. Shatalov, and M. S. Shur,
J. Appl. Phys.
108, 093113 (2010).
[49] F. Bernardini, V. Fiorentini, and D. Vanderbilt, Phys. Rev. B 56, R10024
(1997).
[50] P. Lefebvre, J. Allègre, B. Gil, H. Mathieu, N. Grandjean, M. Leroux, J.
Massies, and P. Bigenwald, Phys. Rev. B 59, 15363 (1999).
[51] A. Pinos, Ph.D. thesis, Royal Institute of Technology, 2011.
[52] V. Fiorentini, F. Bernardini, F. Della Sala, A. Di Carlo, and P. Lugli, Phys.
Rev. B 60, 8849 (1999).
[53] D. Queren, A. Avramescu, G. Brüderl, A. Breidenassel, M. Schillgalies, S.
Lutgen, and U. Strauß, Appl. Phys. Lett.
94, 081119 (2009).
[54] C. Weisbuch and B. Vinter, Quantum Semiconductor Structures: Fundamentals and Applications (Academic Press, 1991).
[55] M. Fox, Optical Properties of Solids , 2nd ed. (Oxford University Press, Oxford, 2010).
[56] M. Pourfath, H. Kosina, and S. Selberherr, J. Comput. Electron.
5, 155
(2006).
[57] E. Kuokštis, W. H. Sun, C. Q. Chen, J. W. Yang, and M. A. Khan, J. Appl.
Phys.
97, 103719 (2005).
[58] V. Y. Kachorovskii and M. S. Shur, Appl. Phys. Lett.
86, 012101 (2005).
[59] S. Ghosh, P. Waltereit, O. Brandt, H. T. Grahn, and K. H. Ploog, Phys. Rev.
B 65, 075202 (2002).
[60] H. Masui, H. Yamada, K. Iso, H. Hirasawa, N. N. Fellows, J. S. Speck, S.
Nakamura, and S. P. DenBaars, Phys. Status Solidi A 205, 1203 (2008).
[61] Y. Zhao, S. Tanaka, Q. Yan, C.-Y. Huang, R. B. Chung, C.-C. Pan, K. Fujito,
D. Feezell, C. G. Van de Walle, J. S. Speck, S. P. DenBaars, and S. Nakamura,
Appl. Phys. Lett.
99, 051109 (2011).
[62] J. Zhang, X. Hu, A. Lunev, J. Deng, Y. Bilenko, T. M. Katona, M. S. Shur,
R. Gaska, and M. A. Khan, Jpn. J. Appl. Phys.
44, 7250 (2005).

BIBLIOGRAPHY 59
[63] M. S. Shur and R. Gaska, Proc. SPIE 6894, 689419 (2008).
[64] R. B. Chung, O. Bierwagen, F. Wu, S. Keller, S. P. DenBaars, J. S. Speck, and S. Nakamura, Jpn. J. Appl. Phys.
50, 101001 (2011).
[65] R. Jain, W. Sun, J. Yang, M. Shatalov, X. Hu, A. Sattu, A. Lunev, J. Deng, I.
Shturm, Y. Bilenko, R. Gaska, and M. S. Shur, Appl. Phys. Lett.
93, 051113
(2008).
[66] J. Speck and S. Rosner, Physica B 273, 24 (1999).
[67] J. Kioseoglou, G. P. Dimitrakopulos, P. Komninou, and T. Karakostas, Phys.
Rev. B 70, 035309 (2004).
[68] F. Hitzel, G. Klewer, S. Lahmann, U. Rossow, and A. Hangleiter, Phys. Rev.
B 72, 081309 (2005).
[69] A. Mouti, J.-L. Rouvière, M. Cantoni, J.-F. Carlin, E. Feltin, N. Grandjean, and P. Stadelmann, Phys. Rev. B 83, 195309 (2011).
[70] F. Rotermund and V. Petrov, Opt. Lett.
23, 1040 (1998).
[71] C. Netzel, V. Hoffmann, T. Wernicke, A. Knauer, M. Weyers, M. Kneissl, and N. Szabo, Journal of Applied Physics 107, 033510 (2010).
[72] Guide to Streak Cameras , http://www.hamamatsu.com, Hamamatsu Photonics K.K., System Division, 812 Joko-cho, Hamamatsu City, Japan (2002).
[73] W. Uhring, C. V. Zint, P. Summ, and B. Cunin, Rev. Sci. Instrum.
74, 2646
(2003).
[74] J. Shah, Ultrafast spectroscopy of semiconductors and semiconductor nanostructures (Springer, 1996).
[75] E. Abbe, Archiv f. Mikroskop. Anat.
9, 413 (1873).
[76] M. Ohtsu and K. Kobayashi, Optical Near Fields: Introduction to Classical and Quantum Theories of Electromagnetic Phenomena at the Nanoscale
(Springer, 2004).
[77] E. Synge, Phil. Mag.
6, 356 (1928).
[78] D. W. Pohl, W. Denk, and M. Lanz, Appl. Phys. Lett.
44, 651 (1984).
[79] A. Lewis, M. Isaacson, A. Harootunian, and A. Muray, Ultramicroscopy 13,
227 (1984).
[80] A. Kaneta, T. Hashimoto, K. Nishimura, M. Funato, and Y. Kawakami, Appl.
Phys. Express 3, 102102 (2010).

