many body effects in heavily doped gan - tu berlin · many body effects in heavily doped gan...

ManybodyeffectsinheavilydopedGaN

vorgelegt von Diplom-Physiker

Marc Christian Siegfried Nenstiel geb. in Berlin

von der Fakultät II – Mathematik und Naturwissenschaften

der Technischen Universität Berlin

zur Erlangung des akademischen Grades

Doktor der Naturwissenschaften – Dr. rer. nat. –

genehmigte Dissertation

Promotionsausschuss: Vorsitzende: Prof. Dr. Ulrike Woggon

Gutachter: Prof. Dr. Axel Hoffmann

Gutachter: Prof. Dr. Matthew Phillips

Gutachter: Prof. Dr. Tadeusz Suski

Gutachterin: Prof. Dr. Janina Maultzsch

Tag der wissenschaftlichen Aussprache:

13.03.2017

Berlin 2017

ii

Zusammenfassung

In der vorliegenden Arbeit werden die grundlegenden optischen Eigenschaften von hoch

Germanium dotierten GaN Epischichten untersucht und diskutiert. Diese Arbeit liefert

neue Erkenntnisse über Wechselwirkungen von Ladungsträgern in einem

dreidimensionalen degenerierten Elektronengas.

Der erste Teil dieser Doktorarbeit befasst sich mit Kompensationsmechanismen in

GaN:Ge Proben bis in den Dotierungsbereich von 1021 cm‐3. Die Kombination von

Halleffekt‐Messungen, Sekundärionen‐Massenspektrometrie, Ramanspektroskopie,

sowie temperaturabhängiger und zeitaufgelöster Photolumineszenzspektroskopie

veranschaulicht den Einfluß von Kompensationsmechanismen auf die strukturellen,

elektrischen und optischen Eigenschaften des Kristalls.

Der zweite Teil dieser Doktorarbeit vergleicht Ge‐ und Si‐dotierte GaN Epischichten auf

ihre Kompensationseigenschaften. Kompensationsarme (GaN:Ge) Proben zeigen im

Vergleich zu kompensationsreichen (GaN:Si) neue, bisher unbekannte bindende

Vielteilchenwechselwirkungen zwischen einem dreidimensionalem Elektronengas und

Elektron‐Loch‐Paaren. Diese Vielteilchenwechselwirkung führt zur Einführung eines

neuen Quasiteilchens, dem Collexon. Dabei handelt es sich um einen exzitonartigen

Vielteilchenkomplex, der durch ein dreidimensionales Elektronengas stabilisiert wird. Je

höher die Elektronengasdichte, desto stärker ist die Stabilisierung des Collexons. Diese

Stabilisierung zeichnet sich durch die kleine spektrale Halbwertsbreite (≈10 meV),

Temperaturstabilität bis zu 100 K und Langlebigkeit (≈250 ps) aus. Im Gegensatz zu

anderen Vielteilchenkomplexen, wie z.B. Trionen und Biexzitonen, wird das Collexon nur

in einer dreidimensionalen Struktur beobachtet, aber nicht in einer niederdimensionalen,

wie Quantengräben oder Quantenpunkten. Im Gegensatz zum üblichen Verhalten von

Lumineszenzen solcher Vielteilchenkomplexe in stark dotierten Halbleitern nimmt die

Intensität und Lebenszeit der Lumineszenz des Collexons mit steigender

Ladungsträgerkonzentration zu und die Halbwertsbreite ab, solange keine Kompensation

eintritt.

iii

Abstract

In this work the fundamental optical signatures of highly Germanium doped GaN are

investigated. This work gives fundamental new insights in interactions of carriers in a

three dimensional degenerated electron gas.

The first part of this thesis is dedicated to the investigation of compensation mechanisms

in GaN samples with Germanium concentrations up to 1021 cm‐3. The combination of Hall

effect measurements, secondary ion mass spectrometry (SIMS), Raman spectroscopy and

temperature‐dependent and time‐resolved photoluminescence spectroscopy shows the

influence of compensation mechanisms on the structural, electrical and optical properties

of the crystal.

In the second part of this doctoral thesis, Germanium doped GaN samples are examined,

which do not show any compensation effects. This fact is illustrated through comparison

with Silicon doped GaN samples, which were grown with the same growth conditions as

the Germanium doped samples. By the means of photoluminescence spectroscopy a new

quasi particle was discovered, the collexon. The collexon is an exciton‐like many‐particle

complex, which is stabilized by a three dimensional electron gas. The stabilization

enhances with increasing electron gas density and is featured by small spectral half‐width

(≈10 meV), temperature stability up to 100 K and longevity (≈250 ps). In contrast to other

many‐particle complexes, e.g. trions and biexcitons, the collexon is observed in a three‐

dimensional structure and not in a low‐dimensional structure, such as quantum wells and

quantum dots. Contradicting the common behavior of heavily doped semiconductors, the

entire emission intensifies and its decay time slows towards increasing free electron

concentration, while the collexonic emission narrows spectrally, as long as no

compensation occurs.

iv

Contents

1 Introduction .................................................................................................................. 1

1.1 Scope of thesis....................................................................................................... 2

2 Theory ........................................................................................................................... 4

2.1 Raman spectroscopy ............................................................................................. 4

2.1.1 Lattice dynamics and phonons in wurzite crystals ........................................ 4

2.1.2 Determination of strain values ...................................................................... 7

2.1.3 Longitudinal optical phonon plasmon coupling ............................................ 8

2.2 Deformation potentials of excitons .................................................................... 10

2.3 Optical and electronic properties of semiconductors with high free carrier

concentrations ............................................................................................................... 12

3 Experimental details ................................................................................................... 18

3.1 Raman setup ........................................................................................................ 18

3.2 Macro PL/PLE setup ............................................................................................ 18

3.3 Micro PL/TRPL/PLE setup .................................................................................... 19

4 Compensation in highly Germanium doped GaN ....................................................... 21

4.1 Investigated samples ........................................................................................... 21

4.2 Compensation mechanisms in doped semiconductors ...................................... 22

4.3 Raman spectroscopy results ............................................................................... 24

4.4 PL of compensated semiconductors ................................................................... 28

4.5 Dynamics and time resolved analysis ................................................................. 33

4.6 Discussion ............................................................................................................ 36

5 Electronic excitations stabilized by a degenerate electron gas in highly Ge doped GaN

38

5.1 Investigated samples ........................................................................................... 38

5.2 Raman spectroscopy results ............................................................................... 39

v

5.2.1 Investigation of strain and crystalline quality by the means of Raman

spectroscopy ............................................................................................................... 39

5.2.2 Comparison of structural properties in Si and Ge doped GaN .................... 41

5.3 Optical properties of uncompensated degenerated electron gas in GaN .......... 51

5.3.1 Low temperature Photoluminescence spectroscopy .................................. 52

5.3.2 Cross section and depth resolved cathodoluminescence ........................... 57

5.3.3 Absorption deep defects .............................................................................. 59

5.3.4 Dynamics and time resolved analysis .......................................................... 61

5.3.5 Excitation channels ...................................................................................... 65

5.3.6 Strain introduced shift ................................................................................. 68

5.3.7 Excitation energy dependent photoluminescence and reflection .............. 70

5.3.8 Thermalization ............................................................................................. 73

5.4 Discussion ............................................................................................................ 74

5.5 Theoretical model ............................................................................................... 78

6 Summary ..................................................................................................................... 83

7 Publications ................................................................................................................ 86

8 Bibliography ................................................................................................................ 90

9 Acknowledgement ...................................................................................................... 97

1 Introduction

1

1 Introduction

GaN‐based electronic and optoelectronic devices have gained considerable interest in

the recent years. The wide band gap, superior structural and thermal stability, high

breakdown electrical field and high saturation drift velocity make AlGaN/GaN an

excellent candidate for high electron mobility transistors (HEMTs), high power switches,

hetero bipolar transistors and many other high power devices. [1–3] Up to date the

ternary systems AlGaN and InGaN are the only technologically relevant candidates for

blue and near UV light emitting diodes (LEDs) [4,5] and laser diodes (LDs) [6,7]. GaN

based devices are built into many common technologies like blue‐ray disc players or

white light flashes in photo cameras.

Most of electronic and optoelectronic devices strongly benefit from high free carrier

concentrations. Such benefits do not only directly affect the pivotal ratio between

majority and minority carrier concentrations, but also other, less obvious advantages

are just as significant. For instance, tailoring the refractive index by means of doping can

lead to a better optical mode confinement in InGaN laser diodes [8] or screening

polarization induced electric fields in InGaN light emitting devices. [9]

Despite the subsequent emergence of various electronic and optoelectronic devices,

high doping concentrations are generally accompanied by compensation effects, and

the incorporation of deep level point defects, structural defects and unintentionally

dopants, which collectively diminish the crystalline quality and limit the device

performance. Studying the influence of doping on the luminescence has been proven as

an effective tool for deepening the understanding of the nature of compensation

mechanisms.

The effect of high free carrier concentrations on the electronic structure of a

semiconductor is well known. [10] Additionally, the interaction of high free carrier

concentrations and excitons has been studied for many years and is also well

understood. [11] Nevertheless, high doping concentrations are generally highly

1 Introduction

2

correlated with high free carrier concentrations, but also compensation mechanisms

and a reduction of crystalline quality. This makes it difficult to observe either the effects

of high free carrier concentrations on the electronic structure or excitons separately.

However, in n‐type GaN high free carrier concentrations above 1019 cm‐3 are achievable

with minor compensation issues. Additionally, GaN is a direct semiconductor with a high

exciton binding energy, which makes it an ideal semiconductor to study the influence of

high free carrier concentrations on the electronic structure and excitons in detail.

1.1 Scopeofthesis

The aim of this thesis is to give the reader a deeper understanding of compensation

mechanisms in n‐type GaN and the effect of high free carrier concentrations on

luminescence and excitons. The operation of most semiconductor based devices require

free carriers within the crystal. However, doping of semiconductors is generally

accompanied by unwanted compensation effects. Compensation leads to a reduction of

free carrier concentration and mobility, increased point defect concentration,

decreased crystalline quality and optical transparency. Currently it is very challenging to

grow semiconductors with high free carrier concentrations without any compensation

effects. Commercially available transparent conducting oxides (TCO) are generally highly

compensated. In these materials, it is known that for free carrier concentrations in the

range of 1019 ‐ 1020 cm‐3 doping concentrations in the range of 1021 ‐ 1022 cm‐3 are

necessary, which means only every approx. 100th dopant atom contributes an electron

to the system. To overcome this challenge one has to understand the limits of free

carrier concentrations as well as under which circumstances compensation starts and

how it affects the crystalline quality. This matter is discussed in chapter 4 for the case of

compensation in Ge doped GaN and in chapter 5.2 for the case of Si doped GaN.

The free carrier concentration and the defect concentration due to compensation have

a strong impact on excitons. High free carrier concentrations in the range of 1020 cm‐3

1 Introduction

3

are generally accompanied by poor crystalline quality, which makes it difficult to

investigate the effect of high free carrier concentrations on excitons on its own.

In order to address these issues, several GaN samples are investigated with different

growth techniques and dopants. The outcome of this work provides crucial information

for producing GaN layers with high free carrier concentrations without any measurable

compensation, which makes them perfect for optoelectronic devices that require high

free carrier concentrations and transparency. Additionally, this work gives a new insight

in the interaction of excitons with high free carrier concentrations, which is shown on

chapter 5.3.

2 Theory

4

2 Theory

2.1 Ramanspectroscopy

2.1.1 Latticedynamicsandphononsinwurzitecrystals

Wurtzite crystals, like GaN, belong to the space group C46v with n = 4 atoms in the

primitive unit cell. This leads to 3n = 12 vibrational Eigenmodes. Three of these modes

are acoustical; nine are optical. Group theory delivers the following irreducible

representation. [12]

1 1 1 22 (1) 2 (1) 2 (2) 2 (2)A B E E

The A1 and B1 modes are onefold; the E1 and E2 modes are twofold degenerated. The

three acoustical modes consist of one A1 mode and one E1 pair mode. The other nine

modes are optical, which are illustrated in Figure 1.

Figure 1: Schematic illustration of the optical phonon modes in GaN.

The A1 and B1 modes oscillate in the direction of the c‐axis, the E1 and E2 modes oscillate

perpendicular to it. Both B1 and E2 modes are distinguished by their energy and named

2 Theory

5

“high” and “low”. In the case of GaN the prominent displacement of the low modes

occurs from the Gallium and for the high modes from the nitride. The B1 modes are silent

and not measurable by the means of Raman spectroscopy. The A1 and E1 modes are

polar modes, because of opposite directions of vibrations of anions and cations. That

means they can be influenced by electric fields. The wurtzite structure of GaN allows for

spontaneous polarization and piezoelectric fields and generate a splitting into

longitudinal optical (LO) and transversal optical (TO) A1 and E1 modes. This splitting is

not predicted by the group theory as the description is limited to purely vibrational non‐

propagating Eigenmodes at the point with q = 0. The size of the splitting is expressed

by the Lyddane‐Sachs‐Teller relation. [13] The B1 and E2 modes are not polar, so they do

not exhibit such a splitting. That means a shift of the E2 modes can only occur from strain

within a crystal and is independent of the free carrier concentration. The intensity or

even observability of all modes depends on the geometry used in the measurement,

where ∝ | ∙ ∙ |. The intensities consists of orientation of the polarization and

direction of the incoming light and orientation of the polarization and direction of the

detected light . The elements of the Raman tensors, R, determine if the modes are

allowed or forbidden in certain geometries. The Raman tensors for phonon modes in

wurtzite structures are given by:

1

aA z a

b

1

cE x

c

1E y cc

2

d dE d d

The scatter geometry for Raman measurements is represented in the so called Porto

notation [14], , where and indicate the direction and and the

polarization the incident and scattered light. In this work only backscattering geometry

is applied, so the incident and the detection of the scattered light is in opposite

directions. For simplification the Cartesian coordinate system is used whereas the z‐axis

is chosen to be parallel to the c‐axis of the wurtzite crystal. Table 1 shows all allowed

modes for backscattering geometry. In these configurations all modes can be visualized.

2 Theory

6

The E1(LO) mode is theoretically forbidden in backscattering geometry, but the Fröhlich

interaction allows for the visualization in x(zz)x geometry. [15]

Table 1: Selection rules for Raman active modes in wurtzite crystals.

Scatter geometry Allowed modes

z(xx)z E2(low, high), A1(LO)

z(xy)z E2(low, high)

x(yy)x E2(low, high), A1(TO)

x(zz)x A1(TO), E1(LO) – Fröhlich interaction

x(yz)x E1(TO)

500 550 600 650 700 750

0.0

0.5

1.0

1.5

E1(LO)

E1(TO)A1(LO)

Inte

nsity

(arb

. uni

ts)

Ramanshift (cm-1)

m-plane / x(..)x c-plane / z(..)z

A1(TO)E2(high)

Figure 2: Room temperature Raman spectroscopy spectra of in backscattering geometry of a GaN sample. The spectra are taken in c‐plane (black line) and m‐plane (red line) direction. The spectra are normalized and shifted vertically for clarification.

Figure 2 shows two Raman spectra of the same undoped c‐plane GaN sample measured

in two different orientations without polarization. The black and red spectra represent

2 Theory

7

Raman measurements of the surface in‐plane and in cross section. Both spectra show

all expected phonon modes, except for the E2(low) modes, which is outside the given

measurements range. A faint signature of forbidden phonon modes for given scatter

geometries is generally visible due to low crystalline quality, scattering at defects and

the broad light collection angle of the applied objective.

2.1.2 Determinationofstrainvalues

GaN is commonly grown on foreign substrates like Sapphire, Silicon and Silicon Carbide

for cost reduction. [16,17] The lattice mismatch and different thermal expansion

coefficients of substrate and GaN cause biaxial strain in the grown GaN layer. Raman

spectroscopy is the standard nondestructive method for the determination of the strain

in bulk and nanostructured materials. In this thesis the investigated samples are grown

on (0001) Sapphire substrates (lattice mismatch 13.5%), which cause biaxial

compressive strain in the GaN layer. This experimental technique relies on the precise

knowledge of phonon deformation potentials which in turn allows the determination of

the strain levels from the phonon frequencies. The shift of phonon modes in comparison

to the relaxed values (given in Table 2) allows for determination of strain values. Non‐

polar modes as e.g. the E2(high) in a wurtzite crystal like GaN serve as an ideal tool for

stress determination, because the position of the E2(high) only depends on strain and is

independent of free carrier concentration. All other modes are much less suitable for

strain determination. The LO modes interact with free carriers and the TO modes are

not present in every scatter geometry. Even if the E2(low) mode is a non‐polar mode and

therefore appropriate for strain determination, its strain coefficient is very small in

comparison to the E2(high) mode’s.

Table 2: Relaxed values for all Raman modes in GaN [18]

Mode E2(low) A1(TO) E1(TO) E2(high) A1(LO) E1(LO)

Wavenumber (cm‐1) 145 533 560 567 735 742

2 Theory

8

The pressure induced frequency shift ∆ of a phonon mode for the case of biaxial

isotropic strain ( ) within the frame of Hooke’s law is:

∆ 2 2

, 2

The measured frequency shift ∆ provides the phonon deformation potentials and

for a known strain or the phonon pressure coefficients and for a known stress .

The elastic stiffness constants for GaN were taken from Brillouin scattering

experiments. [19] Uniaxial pressure ( 0 and 0) dependent Raman

spectroscopy measurements allow for direct determination of , whereas hydrostatic

pressure ( ) dependent Raman spectroscopy measurements allow for

determination of 2 . With the knowledge of the hydrostatic pressure coefficient of

the E2(high) mode (4.24 cm‐1/GPa [20]) and the corresponding uniaxial pressure

coefficient (1.38 cm‐1/GPa [21]) one can calculate the biaxial pressure coefficient of

2.86 cm‐1/GPa.

The relative change of the a‐ and ‐lattice constant ( and ) due to strain can be

calculated from

∆E high / 2.86 ∙ and 2 / ,

where M is the biaxial modulus of a wurtzite crystal, which is defined based on the elastic

constants :

2 /

The values taken from Polian et al. [19], yield a value of M = 479 GPa.

2.1.3 Longitudinalopticalphononplasmoncoupling

In polar semiconductors like GaN long range electrostatic fields generate a LO‐TO

splitting of the polar A1 and E1 modes. These LO‐modes can interact and couple with

plasmons and create longitudinal optical phonon plasmon (LPP) modes. In doped

2 Theory

9

semiconductors, instead of a single LO mode, two coupled LPP modes are visible, a LPP‐

and a LPP+. Disregarding the damping of phonons and plasmons, the position of these

LPP modes depend only on the position of the LO mode LO, the TO mode TO and the

plasma frequency p.

2 2

Whereas the LO and TO modes are fixed and known for the given semiconductor, the

plasma frequency depends on the free carrier concentration n.

4∗

Here is the elementary charge, the high frequency dielectric constant for the given

material and ∗ the effective mass for the given material and charge carrier. With these

formulas it is possible to calculate the free carrier concentration within the investigated

sample by measuring the position of at least one of the LPP modes.

1017 1018 1019 1020

500

1000

1500

2000

2500

A1(TO)

LPP-

Freq

uenc

y (c

m-1)

Free carrier concentration (cm-3)

LPP+

p

A1(LO)

Figure 3: Values of the undamped LPP‐ mode, LPP+ mode and plasma frequency over free carrier concertation in GaN.

2 Theory

10

Figure 3 displays the frequency of the LPP modes versus the free carrier concentration.

The LPP‐ mode shifts from zero wavenumbers towards the position of the A1(TO) mode

and the LPP+ mode starts at the position of the A1(LO) and shifts towards higher

wavenumbers with increasing free carrier concentration. Not shown in this figure are

the increase of FWHM and decrease of intensity with free carrier concentration of the

LPP+ in addition to the decrease of FWHM and increase of intensity of the LPP‐ at the

same time. With increasing free carrier concentration the LPP‐ gains a more phonon like

character and the LPP+ gains a more plasmon like character. The plasmon damping is

two to three orders of magnitudes stronger than the phonon damping. This means the

stronger the plasmon‐like character of a mode the broader and weaker the mode,

becomes in Raman spectroscopy [22]. The damping also influences the position of the

modes [23]. However, as the focus is on the LPP‐ for free carrier concentrations above

1019 cm‐3, the influence of the plasmon damping on the position is negligibly small due

to its strong phonon‐like character.