60 BIBLIOGRAPHY
[81] A. Kaneta, T. Mutoh, Y. Kawakami, S. Fujita, G. Marutsuki, Y. Narukawa, and T. Mukai, Appl. Phys. Lett.
83, 3462 (2003).
[82] R. C. Reddick, R. J. Warmack, and T. L. Ferrell, Phys. Rev. B 39, 767
(1989).
[83] S. I. Bozhevolnyi and B. Vohnsen, J. Opt. Soc. Am. B 14, 1656 (1997).
[84] R. Müller and C. Lienau, Appl. Phys. Lett.
76, 3367 (2000).
[85] M. Geller, I. Manke, K. Hodeck, R. Heitz, and M. Dähne, Phys. Rev. B 64,
233312 (2001).
[86] H. Nishikawa, T. Shiroyama, R. Nakamura, Y. Ohki, K. Nagasawa, and Y.
Hama, Phys. Rev. B 45, 586 (1992).
[87] R. Stöckle, C. Fokas, V. Deckert, R. Zenobi, B. Sick, B. Hecht, and U. P.
Wild, Appl. Phys. Lett.
75, 160 (1999).
[88] H. A. Bethe, Phys. Rev.
66, 163 (1944).
[89] T. Yatsui, M. Kourogi, and M. Ohtsu, Appl. Phys. Lett.
73, 2090 (1998).
[90] S. Madsen, N. Holme, P. Ramanujam, S. Hvilsted, J. M. Hvam, and S. Smith,
Ultramicroscopy 71, 65 (1998).
[91] G. A. Valaskovic, M. Holton, and G. H. Morrison, Appl. Opt.
34, 1215 (1995).
[92] C. Lehrer, L. Frey, S. Petersen, T. Sulzbach, O. Ohlsson, T. Dziomba, H.
Danzebrink, and H. Ryssel, Microelectron. Eng.
57–58, 721 (2001).
[93] F. Hitzel, G. Klewer, S. Lahmann, U. Rossow, and A. Hangleiter, Phys. Status
Solidi C 1, 2520 (2004).
[94] G. Behme, A. Richter, M. Suptitz, and C. Lienau, Rev. Sci. Instrum.
68,
3458 (1997).
[95] C. Durkan and I. V. Shvets, J. Appl. Phys.
79, 1219 (1996).
[96] M. Albrecht, L. Lymperakis, J. Neugebauer, J. E. Northrup, L. Kirste, M.
Leroux, I. Grzegory, S. Porowski, and H. P. Strunk, Phys. Rev. B 71, 035314
(2005).
[97] M. Gao, S. T. Bradley, Y. Cao, D. Jena, Y. Lin, S. A. Ringel, J. Hwang,
W. J. Schaff, and L. J. Brillson, J. Appl. Phys.
100, 103512 (2006).
[98] Z. Liliental-Weber, Y. Chen, S. Ruvimov, and J. Washburn, Phys. Rev. Lett.
79, 2835 (1997).
[99] M. Marques, L. K. Teles, and L. G. Ferreira, Phys. Rev. B 75, 033201 (2007).