2.2 Deformationpotentialsofexcitons

The three valence bands of wurtzite GaN have the geometries HH Γ , LH Γ and

CH Γ with the energy maxima given by

Δ Δ ,

/ 2∆ ,

with

,

. [24]

Here, Δ describes the crystal‐field splitting of the Γ and Γ orbital states, while Δ and

Δ describe the spin‐orbit interaction parallel and perpendicular to the c axis,

respectively. are components of the strain tensor and are valence band

2 Theory

11

deformation potentials. It should be noted that herein a negative sign indicates

compressive strain, a positive sign indicates tensile strain and Δ are strain

independent. [25] For the conduction band the minimum energy is given by

Δ Δ .

Where, is the strain free or relaxed value for the band gap and and are the

conduction band deformation potentials. The combination of the formula for the strain

dependent shifts of the valence band and conduction band results into the band gap

shifts due to strain:

Δ

Δ // Δ 3Δ

2 2 2

∓Δ Δ

22∆

For the case of biaxial strain holds 2 / and 0.

-0.003 -0.002 -0.001 0.000 0.001 0.002 0.003

3.40

3.45

3.50

3.55

3.60

GaN

:nid

zz

FXA

FXB

FXC

FXn=2A

rela

xed

posi

tion

Ene

rgy

(eV

)

Figure 4: Shifts of the excitonic luminescence related to the A, B and C valence band versus c‐axis lattice parameter based on Ref. [26]. The dashed lines mark positions of a relaxed GaN layer and the position of the investigated GaN:nid layer.

2 Theory

12

Figure 4 shows the shift of A‐, B‐, C‐excitons and the excited state of the A‐exciton (n=2)

versus the c‐lattice constant due to biaxial strain, assuming that the binding energy of

excitons is strain independent. The elastic constants values are taken from Polian et

al. [19], the deformation potentials and Δ values from Alemu et al. [26], the exciton

transition energies from Monemar et al. [27] and the energy shift of the excited state

from Volm et al. [28].

The centered dashed line at 0 marks the position of all excitonic transitions in

strain free GaN. The second dashed line at 0.0014 marks the position of all

excitonic transitions for a compressive strained GaN:nid sample grown on sapphire,

which is investigated in detail in chapter 5.3.1. As the biaxial compressive strain

perpendicular to the c‐axis of the crystal becomes greater, both the c‐lattice constant

and the energies of the A‐, B‐ and C‐exciton increase. The energetic difference between

the A‐, B‐ and C‐excitons changes with strain due to the non‐linear nature of the shift of

the B‐ and C‐valence band. This behavior also allows for a crossing of the A‐ and B‐

valence band (heavy and light hole) for tensile strain conditions.

It is important to note that the strain values of the GaN measured by Raman

spectroscopy at room temperature cannot be used to determine the energetic positions

of excitons at liquid Helium temperatures. Due to the higher thermal expansion

coefficient of sapphire in comparison to GaN [29], the GaN layer is more compressively

strained at Helium temperatures than at room temperature. Consequently an additional

compressive stress of 0.21 GPa has to be taken into account. [30]

2.3 Optical and electronic properties of semiconductors with high free

carrierconcentrations

The influence of high free carrier concentrations on electronic states in heavily doped

semiconductors is well known for many decades [10]. High excitation densities or heavy

doping of semiconductors changes the band structure of the intrinsic material, which

2 Theory

13

can be observed e.g. by absorption and photoluminescence measurements. This effect

has been observed experimentally for optically pumped ZnO [31] and GaN [32–34] and

highly n‐doped Si [35], ZnO [36] and GaN [37–40]. Besides band filling, increasing free

carrier densities cause a band gap renormalization (BGR) [35] or band gap shrinkage due

to electron‐electron and electron‐ion interaction.

∆ ∆ ∆

These two contributions can be approximated by: [41]

Electron‐electron interaction: ∆ 1 arctan

Electron‐ion interaction: ∆ ∗

Fermi vector: √3

Thomas‐Fermi screening length: 2 ∗

Screened Bohr radius: ∗∗

The Burstein‐Moss shift (BMS) describes the shift of the absorption edge to higher

energies in highly doped semiconductors. [42,43] Doping introduced high free carrier

densities cause population of band‐states by carriers, which relax to the lowest available

energy. Due to Pauli‐blocking this energy increases with increasing carrier density,

making a semiconductor transparent up to the degenerate Fermi energy level. The BMS

in the conduction band follows the law:

∆2 ∗

2 Theory

14

Figure 5: (a) Band structure of an undoped semiconductor and (b) in the presence of a degenerated electron gas with density ne. [44] (c) Shift of the lowest not occupied state of the conduction band for GaN versus free carrier concentration.

The effect of the band gap narrowing due to the BGR and band filling due to the BMS on

the band structure is depicted in Figure 5. The lower part of Figure 5 shows the

characteristic of the lowest not occupied state of the conduction band. For free carrier

2 Theory

15

concentrations below 1018 cm‐3 the BGR has a much stronger influence on the CB than

the BMS and causes a reduction of the bandgap of more than 30 meV. At higher free

carrier concentrations the influence of the BMS overtakes the BGR effect and the lowest

not occupied state of CB shifts towards higher energies due to filling of the CB states.

Undoped semiconductors show a dominant absorption of excitonic states below the

fundamental band gap. [45–48] They are caused by the Coulomb attraction between

electrons and holes that stabilizes the excited electron‐hole pairs. High free carrier

concentrations screen the electron‐hole interaction significantly. However, free carrier

densities as low as n < 1018 cm‐3 can cause a reduction of the Wannier‐Mott exciton

binding energy. With increasing free carrier concentration, the binding energy decreases

further until the Mott density [49] is reached and the exciton binding energy approaches

zero, when the critical carrier density is about n = 7 x 1018 cm‐3 – 1 x 1019 cm‐3 in

GaN. [32,50]

Figure 6: Exciton binding energy (black) and relative oscillator strength (red), normalized with respect to the oscillator strength of the A‐exciton of undoped ZnO, versus free electron concentration n. The dotted line indicates the density calculated from the Mott criterion. [11]

2 Theory

16

Figure 6 shows the binding energy of the lowest excitonic bound state for doped ZnO

along with the corresponding oscillator strength. The binding energy and the oscillator

strength both decrease by orders of magnitude with increasing density of the

degenerate electron gas. Nevertheless, excitonic state with binding energies of approx.

1 meV and oscillator strengths of approx. 10%, in comparison to the undoped case, exist

for free carrier densities beyond the Mott transition. Schleife et al. [11] showed that

instead of the concept of an exciton Mott transition, a continuous transition from

Wannier‐Mott excitons to Mahan excitons [51] appears.

3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.00.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

D(E

)

F(E)

Energy (eV)

Fermi-Dirac funktion DOS GaN:nid DOS highly doped GaN theoretical PL spectrum

Fermi energy

I(E) ~ F(E) * D(E)

Figure 7: Calculated DOS for GaN:nid and highly doped GaN. The theoretical PL spectrum is calculated by the Fermi‐Dirac function and the DOS at a Fermi energy of 3.6 eV and a temperature of 100 K.

The influence of the degenerated electron gas on the density of states (DOS) is

depicted in Figure 7. The DOS of undoped semiconductors exhibits a square root

like character, the DOS of highly doped semiconductors have a square root like

character for energies above the band gap energy of the undoped semiconductor

as well, but for lower energies a band tail with exponential character appears. The

change of the shape of the DOS characteristic is based on approximations of Kane

2 Theory

17

et al. [52] The resulting PL spectrum can be calculated from the DOS function and

the Fermi‐Dirac function. An increase of the free carrier concentration or a higher

Fermi energy results in a longer band tail at lower energies due to the BGR and a

shift of the maximum of the PL to higher energies due to the BMS. Additionally,

Figure 7 shows that the saddle point of the PL spectra matches the Fermi energy.

Consequently the Fermi energy can be estimated from PL measurements (at least

for low excitation densities, where the number of additional free carriers in the CB

through excitation is negligible small). This energy is comparable with the lowest

not occupied state in the CB, which allows for the calculation of the free carrier

concentration as depicted in Figure 5c.

3 Experimental details

18

3 Experimentaldetails

3.1 Ramansetup

The micro‐Raman measurements were performed in backscattering geometry z(..)z,

utilizing a Horiba LabRAM HR 800 equipped with a 1200 l/mm grating (500 nm blaze),

an Olympus BX41 optical microscope, and a 532 nm DPSS laser as excitation source. For

measurements at room temperature, a 100x objective was applied, while cryogenic

temperatures requested an 80x long distance objective as the sample was placed in an

Oxford liquid‐He microstat cryostat. The sample was placed on a 3‐axes motorized stage.

In order to investigate the homogeneity of the samples, the Raman microprobe was

performed for up to 16 sample positions. All Raman mode positions and full widths at

half maximum (FWHM) were obtained by a careful manual peak fitting routing applying

Lorentzian fit functions.

3.2 MacroPL/PLEsetup

The photoluminescence (PL) and photoluminescence excitation (PLE) experiments were

performed in 90° geometry in one setup comprising a He‐bath‐cryostat providing a base‐

temperature of 1.8 K. For the continuous wave PL measurements, a HeCd‐laser (325 nm)

was used, while the PLE measurements were conducted with a dye‐laser (100 Hz

repetition rate, 20 ns pulse width, pumped by a XeCl excimer laser) containing 2‐methyl‐

5‐t‐butyl‐p‐quaterphenyl (DMQ) as active medium. The luminescence signal was

dispersed by an additive double monochromator (Spex 1404 ‐ 0.85 m focal length,

1200 groves/mm, 500 nm blaze) equipped with an ultra‐bialkali photomultiplier tube

(Hamamatsu, H10720‐210). This PL setup is specialized for low excitation power

(~100 W/cm2) measurements at very low temperature (1.8 K) and very high spectral

resolution of the detection (< 20 pm) and the PLE excitation (< 0.15 nm).


19

3.3 MicroPL/TRPL/PLEsetup

Figure 8: Scheme of the complete micro‐photoluminescence setup. Most of the applied optics are enhanced for the ultraviolet spectral range. Optics symbolized with solid contour lines are constantly present in the beam path, whereas dashed contour lines symbolize optional or flip mounted optics.

A scheme of the complete micro‐photoluminescence (µ‐PL) setup is shown in Figure 8.

This setup contains four possible excitation sources. A frequency doubled Ar2+ laser

(244 nm/257 nm) and a HeCd laser (325 nm/441.6 nm) provided continuous wave (cw)

excitation. For time resolved PL (TRPL) measurements the fourth harmonic of a

picosecond Nd:YAG laser (266 nm, 76 MHz repetition rate, 55 ps pulse width) was used.

For PLE measurements a 500 W XBO lamp was employed and the excitation wavelength

was selected by a 0.22 m double monochromator with switchable gratings, which allow

for a broad excitation wavelength range from about 200 nm to 1000 nm. The excitation


20

power of the selected excitation source can be limited by two motorized continuous

optical grey filter wheels.

The sample was situated in a helium flow micro cryostat ST‐500 manufactured by Janis,

enabling a temperature variation from 4 K to 400 K. The measurements are performed

in backscattering geometry. The light was focused and collected by an UV enhanced

objective manufactured by Thorlabs with 20x magnification and a numerical aperture of

0.4. A customized Köhler illumination along with a charge coupled device (CCD) camera

module allow an imaging of the sample surface. The entire cryostat is mounted on a 2‐

axes motorized stage with a minimum step size of 200 nm. The objective is mounted on

a closed loop 3‐axes piezo stage with a minimum step size of 1 nm and a maximum scan

range of 100 µm, which allows for a detailed mapping of the samples surface.

The PL was spectrally analyzed by a single monochromator (Spex 1704 ‐ 1 m focal length,

1200 groves/mm, 500 nm blaze) equipped with a nitrogen cooled UV‐enhanced CCD

array and temporally analyzed by a subtractive double monochromator (McPherson

2035 – 0.35 m focal length, 2400 groves /mm, 300 nm blaze) equipped with a

multichannel‐plate (MCP) photomultiplier (Hamamatsu R3809U‐52) limiting the

temporal resolution to ≈ 55 ps. Standard photon counting electronics were applied in

order to derive the final histograms.

This setup combines PL, PLE and TRPL measurements including tunable temperatures

and excitation powers with the option of mapping the samples surface.

4 Compensation in highly Germanium doped GaN

21

4 CompensationinhighlyGermaniumdopedGaN

Within this thesis are two sets of Ge doped GaN samples investigated. Both sets were

grown by metal organic chemical vapor deposition (MOCVD), but with different growth

conditions. These two sets of samples show very different measurement results, despite

having similar Ge contents. The first set is described and investigated in this chapter (4),

the second set in chapter 5.

4.1 Investigatedsamples

The samples were grown by MOCVD in a vertical showerhead reactor. All structures

consist of a low temperature AlN nucleation layer on (0001) sapphire substrate

overgrown with an 1.3 µm undoped GaN buffer layer and 1.1 µm Ge doped GaN layer.

The doped layers were grown under a V/III ratio of 2000 and reactor pressure of 20 Torr.

The TEGa flow was 134 µmol/min with a total gas flow of 7.2 slm. The Ge flow rates

ranged from 0.135 to 4.5 µmol/min. The growth temperature for the undoped buffer

layer was 1040 °C and lowered to 960 °C for the doped layer. This reduction of growth

temperature for the doped layers allows for an efficient incorporation of Ge atoms into

the crystal [53]. The disadvantage of the decreased growth temperature is a lower

crystalline quality. Table 3 shows: the Ge flow rates during the growth, the free carrier

concentrations determined by Hall effect measurements and Ge concentrations from

secondary ion mass spectrometry (SIMS). The Ge concentrations in brackets (sample

M001 and AR960) were not measured, but estimated by linear approximation.


22

Table 3: Free carrier and dopant concentrations measured by Hall effect and SIMS measurements, Germane flow rates of the investigated samples.

Sample Germane flow rate Free carrier

concentration (Hall)

Ge concentration

(SIMS)

M001 0.135 µmol/min 3.0 x 1019 cm‐3 (1.3 x 1019 cm‐3)

AR962 0.225 µmol/min 5.0 x 1019 cm‐3 4.7 x 1019 cm‐3

AR960 0.450 µmol/min 1.2 x 1020 cm‐3 (1.1 x 1020 cm‐3)





4.2 Compensationmechanismsindopedsemiconductors

Free carrier concentrations in semiconductors are not solely determined by their doping

concentrations. In general not every dopant atom contributes a free carrier to a crystal,

especially for high doping concentrations. The growth conditions and the Fermi level

during the growth have a strong influence on the formation of point defects. Point

defects can strongly influence the electronic and optical properties of semiconductors.

The incorporation of point defects by doping can lead to compensation, which limits or

even reduces the electrical conductivity. The formation energy and therefore the

density of point defects is a function of the Fermi energy during the growth of a sample.

The lower the formation energy of a point defect the higher the possibility of its

formation and therefore its concentration within a crystal. This model accounts for the

dopant on a lattice sites, interstitials, complexes and native point defects. With

increasing doping concentration the formation energies of charged defects decrease,

which can cause compensation of dopants in the film. A typical example for this

mechanism is Mg doped p‐type GaN where for doping concentrations above

2 x 1019 cm‐3 the formation energy of nitrogen vacancies (VN) is reduced drastically which

leads to highly compensated layers. [54] Doping concentrations above 3 x 1019 cm‐3 with


23

Mg leads to an increase of the resistivity of GaN layers [55] and at sufficient high doping

in the 1020 cm‐3 regime the material can even become n‐type conductive. [56] The

reduction of the conductivity (i.e. a decrease of the free carrier concentration) is caused

by self‐compensation. [57] Self‐compensation means that from a specific doping

concentration a further increase of the doping concentration leads to a constant (or

decreasing) free carrier concentration. [58–60] Amphoteric or interstitial behavior of

the dopant can cause a saturation of the free carrier concentration, [60] but a decrease

of the free carrier concentration needs a compensating defect of the opposite charge

species. For Mg doped GaN the compensating defects are nitrogen vacancies

(VN) [55,57] and complexes containing VN. [61,62]

In n‐type GaN, Silicon is the most common n‐type dopant, where free carrier

concentrations above 1019 cm‐3 can be achieved. Higher free carrier concentrations are

inhibited by self‐compensation, which is caused by the reduction of the formation

energy of Gallium vacancies (VGa) [54,63] and residual Carbon impurities [64]. Carbon

can also be deliberately incorporated into GaN to create semi‐insulating buffer layers

that improve the performance of AlGaN/GaN heterojunction field‐effect

transistors. [65–67] This is possible because Carbon substitutes the Nitrogen site (CN) in

GaN and acts as a deep acceptor. [68] As deep centers, the carbon acceptor and the

Gallium vacancies, are both possible candidates for the widely observed yellow

luminescence (YL). It is still of debate whether Carbon [68–71], VGa [72–75] or even a

complex both centers are the origin of the YL in GaN. Nonetheless, the YL in n‐type GaN

indicates the presence of compensation, because in all cases the YL is caused by

acceptors. Consequently by increasing doping concentration the density of VGa and

Carbon can increase, which can limit the free carrier concentration.

Besides the compensation, another challenge in Silicon doped GaN is the occurrence of

significant tensile strain levels in epilayers, [37,76,77] which is discussed in detail in

chapter 5.2.2. Every type of point defect influences the strain within the crystal, and


24

depending on their sizes and bonding lengths, point defects can affect the lattice

parameters [77,78].

The following section presents the influence of high Germanium (Ge) concentrations, in

the range of Ge = 1.3 x 1019 cm‐3 to 1.2 x 1021 cm‐3, on the free carrier concentration,

strain and crystal quality within the doped GaN layer, measured by Raman spectroscopy.

4.3 Ramanspectroscopyresults

480 500 520 540 560 580 600

Inte

nsity

(arb

. uni

ts)

Ramanshift (cm-1)

Ge = 1.2 x 1021 cm-3, n = 6.1 x 1019 cm-3

Ge = 6.9 x 1020 cm-3, n = 1.1 x 1020 cm-3

Ge = 4.0 x 1020 cm-3, n = 1.3 x 1020 cm-3

Ge = 2.5 x 1020 cm-3, n = 2.3 x 1020 cm-3

Ge = 1.1 x 1020 cm-3, n = 1.2 x 1020 cm-3

Ge = 4.7 x 1019 cm-3, n = 5.0 x 1019 cm-3

Ge = 1.3 x 1019 cm-3, n = 3.0 x 1019 cm-3

Figure 9: Raman spectra of all investigated samples. The results of SIMS and Hall effect measurements are given in the legend. The influence of the free carrier concentration on the shift of the LPP‐ mode is depicted by the arrow.

To investigate the dependence of free carrier concentration on the doping level in

Germanium doped GaN by the means of Raman spectroscopy (see chapter 2.1.3), seven

samples have been grown. The Ge content of these samples has been varied by using

different GeH4 flow rates during the growth. The flow has been varied from 0.135 to

4.5 µmol/min, which resulted in a Ge content ranging from 1.3 x 1019 cm‐3 to

1.2 x 1021 cm‐3. Hall effect measurements revealed an increase of the free carrier


25

concentration from n = 3 x 1019 cm‐3 at Ge = 1.3 x 1019 cm‐3 to n = 2 x 1020 cm‐3 at

Ge = 2.5 x 1020 cm‐3. A further increase of the Germanium concentration leads to a

decrease of the free carrier concentration to n = 1.3 x 1020 cm‐3 and lower. Figure 9

shows the Raman spectra of these seven Germanium doped samples, featuring two

E2(high) modes and a LPP‐ mode in each spectra. The E2(high) modes at 568 cm‐1

originate from the buffer layer and is stable in position and half width for all samples.

The E2(high) modes at lower wavenumbers originate from the Ge doped layer, which

show a shifting and broadening with increasing Germanium content. The LPP‐ modes

show a similar trend to the Hall effect measurements. A detailed trend of the E2(high)

and LPP‐ modes originating from the Germanium doped layer is shown in Figure 10.

565

566

567

E 2 (hi

gh) (

cm-1)

2345

FWH

M E

2 (hi

gh) (

cm-1)

0.0 2.0x1020 4.0x1020 6.0x1020 8.0x1020 1.0x1021 1.2x1021500

510

520

530

LPP- (c

m-1)

Ge content (cm-3)

Figure 10: Compilation of positions and FWHMs of E2(high) modes and positions of LPP‐ Raman modes depending on the Ge content.

With increasing Ge content in the range of Ge = 4.7 x 1019 to Ge = 4.0 x 1020 cm‐3 shifts

the position of the E2(high) mode from 566.3 to 567.2 cm‐1, the FWHM increases from

3.5 to 3.9 cm‐1. For the two higher Ge concentrations the positions of the E2(high) modes

are shifted to lower wavenumbers with a minimum of 565.7 cm‐1 and the FWHM is

considerably increased with a maximum of 4.8 cm‐1. These results show that the layers


26

with low Ge content are tensile‐strained by 0.24 GPa (see chapter 2.1.2) with a good

crystalline quality. With increasing Ge content the layers become compressive‐strained

and are almost relaxed at a Ge content of 4.0 x 1020 cm‐3 with a compressive strain of

0.07 GPa and a slightly decreased crystalline quality. For higher Ge contents the layers

are tensile strained up to 0.45 GPa and the crystalline quality is drastically decreased.

This indicates a drastic change in the defect concentration or crystalline structure within

the sample and is a hint for an onset of compensation.

The bottom graph in Figure 10 shows the process of the positions of the LPP‐ modes. It

starts at 513 cm‐1 which corresponds to a free carrier concentration of 4.8 x 1019 cm‐3

(see chapter 2.1.3). With increasing Ge concentration the position of the LPP‐ mode

shifts towards higher wavenumbers and therefore indicates an increase of the free

carrier concentration. The maximum is reached at a Ge content of 2.5 x 1020 cm‐3 with a

position of 522 cm‐1 which corresponds to a free carrier concentration of 8.6 x 1019 cm‐3.

A further increase of the Ge content leads to a shift of the LPP‐ mode towards lower

wavenumbers. At the highest Ge content of 1.2 x 1021 cm‐3 the LPP‐ mode amounts to

518 cm‐1 which corresponds to a free carrier concentration of 6.1 x 1019 cm‐3. The

possible reasons for the decrease of free carrier concentration with increasing Ge

content are due to compensation and passivation effects.


27

0.0 2.0x1020 4.0x1020 6.0x1020 8.0x1020 1.0x1021 1.2x1021

1x1020

2x1020

3x1020

4x1020

5x1020

6x1020

7x1020

Free

car

rier c

once

ntra

tion

(cm

-3)

Ge content (cm-3)

Hall effect LPP-

Compensation

Idea

l beh

avior

Figure 11: Comparison free carrier concentrations estimated by Hall effect measurements and LPP coupling over doping concentrations estimated by SIMS. The solid black lines depicts the ideal behavior for the case that Ge content and free carrier concentration are the same.

A direct comparison of the free carrier concentration, measured by Hall effect and

Raman spectroscopy, with Ge content, measured by SIMS, shows the efficiency of the

doping procedure. Figure 11 shows that the free carrier concentrations estimated from

Hall effect and Raman spectroscopy measurements show similar results with the same

trend. The large offset of the free carrier concentration between the measurement

results of the sample with a Ge content of 2.5 x 1020 cm‐3 most likely originates from an

over estimation of the free carrier concentration by the Hall effect measurement or an

inhomogeneity of the free carrier concentration over the as‐grown wafer, because the

measurements have been undertaken on two different pieces of the same wafer.

Nevertheless the increasing offset between the doping and free carrier concentration

with increasing Ge‐content provides a clear evidence of compensation.


28

4.4 PLofcompensatedsemiconductors

Photoluminescence is an excellent tool for studying the influence of high free carrier

concentrations on the electronic structure of a semiconductor. It supports the

understanding of compensation mechanisms and helps to interpret the influence of high

Ge doping with the given growth conditions on the incorporation of defects. Free carrier

concentrations beyond the Mott transition are known to have a strong influence on the

band structure of a semiconductor.

3.30 3.35 3.40 3.45 3.50 3.55 3.60 3.65

Inte

nsity

(arb

. uni

ts)

Energie (eV)

1.2 x 1021 cm-3

6.9 x 1020 cm-3

4.0 x 1020 cm-3

2.5 x 1020 cm-3

1.1 x 1020 cm-3

4.7 x 1019 cm-3

1.3 x 1019 cm-3

3.50 3.55 3.60 3.650.0

0.2

0.4

Energy (eV)

Figure 12: PL spectra of all investigated Germanium doped samples at 1.8 K. The spectra are vertically shifted and normalized for clarification. The inset presents the BMS caused high energy part of the PL spectra of the four lowest doped samples. The legend shows the Ge content measured by SIMS for the investigated samples.

Figure 12 shows the evolution of the PL spectra of Ge doped samples with increasing Ge

content. The samples have been excited by a HeCd laser with an excitation power

density of 100 W/cm2. With a Ge content of 1.3 x 1019 cm‐3 (black line) a typical spectrum

for a degenerate semiconductor is visible. The broad luminescence at 3.43 eV and

3.468 eV are caused by Ge related point defects. The luminescence at higher energies

originates from direct transitions of the conduction band (CB) to the valence band (VB),

so called band‐to‐band transitions. The high doping concentration leads to a high free


29

carrier concentration, in this case electrons, which causes a filling of the CB states. High

free carrier concentrations, especially beyond the Mott transition [49] at about

n = 7 x 1018 cm‐3 – 1 x 1019 cm‐3 [32,50], lead to a deformation of the CB caused by the

band gap renormalization [35] (BGR). In addition, the Burstein‐Moss shift [42,43] (BMS)

causes a shift of the luminescence to higher energies depending on the Fermi

energy. [79]

A further increase of the Ge content to 4.7 x 1019 cm‐3 and 1.2 x 1020 cm‐3 (red and blue

line) shows a decrease of the intensity and strong quenching of the defect related

luminescence at 3.43 eV. The second, higher energetic defect related luminescence

increases in intensity compared to the energetic higher and lower parts of the spectra

and shifts to lower energies of 3.567 eV. The BMS related luminescence shifts strongly

towards higher energies (as shown in the inset of Figure 12) due to the filling of the CB

states.

In the spectrum with a Ge content of 2.5 x 1020 cm‐3 (magenta line) the BMS related

luminescence is no longer visible and only the defect related luminescence remains,

which shifts further to lower energies of 3.465 eV. Any further increase of the Ge content

leads to a strong broadening of the luminescence and a shift to higher energies until it

reaches 3.48 eV for the highest Ge concentration of 1.2 x 1021 cm‐3 (violet line). The shift

of the defect related luminescence over the whole range of Ge concentrations is mainly

related to the BGR and less to strain, as it can be seen in the Raman spectroscopy

measurements in Figure 10. Stress in compressive direction causes an increase of the

band gap and therefore a shift of the luminescence to higher energies. Stress in tensile

direction would cause the opposite behavior. An increase of the free carrier

concentration causes a decrease of the band gap and therefore a shift of the

luminescence to lower energies. A comparison of the shift behavior of the defect related

luminescence with the free carrier concentrations shows the direct interaction.

The strong broadening of the defect related luminescence with increasing Ge content

can be explained by an energetic broadening of a Ge related defect band within the band


30

gap. The defect band evolves into band‐to‐band transitions of a GaNGe alloy. This effect

is already known, e.g. from GaAs:N. [80]

0.0 5.0x1019 1.0x10200.0

5.0x1019

1.0x1020

1.5x1020

Free

car

rier c

once

ntra

tion

(cm

-3)

Ge-content (cm-3)

Hall effect LPP-

PL

Figure 13: Comparison of free carrier concentrations estimated by Hall effect measurements, LPP coupling and shift of the Fermi edge estimated by PL measurements over Ge content given by SIMS measurements.

The BMS related luminescence and the knowledge of the strain values of the layers allow

for an estimation of the free carrier concentration from the PL measurements. The

saddle point of the high energy side of the above band gap luminescence shows the

maximum filling of the CB and is equal to the Fermi energy. The strain within the Ge

doped layers is given by the shift of the E2(high) mode in the Raman spectroscopy

measurements shown in chapter 4.3 and allows for the calculation of the band gap

value. The knowledge of the band gap and Fermi energy values allow for the calculation

of the free carrier concentration (see chapter 2.1.3).

A comparison of the free carrier concentrations estimated by PL, Raman and Hall effect

measurements of the three lowest Ge doped samples is shown in Figure 13.


31

3.40 3.45 3.50 3.55 3.60 3.650

10000

20000

30000

40000

50000

60000

70000

80000

1.2 x 1021 cm-3

6.9 x 1020 cm-3

4.0 x 1020 cm-3

2.5 x 1020 cm-3

1.1 x 1020 cm-3

4.7 x 1019 cm-3

1.3 x 1019 cm-3

Inte

nsity

(arb

. uni

ts)

Energie (eV)

1019 1020 10210.02.0x106

4.0x106

6.0x106

8.0x106

1.0x107

Inte

grat

ed in

tens

ity (a

rb. u

nits

)

Ge content (cm-3)

Figure 14: High excitation power density PL spectra of all investigated Germanium doped samples at 1.8 K. The spectra are vertically shifted for clarification. The inset presents the relative change of the integrated luminescence of all spectra. The legend shows the Ge content measured by SIMS for the investigated samples.

High excitation power density PL measurements have been performed to saturate most

non‐radiative recombination channels and defect luminescences to enhance the

visibility of band‐to‐band transitions, as shown in Figure 14. Here, the excitation power

density of 50 kW/cm2 is 500 times higher than for the performed PL measurements

shown in Figure 12.

While the defect luminescence at 3.43 eV almost disappeared and saturated the

intensity of the luminescence at 3.468 eV is drastically enhanced and became even

stronger than the BMS related luminescence. This result indicates the origin of the

emission at 3.468 eV has a very high density of states.

The relative PL intensities in comparison to each other, as shown in the inset of Figure

14, displays that the overall near band edge luminescence first increases at a free carrier

concentration of 1.1 x 1020 cm‐3, then constantly decreases with increasing Ge content

and increases again for the highest Ge content. A decrease of the intensity of the near


32

band edge luminescence can be caused by an increase of non‐radiative recombination

or an increase of luminescences at lower energies.

1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.610-4

10-3

10-2

10-1

100 1.2 x 1021 cm-3

6.9 x 1020 cm-3

1.1 x 1020 cm-3

4.7 x 1019 cm-3

Inte

nsity

(arb

. uni

ts)

Energy (eV)

Figure 15: PL spectra of all investigated Germanium doped samples at 5 K. The data is normalized for better comparison. The legend shows the Ge content measured by SIMS for the investigated samples.

Figure 15 shows the PL emission from 1.8 eV to 3.7 eV of four selected Ge doped samples

on a logarithmic scale. The spectra are normalized to compare the intensity ratio

between the near band edge luminescence and the yellow luminescence. For Ge

concentrations of 1.1 x 1020 cm‐3 and below the Ge has almost no effect on the yellow

luminescence. With increasing Ge content, the luminescence of band‐to‐band

transitions disappears and the yellow luminescence increases relative to the normalized

intensity of the near band edge luminescence. An increase of the yellow luminescence

is an indicator for compensation although its exact origin and the nature of the

compensation are not known. The different maxima of the yellow luminescence for

different Ge concentrations with a difference of up to 300 meV and the not symmetric

shapes indicate that multiple transitions are involved.


33

4.5 Dynamicsandtimeresolvedanalysis

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

101

102

103

104

2.5 x 1020 cm-3

1.1 x 1020 cm-3

4.7 x 1019 cm-3

1.3 x 1019 cm-3

Inte

nsity

(arb

. uni

ts)

Time (ns)

0

100

200

300

Dec

aytim

e (p

s)

1019 1020 10210.00.51.0

Ge content (cm-3)

Figure 16: Transients (symbols) and biexponential fits (solid lines) of the emissions of the four lowest Ge doped samples measured at the high energy flank of each spectrum. The two insets show the results of the applied fit model (two decay times and the ratio of the amplitudes).

Figure 16 shows the change of the decay of the luminescence at the high energy flank

for the four lowest doped samples. The rise times of the emission are below the

detection limit (<10 ps) and the decay times are multi‐ or stretched exponential. A non‐

monoexponential decay of the emission is commonly observed in inhomogeneous

materials like InGaN. In this case the inhomogeneous distribution of Ge in GaN causes

an inhomogeneous distribution of free carrier concentrations. The value of the beta (β)

factor (blue triangles in the inset of Figure 16) indicates how strong the distribution of

Ge atoms is disordered. For the usual monoexponential function is β = 1. Disordered

systems have values between 0 and 1, where the lower the value, the higher the

disorder or clustering. The fitting results of the measured transients reveal a β value of

0.94 for the Ge concentration of Ge = 1.3 x 1019 cm‐3, which is a value close to a

monoexponential decay. With increasing Ge content decreases the β value and reaches

0.53 at Ge = 2.5 x 1020 cm‐3.


34

Increasing free carrier concentrations cause screening of excitons and a decrease of

their binding energy. [11] Once the Mott density is reached, band‐to‐band transitions

dominate the emission. Since the band‐to‐band transition probability is smaller than the

exciton transition probability, the decay time increases with increasing free carrier

concentration. This effect generally competes with an increase of other recombination

channels caused by compensation and energetically deep defects. The decay time

decreases with increasing Ge content despite the fact that the free carrier concentration

further increases. This effect could be explained by an increasing number of other

recombination probabilities, like transitions over deep defects, e.g. the yellow

luminescence, and non‐radiative recombination. The increase of the NBE intensity in

Figure 14 and the stable intensity of the YL in Figure 15 with increasing Ge content show

that this is not the case. The only remaining transition is the strong Ge related peak at

3.468 eV.


35

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

101

102

103

104

1.2 x 1021 cm-3

4.0 x 1020 cm-3

1.1 x 1020 cm-3

1.3 x 1019 cm-3

Inte

nsity

(arb

. uni

ts)

Time (ns)

100

150400

600

t3rise

Dec

aytim

e (p

s)

t2decay

t1decay

1019 1020 10210.000.020.04

Am

plitu

de

ratio

A1/

A2

Ge content (cm-3)

Figure 17: Transients (symbols) and triexponential fits (solid lines) of the emissions of the four selected Ge doped samples measured at the maximum of the strongest peak around 3.468 eV. The two insets show the results of the applied fit model for all Ge doped samples (one rise time, two decay times and the ratio of the amplitudes of the decay times). Only the highest Ge doped sample shows a mono exponential transient.

Figure 17 shows the transients of the strong peak close to 3.468 eV of four selected

samples. The decay time decreases with increasing Ge content and all transitions show

an increased rise time except for the highest doped sample. Theoretically all transients

contain a stretched exponential rise time and a stretched exponential decay time, but

such a fit model requires too much computing power. Therefore a simpler triexponential

(one rise and two decay times) fit has been applied. The resulting fit parameters are

shown in the inset of Figure 17. With increasing Ge content the first decay time

decreases, while the second decay time and their amplitude ratio decreases. The rise

times increase from Ge concentrations of 1.3 x 1019 cm‐3 to 1.1 x 1020 cm‐3 and decrease

afterwards with increasing Ge content.

The increase in the difference of the two decay times and the decrease of their

amplitude ratio is basically the same as a decrease of a β value in a stretched exponential

function. This indicates an increase of disorder or clustering of Ge atoms.


36

A decrease of the decay time with increasing Ge content is caused by the increase of the

yellow luminescence and non‐radiative recombination. While increased rise times are

caused by the transition from the filled conduction band to the Ge related defect band.

The increase and decrease of the rise times are consistent with the observation of the

BMS related luminescence in Figure 14. The monoexponential transient of the highest

Ge doped sample indicates that this emission does not originate from a defect band in

GaN, but from the conduction band of the ternary semiconductor GeGaN.

4.6 Discussion

The results of Hall effect, Raman spectroscopy and PL measurements (Figure 11 and

Figure 13) show that free carrier concentrations above 1 x 1020 cm‐3 are achievable in

Ge doped GaN. The maximum free carrier concentration is achieved at a Ge

concentration of 2.5 x 1020 cm‐3. Compensation effects are barely measurable up to a Ge

concentration of 1.1 x 1020 cm‐3. From Ge = 1.3 x 1019 cm‐3 to 1.1 x 1020 cm‐3 the free

carrier concentration increases almost linear with the doping concentration and is

roughly the same. Raman spectroscopy measurements show a shift of the E2(high) by

only 0.5 cm‐1 to higher wavenumbers, which corresponds to 0.17 GPa strain in the

compressive direction, and a slight increase of the FWHM by 0.4 cm‐1, which

corresponds to a slight decrease of the crystalline quality. PL measurements show an

increase of the intensity of the NBE luminescence with increasing Ge concentration

while the intensity of the YL does not increase. That indicates the number of transitions

over energetically deep defects does not increase. Also the time resolved measurements

show no hint of compensation mechanisms. The decay time of the high energy flank

decreases with increasing Ge content, but the intensity of the YL does not increase in

comparison to the NBE luminescence. While the high excitation density measurement

(Figure 14) shows that the intensity of the NBE luminescence increases with Ge content,

especially the intensity of the peak at 3.468 eV. These results indicate that the non‐


37

radiative recombination does not increase, but the recombination probability through

the Ge related feature at 3.468 eV.

From a Ge concentration of 2.5 x 1020 cm‐3 a reduction of the intensity NBE luminescence

is observable, which marks the onset of compensation. Higher Ge concentrations cause

a decrease of the free carrier concentration due to compensation. The most likely

candidates for compensation acceptors are VGa and Carbon, due their low formation

energies at high Fermi energies. Other acceptors like Germanium atoms on a Nitrogen

lattice site (GeN) are very unlikely to be formed by compensation because of their very

high formation energy [81]. An increase of VGa and/or Carbon impurities is confirmed by

the increase of the YL (Figure 15). The sudden shift of the E2(high) mode by 1.5 cm‐1 to

lower wavenumbers and the increase of the corresponding FWHM by 0.8 cm‐1 at

6.9 x 1020 cm‐3, indicates a change in the crystalline structure. The transients of the peak

maximum around 3.468 eV (depending on the Ge concentration) show a rise times

between 74 and 89 ps, except for the sample with the highest Ge concentration of

1.2 x 1021 cm‐3. The rise times originate from transitions of carriers from the CB to the

Ge related defect band. The absence of a measurable rise time, the increase of the

intensity of the NBE luminescence, the broadening and shift of the NBE luminescence to

higher energies are collectively characteristic of a degenerated semiconductor. The CB

states are filled due to the effect of the BMS. The Ge related defect band at 3.468 eV

broadens and its DOS increases with increasing Ge content. At the two highest Ge

concentrations of 6.9 x 1020 cm‐3 and 1.2 x 1021 cm‐3 this defect band disappears and

transforms into a GeGaN band‐to‐band transitions.


38

5 Electronicexcitationsstabilizedbyadegenerateelectrongasin

highlyGedopedGaN

In chapter 4 it was shown that Germanium is an outstandingly efficient dopant in n‐type

GaN, but at very high Ge concentrations are compensation effects and a transfer to the

ternary semiconductor GeGaN observable. The question arises, what happens if no

compensation effect occurs and how do excitons interact with a three dimensional

electron gas. To clarify this matter a second set of Ge doped GaN has been investigated.

This set of Ge doped samples is compared by the means of Raman spectroscopy (chapter

5.2) to a set of Si doped samples, which has been grown with the same growth

conditions. The interaction of excitons and other discovered quasiparticles with a

degenerated electron gas are shown Chapter 5.3.

5.1 Investigatedsamples

These samples were grown by MOVPE in an AIXTRON AIX 200/4 RF‐S reactor. The (0001)

sapphire substrates were 0.25° off‐oriented towards the m‐direction. All structures

consist of an AlN seed layer and an AlGaN buffer layer followed by an 1.3 µm thick

undoped GaN buffer layer. Subsequently, an approx. 1 µm thick doped GaN layer was

grown. The doped GaN layer was grown at 1075 °C with a trimethylgallium (TMGa) flow

of 96 µmol/min and an ammonia flow of 89 mmol/min at 200 mbar in hydrogen

atmosphere with a total gas flow of 7.5 standard litre per minute (slm). Silane (SiH4,

100 ppm in 6H H2), germane (GeH4, 10% in 6H H2), or isobutylgermane (IBGe) were used

as dopant sources. The achieved free carrier concentrations were determined by Hall

effect measurements at room temperature. For three samples the dopant

concentrations were determined by secondary ion mass spectrometry (SIMS). The

results are listed in Table 4.


39

Table 4 Free carrier and dopant concentrations of the investigated samples measured by Hall effect and SIMS measurements.