BIBLIOGRAPHY 61
[100] L. Chang, S. K. Lai, F. R. Chen, and J. J. Kai, Appl. Phys. Lett.
79, 928
(2001).
[101] D. Dobrovolskas, J. Mickevičius, E. Kuokštis, G. Tamulaitis, M. Shur, M.
Shatalov, J. Yang, and R. Gaska, J. Phys. D: Appl. Phys.
44, 135104 (2011).
[102] H. Murotani, T. Saito, N. Kato, Y. Yamada, T. Taguchi, A. Ishibashi, Y.
Kawaguchi, and T. Yokogawa, Appl. Phys. Lett.
91, 231910 (2007).
[103] X. L. Wang, D. G. Zhao, D. S. Jiang, H. Yang, J. W. Liang, U. Jahn, and K.
Ploog, J. Phys.: Condens. Matter 19, 176005 (2007).
[104] Q. Sun, Y. Huang, H. Wang, J. Chen, R. Q. Jin, S. M. Zhang, H. Yang, D. S.
Jiang, U. Jahn, and K. H. Ploog, Appl. Phys. Lett.
87, 121914 (2005).
[105] S. Kim, J. Oh, J. Kang, D. Kim, J. Won, J. W. Kim, and H.-K. Cho, J.
Cryst. Growth 262, 7 (2004).
[106] Y. J. Sun, O. Brandt, M. Ramsteiner, H. T. Grahn, and K. H. Ploog, Appl.
Phys. Lett.
82, 3850 (2003).
[107] N. Garro, A. Cros, J. A. Budagosky, A. Cantarero, A. Vinattieri, M. Gurioli,
S. Founta, H. Mariette, and B. Daudin, Appl. Phys. Lett.
87, 011101 (2005).
[108] C. Jia, T. Yu, S. Mu, Y. Pan, Z. Yang, Z. Chen, Z. Qin, and G. Zhang, Appl.
Phys. Lett.
90, 211112 (2007).
[109] P. G. Eliseev, P. Perlin, J. Lee, and M. Osiński, Appl. Phys. Lett.
71, 569
(1997).
[110] P. G. Moses and C. G. Van de Walle, Appl. Phys. Lett.
96, 021908 (2010).
[111] S. Rudin, T. L. Reinecke, and B. Segall, Phys. Rev. B 42, 11218 (1990).
[112] V. Y. Davydov, Y. E. Kitaev, I. N. Goncharuk, A. N. Smirnov, J. Graul,
O. Semchinova, D. Uffmann, M. B. Smirnov, A. P. Mirgorodsky, and R. A.
Evarestov, Phys. Rev. B 58, 12899 (1998).
[113] A. Kasic, M. Schubert, Y. Saito, Y. Nanishi, and G. Wagner, Phys. Rev. B
65, 115206 (2002).
[114] S. J. Xu, L. X. Zheng, S. H. Cheung, M. H. Xie, S. Y. Tong, and H. Yang,
Appl. Phys. Lett.
81, 4389 (2002).
[115] P. Ebert, L. Ivanova, S. Borisova, H. Eisele, A. Laubsch, and M. Dähne, Appl.
Phys. Lett.
94, 062104 (2009).
[116] D. C. LeBlanc, Statistics: Concepts and Applications for Science (Jones and
Bartlett, 2004).
[117] A. Minj, D. Cavalcoli, and A. Cavallini, Appl. Phys. Lett.
97, 132114 (2010).