Sample Dopant Free carrier concentration (Hall)

Dopant concentration (SIMS)

MD6968 Reference sample ~ 1.0 x 1017 cm‐3

MD6960 Germane 2.0 x 1018 cm‐3

MD6957 Germane 1.5 x 1019 cm‐3 2.2 x 1019 cm‐3




MD6981 Silane 1.2 x 1018 cm‐3

MD6977 Silane 1.4 x 1019 cm‐3

MD6979 Silane 2.2 x 1019 cm‐3

5.2 Ramanspectroscopyresults

5.2.1 InvestigationofstrainandcrystallinequalitybythemeansofRaman

spectroscopy

The non‐polar character of the E2 modes makes them an excellent choice for

determination of strain within a wurtzite crystal. Due to higher intensities of the E2(high)

mode and a stronger stress introduced shift compared to the E2(low) mode in GaN,

strain in GaN crystals is usually investigated with the E2(high) mode. The position of the

E2(high) mode of relaxed GaN at room temperature is at 567 cm‐1. [82] If the crystal is

tensile strained this mode shifts to lower wavenumbers. In the case of compressive

strain, it shifts to higher wavenumbers. The amount of strain can be determined by

pressure coefficients for the applied strain in a material. The E2(high) mode gives no

direct information about strain values in a‐ or c‐direction in the crystal. The given strain

conditions are generally determined by the used growth procedure of the crystal. All

investigated samples were MOCVD grown on sapphire and c‐plane. That means the

given strain is biaxial, perpendicular to the c‐axis. The shift of the E2(high) mode in biaxial


40

strain in GaN is 2.86 cm‐1 per GPa. [21] With the knowledge of these values, the strain

can be calculated from the difference between the relaxed and measured value of the

E2(high) mode.

The investigation of the FWHM of the E2(high) mode gives information about the

crystalline quality of the sample. The smaller the FWHM the lower the number of

structural and point defects. It is not possible to get an exact value for the number of

defects, but a qualitative evaluation of the impact of growth technique and doping on

the crystalline quality.

400 450 500 550 600 650 700 750 80010-3

10-2

10-1

100

Sapphire

E1(LO)E1(TO)

A1(TO)

Inte

nsity

(arb

. uni

ts)

Raman shift (cm-1)

E2(high)

A1(LO)

Sapphire

*

z(..)z

Figure 18: Room temperature Raman spectroscopy spectra of the reference sample in backscattering geometry.

Figure 18 shows the Raman spectra of the undoped reference sample at room

temperature. In the z(..)z geometry, over the 400 to 800 cm‐1 range for GaN, only the

E2(high) and A1(LO) mode are expected. The additional sapphire modes originate from

the substrate. The crystal is transparent for the wavelength of the applied laser and the

focal depth of the applied objective of approx. 4 µm is larger than the thickness of the

epitaxial grown layers. Therefore the substrate modes are visible. The A1(TO), E1(TO) and

E1(LO) are forbidden in this geometry. Nevertheless, they are visible due to multiple light


41

scatterings within the sample and the large numerical aperture of 0.95 of the objective

lens. The additional * marked peak is most likely a local mode related structural

defects [83,84]. The E2(high) mode lies at 569.8 cm‐1 with a FWHM of 2.7 cm‐1. This

observation indicates a compressive stress of 0.98 GPa and a superior crystalline quality

if compared to literature values of heteroepitaxial growth.

5.2.2 ComparisonofstructuralpropertiesinSiandGedopedGaN

5.2.2.1 InvestigationoftheE2(high)modesandstress

564 566 568 570 572 574 564 566 568 570 572 574

2.2 x 1019 cm-3

1.4 x 1019 cm-3

1.2 x 1018 cm-3

GaN:nid

8.9 x 1019 cm-3

5.1 x 1019 cm-3

3.4 x 1019 cm-3

1.5 x 1019 cm-3

2.0 x 1018 cm-3

GaN:nid

rela

xed

GaN

Inte

nsity

(arb

. uni

ts)

Raman shift (cm-1)

GaN:Ge GaN:Si

rela

xed

GaN

Figure 19: Depiction of E2(high) Raman modes of Ge (left) and Si (right) doped GaN epilayers. The free carrier concentrations ranges from GaN:nid (black curve) to up to 3 x 1019 cm‐3 for the Si and 8.9 x 1019 cm‐3 for the Ge doped layers.

The influence of Ge and Si doping on the E2(high) modes and therefore on crystalline

quality and strain is depicted in Figure 19. The two black spectra at the bottom present

the undoped reference sample. The other lines originate from Ge (left) and Si (right)

doped samples with increasing free carrier and doping concentration from bottom to

top. Comparable carrier densities in the Si and Ge doped GaN layers are displayed next

to each other with matching colors. Free carrier concentrations for the Si doped samples

range from 1.2 x 1018 cm‐3 to 2.2 x 1019 cm‐3 and for the Ge doped samples from


42

2 x 1018 cm‐3 to 8.9 x 1019 cm‐3. This picture clearly shows that the influence of Ge doping

on the E2(high) mode is much smaller compared to Si. The Ge doped samples exhibit a

small increase from 2.7 cm‐1 to 3.4 cm‐1 of the FWHM and almost no impact (less than

0.6 cm‐1) on the position. This indicates that Ge doping has almost no influence on the

strain in the doped layer. The increase of the FWHMs show a minor decrease of the

crystalline quality, which is expected with implantation of impurities. The Si doped

samples, on the other hand, separate into two peaks, which is most clearly illustrated in

the green spectra with a free carrier concentration of 1.4 x 1019 cm‐3. With increasing Si

content one peak shifts towards higher wavenumbers maintaining its FWHM, the other

shifts towards lower wavenumbers with a strong increase of the FWHM up to 4.5 cm‐1.

The peak at lower wavenumbers originates from the Si doped layer and the peak at

higher wavenumbers originates from the undoped buffer layer. Si induces strain in

tensile direction and deterioration of the crystalline quality. The strain in tensile

direction in Si doped GaN layers is commonly attributed to a continuous variation of

edge‐type dislocations angles with increasing doping concentration [37,76,77,85–87].

The high energy shift of the E2(high) mode in the buffer layer of the Si doped samples is

caused by compressive stress that is introduced by the Si doped top layer during the

growth.


43

564 566 568 570 5720.00.20.40.60.81.01.21.41.61.82.0

BufferGaN:nid

Posi

tion

(µm

)

Raman shift (cm-1)

GaN:Ge

a)

564 566 568 570 572

b)

BufferGaN:nid

GaN:Si

Figure 20: Comparison of depth profile of E2(high) Raman modes of a Ge doped GaN epilayer with a free carrier concentration of 5.1 x 1019 cm‐3 and a Si doped GaN epilayer with a free carrier concentration of 2.2 x 1019 cm‐3. These data reflects the development of strain and crystalline quality throughout the deposited layers.

The Si introduced tensile strain becomes clearly visible in the cross section scan in Figure

20. It shows the evolution of the E2(high) mode over sample thickness. The

corresponding spatially resolved spectra are normalized and color‐coded in blue to red,

indicating a rise in intensity. The measured Raman shift in cross section is not

representative for the volume crystal due to relaxation towards the edge in a strained

crystal. The broken edge leads also to additional structural defects which increase the

FWHM compared to the volume crystal. However, the relative change of the peak

positions and their FWHMs still reflects the strain variation over the depth of each film.

The buffer layer of the Si doped sample with a carrier concentration of 2.2 x 1019 cm‐3

shows a variation of the E2(high) mode’s FWHM in between 3.4 cm‐1 and 5.0 cm‐1, while

its energetic position close to the nucleation layer yields 569.5 cm‐1 and reaches

568.2 cm‐1 as soon as the interface to the Si‐doped layer is approached. Within the Si

doped layer the shift continues to 567 cm‐1.


44

Consequently, the Ge doped sample, with a free carrier concentration of 5.1 x 1019 cm‐3,

shows a small increase of the FWHM from 3.1 cm‐1 to 3.9 cm‐1 and a stable position of

the E2(high) mode over the whole sample thickness at 569.5 cm‐1. Even though these

absolute values are not representative for the volume crystal, the relative change of the

FWHM and position of the E2(high) over thickness shows the differences of Si and Ge

doping. Si doping does not only affect the doped layer by strain in tensile direction and

decreased crystalline quality, it also affects the buffer layer. That means the change of

the inclination angle of edge type dislocations influence the strain conditions in the layer

below. Ge doping on the other hand has no influence at all on the strain conditions

within the buffer layer.

568

570

572

1017 1018 1019 1020

2.53.03.54.04.5

1 cm-1Relaxed GaN 10 K

GaN:nid GaN:Si 300 K GaN:Ge 300 K GaN:Ge 10 K

Pos

ition

E 2(

high

) (cm

-1)

Relaxed GaN 300 K

1.6 cm-1

FWH

M

E 2(hi

gh) (

cm-1)


Figure 21: Compilation of positions and FWHMs of E2(high) Raman modes depending on free carrier concentration for Ge and Si doped GaN epilayers.

The fitting results of the E2(high) modes originating from the doped layer of the

investigated Ge and Si doped samples at room temperature (Figure 19) and also at 10 K

for the Ge doped samples are depicted in Figure 21. The upper part shows the evolution

of the position of the E2(high) and therefore the strain over free carrier concentration.

While the lower part shows the evolution of the FWHM of the E2(high) and therefore


45

the crystalline quality over free carrier concentration. Each data point in the room

temperature measurements represents the average of 16 peak positions originating

from 16 different sample positions. The error bars represent the standard deviation. The

dashed lines indicate the position of the E2(high) in fully relaxed and undoped GaN at

300 K (567 cm‐1) and 10 K (568 cm‐1) [82]. The Ge doped samples show a small shift to

higher wave numbers over free carrier concentration, except for the sample with the

highest free carrier concentration. The maximum shift (at 5.1 x 1019 cm‐3) in comparison

to the undoped sample is 0.6 cm‐1, which corresponds to an increase of the compressive

strain by 0.21 GPa. The FWHM is only increased by 0.3 cm‐1 at this point. The layer with

the highest free carrier concentration (8.9 x 1019 cm‐3) is 3.6 µm thick, which means

2.9 µm thicker than the doped layers of all other samples. The increased thickness of

this layer leads to a stronger curvature of the surface accompanied by a higher number

of structural defects and therefore the Ge doped layer of this sample is slightly more

relaxed than the thinner ones. The position of the E2(high) mode for this thick layer is

569.7 cm‐1, which within the error of this measurement is the same value of the

undoped sample with 569.8 cm‐1. The increased value for the FWHM of the E2(high)

mode to 3.4 cm‐1 for the thick layer is caused by the increased number of defects and

the change of strain over its thickness. The general small increase of the FWHMs E2(high)

mode in the Ge doped layers is an indicator for a very small or almost no compensation

within this material. The error bars, which indicate the standard variation of the

positions and FWHMs for all the measured spots on each sample, are mostly below

0.1 cm‐1. Only the thicker sample has a value of 0.2 cm‐1. That means defect distribution

within these layers is absolutely homogenous and is not effected by the Ge doping.

The Si doped samples show a shift to lower wave numbers and an increase of the FWHM

of the E2(high) mode over free carrier concentration. With increasing Si content the

strain in tensile direction, originating from the change of the inclination angle of edge

type directions, increases and the crystalline quality decreases drastically as well. The

highest Si doped sample shows a shift back to higher wave numbers and a decrease of

the FWHM. Doping with Si concentrations beyond 1019 cm‐3 even evokes a more


46

dramatic deterioration of the GaN layer due to the effect of back etching, yielding an

immediate surface roughening [88,89]. This indicates that the shift back to higher wave

numbers is directly correlated to the onset of island growth and surface roughening. A

comparison of the error bars and therefore the standard deviation of the peaks positions

and FWHM over the samples show that the Si doped samples generally have a larger

difference than the Ge doped samples. Consequently the Si doped samples are much

more inhomogeneous compared the Ge doped ones. This inhomogeneity increases with

Si content.

The Raman measurements at 10 K show a shift of the E2(high) mode to lower wave

numbers by 1.6 cm‐1 for all Ge doped samples. The change of the position of the E2(high)

mode due to the change of the lattice constant over temperature from 300 K to 10 K in

GaN is 1 cm‐1. The additional shift of 0.6 cm‐1 comes from additional stress originating

from the sapphire substrate, because of the higher thermal expansion coefficient

compared to GaN. [29] This result shows that the GaN layers are 0.2 GPa more strained

at 10 K compared to room temperature. This additional strain has to be taken into

account for calculating the relaxed positions of luminescence peaks in low temperature

photo luminescence spectroscopy.


47

5.2.2.2 InvestigationoffreecarrierconcentrationsandLPPcoupling

400 420 440 460 480 500 520 540 560 580 600

8.9 x 1019 cm-3

5.1 x 1019 cm-3

3.4 x 1019 cm-3

1.5 x 1019 cm-3

2 x 1018 cm-3

GaN:nid

Inte

nsity

(arb

. uni

ts)

Ramanshift (cm-1)Figure 22: Raman spectra of all investigated uncompensated Germanium doped GaN samples. The behavior of the LPP‐ mode with free carrier concentration is clearly visible.

Figure 22 shows Raman spectra of the first set of Germanium doped GaN samples and

the evolution of the LPP‐ mode with increasing free carrier concentration. The free

carrier concentrations for each sample are given by Hall effect measurements. In Figure

23 the position of the LPP‐ mode is drawn against the free carrier concentration for the

four highest doped samples. The solid lines show the calculated position of the LPP‐

mode for a relaxed sample (black line) and for a 1 GPa compressively strained layer (red

line). The compressive stress has almost no effect on the position of the LLP‐ mode in

the range of 1019 cm‐3 and results into a shift of less than 2 cm‐1 towards higher

wavenumbers at 1020 cm‐3. The comparison of the Hall effect and Raman spectroscopy

results shows a strong agreement in the estimated free carrier concentrations,

especially for the highest doped samples.


48

1019 1020440450460470480490500510520530

LPP- (c

m-1)


relaxed 1 GPa compressive strained

Figure 23: The solid lines show the calculated undamped position of the LPP‐ mode over free carrier concentration for the relaxed case (black line) and the by 1 GPa compressively strained case (red line). The black dots show of the measured LPP‐ position over free carrier concentration measured by Hall effect measurements.

This indicates that the Hall measurements were not influenced by other layers or surface

effects and that the influence of plasma damping and strain on the determination of the

free carrier concentration by LPP coupling is negligible small.

To determine how efficient the doping of a layer is for a given growth procedure,

secondary ion mass spectrometry (SIMS) has been performed. SIMS measurements

allow for determination of Germanium concentration within the GaN layers over

thickness as shown in Figure 24.


49

0 500 1000 1500 2000 2500 3000 35001017

1018

1019

1020 ~1.8x1020 cm-3

7.0x1019 cm-3

4.6x1019 cm-3

Ge

conc

entra

tion

(cm

-3)

Depth (nm)

MD6957 MD6958 MD6983 MD7906

2.2x1019 cm-3

Figure 24: SIMS measurement results of the four highest Ge doped GaN layers.

The sample with the lowest Ge concentration (green line) of 2.2 x 1019 cm‐3 shows a

homogeneous Ge concentration over the whole thickness. The two samples with the

medium Ge concentrations (blue and teal) have a decreasing Ge concentration over

thickness with a difference of approx. 2 x 1019 cm‐3 between surface and the interface

to the buffer layer. The presented Ge concentrations result from the average Ge

concentration over the whole thickness. The sample with the highest Ge content

(magenta line) is five times thicker than the other samples and seems to be very

inhomogeneous over thickness. This sample suffers from a higher concentration of

structural defect such as v‐pits. These structural defects lead to a more porous and

amorphous crystalline structure, which causes a huge error in the determination of the

Ge concentration (in the range of 4 x 1019 cm‐3). Additionally, the ion beam cuts faster

in the thicker porous sample than in the thinner ones with a higher crystalline quality.

This makes the calibration of the correct depth very difficult.


50

0 2 4 6 8 10 12 14 16 18 200

2

4

6

8

10

12

14 Ideal behavior of negligible compensation

LPP- (Raman) Hall effect

Free

car

rier c

once

ntra

tion

(1019

cm

-3)

Germanium concentration (1019 cm-3)

Figure 25: Comparison of Hall effect, Raman with SIMS measurements.

The actual Ge concentration measured by SIMS compared with the free carrier

concentration measured by Hall effect and Raman spectroscopy clearly demonstrate the

degree of compensation within the grown layers, which is shown in Figure 25. A direct

comparison of the free carrier concentrations measured by Hall effect and Raman

spectroscopy shows that both results are very close to each other and within the given

error of the measurements. The solid line shows the ideal behavior of an

uncompensated semiconductor, where every single dopant atom generates a free

electron. The three thin samples with Germanium concentrations below 8 x 1019 cm‐3

show almost this ideal behavior. Only the thicker sample with a Ge concentration of

approx. 18 x 1019 cm‐3, has a much higher doping than free carrier concentration, which

is caused by compensation. The SIMS measurements give no insight into the energetic

positions of the Ge impurities within the band gap. The Ge atoms which do not

contribute a free electron can be deeply bound or even present in the lattice as

acceptors. A clearer insight into the nature of these kind of impurities is given by

photoluminescence measurements in chapter 5.3.


51

5.3 OpticalpropertiesofuncompensateddegeneratedelectrongasinGaN

The influence of high free carrier concentrations on electronic states in heavily doped

semiconductors is well‐known. [10] Rising free carrier concentrations alter the optical

characteristics of the semiconductor through the Burstein‐Moss shift [42,43] (BMS),

band gap renormalization [35] (BGR), and Pauli blocking of the optical transitions. [44]

In the limit of high free carrier concentrations, decreasing exciton binding energies

facilitate a transition to Mahan‐like excitons [51] at the Fermi edge singularity. [11]

Excitonic excitations above the band edge have already been discussed for bulk

systems [90] above the Mott density [49], the onset of a metal like state. The quasi‐

bosonic character of excitons is lost upon their dissociation into purely fermionic

constituents (electron and hole). The spatial confinement inherent to

nanostructures [91–93] like quantum dots [94,95] and quantum wells [96,97] stabilises

quasiparticle complexes causing distinct optical features even under intermediate free

carrier concentrations, including trions in two‐dimensional crystals like MoS2

layers. [98,99]

In this chapter a novel class of exciton like particles, named collexons, will be presented

and discussed. Collexons can increasingly be stabilized with rising density of a

degenerate electron gas in a bulk semiconductor (GaN) up to 100 K due to many‐particle

effects. Upon optical excitation long‐lived (> 250 ps), strongly localized quasiparticles

are formed, which do not only comprise a single electron‐hole pair, but also the

collective excitation of the entire degenerate electron gas in the high‐density limit.

Contradicting the commonly observed behavior of heavily doped semiconductors, the

entire emission intensifies and its decay time slows towards increasing free electron

concentration, while the collexonic emission narrows spectrally, as long no

compensation occurs. Based on theoretical modelling of this intricate many‐particle

problem, a model that is consistent with all experimental findings is presented,

demonstrating an increasing bosonization of collexons with rising electron

concentration. The results confirm that, collexons are fundamental, many‐particle


52

excitations in the presence of a degenerate electron gas irrespective of the particular

semiconductor host material.

5.3.1 LowtemperaturePhotoluminescencespectroscopy

To study the optical properties of a degenerated electron gas in a highly Ge doped GaN

all samples have been investigated by photoluminescence (PL) measurements at 1.8K.

At first, the near band edge luminescence of the not intentionally doped (nid) GaN

reference sample is shown in Figure 26.

3.46 3.47 3.48 3.49 3.50 3.51 3.52 3.53 3.54

0.01

0.1

1

FXn=2A FXC

FXB

FXA

DBX2

Inte

nsity

(arb

. uni

ts)

Energy (eV)

DBX1

Figure 26: PL spectra of the undoped GaN sample at 1.8 K in logarithmic scale featuring at least two donor bound excitons (DBX) and two free excitons (FX) originating from the A‐ and B‐valence band. The two free excitons originating from the excited sate of the A‐valence band and the C‐valence band are weakly visible.

The luminescence of the reference GaN:nid sample is dominated by the luminescence

of donor bound excitons (DBX). Additionally, two free excitonic contributions are visible

at 3.5015 eV and 3.5098 eV originating from the A‐ and B‐valence bands. Luminescence

from the C‐valence (3.533 eV) band and the excited state of the A‐valence band


53

(3.521 eV) are barely visible and just roughly estimated. The donor bound excitons in

this energetic regime in MOCVD grown GaN samples typically originate from Oxygen,

Silicon, Nitrogen vacancies (VN) or Nitrogen vacancies Hydrogen complexes

(VN‐H). [63,100–103] The positions of the free excitons are shifted to higher energies

compared to the relaxed values in GaN (FXA = 3.4751 eV, FXB = 3.4815 eV,

FXC = 3.493 eV). [104]

The biaxial strain within the GaN layer has a strong influence on the band gap value.

With increasing compressive strain increases the energetic difference between the A‐,

B‐, and C‐valence band and the conduction band. With the knowledge of the exciton

deformation potentials [26] it is possible to calculate the shift of the excitonic

luminescence as discussed later in detail with Figure 36.

The intensity of the free excitons is only two orders of magnitude smaller in comparison

to the bound excitons and the full width at half maximum (FWHM) of the FXA is below

3 meV. This result confirms the outstanding crystalline quality of the MOCVD grown

samples. These superior crystalline qualities are generally achieved by hydride vapor

phase epitaxy (HVPE) or epitaxial lateral overgrowth (ELOG) grown samples. [28,105]


54

3.1 3.2 3.3 3.4 3.5 3.6 3.70

1

2 BMS

Inte

nsity

(arb

. uni

ts)

Energy (eV)

8.9 x 1019 cm-3

5.1 x 1019 cm-3

3.4 x 1019 cm-3

1.2 x 1019 cm-3

2.0 x 1018 cm-3

GaN:nid

BGR

Figure 27: PL spectra of all investigated Germanium doped samples at 1.8 K. The spectra are vertically shifted for clarification. The legend shows the free carrier concentrations for the investigated samples given by Hall effect measurements.

Figure 27 shows the PL spectra at 1.8 K of the undoped reference sample and five Ge

doped samples with free carrier concentrations ranging from 3 x 1018 cm‐3 to

8.9 x 1019 cm‐3 acquired by Hall effect measurements. While the luminescence of the

not‐intentionally doped sample is still dominated by donor bound excitons, [100] any

rise of the Ge concentration (nGe) by intentional Ge doping causes a spectral broadening

as frequently observed for various semiconductor compounds. [79,106] At a sufficiently

high free electron concentration, the corresponding Fermi energy passes the conduction

band and causes above band gap luminescence due to the BMS [42,43], while the

simultaneous extension of the luminescence towards lower energies is caused by the

BGR [47]. The corresponding critical carrier density is the Mott‐density, which is about

n = 7 x 1018 cm‐3 – 1 x 1019 cm‐3 in GaN. [32,50]


55

3.485 3.490 3.495 3.500 3.505 3.5100.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

XU

Inte

nsity

(arb

. uni

ts)

Energy (eV)

8.9 x 1019 cm-3

5.1 x 1019 cm-3

3.4 x 1019 cm-3

XL

Figure 28: PL spectra of the two peaks XL and XU of the three highest Ge doped samples without the PL of band‐to‐band transitions. The spectra are shifted vertically for clarity.

Despite the present high doping level, two pronounced peaks stand out at free carrier

concentrations from 3.4 x 1019 cm‐3 and above, labelled XL and XU (see Figure 28). The

energetic positions, FWHMs and energetic differences are given in Table 5. The emission

lines with such low FWHM values between 4 and 9 meV, represent a highly unusual

observation at these high doping levels. Contributions from any luminescence from the

underlying buffer layer of pure GaN can be excluded as shown in the following chapter

by cross‐section and depth‐resolved cathodoluminescence. Inhomogeneities in the

doped layer, which cause local doping free areas, can be excluded as well, as evidenced

in the Raman spectroscopy results.


56

Table 5: Stress values, energetic positions, FWHM and difference of XL and XU for the three highest doped samples. The stress values are extracted from the Raman spectroscopy measurements, the PL values are extracted from Figure 28.

Free carrier concentration n (× 1019 cm‐3)

Stress at 10K (GPa)

XL position

Ε ∗(eV)

XU position

E (eV)

E ‐ Ε ∗ (meV) ↓

XL FWHM (meV) ↓

XU FWHM (meV) ↓

3.4 1.26 3.4914 3.5045 13.1 ± 0.3 8.8 ± 0.1 6.8 ± 0.2 5.1 1.33 3.4923 3.5052 12.9 ± 0.3 9 ± 0.1 6.0 ± 0.1 8.9 1.12 3.4897 3.5014 11.7 ± 0.3 7.8 ± 0.1 4.0 ± 0.1

A comparison of the positions of XL and XU, their energetic differences and FWHM for

the three samples given in Table 4 reveals their dependence on free carrier

concentration and stress. The positions of XL and XU are mainly influenced by the stress

within the Ge doped layer, but their energetic difference shrinks with increasing free

carrier concentrations. That means both peaks are influenced by stress and free carrier

concentration or alternatively just one peak is influenced by stress while the position of

the second is determined by the free carrier concentration. The FWHM of both peaks

show a similar behavior. Whereas the FWHM of XL seems mainly be related to stress XU

is mostly related to the free carrier concentration, where the FWHM of XL increases with

increasing stress and the FWHM of XU decreases with increasing free carrier

concentration. The strong narrowing of the linewidth of XU from 6.8 meV to 4.0 meV

indicates a stabilization mechanism of XU by the electron gas. This means the effect of

the varying strain level in all samples is partially reflected by the linewidths trend related

to XL that is influenced by the strain level in the GaN:Ge layers and the free electron

concentration.

Within the present work, minor strain variation the energetic FWHM of XL rises with

increasing biaxial stress. Wagner et al. found a similar behavior for bound and free

excitons in ZnO. [107,108] Therefore, the linewidths of XL exhibits a maximum for the

sample with a free electron concentration of 5.1 x 1019 cm‐3 as a maximal bi‐axial

compressive stress of 1.33 GPa occurs. Any further rise of the free electron


57

concentration leads to a decrease in the linewidths as the beneficial stabilization effect

by exchange overcompensates the detrimental strain effects.

5.3.2 Crosssectionanddepthresolvedcathodoluminescence

3.2 3.3 3.4 3.5 3.60.0

0.5

1.0

1.5

Inte

nsity

(arb

. uni

ts)

Energy (eV)

PL low excitation power PL high excitation power CL cross-section

(close to GaN:Ge surface) CL cross-section

(GaN buffer layer)T = 8 K

XL

XU

x10

DBX

Figure 29: Comparison of cross section CL of the buffer layer and the Ge doped layer with in‐plane PL of the Ge doped layer at 8 K.

To prove that the luminescence of XU and XL originates from the Ge doped layers

cathodoluminescence (CL) spectroscopy has been applied. Figure 29 shows a

comparison of cross section CL measurements of the Ge doped layer and the undoped

buffer layer and in‐plane PL measurements with high and low excitation power densities

of the highest doped and thicker sample. The undoped buffer layer exhibits a strong

emission originating from donor bound excitons (DBX), similar to the GaN:nid reference

sample in Figure 26. The positions of the DBX have an energetically lower spectral

emission compared to XU and XL. This proofs that to XU and XL do not originate from the

buffer layer besides the face that the penetration depth of PL measurements with an

excitation wavelength of 325 nm is below 80 nm. [109] Nevertheless the peaks XU and

XL are much less pronounced in comparison to the PL measurements with low excitation


58

power. Also the broad emission band at 3.42 eV disappeared. These differences can be

explained by the high excitation power PL measurements with a quadrupled, pulsed

Nd:YAG laser. The emission at 3.42 eV does not increase in intensity with exaction

power, that means it saturates and originates most likely from structural and/or multiple

point defects. The background luminescence of band‐to‐band transitions gains more

intensity over excitation power density as the XU and XL.

3.20 3.25 3.30 3.35 3.40 3.45 3.50 3.55 3.60

XL

30 kV20 kV10 kVIn

tens

ity (a

rb. u

nits

)

Energy (eV)

2 kV

XUCL

sign

al s

treng

th

0 500 1000 1500 2000

sapp

hire

GaN

Buf

fer

GaN

:Ge

Depth (nm)

surfa

ce

Figure 30: Depth‐resolved cathodoluminescence spectra of the Ge‐doped sample with free carrier concentration of 5.1 x 1019 cm‐3 at 77 K. The excitation voltage has been varied from 2 to 30 kV. The spectra are normalized and stacked. The inset shows the CL signal strength over depth depending on the excitation voltage.

With depth resolved CL measurements it is possible to investigate the change of

luminescence in‐depth within a sample. The inset of Figure 30 shows the strength of the

CL signal in‐depth for the investigated sample structure. For voltages below 10 kV only

CL from the Ge doped layer is measurable. CL originating from layers below the Ge

doped layers can only be observed with higher voltages as long as it is not absorbed by

the Ge doped layer. An increase from 2 to 30 kV does not show any change in the


59

position or intensity of the peaks XU and XL and no additional peaks are visible. Only a

small change in the slope of the high energy flank of the luminescence is observed which

becomes steeper with increasing excitation voltage.

The unchanged peak intensity of XU and XL in comparison to the band‐to‐band transition

with depth proves that these peaks originate from the Ge doped layer. They do neither

originate from the buffer layer nor are they related to surface effects. If they would

originate from the buffer layer, they were not visible for excitation voltages below 10 kV.

Additionally, if they were related to surface effects their intensity would diminish with

increasing excitation voltage.

The change of the slope on the high energy flank of the luminescence can be explained

by the change of Ge concentration in‐depth as shown in Figure 24. With decreasing Ge

concentration in‐depth decreases as well as the free carrier concentration in‐depth and

the BMS. The excitation volume of the luminescence origin in‐depth increases with the

accelerating voltage. Consequently, the change of the BMS in‐depth leads to a smeared

out high energy flank of the luminescence at higher excitation voltages.

5.3.3 Absorptiondeepdefects

The investigated samples show no sign of any compensation mechanism up to a Ge

concentration of 7 x 1019 cm‐3 and only the onset of such an effect at 1.8 x 1020 cm‐3, as

illustrated in Figure 25. The Hall effect measurements were applied at room

temperature, but the increase of free carrier concentration over temperature is

negligible for degenerated semiconductors. [30] That means almost every Ge‐atom

contributes a free electron up to a Ge concentration of 7 x 1019 cm‐3, before the onset

of compensation and passivation occurs.

Further evidence for the low compensation level in the GaN:Ge samples is provided by

photoluminescence and transmission measurements over a broad energy interval as

shown in Figure 31. The sample with the second highest Ge concentration with a free


60

electron concentration of 5.1 x 1019 cm‐3 exhibits an almost identical transmission up to

the Fermi‐edge when compared to the GaN:nid reference sample with the same

thickness. The transmission modulation in Figure 31 is caused by Fabry‐Pérot oscillations

due to the particular layer stack within samples. No trace of any defect‐induced

absorption in the visible range is noticeable [110] for this free electron concentration

(5.1 x 1019 cm‐3) as the transmission features an almost flat characteristic over the broad

energy interval. A multitude of doping approaches is known to produce the defects

responsible for the so‐called, yellow and blue defect luminescence in GaN, [63] which

are both not pronounced in this sample. The GaN:Ge sample with the highest free

electron concentration exhibits a 15 % reduced transmission due to the onset of

compensation and defect formation (e.g. induced by V‐pits).

2.0 2.5 3.0 3.5

1

10

100

Nor

m. P

L in

tens

ity (a

rb. u

nits

)

Tran

smis

sion

(%)

Energy (eV)

Transmission (GaN:nid)

Transmission (GaN:Ge - 5.1 x 1019 cm-3

Transmission (GaN:Ge - 8.9 x 1019 cm-3

Photoluminescence (5.1 x 1019 cm-3)

T = 300 K

Figure 31: Transmissions of the GaN:nid (red line) and two highly Ge doped sample (blue and green line) at 300 K, and a PL spectra of one Ge doped sample at 300 K.

Nevertheless, the entire transmission characteristic is still comparably flat over a wide

energy interval without any distinct traces of defect‐related bands that would lower the

transmission. The level of compensation due to the influence of defects is minor in all


61

samples except for the sample with the highest doping concentration, which marks the

onset of compensation.

5.3.4 Dynamicsandtimeresolvedanalysis

1E18 1E19 1E200

50

100

150

200

250

300

350

T = 1.8 K

10 GW/cm2

10 MW/cm2

Fermi-edge energy (100 W/cm2)

Dec

ay ti

me BM

S (p

s)

Free electron concentration (cm-3)

Fermi edge sh

ift

decay t

ime tre

nd

3.45

3.50

3.55

3.60

3.65

3.70

Ferm

i-edg

e en

ergy

(eV

)

Figure 32: Decay times of band‐to‐band transitions at excitation power densities of 10 GW/cm² and 10 MW/cm² and the shift of the Fermi‐edge over free carrier concentration.

Since the band‐to‐band transition probability is smaller than the exciton transition

probability, slower decay times are expected. [111] This change is visible in the time‐

resolved photoluminescence (TRPL) measurements below and above the Mott density.

With increasing free carrier concentration, excitons become screened and the

recombination of free electrons with free holes becomes more plausible. [11] This leads

to an increase of the decay time with free carrier concentration. The temporal evolution

of the high energy luminescence affected by the Burstein‐Moss‐shift (BMS) can be most

efficiently compared among all analyzed samples if the Fermi‐edge is chosen as the

spectral measurement position. Figure 32 depicts a comparison of the decay times close

to the Fermi edge for two excitation powers (10 MW/cm² and 10 GW/cm²) and the shift


62

of the Fermi edge over free carrier concentration. The position of the Fermi edge is

estimated from the shift of the high energy flank in PL measurements Figure 27. For the

low excitation regime (in this case 100 W/cm²) the Fermi energy can be estimated by

the saddle point of the high energy flank of each PL spectra. [79] Higher excitation power

densities lead to an additional blue shift (not shown) and an increased decay time

caused by band filling. The constantly increasing decay time and BMS shift over free

carrier concentration shows that there is no or a negligible small increase of impurities,

which would increase the non‐radiative recombination or compensation.

0 1 2 3 4 510

100

1000

10000

100000

Inte

nsity

(arb

. uni

ts)

Time (ns)

3.45 3.50 3.55 3.60

Inte

nsity

(arb

. u.)

Energy (eV)

0

50

100

150

200

250

300

Dec

ay ti

me

(ps)

Figure 33: Transients of three emission regimes (XL in red, XU in blue and band‐to‐band transmission background in green) of one GaN:Ge sample (5.1 x 1019 cm‐3) and the energy dependent life times over the whole emission spectra in the inset

A comparison of transients of different energetic regions of the spectral emission of the

GaN:Ge sample with the free carrier concentration of 5.1 x 1019 cm‐3 is shown in Figure

33. All three transients seem to be mono exponential and transient of XU (blue) and the

BMS background (green) are almost identical. Only the transient of XL (red) has a slightly

longer decay time. The inset shows the corresponding PL spectra for taken TRPL

measurements. The black squares mark the energy dependent decay times for a mono


63

exponential decay of the luminescence. In the green area the decay times are almost

stable and lie at 307 ps (±2 ps), for blue at 305 ps and for red at 319 ps. In the spectral

BGR regime the decay times decrease slowly towards lower energies. Towards the Fermi

edge the decay times decrease much more drastically and reach 216 ps near the Fermi

edge.

The absence of an obvious visible second decay time means that XL and XU must have

similar decay times to the background luminescence originating from the band‐to‐band

transitions. To extract the pure transient of XL and XU without the contribution of the

band‐to‐band transitions, the transient from the BMS (green area) is subtracted from

the transients measured at the positions of XL and XU (red and blue). The resulting

transients including the fit results are plotted in Figure 34.

0.0

0.5

1.0

T = 5 K, P = 0.3 kW/mm2

XU: Udecay = 267 ps

XL: Ldecay = 303 ps

Lrise = 104 ps

Inte

nsity

(arb

. uni

ts)

n = 3.4 x 1019 cm-3

0.0

0.5

1.0 n = 5.1 x 1019 cm-3 XU: Udecay = 281 ps

XL: Ldecay = 344 ps

Lrise = 68 ps

0.50 0.75 1.00 1.25 1.50 1.750.0

0.5

1.0 n = 8.9 x 1019 cm-3 XU: Udecay = 305 ps (2nd decay: 31 ps)

XL: Ldecay = 303 ps (2nd decay: 46 ps)

Lrise = <10 ps

Decay time (ns)Figure 34: Time‐resolved Photoluminescence analysis of XU and XL for three doping concentrations. The transients are plotted on a linear scale for a clear depiction of the rise times.

The extracted transients of XL and XU reveal a clear trend over free carrier concentration.

Not only the decay times of the BGR‐ and BMS‐luminescence increase with rising doping

concentration (see Figure 32), also the decay times of XU (τ ) and XL (τ ) follow


64

this trend. While the high energy peak (XU) features a fast rise time (τ < 10 ps) in all

transients, its low energy counterpart (XL) exhibits a well‐resolvable rise time τ that

diminishes with rising doping concentration. The solid lines in Figure 34 represent the

results of biexponential (consideration of rise and decay times) fitting model

(n = 3.4 x 1019 cm‐3 and 5.1 x 1019 cm‐3), while only the highest doped sample required

the inclusion of a second decay time due to the onset of compensation and structural

defect formation. [100] Also the reduced decay time of the highest doped sample

(n = 8.9 x 1019 cm‐3) in comparison to the lower doped sample (n = 5.1 x 1019 cm‐3) is

most likely caused by the increased compensation (cf. chapter 5.2.2.2). The decay times

of the XL and XU excitations surpass the value of the background BGR‐ and BMS‐

luminescence, indicating a localization mechanism induced by the Fermi sea of

electrons.


65

5.3.5 Excitationchannels

3.40 3.45 3.48 3.49 3.50 3.51 3.52 3.53 3.540.60.70.80.91.0

XUA (n=2)

XUC

Energy (eV)

T = 1.8 K,P = 0.1 W/mm2,n = 8.9 x 1019 cm-3

Inte

nsity

(arb

. uni

ts)

XUB

-100 -50 -20 -10 0 10 20 30

1.0

2.0

3.0

4.0XU

A (n=2)

PL

3.4 x 1019 cm-3

5.1 x 1019 cm-3

Energy (meV)

XU

XL

8.9 x 1019 cm-3

x 2.5

x 10

PLE

Figure 35: Photoluminescence excitation (PLE) spectra (top) along with the corresponding photoluminescence spectrum (bottom). All PLE spectra are shown in energetic difference in regard to the energetic position of XU at the given free electron concentration in order to eliminate the influence of strain. The PLE spectra are color matched with their corresponding detection energy (dashed lines).

The lower part of Figure 35 shows a section of the PL spectrum of the highest doped

sample including XL, XU and the near band edge broad defect band luminescence. The

upper part shows PLE spectra of the highest doped sample with detection energies

indicated by the dashed lines in the PL spectrum (green = defect luminescence, red =

spectral position of XL, blue = spectral position of XU). Additionally, there are two PLE

spectra of the two lower doped samples (n = 3.4 x 1019 cm‐3 and 5.1 x 1019 cm‐3). All PLE

spectra are depicted in energetic difference in regard to energetic position of XU at the

given free electron concentration in order to eliminate the influence of strain. The

influence of varying strain levels is explained in the following Figure 36. The green and


66

the blue PLE spectra are increased by the factors 2.5 and 10 respectively. None of the

spectra are shifted vertically.

The PLE spectrum of XU (blue) features only one excitation channel ( ) comprising

a hole from the A‐valence band and an electron in the first excited state of the exciton‐

like complex, a situation similar to the excited A‐exciton in high‐quality, undoped GaN

samples. [26,28,112] In contrast, the PLE spectra related to XL (red ‐ color gradient) are

much less perturbed due to the enhanced localization. That means, in the high energy

regime, two additional excitation channels appear, comprising holes from the B‐ ( )

and C‐ ( ) valence band along with an electron in the ground state.

In addition, close to resonant excitation of XU boosts the intensity of XL, driving a

quasiparticle transformation process, whose dynamics is quantified by the rise time of

XL (τ extracted from the time‐resolved analysis (Figure 34). Interestingly, the

intensity of the excitation channel related to XU rises with doping concentration, while

the entire PLE spectrum of XL gets more distinct. This observation, in addition to the

particular scaling behavior of the rise time with doping concentration, supports the

identification of XU and XL as the exciton‐like collexon, and the four‐particle complex,

the bicollexon, both stabilized by the Fermi sea.