1.1
Wavelengths of Al x
Ga
1-x
N QWs versus Al molar fraction
. . . . . . . .
3
1.2
Typical efficiencies of visible (blue) and UV LEDs
. . . . . . . . . . . .
4
2.1
Crystal structure of wurtzite GaN
. . . . . . . . . . . . . . . . . . . . .
8
2.2
Arrangement of atoms for InN, AlN, and GaN
. . . . . . . . . . . . . .
8
2.3
Arrangement of atoms for In
0.25
Al
0.75
N . . . . . . . . . . . . . . . . . .
9
2.4
Energy band gap of III-nitride alloys . . . . . . . . . . . . . . . . . . . .
9
2.5
Bond distribution in the wurtzite, Ga-face GaN . . . . . . . . . . . . . .
10
2.6
Screening of built-in electric field in AlGaN MQWs . . . . . . . . . . . .
15
2.7
Polar and nonpolar planes of a wurtzite structure . . . . . . . . . . . . .
16
2.8
Energy level diagram of compressively strained InGaN . . . . . . . . . .
17
2.9
Crystal planes (20¯ 2¯ m -plane . . . . . . . . . . . . . . . . .
18
2.10 Threading dislocations in GaN . . . . . . . . . . . . . . . . . . . . . . .
19
2.11 Stacking fault in wurtzite structure . . . . . . . . . . . . . . . . . . . . .
20
3.1
Two stage generation of the third harmonic . . . . . . . . . . . . . . . .
22
3.2
Operating principle of the streak tube . . . . . . . . . . . . . . . . . . .
24
3.3
Example of an acquired streak image . . . . . . . . . . . . . . . . . . . .
25
3.4
Differential transmission pump-probe setup . . . . . . . . . . . . . . . .
26
3.5
Aperture SNOM configurations . . . . . . . . . . . . . . . . . . . . . . .
29
3.6
Self luminescence from a pure silica core fiber . . . . . . . . . . . . . . .
30
3.7
Tip manufacturing using tube etching method
. . . . . . . . . . . . . .
31
3.8
SEM images of etched fiber probes . . . . . . . . . . . . . . . . . . . . .
32
3.9
SNOM setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.10 Analysis of an acquired near-field spectrum . . . . . . . . . . . . . . . .
35
4.1
Near-field PL maps of Al
0.42
Ga
0.58
N . . . . . . . . . . . . . . . . . . . .
39
4.2
Growth schematics of an AlGaN structure . . . . . . . . . . . . . . . . .
40
4.3
Differential transmission transients of AlGaN MQWs . . . . . . . . . . .
41
4.4
Evolution of time-resolved transmission shape for AlGaN MQWs . . . .
42
4.5
Temporal change of the polarization degree in m -InGaN QWs . . . . . .
43
4.6
Near-field PL maps of m -plane In
0.2
Ga
0.8
N QW
. . . . . . . . . . . . .
44
4.7
Gaussian density of states model . . . . . . . . . . . . . . . . . . . . . .
45
63

64 LIST OF FIGURES
4.8
Near-field PL maps of GaN layer in InGaN/GaN QW structure . . . . .
47
4.9
SNOM PL maps of AlInN/GaN heterostructure . . . . . . . . . . . . . .
48
4.10 PL and PLE spectra of AlInN/GaN heterostructure . . . . . . . . . . .
49
4.11 Transient photoreflection of AlInN/GaN heterostructure . . . . . . . . .
50

65

In this article, time-resolved transmission pump-probe spectroscopy and TRPL were used to study Al
0.35
Ga
0.65
N/Al
0.49
Ga
0.51
N multiple quantum wells with different well widths ranging from 1.7 to 5 nm. The carrier dynamics was shown to be governed by non-radiative recombination at excitation into the extended states.
However, in contrast to TRPL measurements, differential transmission traces exhibited very slow decay under excitation into the localized states. The difference was attributed to localization of one type of carriers, presumably electrons. The evolution of the transmission transient shape with varying excitation photon energy in the pump-probe experiment suggested localization potential to be about 80 meV in all the samples and it was attributed to fluctuations in AlN molar fraction.
Additionally, excitation intensity measurements demonstrated intensity dependent saturation of the localized states, allowing to estimate the 2D density of these states at approximately 7 × 10
9 cm
− 2
.
Author contribution: part of the measurements, data analysis and part of the writing.
This paper reports photoluminescence studies of Al
0.35
Ga
0.65
N/Al
0.49
Ga
0.51
N multiple quantum wells with well widths spanning from 1.7 to 5 nm. PL measurements were carried out at both room temperature and 8 K. Low temperature measurements showed that carrier dynamics were governed by strong carrier localization, however, there were signs of localization even at RT, particularly for the narrower wells. Furthermore, AlGaN QWs exhibited clearly pronounced blueshift with increasing excitation power density. The shift was attributed to the screening of built-in electric field by photogenerated free carriers as the widest wells showed the
67