Shifting the PLE detection towards lower energies (e.g. 3.425 eV) yields a PLE spectrum

that exhibits resonances in close energetic vicinity to XU and XL. The shift between the

maxima of the PLE spectra related to XU and the corresponding PL features rises towards

higher energies with increasing free electron concentration. In addition, a similar,

electron‐density‐dependent shift can be observed for all excitation channels related to

X , X and X . The PLE maximum related to XL is also shifted towards higher

energies when compared to its PL counterpart as evidenced in the PLE spectrum

associated to the defect band situated at 3.425 eV. Similar shifts between PLE and PL

maxima have been attributed to self‐absorption processes often related to donor‐

acceptor‐pair luminescence in, e.g., GaN [100] and CdS [113]. In this case, any excitation


67

energy in close resonance to XU promotes the generation of XL related complexes scaling

with free electron concentration due to an intensification of the closely related

scattering process quantified by means of τ . The pronounced generation of XU

complexes at the GaN:Ge sample’s surface also promotes the absorption of light that

arises from the decay of XL by means of XU dissociation, a process similar to infrared

absorption of free excitons in Si and Ge that shows an absorption extension towards the

free excitonic continuum. [114,115] The most efficient generation of XL via the total

resonant excitation of XU complexes is not necessarily accompanied by a maximum in PL

intensity as the entire process is counterbalanced by a self‐absorption process. It is

exactly this subtle balance between close to resonance excitation and self‐absorption

processes that governs the particular structure of the PLE spectra shown in Figure 35. In

this picture it is also possible to understand the electron‐density‐dependency of the

maximum in the PLE spectra related to XL. As long as the scattering process that feeds

XL via XU is of limited efficiency due to a comparably low number of free electrons

(3.4 x 1019 cm‐3), the self‐absorption process is negligible and only a minor shift in

between the maximum in the PLE and PL spectrum occurs. However, as soon as the free

electron concentration rises, the resonant excitation of XL via XU becomes increasingly

efficient as the underlying scattering processes are enhanced (5.1 x 1019 cm‐3 and

8.9 x 1019 cm‐3). A closer look to the PLE spectrum of the defect‐related band (3.425 eV)

in Figure 35 even shows that the offset between the PLE and PL maxima is here more

pronounced for XL. Here, XL constitutes the supreme excitation channel of this defect‐

band as the strongest sign of a self‐absorption process is observed, despite the fact that

both excitation channels related XU and XL show similar intensity in the PLE spectrum. A

detailed analysis of the microscopic origin that balances the intensity ratio for this

defect‐band goes beyond the scope of the present thesis and must remain a task for

future work.

Nevertheless, the assignment of the luminescence peaks to the corresponding

excitation channels is apparent from Figure 35, showing that holes from the three

topmost valence bands contribute to XL, while the complexes related to XU possess an


68

efficient conversion pathway towards XL. Naturally, both complexes can dissociate and

therefore contribute to any luminescence at lower energies as their components

(electrons and holes) relax towards the corresponding defect bands.

5.3.6 Strainintroducedshift

1.3 1.4 1.5 1.6 1.73.49

3.50

3.51

3.52

3.53

3.54

3.55

XU XU

B XU

A (n=2)

XUC

C-VB-related

(n = 2)A-VB-related

B-VB-related

Ref. [X]

8.9

x 10

19 c

m-3

5.1

x 10

19 c

m-3

Ener

gy (e

V)

zz (x 1000)

3.4

x 10

19 c

m-3

GaN

:nid

(free

exc

itons

)

A-VB-related

T = 1.8 K,P = 0.1 W/mm2

Figure 36: A comparison of the theoretical biaxial strain related shift of GaN exciton emissions with exciton emissions and exciton‐like recombination channels measured by PL and PLE. The solid lines represent the shift of the free excitonic emissions related to the A‐, B‐, C‐valence band and first excited state of the A‐exciton over biaxial strain, calculated by exciton deformation potentials. The rectangles show the positions of the free exciton emission of the GaN:nid sample. The circles show the positions of the exciton like emissions and recombination channels of the highly Ge doped samples. The strain values are given by Raman spectroscopy measurements.

The strain variation in the germanium doped samples is discussed in detail in chapter

5.2.2.1 based on Raman spectroscopy. The E2(high) mode can be used to derive, the

corresponding strain tensor component ε based on a set of stiffness constants [26]

and phonon deformation potentials. [21] Figure 36 correlates the change of the c‐lattice

constant ε for the GaN:nid sample (triangles) with the free exciton transition energies


69

related to the A‐ (grey), B‐ (light red), and C‐ (light blue) valence band in addition to the

first excited state of the A‐exciton (light green, n = 2). For the GaN:nid sample a perfect

agreement between the free exciton transition energies derived from PL and PLE

measurements is observed.

Figure 36 summarizes the energy of all XU‐related excitation channels from PLE

measurements (circles) observed for the three highest doped GaN:Ge samples, in direct

relation to the strain tensor element ε . As introduced in Figure 35, these excitation

channels of XU stand in close relation to the three topmost valence bands of GaN and

are labelled accordingly, where X , X and X . X denotes the excited state

excitation channel (n = 2)‚ related to the A‐valence band. Based on Ref. [26], all ε

values can be translated into corresponding free‐exciton emission energies as shown in

Figure 36 (solid lines). The Ge doped samples with free carrier concentrations below

3.4 x 1019 cm‐3 do not show any similar luminescence features. As a result of this

comparison, it is apparent that the absolute energetic positions of all PLE excitation

channels are dominated by the varying strain level in the samples, an observation that

also directly applies for all emission energies derived from PL measurements.

Nevertheless, especially for the sample with the highest free electron concentration

(8.9 x 1019 cm‐3) one observes an additional shift of the excitation channels towards

higher energies in comparison to the expected values based on the strain level

determined by Raman spectroscopy (solid lines). This phenomenon can be related to the

offsets between PL and PLE resonances as shown in Figure 35.

In addition any offsets in Figure 36 are caused by the fundamentally different physical

origin of free excitons and collexons leading to deviating electronic deformation

potentials. In both cases an influence for the strain dependence of the valence bands is

directly demonstrated.


70

5.3.7 Excitationenergydependentphotoluminescenceandreflection

3.46 3.48 3.50 3.52 3.54 3.56 3.58 3.60 3.62 3.6410

100

1000

10000In

tens

ity (a

rb. u

nits

)

Energy (eV)

Figure 37: Photoluminescence excitation spectra and reflected light of the XBO lamp of the highest doped sample. The high intensity peak on the high energy side of each spectrum originates from the excitation source and marks the excitation energy. The change of the whole spectrum and the amount of the reflected light over excitation energy becomes visible.

The change of the whole PL spectra over excitation energy is shown in Figure 37. The

major difference to the PLE spectra shown in Figure 35 is the excitation source. The

monochromatic XBO lamp has an excitation power density of 0.1 W/cm² and is therefore

more than two orders of magnitude weaker than the applied dye Laser or the HeCd Laser

as shown in Figure 35 and Figure 27. Nevertheless has the PL spectrum for excitation

energies above the Fermi edge (violet spectrum) the same shape as for the HeCd Laser,

which means the difference in the excitation power is in this case negligible. Decreasing

excitation energies cause a number of changes to the whole PL spectrum. The high

energy flanks of the PL spectra shift towards lower energies with decreasing excitation

energy, because only transitions below the excitation energy are generated. Excitation

energies close to the Fermi edge (dark and light blue spectra) cause an increase of the

PL intensity. This effect is caused by an increased absorption and is known as the Fermi


71

edge singularity. [51] This singularity of the absorption arises from the growing Fermi

surface and the modification of the effective electron‐hole attraction due to the CB

filling. [11]

Excitation energies below the Fermi edge (yellow and orange spectra) lead to the

disappearance of the band‐to‐band transition luminescence, while the luminescence of

XL and XU remains visible. The intensity of XL stays roughly the same and the intensity of

XU is strongly decreased in comparison to over Fermi edge excitations. The reason for

the change of the intensity is not clear. However, it seems as if XU is more efficiently

generated via over Fermi edge excitation and XL is less dependent of the excitation

energy. The TRPL and PLE measurements as shown in Figure 34 and Figure 35 reveal a

transfer process from XU to XL, including that a decrease of the intensity of XU should

also lead to a decrease of XL. Accordingly this transfer process from XU to XL is strongly

enhanced for under Fermi edge excitation and supports the many particle interaction

theory as explained later in the discussion in 5.4.

Once the excitation energy approaches the energetic position of XU the intensity of XL

increases in the same manner as for the PLE measurements with the dye laser (Figure

35). Additionally, the intensity of the detected excitation source is increased as well, but

by one order of magnitude stronger that the luminescence of XL. The same behavior for

the intensity of the detected light from the excitation source is visible for the energetic

position of XU. Contributions of the luminescence to the increase of the detected light

from the light source are negligible small, because the luminescence is more than two

orders of magnitude smaller. Consequently the reflection from the sample’s surface

increases in vicinity of XU and XL. This effect is known from reflection measurements of

free excitons. [28]

The only difference to conventional reflection spectroscopy is that the angle of the

optical pathway of the excitation and detection to the surface normal is not the same.

This unusual configuration is unfavorable for high quality reflection spectra. However, it


72

facilitates for the omission of Fabry‐Pérot interferences, which are very strong for the

given sample structure, and would strongly interfere with the reflection spectrum.

3.44 3.46 3.48 3.50 3.52 3.54 3.56 3.58 3.60

0.6

0.8

1.0

1.2T = 3.5 K,P = 0.1 W/cm2,n = 8.9 x 1019 cm-3

Inte

nsity

(arb

. uni

ts)

Energy (eV)

Photoluminescence Reflection

XL XU

-300-250-200-150-100-50050100150200250300

R/

E (a

rb. u

nits

)

Figure 38: The change of the reflectivity over Energy of the highest doped sample in comparison to the above Fermi edge excitation PL spectra.

Figure 38 contains a PL spectrum of the highest doped sample with above Fermi edge

excitation (black line) and reflectivity data (green line). As the back reflected lamp

intensity is continuously monitored during the recording of the polychromatic PLE

spectra, a reflectivity spectrum can easily be extracted. A pronounced change in

reflectivity is observed in close energetic vicinity to the spectral position of XU and XL in

the PL spectrum. Such behavior is known for complexes like free excitons causing major

reflectivity changes in contrast to classical bound excitons. [28] This observation

supports the interpretation of XU and XL as an exciton like quasiparticles unaffected by

impurities, like Ge‐donors. Figure 38 indirectly proves that neither XL nor XU arise from

the buffer layer underneath the GaN:Ge layer. Any luminescence contribution of the

GaN buffer layer would always be dominated by bound‐excitonic luminescence that


73

stands in stark contrast to the strong change in reflectivity. At this point, low‐

temperature ellipsometry measurements are a task for future work.

5.3.8 Thermalization

0 5 10 15 20 25 300.01

0.1

1

Inte

nsity

(arb

. uni

ts)

Temperature (100/K)

BMS: 23.6 meV and 7.6 XU: 23.4 meV and 4.4 XL: 18.3 meV and 1.8

n = 8.9 x 1019 cm-3

P = 0.1 W/cm2

3.5 K

75 K

3.46 3.47 3.48 3.49 3.50 3.51 3.52

5

10

15

20

25

Energy (eV)

Inte

nsity

(arb

. uni

ts)

Figure 39: The Intensity of XL, XU and the band‐to‐band transition luminescence over temperature of the highest doped sample in an Arrhenius plot (symbols). For each fit (solid lines) are two activation energies used. The inset shows the development of XL and XU over temperature.

The exciton like character of XL and XU is confirmed by their rapid thermalization

behavior as shown in the inset of Figure 39. Both emission bands vanish simultaneously

with increasing temperature, while the Burstein‐Moss shifted background luminescence

of band‐to‐band transitions remains comparably stable. Figure 39 summarizes the

corresponding thermalization behavior of XL, XU, and BMS in an Arrhenius‐plot. The

trend of diminishing intensity with rising temperature is fitted in accordance to

Ref. [116] under consideration of two separate thermal activation energies. The

background of the band‐to‐band transitions is most robust against all thermalization


74

processes, while XU and XL thermalize more rapidly. This behavior can be understood as

soon as the particular physical origin of XL and XU as a bicollexon and a collexon is taken

into account. The liberation of one of the constituents (electron or hole) of the individual

many‐particle complex always leads to a complete dissociation of XL or XU. Hence, an

increasing number of particles that form the underlying excitonic complex (collexon

bicollexon) increase the susceptibility to thermalization. Hence, XL exhibits smaller

thermal activation energies than XU, while the BMS luminescence constitutes the most

simplistic and therefore the most resistant luminescence with increasing temperature.

More detailed studies of this particular temperature dependency must be undertaken

with focus on the doping and excitation power ‐ another motivation for future research.

5.4 Discussion

Several observations based on Figure 27 and Figure 28 contradict the standard model

for the behavior of highly doped semiconductors. First the high doping level in the

present case causes the appearance of distinct peaks (XL and XU), and second the entire

luminescence intensity in the band edge region rises with nGe due to an extraordinarily

high crystal quality. Additionally, the lifetime τBMS of the BMS‐shifted band‐to‐band

transitions close to the Fermi edge increases with nGe, a clear sign of negligible

compensation. The combination of SIMS, Hall‐effect, Raman, transmission, reflection,

and PL measurements, proves that the present set of highly germanium‐doped GaN

samples exhibits only minor compensation effects as only a well‐tolerable concentration

of extended structural defects and deep carrier traps is introduced. [30,89] The free

electron concentration, n, scales with the dopant concentration, nGe, in an almost linear

fashion, with the highest doped sample as only exception. The main reason for the

overall high crystalline quality is given by the similarity of the Ge4+ and Ga3+ core radii

and bond lengths in GaN. [81] The only significant doping‐induced difference is the

additional valence electron of Ge that leads to the Fermi sea formation a most ideal

textbook‐like doping situation.


75

The effect of the high doping concentration in combination with low compensation is

directly revealed by the particular luminescence characteristics of XL and XU. The spectral

splitting between both peaks, the binding energy of XL (E = E Ε ∗), while XU,

diminishes with rising free electron concentration. Simultaneously, the line‐broadening

(FWHM) of XU and XL peaks decrease indicating a stabilization mechanism. In addition,

the time‐resolved PL (TRPL) measurements from Figure 34 reveal a characteristic scaling

behavior of the associated decay times along with pronounced quasiparticle

transformation processes studied in more detail based on PL excitation (PLE)

spectroscopy (see Figure 35). All these observations underline the novel characteristics

of the elementary excitations that are not inhibited, but instead truly stabilized by the

degenerate electron gas.

All experimental observations correlate with the formation of many‐particle complexes.

The XU peak is identified as a complex interacting with the Fermi sea, achieving its

stabilization by exchange of electrons. The most important energy gain mechanism

providing stability in an electron gas is particle exchange. [44] This effect is illustrated in

the upper, left panel of Figure 40. The second low energy peak XL is correspondingly

interpreted as a more extended, many‐particle complex, capturing an electron‐hole pair

from the Fermi sea (see upper, right panel of Figure 40). In contrast to the interband

electron‐hole pair forming the complex XU, the collexon, the additionally captured

electron‐hole pair is of an intraband nature. The resulting four‐particle complex in

interaction with the excited Fermi sea, the bicollexon, is consequently stabilized by

electron exchange and Coulomb attraction. Its total excitation energy is lowered by the

effective binding energy of the additional intraband electron‐hole pair. Here, this

effective value is the full binding energy minus the excitation energy of an electron from

the Fermi sea above the Fermi level, leaving a hole behind.


76

Figure 40: Depiction of the many particle interactions of collexons and bicollexons in the 3D Fermi sea. The ground state (G) is formed by the degenerate electron gas, the Fermi sea. Upon optical excitation, inter‐band excitations are formed and stabilized by the Fermi sea of electrons giving rise to the first excited state, the collexon (XU) with a decay rate γU. In this high density regime, such collexons scatter with electrons in the Fermi sea (γUL), enabling the formation of the bicollexon (XL). The bicollexon represents a mixed complex of inter‐ and intra‐band electron‐hole pairs that decays (γL), forming the second excited state.

In contrast to trions, [95–97,99,117] the bicollexon is electrically neutral, spinless, and

of bosonic nature. PL measurements in presence of a magnetic field up to 5 T (not

shown) do not reveal any shift or splitting of the collexon lines. The formation of even

larger complexes with more than one intraband electron‐hole pair but only slightly

smaller excitation energies in comparison to the bicollexon cannot be excluded. In

Figure 39 the XL peak exhibits a prominent line‐shape asymmetry below 10 K that may

be traced back to additional, weakly bounded intraband electron‐hole pairs. The

theoretical picture developed for low or moderately doped semiconductor


77

nanostructures such as quantum wells, wires, or dots cannot be applied in the present

case. The measured free electron concentrations (n = 3.4 x 1019 cm‐3, 5.1 x 1019 cm‐3,

8.9 x 1019 cm‐3) correspond respectively to Fermi energies of ε k 2m∗⁄ 166,

217, and 315 meV in the conduction band with a Fermi wave vector of k

3π n ⁄ 1.00, 1.15, and 1.38 nm‐1, if an isotropic electron mass m∗ 0.231 m is

assumed. [50] Direct comparison with the Bohr radius of an A‐exciton in GaN

(a 2.16 nm) shows that the high‐density limit a k ≫ 1 beyond the Mott density is

reached, where free excitons and their charged, trionic counterparts are completely

dissociated, nullifying their relevance for optical data in the high‐density regime.

The extraordinary role of the degenerate electron gas for the formation of XU and XL is

illustrated in Figure 40. The ground state (G) of the system is given by the Fermi sea in

the Γ7c conduction band of GaN:Ge. Upon optical inter‐band excitation, the collexon is

formed as an electron‐hole pair well below the Burstein‐Moss edge, but strongly

coupled to the degenerate electron gas and scattered by the N electrons present in G

state. The collexon forms a first excited state stabilized by the Fermi sea of electrons

through unscreened electron exchange (illustrated by the arrows in Figure 40). It is

protected against interactions with polar optical phonons, since the Fröhlich

coupling [118,119] is also screened. The formation of the larger complex XL as a second

excited state with a lower excitation energy is consequently explained by a two‐step

procedure. First, a collexon is excited. Second, scattering of the collexon in the Fermi sea

excites an electron and leaves a hole behind, yielding a bicollexon complex, which emits

at lower energies with respect to the collexon. XL is a highly unconventional, mixed

complex constituting intra‐ as well as interband excitations that originate from the

optical excitation and the subsequent exciton‐electron scattering (γUL) and coupling

mechanism to the Fermi sea. The optical decay of a bicollexon leaves an intra‐band

excitation behind that will relax rapidly towards the G state, the unperturbed Fermi sea

in the conduction band.


78

The particular formation process of the bicollexon XL explains the occurrence of the rise

time (τ ) in the corresponding PL transients in Figure 34 and the particular excitation

channel in close energetic vicinity to the collexon XU (see Figure 35). The rising free

electron concentration promotes the excitation channel associated to XU as the

probability for electron exchange events is enhanced. The PLE spectra of XL (red ‐ color

gradient) from Figure 35 confirm their correlation. The increasing stabilization

mechanism via electron exchange is quantified byτ and τ and their particular

scaling behavior with rising doping concentration in qualitative agreement with the

spectral narrowing in Figure 28.

5.5 Theoreticalmodel

The theoretical description of the novel many‐particle complexes, collexon and

bicollexon, stabilized by the presence of a degenerate electron gas from first principles

requires the calculation of the spectral behavior of the two‐ and four‐particle Green

functions, including the correct occupation of the quasiparticle bands. For the two‐

particle case this is possible by solving a corresponding Bethe‐Salpeter equation

(BSE). [44] Typically, Hedin’s GW approximation [120] is applied, where for the

description of quasi‐electrons and quasi‐holes vertex corrections to the single‐particle

self‐energy are neglected, and only a statically screened Coulomb attraction between

electrons and holes is taken into account. [10,121] In this limit, the BSE can also be

solved in the presence of the degenerate electron gas. [11] This solution contains the

band‐gap renormalization and the Burstein‐Moss shift on the single‐particle

level [35,50,11,92] as well as the additional screening of the electron‐hole interaction,

which may lead to a dissociation of the exciton bound states and, hence, to the Mott

transition [49], Fermi edge singularity, and Mahan excitons. [51] However, even this

theory is unable to describe bound electron‐hole excitations coupled to the degenerate

electron gas with an excitation energy close to that of excitons in the absence of free

carriers, [11] mainly due to the omission of vertex corrections in the BSE kernel. [44] The


79

inclusion of the vertex function modified by the degenerate electron gas is a challenge

for future theoretical investigations of the novel complexes. For more particle

excitations such as in the case of biexcitons and trions in the low‐density limit, the

situation becomes worse as many‐particle Green functions are required. Typically, such

problems are approached constructing model Hamiltonians (see Refs. [122,123]). A

T‐matrix approximation containing particle‐particle and particle‐hole channels allows

the description of charged excitons. Such approximations are also applicable in the

biexciton case, [124] in particular to describe the additional binding between the two

electron‐hole pairs. Until now, there is no attempt to study the influence of a degenerate

electron gas on such a four‐particle excitation.