68 GUIDE TO THE ARTICLES largest shift. The screening dynamics was in fair agreement with a semi-classical numerical calculation of coupled Schrödinger and Poisson equations, however the best fit was achieved with the built-in electric field strength of only 300 kV/cm whereas theoretical estimates exceed 1 MV/cm for the given QWs. The possible origins of this discrepancy are discussed in the article.
Author contribution: the built-in electric field screening calculations and part of writing reflecting these calculations.
m
This article presents a TRPL study of 2.5 nm thick m -plane InGaN/GaN SQWs with estimated indium molar fractions at 17 % and 20 %. Nonpolar m -plane QWs are not affected by the intrinsic piezoelectric field, however they exhibit significant hole level splitting due to anisotropic strain, which gives rise to strongly polarized luminescence. In this article, PL emission was investigated at both perpendicular (dominant) and parallel (with respect to c axis) polarizations over a range of temperatures from 80 K to 320 K. At higher temperatures, PL decay was governed by non-radiative recombination, although a very slowly decaying component from localized states was also observed. At lower temperatures, the PL signal from the parallel polarization showed initial rise instead of the expected decay. Effectively, the degree of polarization was decreasing over time after the excitation, which was attributed to ongoing carrier localization. Localization to potential minima was further evidenced by the temporal redshift of the PL peak with a characteristic time similar to that of the polarization ratio. Lesser polarization degree of emission from localization centers was associated with local strain relaxation.
Author contribution: TRPL measurements, data analysis and major part of writing. The samples were grown at the University of California, Santa Barbara.
m
In this paper, m -plane InGaN/GaN SQWs were investigated by scanning near-field microscopy in illumination-collection mode. Dual localization profile was evidenced by SNOM. Larger scale potential fluctuations were observed directly by near-field spectral mapping as varying emission wavelength and attributed to inhomogeneous
In distribution. The smaller localization centers were distributed closer apart than the experiment resolution (approx. 100 nm) and could not be directly resolved, however nanoscopic localization depth was indirectly calculated from near-field spectral width and was estimated to range from 35 to 55 meV for different positions. The nanoscopic localization was assigned to partial strain relaxation at the monolayer

69 steps. Furthermore, the top GaN layers were also studied in the same setup by tuning the excitation wavelength to UV. GaN PL maps from GaN showed only micrometer scale linelike features, presumably originating from stronger non-radiative recombination at stacking faults or pyramidal hillock borders.
Author contribution: SNOM measurements, data analysis and major part of writing. The samples were grown at the University of California, Santa Barbara.
In this work, Al
0.3
Ga
0.7
N, Al
0.42
Ga
0.58
N and Al
0.5
Ga
0.5
N epitaxial layers were studied by near-field PL spectroscopy. Dual localization pattern was observed. Larger scale variations, most clearly pronounced in the layers with a lower AlN molar fraction, were associated with Ga-rich regions close to grain boundaries or atomic layer steps. The lower bandgap domains were clearly shown to correlate with the regions of the strong non-radiative recombination. The small-scale (<100 nm, beyond the experiment resolution) potential fluctuations were evaluated indirectly from the width of NF spectra and were found to increase with increased Al percentage, with the localization depth up to 51 meV for Al
0.5
Ga
0.5
N. These potential variations were assigned to small-scale compositional fluctuations occurring due to stress variations, dislocations, and formation of Al-rich grains during growth.
Author contribution: part of the setup construction and some preliminary measurements, part of the data analysis and discussions. The AlGaN layers were grown at SET, Inc.
This article presents a SNOM study of light extraction from photonic crystal patterned InGaN MQW LEDs, complemented by a TRPL assessment of the MQW layers. In addition, otherwise identical conventional LEDs without photonic crystals were investigated for comparison. Electroluminescence emission was collected in both near- and far- field conditions. Interestingly, the near-field distribution of emission did not precisely follow the photonic crystal rectangular grid arrangement and periodicity. Moreover, photonic crystals showed not only improved light extraction, but also made the emission more homogeneous across the surface with respect to the unpatterned devices.
Author contribution: measurements, data analysis and writing. The devices were fabricated by S. M. Kim’s group at Korea Photonics Technology Institute.

70 GUIDE TO THE ARTICLES
In this work, photoexcited carrier dynamics in Al
0.86
In
0.14
N/GaN heterostructures were examined by both TRPL and SNOM. AlInN exhibited very broad PL spectra and large Stokes shift (up to 0 .
9 eV), suggesting very large potential fluctuations.
The spectral shape was the same in the NF, indicating that localization sites were densely packed in comparison to the experiment resolution (approx. 100 nm). However, huge potential fluctuations in AlInN did not seem to confine carriers as efficient photoexcited hole transfer was observed from AlInN to GaN via sub-band-gap states. The transport was indicated by a fast decay component of AlInN PL transient and corresponding rise in GaN PL. In addition, GaN showed intense emission in comparison to AlInN throughout the whole experiment temperature range (5 K to RT). The hole transport was associated with high conductivity channels related to V-shape defects and In clusters. In contrast to AlInN, SNOM PL mapping of the underlying GaN layer revealed clearly defined islands of different peak wavelength and emission intensity on a submicrometer scale, typically not observed in pure
GaN. These spatial variations were attributed to electric field inhomogeneities at the heterostructure interface.
Author contribution: SNOM, PL and PLE measurements and data analysis, part of TRPL measurement and analysis, and minor part of the writing.