In order to describe collexon and bicollexon complexes additional approximations,

based mainly on physical, but less on mathematical grounds, have to be introduced.

For the investigation of near band edge optical phenomena of n‐doped wurtzite GaN,

only a two band model is considered. Near the ‐point only the 7c conduction band and

the uppermost valence band 9v are taken into account. In the undoped case this

restriction would be consistent with a study of the A‐exciton. The two bands are

assumed to be isotropic as well as parabolic with the effective masses m∗ and m∗ and

to be separated by a density‐dependent direct gap Eg(n). [92] Details regarding the band

structure as the differences between the effective masses parallel and perpendicular to

the c‐axis [125] and the band coupling giving rise to the non‐parabolicity [126] can be

found elsewhere. The optical transition 9v 7c is dipole‐allowed at the ‐point for

light polarization perpendicular to the c‐axis. [125] This polarization anisotropy will not

be discussed further in the following discourse.

In the low excitation limit, the absorption and emission properties are dominated by

bound states, e.g., of free 1s‐excitons [44,124] or trions [122,123,127] and can be

simulated by the imaginary part of a frequency‐dependent optical susceptibility as

a sum of such quantities in oscillator form. Following this approach it is possible to

describe the collexon and bicollexon by:


80

χ ω |M| Nϕ 0

E n ω iγ nN

ϕ ∗ 0

E n ∑ n∗ ω iγ n

E n and γ n denote the density‐dependent excitation energy and the line‐

broadening of the collexon (XU). The capture process of an additional electron‐hole pair

constituting the bicollexon (XL) is described by a self‐energy ∑ n∗ . [123,127] Its

negative real part E n Re∑ n∗ is the effective binding energy of the

additional intraband electron‐hole pair affected by the degenerate electron gas,

Ε ∗ n Ε n Re∑ n∗ , while the negative imaginary part Im∑ n∗

characterizes the modification of the bicollexon's line‐broadening γ ∗ n γ n

Im∑ n∗ . The density‐dependent envelope functions ϕ 0 and ϕ ∗ 0 of the many‐

particle complexes XU and XL appear in real space at the coordinate centre because the

optical transition matrix element M is approximated by its value at the Γ‐point. [44,124]

Their squares characterize the localization of the many‐particle complexes with rising

electron density. The index N labels the stabilizing degenerate gas with N electrons, the

corresponding index N∗ indicates that the intraband electron gas is excited with an

electron above the Fermi level and a hole in the Fermi sea, while the pre‐factors N and

N describe the number of corresponding complexes. Only the rate γUL, introduced in

Figure 40, is directly accessible via TRPL measurements yieldingτ .

To describe PL and PLE spectra it is assumed that the density‐dependent envelope

functions of the two complexes at vanishing particle distances are ϕ ∗ 0 and ϕ 0 .

This fact indicates that the sum of the oscillator strengths of the two complexes is not

conserved for varying electron density as observed experimentally for trions and

excitons. [128,129] In the collexon case it is described by the density‐dependent Fourier‐

transformed envelope function, here given by: [10,123]

A k ~ d xϕ x e


81

One may approximate this expression by the Fourier transform of a hydrogenic 1s‐wave

function

A k ~ 1 a n k

with a characteristic radius a n . However, its density dependence strongly deviates

from that of a screened Wannier‐Mott exciton, which underlies a Mott transition in

contrast to the collexon. This is a complex whose characteristic extent is reduced by the

exchange interaction with the electron gas. For the four‐particle bicollexon complex,

three different relative motions appear. The most important one is the relative motion

of the electron with respect to the corresponding hole in different bands. In order to

illustrate the small spatial extent of the complex, this matter is similarly specified by

assuming vanishing relative particle distances as indicated by ϕ ∗ 0 .

The localization of the complexes in the presence of the electron gas leads to excitation

energies E n and E ∗ n of the complexes, which are not coupled to the Burstein‐

Moss edge E n ε as in the high‐density limit of Mahan excitons. On the contrary,

collexon and bicollexon appear below the excitation energy of an exciton in the absence

of free carriers. This observation may be interpreted as a consequence of the

unscreened particle‐particle exchange with energies of the order of ε for the electron‐

hole pairs excited virtually in the degenerate electron gas, thereby almost compensating

the Burstein‐Moss shift.

Taking the collexon and bicollexon into account, their spectra can be approximated in

the high density regime by

Imχ ω |M| NUϕNU 0

2γ n

E n ω γ nNL

ϕ ∗L 0

2γ ∗ n

E ∗ n ω γ ∗ n

where collexon and bicollexon parameters, are functions of the free carrier density.


82

The relation of the two many‐particle complexes, collexon and bicollexon, can be

discussed in two ways. The collexon may be considered as a dissociated bicollexon,

where the intraband electron‐hole pair recombines. The more intuitive picture is a

bound interband electron‐hole pair stabilized by the electron gas, the collexon. Due to

the strong interaction with the Fermi sea, the collexon may capture a low‐energy

electron‐hole pair, whose binding in the complex overcomes its virtual excitation

energy. This capturing is described by a pair self‐energy ∑ nN∗ . The self‐energy ∑ nN∗

may be approximated in the framework of a T‐matrix approximation with an effective

attracting potential, similar to the treatment for quantum wells. [123,127]

Additionally, especially in GaN, it is important to note that the emission linewidth of

excitons is not trivially related to the inherent decay time. The particular spectral line

shape depends on many factors such as, excitation energy, strain, and impurity

density. [130] Specifically high doping densities lead to a pronounced broadening of

spectral emission lines in bulk GaN as well as related nanostructures. [131,132] Starting

at a free electron concentration of n = 3.4 x 1019 cm‐3, the emission linewidth of XU and

XL diminishes as all detrimental effects of the host lattice ‐ dictating the initial linewidths

of XU and XL ‐ are overcompensated by the beneficial stabilization action of the Fermi

sea of electrons. By increasing the free electron concentration, the weight for the origin

of the observable linewidth is shifted from the crystal lattice to the degenerate electron

gas. The rising doping concentration continuously weakens the quasi‐bosonic character

of all excitonic complexes until a re‐bosonization is achieved in the form of XU and XL

with rising free electron density and diminishing emission linewidths.

6 Summary

83

6 Summary

The main topic of this work is the detailed investigation of electrical, optical and

structural properties of highly Germanium doped GaN. The correlative analysis of two

series of Germanium doped samples reveals the strong influence of the growth

conditions on point defects and compensation mechanisms in semiconductors. A low

growth temperature of 960 °C facilitates a very efficient incorporation of Germanium

atoms into the GaN crystal. The maximum Germanium concentration of 1.2 x 1021 cm‐3

corresponds to more than 1% Germanium content in the GaN crystal. The maximum

free carrier concentration of 2.3 x 1020 cm‐3 is achieved at a Germanium concentration

of 2.5 x 1020 cm‐3, which also marks the onset of measureable compensation effects. The

strength of these compensation effects increases with increasing Germanium

concentration. The higher the Germanium content the higher is the FWHM of the

E2(high) mode of the Germanium doped layer, which corresponds to a decrease of the

crystalline quality. The reduction of the intensity and decay time of the NBE

luminescence while the YL increases is attributed to an increase of VGa and/or Carbon

impurities. Other point defects are less likely to play a role in this compensation

mechanism due to their relatively high formation energy. PL measurements show an

energetically shallow Germanium related strong luminescence. This luminescence

evolves with increasing Germanium content from a Germanium related defect band into

the conduction band of the ternary semiconductor GeGaN.

Higher growth temperatures in the range of 1075 °C lead to better crystalline quality of

GaN layers, which is shown by the remarkably small FWHM of the E2(high) of 2.7 cm‐1.

However, two to three orders of magnitude higher flow rates of the precursor of the

dopant are necessary to achieve similar Germanium concentrations as for lower growth

temperatures. For comparison Silicon doped samples with similar doping concentrations

were grown with the same growth conditions. High Silicon concentrations in GaN cause

a change of the inclination angle of edge type dislocations, which results in tensile stress

and a strong decrease of the crystalline quality. Irregular strain fields in Silicon doped

6 Summary

84

GaN samples increase the tendency of inhomogeneous incorporation of the dopant and

significantly increase the concentration of compensating defects.

The Raman spectra of the Ge‐doped samples exhibit a homogenous strain distribution

over the entire epitaxial layer in strong contrast to the observations for any Silicon

doped samples. Germanium doping in GaN has a negligible small impact on the strain

and a remarkably small influence on the crystalline quality in comparison to Silicon

doping. A comparison of the Germanium concentration, measured by SIMS, with the

free carrier concentrations, measured by Hall effect and Raman spectroscopy, reveals

that only the highest Germanium doped, thicker sample shows compensation. The

excellent crystalline quality of the GaN crystal is also shown by the clear signature

luminescence of free excitons. With increasing Germanium content samples reveal a

text book like behavior of the NBE luminescence, which broadens and shifts towards

higher energies due to the BMS and BGR. From a free carrier concentration of

3.4 x 1019 cm‐3 two pronounced narrow peaks appear, which originate from an, prior to

this work, unknown quasi particle, the collexon. With increasing free carrier

concentration the intensity and decay time of the collexon luminescence increases,

while the FWHM decreases. The increasing density of the three dimensional electron

gas stabilizes the many particle complex of the collexon. Scattering of the collexon in the

Fermi sea excites an electron and leaves a hole behind, yielding a bicollexon complex,

which emits at lower energies with respect to the collexon. The bicollexon is a highly

unconventional, mixed complex constituting intra‐ as well as interband excitations that

originate from the optical excitation and the subsequent exciton‐electron scattering and

coupling mechanism to the Fermi sea. Significantly, the previously unreported two

fundamental excitations discovered in this work should occur in the optical signature of

any highly doped and at the same time lowly compensated semiconductor with an

extraordinary high crystalline quality. No additional spatial confinement of carriers, as

in quantum well and quantum dot structures, is needed. Dopants that achieve a close‐

to‐perfect substitution of the corresponding host atom should lead to near perfect,

6 Summary

85

doped crystals comprising a degenerated electron gas, which will stabilize with

collexons.

7 Publications

86

7 Publications

19. Charge state switching of Cu acceptors in ZnO nanorods

M. A. Rahman, M. T. Westerhausen, C. Nenstiel, S. Choi, A. Hoffmann, A. Gentle, M. R. Phillips, C. Ton‐That

Appl. Phys. Lett. 110, 121907 (2017)

18. Intrinsic electronic properties of high‐quality wurtzite InN

H. Eisele, J. Schuppang, M. Schnedler, M. Duchamp, C. Nenstiel, V. Portz, T. Kure, M. Bügler, A. Lenz, M. Dähne, A. Hoffmann, S. Gwo, S. Choi, J. S. Speck, R. E. Dunin‐Borkowski, Ph. Ebert

Phys. Rev. B 94, 245201 (2016)

17. Electronic excitations stabilised by a degenerate electron gas in semiconductors

C. Nenstiel, G. Callsen, F. Nippert, T. Kure, M. R. Wagner, S. Schlichting, N. Jankowski, M. P. Hoffmann, A. Dadgar, S. Fritze, A. Krost, A. Hoffmann, and F. Bechstedt

arXiv preprint arXiv: 1610.01673 (2016)

16. Temperature‐dependent recombination coefficients in InGaN light‐emitting diodes: hole localization, Auger processes, and the green gap

F. Nippert, S. Karpov, G. Callsen, B. Galler, T. Kure, C. Nenstiel, M. Wagner, M. Straßburg, H. J. Lugauer, and A. Hoffman

Appl. Phys. Lett. 109, 161103 (2016)

15. Evaluation of local free carrier concentrations in individual heavily‐doped GaN:Si micro‐rods by micro‐Raman spectroscopy

M. S. Mohajerani, S. Khachadorian, T. Schimpke, C. Nenstiel, J. Hartmann, J. Ledig, A. Avramescu, M. Strassburg, A. Hoffmann, and A. Waag

Appl. Phys. Lett. 108, 91112 (2016)

14. Determination of recombination coefficients in InGaN quantum‐well light‐emitting diodes by small‐signal time‐resolved photoluminescence

F. Nippert, S. Karpov, I. Pietzonka, B. Galler, A. Wilm, T. Kure, C. Nenstiel, G. Callsen, M. Straßburg, H. J. Lugauer, and A. Hoffmann

Jpn. J. Appl. Phys. 55, 05FJ01 (2016)

7 Publications

87

13. Polarization‐induced confinement of continuous hole‐states in highly pumped, industrial‐grade, green InGaN quantum wells

F. Nippert, A. Nirschl, T. Schulz, G. Callsen, I. Pietzonka, S. Westerkamp, T. Kure, C. Nenstiel, M. Strassburg, M. Albrecht, and A. Hoffmann

J. Appl. Phys. 119, 215707 (2016)

12. Nature of red luminescence in oxygen treated hydrothermally grown zinc oxide nanorods

S. Anantachaisilp, S. M. Smith, C. Ton‐That, S. Pornsuwan, A. R. Moon, C. Nenstiel, A. Hoffmann, and M. R. Phillips

J. Lumin. 168, 20–25 (2015)

11. Analysis of the exciton‐LO‐phonon coupling in single wurtzite GaN quantum dots

G. Callsen, G. M. O. Pahn, S. Kalinowski, C. Kindel, J. Settke, J. Brunnmeier, C. Nenstiel, T. Kure, F. Nippert, A. Schliwa, A. Hoffmann, T. Markurt, T. Schulz, M. Albrecht, S. Kako, M. Arita, and Y. Arakawa

Phys. Rev. B ‐ Condens. Matter Mater. Phys. 92, 235439 (2015)

10. Germanium ‐ The superior dopant in n‐type GaN

C. Nenstiel, M. Bügler, G. Callsen, F. Nippert, T. Kure, S. Fritze, A. Dadgar, H. Witte, J. Bläsing, A. Krost, and A. Hoffmann

Phys. Status Solidi ‐ Rapid Res. Lett. 9, 716–721 (2015)

9. Shallow carrier traps in hydrothermal ZnO crystals

C. Ton‐That, L. L. C. Lem, M. R. Phillips, F. Reisdorffer, J. Mevellec, T.‐P. Nguyen, C. Nenstiel, and A. Hoffmann

New J. Phys. 16, 83040 (2014)

8. Structural and optical investigation of non‐polar (1‐100) GaN grown by the ammonothermal method

D. Gogova, P. P. Petrov, M. Buegler, M. R. Wagner, C. Nenstiel, G. Callsen, M. Schmidbauer, R. Kucharski, M. Zajac, R. Dwilinski, M. R. Phillips, A. Hoffmann, and R. Fornari

J. Appl. Phys. 113, 103504 (2013)

7 Publications

88

7. Ge doped GaN with controllable high carrier concentration for plasmonic applications

R. Kirste, M. P. Hoffmann, E. Sachet, M. Bobea, Z. Bryan, I. Bryan, C. Nenstiel, A. Hoffmann, J. P. Maria, R. Collazo, and Z. Sitar

Appl. Phys. Lett. 103, 242107 (2013)

6. Compensation effects in GaN:Mg probed by Raman spectroscopy and photoluminescence measurements

R. Kirste, M. P. Hoffmann, J. Tweedie, Z. Bryan, G. Callsen, T. Kure, C. Nenstiel, M. R. Wagner, R. Collazo, A. Hoffmann, and Z. Sitar

J. Appl. Phys. 113, 103504 (2013)

5. Effect of TMGa preflow on the properties of high temperature AlN layers grown on sapphire

R. Kirste, M. R. Wagner, C. Nenstiel, F. Brunner, M. Weyers, and A. Hoffmann

Phys. Status Solidi Appl. Mater. Sci. 210, 285–290 (2013)

4. Optical signature of Mg‐doped GaN: Transfer processes

G. Callsen, M. R. Wagner, T. Kure, J. S. Reparaz, M. Bügler, J. Brunnmeier, C. Nenstiel, A. Hoffmann, M. Hoffmann, J. Tweedie, Z. Bryan, S. Aygun, R. Kirste, R. Collazo, and Z. Sitar

Phys. Rev. B 86, 075207 (2012)

3. Phonon deformation potentials in wurtzite GaN and ZnO determined by uniaxial pressure dependent Raman measurements

G. Callsen, J. S. Reparaz, M. R. Wagner, R. Kirste, C. Nenstiel, A. Hoffmann, and M. R. Phillips

Appl. Phys. Lett. 98, 61906 (2011)

2. Bound excitons in ZnO: Structural defect complexes versus shallow impurity centers

M. R. Wagner, G. Callsen, J. S. Reparaz, J. H. Schulze, R. Kirste, M. Cobet, I. A. Ostapenko, S. Rodt, C. Nenstiel, M. Kaiser, A. Hoffmann, A. V. Rodina, M. R. Phillips, S. Lautenschläger, S. Eisermann, and B. K. Meyer


8

89

1. Identification of a donor‐related recombination channel in ZnO thin films

M. Brandt, H. Von Wenckstern, G. Benndorf, M. Lange, C. P. Dietrich, C. Kranert, C. Sturm, R. Schmidt‐Grund, H. Hochmuth, M. Lorenz, M. Grundmann, M. R. Wagner, M. Alic, C. Nenstiel, and A. Hoffmann


8 Bibliography

90

8 Bibliography

[1] Y. Ohno and M. Kuzuhara, IEEE Trans. Electron Devices 48, 517 (2001).

[2] S. Yoshida and H. Ishii, 246, 243 (2001).

[3] S. J. Pearton, B. S. Kang, S. Kim, F. Ren, B. P. Gila, C. R. Abernathy, J. Lin, and S. N. G. Chu, J. Phys. Condens. Matter 16, R961 (2004).

[4] S. Yoshida, S. Misawa, and S. Gonda, Appl. Phys. Lett. 42, 427 (1983).

[5] F. A. Ponce and D. P. Bour, Nature 386, 351 (1997).

[6] S. Nakamura, M. Senoh, S. ichi Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, and Y. Sugimoto, Japanese J. Appl. Physics, Part 2 Lett. 35, L74 (1996).

[7] S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, and Y. Sugimoto, Jpn. J. Appl. Phys. 35, L217 (1996).

[8] S. Stańczyk, T. Czyszanowski, A. Kafar, R. Czernecki, G. Targowski, M. Leszczyński, T. Suski, R. Kucharski, and P. Perlin, Appl. Phys. Lett. 102, 151102 (2013).

[9] G. Franssen, T. Suski, P. Perlin, R. Bohdan, A. Bercha, W. Trzeciakowski, I. Makarowa, P. Prystawko, M. Leszczyński, I. Grzegory, S. Porowski, and S. Kokenyesi, Appl. Phys. Lett. 87, 85 (2005).

[10] B. I. Shklovskii and A. L. Efros, Springer Verlag, Berlin (1984).

[11] A. Schleife, C. Rödl, F. Fuchs, K. Hannewald, and F. Bechstedt, Phys. Rev. Lett. 107, (2011).

[12] R. Loudon, The Raman Effect in Crystals (1964).

[13] R. H. Lyddane, R. G. Sachs, and E. Teller, Phys. Rev. 59, 673 (1941).

[14] T. Damen, S. Porto, and B. Tell, Phys. Rev. 142, 570 (1966).

[15] M. Cardona, in Light Scatt. Solids II (1982), pp. 19–178.

[16] A. Dadgar, a. Strittmatter, J. Bläsing, M. Poschenrieder, O. Contreras, P. Veit, T. Riemann, F. Bertram, a. Reiher, a. Krtschil, A. Diez, T. Hempel, T. Finger, a. Kasic, M. Schubert, D. Bimberg, F. a. Ponce, J. Christen, and A. Krost, Phys. Status Solidi 1606, 1583 (2003).

[17] L. Liu and J. H. Edgar, Mater. Sci. Eng. R Reports 37, 61 (2002).

[18] U. Haboeck, H. Siegle, A. Hoffmann, and C. Thomsen, Phys. Status Solidi C Conf. 0, 1710 (2003).

[19] A. Polian, M. Grimsditch, and I. Grzegory, J. Appl. Phys. 79, 3343 (1996).

[20] A. Goñi, H. Siegle, K. Syassen, C. Thomsen, and J. M. Wagner, Phys. Rev. B 64, (2001).

[21] G. Callsen, J. S. Reparaz, M. R. Wagner, R. Kirste, C. Nenstiel, A. Hoffmann, and M. R. Phillips, Appl. Phys. Lett. 98, 61906 (2011).

8 Bibliography

91

[22] G. Irmer, V. V Toporov, B. H. Bairamov, and J. Monecke, Phys. Status Solidi 119, 595 (1983).

[23] M. Klose, R. Dassow, M. Gross, and H. Schröder, J. Cryst. Growth 190, 666 (1998).

[24] H. Y. Peng, M. D. McCluskey, Y. M. Gupta, M. Kneissl, and N. M. Johnson, Phys. Rev. B - Condens. Matter Mater. Phys. 71, 1 (2005).

[25] J. E. Rowe, M. Cardona, and F. H. Pollak, Solid State Commun. 6, 239 (1968).

[26] A. Alemu, B. Gil, M. Julier, and S. Nakamura, Phys. Rev. B 57, 3761 (1998).

[27] B. Monemar, Phys. Rev. B 10, 676 (1974).

[28] D. Volm, K. Oettinger, T. Streibl, D. Kovalev, M. BenChorin, J. Diener, B. K. Meyer, J. Majewski, L. Eckey, A. Hoffmann, H. Amano, I. Akasaki, K. Hiramatsu, and T. Detchprohm, Phys. Rev. B 53, 16543 (1996).

[29] M. Leszczynski, T. Suski, H. Teisseyre, P. Perlin, I. Grzegory, J. Jun, S. Porowski, and T. D. Moustakas, J. Appl. Phys. 76, 4909 (1994).

[30] C. Nenstiel, M. Bügler, G. Callsen, F. Nippert, T. Kure, S. Fritze, A. Dadgar, H. Witte, J. Bläsing, A. Krost, and A. Hoffmann, Phys. Status Solidi - Rapid Res. Lett. 9, 716 (2015).

[31] M. A. M. Versteegh, T. Kuis, H. T. C. Stoof, and J. I. Dijkhuis, Phys. Rev. B - Condens. Matter Mater. Phys. 84, 1 (2011).

[32] F. Binet, J. Y. Duboz, J. Off, and F. Scholz, Phys. Rev. B 60, 4715 (1999).

[33] T. Nagai, T. J. Inagaki, and Y. Kanemitsu, Appl. Phys. Lett. 84, 1284 (2004).

[34] D. C. Reynolds, D. C. Look, and B. Jogai, J. Appl. Phys. 88, 5760 (2000).

[35] K. F. Berggren and B. E. Sernelius, Phys. Rev. B 24, 1971 (1981).

[36] B. E. Sernelius, K. F. Berggren, Z. C. Jin, I. Hamberg, and C. G. Granqvist, Phys. Rev. B 37, 10244 (1988).

[37] Z. Chine, A. Rebey, H. Touati, E. Goovaerts, M. Oueslati, B. El Jani, and S. Laugt, Phys. Status Solidi 203, 1954 (2006).

[38] H. P. D. Schenk, S. I. Borenstain, A. Berezin, A. Schön, E. Cheifetz, S. Khatsevich, and D. H. Rich, J. Appl. Phys. 103, (2008).

[39] M. Yoshikawa, M. Kunzer, J. Wagner, H. Obloh, P. Schlotter, R. Schmidt, N. Herres, and U. Kaufmann, J. Appl. Phys. 86, 4400 (1999).

[40] I. H. Lee, J. J. Lee, P. Kung, F. J. Sanchez, and M. Razeghi, Appl. Phys. Lett. 74, 102 (1999).

[41] R. A. Abram, G. J. Rees, and B. L. H. Wilson, Adv. Phys. 27, 799 (1978).

[42] E. Burstein, Phys. Rev. 93, 632 (1954).

[43] T. S. Moss, Proc. Phys. Soc. Sect. B 67, 775 (1954).

[44] F. Bechstedt, Many-Body Approach to Electronic Excitations (Springer, Berlin, 2015).

8 Bibliography

92

[45] G. Onida, L. Reining, and A. Rubio, Rev. Mod. Phys. 74, 601 (2002).

[46] R. J. Elliott, Phys. Rev. 108, 1384 (1957).

[47] P. Y. Yu and M. Cardona, Fundamentals of Semiconductors (2010).

[48] C. Klingshirn, Semiconductor Optics (2007).

[49] N. F. Mott, Rev. Mod. Phys. 40, 677 (1968).

[50] M. Feneberg, S. Osterburg, K. Lange, C. Lidig, B. Garke, R. Goldhahn, E. Richter, C. Netzel, M. D. Neumann, N. Esser, S. Fritze, H. Witte, J. Bläsing, A. Dadgar, and A. Krost, Phys. Rev. B 90, 75203 (2014).

[51] G. D. Mahan, Phys. Rev. 153, 882 (1967).

[52] E. Kane, Phys. Rev. 131, 79 (1963).

[53] R. Kirste, M. P. Hoffmann, E. Sachet, M. Bobea, Z. Bryan, I. Bryan, C. Nenstiel, A. Hoffmann, J. P. Maria, R. Collazo, and Z. Sitar, Appl. Phys. Lett. 103, 2011 (2013).

[54] C. G. Van De Walle and J. Neugebauer, J. Appl. Phys. 95, 3851 (2004).

[55] H. Obloh, K. H. Bachem, U. Kaufmann, M. Kunzer, M. Maier, A. Ramakrishnan, and P. Schlotter, J. Cryst. Growth 195, 270 (1998).

[56] Z. Liliental-Weber, J. Jasinski, M. Benamara, I. Grzegory, S. Porowski, D. J. H. Lampert, C. J. Eiting, and R. D. Dupuis, Phys. Status Solidi Basic Res. 228, 345 (2001).

[57] U. Kaufmann, P. Schlotter, H. Obloh, K. Köhler, and M. Maier, Phys. Rev. B - Condens. Matter Mater. Phys. 62, 10867 (2000).

[58] D. C. Look, K. D. Leedy, L. Vines, B. G. Svensson, A. Zubiaga, F. Tuomisto, D. R. Doutt, and L. J. Brillson, Phys. Rev. B - Condens. Matter Mater. Phys. 84, 1 (2011).

[59] G. Mandel, Phys. Rev. 134, (1964).

[60] G. F. Neumark, J. Appl. Phys. 51, 3383 (1980).

[61] H. Alves, M. Böhm, A. Hofstaetter, H. Amano, S. Einfeldt, D. Hommel, D. M. Hofmann, and B. K. Meyer, Phys. B Condens. Matter 308–310, 38 (2001).

[62] U. Kaufmann, M. Kunzer, H. Obloh, M. Maier, C. Manz, A. Ramakrishnan, and B. Santic, Phys. Rev. B 59, 5561 (1999).

[63] M. A. Reshchikov and H. Morkoç, J. Appl. Phys. 97, 61301 (2005).

[64] A. F. Wright, J. Appl. Phys. 92, 2575 (2002).

[65] S. W. Kaun, P. G. Burke, M. Hoi Wong, E. C. H. Kyle, U. K. Mishra, and J. S. Speck, Appl. Phys. Lett. 101, 19 (2012).

[66] A. Armstrong, A. R. Arehart, B. Moran, S. P. DenBaars, U. K. Mishra, J. S. Speck, and S. A. Ringel, Appl. Phys. Lett. 84, 374 (2004).

[67] C. Poblenz, P. Waltereit, S. Rajan, S. Heikman, U. K. Mishra, and J. S. Speck, J. Vac. Sci. Technol. B Microelectron. Nanom. Struct. 22, 1145 (2004).

8 Bibliography

93

[68] J. L. Lyons, A. Janotti, and C. G. Van De Walle, Appl. Phys. Lett. 97, 0 (2010).

[69] C. H. Seager, A. F. Wright, J. Yu, and W. Götz, J. Appl. Phys. 92, 6553 (2002).

[70] T. Ogino and M. Aoki, Jpn. J. Appl. Phys. 19, 2395 (1980).

[71] S. O. Kucheyev, M. Toth, M. R. Phillips, J. S. Williams, C. Jagadish, and G. Li, J. Appl. Phys. 91, 5867 (2002).

[72] J. Neugebauer and C. G. Van de Walle, Appl. Phys. Lett. 69, 503 (1996).

[73] K. Saarinen, T. Laine, S. Kuisma, J. Nissilä, P. Hautojärvi, L. Dobrzynski, J. Baranowski, K. Pakula, R. Stepniewski, M. Wojdak, a. Wysmolek, T. Suski, M. Leszczynski, I. Grzegory, and S. Porowski, Phys. Rev. Lett. 79, 3030 (1997).

[74] R. Armitage, W. Hong, Q. Yang, H. Feick, J. Gebauer, E. R. Weber, S. Hautakangas, and K. Saarinen, Appl. Phys. Lett. 82, 3457 (2003).

[75] A. Sedhain, J. Li, J. Y. Lin, and H. X. Jiang, Appl. Phys. Lett. 96, 1 (2010).

[76] A. Dadgar, P. Veit, F. Schulze, J. Bläsing, A. Krtschil, H. Witte, A. Diez, T. Hempel, J. Christen, R. Clos, and A. Krost, Thin Solid Films 515, 4356 (2007).

[77] L. T. Romano, C. G. Van de Walle, J. W. Ager, W. Götz, and R. S. Kern, J. Appl. Phys. 87, 7745 (2000).

[78] C. Van de Walle, Phys. Rev. B 68, 1 (2003).

[79] B. Arnaudov, T. Paskova, E. Goldys, S. Evtimova, and B. Monemar, Phys. Rev. B 64, (2001).

[80] T. Makimoto, H. Saito, T. Nishida, and N. Kobayashi, Appl. Phys. Lett. 70, 2984 (1997).

[81] P. Bogusławski and J. Bernholc, Phys. Rev. B 56, 9496 (1997).

[82] A. Hoffmann, H. Siegle, A. Kaschner, L. Eckey, C. Thomsen, J. Christen, F. Bertram, M. Schmidt, K. Hiramatsu, S. Kitamura, and N. Sawaki, J. Cryst. Growth 189–190, 630 (1998).

[83] A. Kaschner, H. Siegle, G. Kaczmarczyk, M. Straßburg, A. Hoffmann, C. Thomsen, U. Birkle, S. Einfeldt, and D. Hommel, Appl. Phys. Lett. 74, 3281 (1999).

[84] W. Limmer, W. Ritter, R. Sauer, B. Mensching, C. Liu, and B. Rauschenbach, Appl. Phys. Lett. 72, 2589 (1998).

[85] D. M. Follstaedt, S. R. Lee, A. A. Allerman, and J. A. Floro, J. Appl. Phys. 105, 83507 (2009).

[86] I. C. Manning, X. Weng, J. D. Acord, M. A. Fanton, D. W. Snyder, and J. M. Redwing, J. Appl. Phys. 106, 23506 (2009).

[87] A. Dadgar, R. Clos, G. Strassburger, F. Schulze, P. Veit, and T. Hempel, Adv. Solids State Phys. B. Kramer 44, 313 (2004).

[88] H. Ishikawa, K. Yamamoto, T. Egawa, T. Soga, T. Jimbo, and M. Umeno, J. Cryst. Growth 189–190, 178 (1998).

8 Bibliography

94

[89] S. Fritze, A. Dadgar, H. Witte, M. Bügler, A. Rohrbeck, J. Bläsing, A. Hoffmann, and A. Krost, Appl. Phys. Lett. 100, 122104 (2012).

[90] G. a. Thomas and T. M. Rice, Solid State Commun. 23, 359 (1977).

[91] K. Kheng, R. Cox, M. d’ Aubigné, F. Bassani, K. Saminadayar, and S. Tatarenko, Phys. Rev. Lett. 71, 1752 (1993).

[92] G. Finkelstein, H. Shtrikman, and I. Bar-Joseph, Phys. Rev. Lett. 74, 976 (1995).

[93] A. J. Shields, M. Pepper, D. A. Ritchie, M. Y. Simmons, and G. A. C. Jones, Phys. Rev. B 51, 18049 (1995).

[94] T. Warming, E. Siebert, A. Schliwa, E. Stock, R. Zimmermann, and D. Bimberg, Phys. Rev. B - Condens. Matter Mater. Phys. 79, 1 (2009).

[95] X. Xu, Sci. Rep. 3, 3228 (2013).

[96] P. Bigenwald, A. Kavokin, and B. Gil, Phys. Status Solidi Appl. Res. 195, 587 (2003).

[97] A. Esser, E. Runge, R. Zimmermann, and W. Langbein, Phys. Rev. B - Condens. Matter Mater. Phys. 62, 8232 (2000).

[98] C. H. Lui, a. J. Frenzel, D. V. Pilon, Y.-H. Lee, X. Ling, G. M. Akselrod, J. Kong, and N. Gedik, Phys. Rev. Lett. 113, 1 (2014).

[99] N. Scheuschner, O. Ochedowski, A. M. Kaulitz, R. Gillen, M. Schleberger, and J. Maultzsch, Phys. Rev. B - Condens. Matter Mater. Phys. 89, 2 (2014).

[100] G. Callsen, M. R. Wagner, T. Kure, J. S. Reparaz, M. Bügler, J. Brunnmeier, C. Nenstiel, A. Hoffmann, M. Hoffmann, J. Tweedie, Z. Bryan, S. Aygun, R. Kirste, R. Collazo, and Z. Sitar, Phys. Rev. B 86, 75207 (2012).

[101] B. Monemar, J. Phys. Condens. Matter 13, 7011 (2001).

[102] D. G. Chtchekine, Z. C. Feng, G. D. Gilliland, S. J. Chua, and D. Wolford, Phys. Rev. B 60, 15980 (1999).

[103] Q. Yang, H. Feick, and E. R. Weber, Appl. Phys. Lett. 82, 3002 (2003).

[104] B. Monemar, P. P. Paskov, J. P. Bergman, A. A. Toropov, T. V. Shubina, S. Figge, T. Paskova, D. Hommel, A. Usui, M. Iwaya, S. Kamiyama, H. Amano, and I. Akasaki, Mater. Sci. Semicond. Process. 9, 168 (2006).

[105] A. Usui, H. Sunakawa, A. Sakai, and A. Y. Atsushi, Jpn. J. Appl. Phys. 36, L899 (1997).

[106] J. Desheng, Y. Makita, K. Ploog, and H. J. Queisser, J. Appl. Phys. 999, 53 (2001).

[107] M. R. Wagner, G. Callsen, J. S. Reparaz, J. H. Schulze, R. Kirste, M. Cobet, I. A. Ostapenko, S. Rodt, C. Nenstiel, M. Kaiser, A. Hoffmann, A. V. Rodina, M. R. Phillips, S. Lautenschläger, S. Eisermann, and B. K. Meyer, Phys. Rev. B - Condens. Matter Mater. Phys. 84, (2011).

[108] M. R. Wagner, G. Callsen, J. S. Reparaz, R. Kirste, A. Hoffmann, A. V Rodina, A. Schleife, F. Bechstedt, and M. R. Phillips, Phys. Rev. B 88, 235210 (2013).

[109] J. F. Muth, J. H. Lee, I. K. Shmagin, R. M. Kolbas, H. C. Casey, B. P. Keller, U.

8 Bibliography

95

K. Mishra, and S. P. DenBaars, Appl. Phys. Lett. 71, 2572 (1997).

[110] Ü. Özgür, Y. I. Alivov, C. Liu, A. Teke, M. A. Reshchikov, S. Doǧan, V. Avrutin, S. J. Cho, and H. Morko̧, J. Appl. Phys. 98, 1 (2005).

[111] I. B. Ch. Fricke, R. Heitz, A. Hoffmann, Phys. Rev. B 49, 5313 (1994).

[112] a. Rodina, M. Dietrich, a. Göldner, L. Eckey, a. Hoffmann, A. Efros, M. Rosen, and B. Meyer, Phys. Rev. B 64, 11 (2001).

[113] I. Broser, J. Gutowski, and R. Riedel, Solid State Commun. 49, 445 (1984).

[114] T. Timusk, Phys. Rev. B 13, 3511 (1976).

[115] T. Timusk, H. Navarro, N. O. Lipari, and M. Altarelli, Solid State Commun. 25, 217 (1978).

[116] D. Bimberg, M. Sondergeld, and E. Grobe, Phys. Rev. B 4, 3451 (1971).

[117] C. H. Lui, a. J. Frenzel, D. V. Pilon, Y. H. Lee, X. Ling, G. M. Akselrod, J. Kong, and N. Gedik, Phys. Rev. Lett. 113, 1 (2014).

[118] H. Fröhlich, H. Pelzer, and S. Zienau, Philos. Mag. 41, 221 (1950).

[119] H. Fröhlich, Adv. Phys. 3, 325 (1954).

[120] L. Hedin, 139, A796 (1965).

[121] M. Rohlfing and S. Louie, Phys. Rev. B 62, 4927 (2000).

[122] a. Esser, R. Zimmermann, and E. Runge, Phys. Status Solidi Basic Res. 227, 317 (2001).

[123] R. A. Suris, in Opt. Prop. 2D Syst. with Interact. Electrons SE - 9, edited by W. Ossau and R. Suris (Springer Netherlands, 2003), pp. 111–124.

[124] G. Haug, Hartmut (Goethe-Universität Frankfurt, Germany), Koch, Stephan W (Philipps-Universität Marburg, Quantum Theory of the Optical and Electronic Properties of Semiconductors (World Scientific, New Jersey, 2009).

[125] L. C. de Carvalho, A. Schleife, and F. Bechstedt, Phys. Rev. B 84, 1 (2011).

[126] J. Furthm, P. H. Hahn, F. Fuchs, F. Bechstedt, N. The, F. Inn, and D.- Lda, Phys. Rev. B 72, 205106 (2005).

[127] R. a. Suris, V. P. Kochereshko, G. V. Astakhov, D. R. Yakovlev, W. Ossau, J. Nürnberger, W. Faschinger, G. Landwehr, T. Wojtowicz, G. Karczewski, and J. Kossut, Phys. Status Solidi Basic Res. 227, 343 (2001).

[128] T. Brunhes, R. André, a. Arnoult, J. Cibert, and a. Wasiela, Phys. Rev. B 60, 11568 (1999).

[129] R. Rapaport, a. Qarry, E. Cohen, a. Ron, and L. N. Pfeiffer, Phys. Status Solidi Basic Res. 227, 419 (2001).

[130] B. Monemar, P. P. Paskov, J. P. Bergman, A. A. Toropov, T. V Shubina, T. Malinauskas, and A. Usui, Phys. Status Solidi B-Basic Solid State Phys. 245, 1723 (2008).

[131] C. Kindel, G. Callsen, S. Kako, T. Kawano, H. Oishi, G. Hönig, A. Schliwa, A.

8 Bibliography

96

Hoffmann, and Y. Arakawa, Phys. Status Solidi - Rapid Res. Lett. 8, 408 (2014).

[132] G. Callsen and G. M. O. Pahn, Phys. Status Solidi - Rapid Res. Lett. 9, 521 (2015).

9 Acknowledgement

97

9 Acknowledgement

I would like to thank everybody who supported my work at the TU Berlin and the UTS

and helped me accomplishing my dissertation. Some people I owe special gratitude:

Prof. Dr. Axel Hoffmann for the opportunity to research in his group. I am very grateful for the scientific freedom, his constant support, personal commitment and care to the development of his students and their future. His endless enthusiasm and impressive knowledge in the field of solid state physics and joy in teaching were always a strong motivator for my research.

Prof. Dr. Matthew Phillips for the opportunity to work in his research group during my co‐tutelle studies at the UTS. I am very thankful for his help organizing my PhD scholarship, and I would like to express my gratitude for his careful proofreading and editing of my thesis. I would also like to thank the UTS Graduate Research School for waiving my PhD tuition fees.

Prof. Dr. Friedhelm Bechstedt for support in theoretical physics and fruitful discussions that lead to the model of the collexon.

Prof. Dr. Janina Maultzsch and Prof. Dr. Tadeusz Suski for examining my thesis

Prof. Dr. Ulrike Woggon for chairing my defense

Dr. Marc Hoffmann, Dr. Ronny Kirste, the group of Prof. Zlatko Sitar at the NCSU and the group of Prof. Alois Krost at the OVGU for suppling the GaN samples.

Dr. Gordon Callsen for the collaboration on the collexon paper and creating the measurement setup in the yellow lab.

Jan Schulze, Dr. Markus Wagner and Dr. Max Bügler for consulting and always having a sympathetic ear for questions.

Marcus Müller for ultra‐fast CL support and discussion.

Felix Nippert for Labview support.

Asmus Vierck, Thomas Kure, Ole Hitzemann, Dr. Dirk Heinrich, Stefan Kalinowski, Amelie Biermann, Nadja Jankowski and Sarah Schlichting for help with measurements.

Special thanks to my parents and my girlfriend Nadja

many body effects in heavily doped gan - tu berlin · many body effects in heavily doped gan...

Documents