Document [original]

Electronic Properties of ALD Zinc

Oxide Interfaces and its Implication for

Chalcopyrite Absorber Materials

vorgelegt von

Diplom-Ingenieur

Eike Janocha

aus Seesen

angefertigt am Helmholtz-Zentrum Berlin f¨

ur Materialien und Energie

eingereicht an der Fakult¨

at IV - Elektrotechnik und Informatik

der Technischen Universit¨

at Berlin

zur Erlangung des Akademischen Grades

Doktor der Ingenieurwissenschaften

Dr.-Ing.

genehmigte Dissertation

Promotionsausschuss:

Vorsitzender: Prof. Dr. Wolfgang Heinrich

Gutachter: Prof. Dr. Bernd Rech

Gutachter: Prof. Dr. Christian Pettenkofer

Gutachter: Prof. Dr. Wolfgang J¨

ager

Tag der wissenschaftlichen Aussprache: 05.03.2012

Berlin 2012

D 83

leer

Abstract

Electronic Properties of ALD Zinc Oxide and its Implication for

Chalcopyrite Absorber Materials

Eike Janocha

With its capacity as a transparent conductive oxide (TCO), zinc oxide (ZnO) is currently

used in a wide field of applications, for instance as the front contact in thin-film solar

cells. Nowadays, in thin-film chalcopyrite solar cells the heterojunction with the p-doped

chalcopyrite absorber is formed by a thin n-type cadmium sulfide buffer layer. Due to

environmental considerations and due to disadvantages in the deposition process, it is

desirable to replace the cadmium sulfide layer by an alternative Cd-free buffer material.

To understand the processes at the interface in detail, well-defined single-crystalline chal-

copyrite absorber films in the technologically relevant (112) orientation are used. Onto

these absorber layers, ZnO was then grown onto the layers by atomic layer deposition

(ALD). Due to its unique self-limiting growth mode, ALD allows a controlled deposition

of single ZnO monolayers. In this work a UHV-compatible ALD deposition chamber was

designed, assembled and commissioned. A constant deposition rate of 3.0˚

A/cycle was

determined in a temperature regime between 200 −225 ◦C, the so-called ALD window.

The electronic and chemical properties of the interfaces were investigated by means of

X-ray photoelectron spectroscopy (XPS), ultraviolet photoelectron spectroscopy (UPS)

and synchrotron-radiation photoelectron spectroscopy (SR-PES) at the BESSY II facil-

ity. Auger parameter analysis of the initial growth of ALD-ZnO on CuInSe2(112) and

comparison to reference samples indicates the formation of an intrinsic ZnIn2Se4(ZISe)

boundary layer having a thickness of only one monolayer. Combined XPS and UPS anal-

ysis allows the investigation of the electronic properties of the CuInSe2|ZnIn2Se4|ZnO

heterostructure, resulting in a detailed picture of the band alignment at the interfaces.

Furthermore, this thesis presents a new water-free ALD process for ZnO deposition using

metal-organic diethylzinc (DEZn) and molecular oxygen (O2) as reacting precursors. Zinc

oxide films usually show significant n-type doping and it is commonly stated that intrin-

sic point defects as well as interstitial hydrogen atoms play a major role in the doping

mechanism. To reduce the amount of hydrogen in the ZnO, an alternative oxygen pre-

cursor to H2O was introduced and successful atomic layer deposition was demonstrated.

The ALD window of the water-free ALD process is shifted slightly towards lower tem-

peratures, located between 185 −210 ◦C and a growth rate of 5.0˚

A/cycle was observed.

This significant increase indicates a completely different reaction mechanism not limited

by steric hindrance effects of the ethyl ligands as in case of standard ALD of ZnO. A de-

tailed comparison of two ZnO films deposited with both, water-free ALD and the standard

precursor combination is presented. Both show an additional oxygen photoemission peak

that is assigned to a surface hydroxide in the case of H2O-ALD and a O2−

2dumbbell defect

in the case of water-free ALD. Additional non-destructive photoemission depth-profiling

and annealing experiments do not show further significant differences in the chemical and

electronic properties of both ZnO films.

leer

Zusammenfassung

Electronic Properties of ALD Zinc Oxide and its Implication for

Chalcopyrite Absorber Materials

Eike Janocha

Zinkoxid (ZnO) in seiner Eigenschaft als transparenter elektrischer Leiter wird heute in

vielen Anwendungen eingesetzt, beispielsweise als Frontkontakt f¨

ur D¨

unnschicht-Solar-

zellen. Dort wird der Hetero¨

ubergang zwischen dem p-dotierten Chalkopyrit Absorber

heutzutage noch mit einer d¨

unnen Schicht aus n-CdS erzeugt. Da es w¨

unschenswert w¨

are

das Schwermetall Cadmium sowohl aus Umweltaspekten, als auch aufgrund unvorteilhaf-

ter Produktionsbedingungen zu vermeiden, sind in den letzten Jahren starke Forschungs-

anstrengungen in dem Feld sogenannter alternativer Pufferschichten zu vermerken.

Um die Prozesse an der Grenzfl¨

ache des Absorbers im Detail zu verstehen werden in

dieser Arbeit einkristalline Chalkopyrit Absorber in der technologisch relevanten (112)

Kristallorientierung hergestellt und verwendet. Auf diese Absorber wurde ZnO mittels

Atomic Layer Deposition (ALD) gewachsen. ALD hat den Vorteil, dass durch sein selbst-

limitierenden Wachstumsmodus einzelne ZnO Monolagen definiert abgeschieden werden

k¨

onnen. Daf¨

ur wurde eine UHV-kompatible ALD Anlage aufgebaut und in Betrieb genom-

men. In einem Temperaturbereich zwischen 200 −225 ◦C, dem sogenannten ALD Fenster,

wurde eine konstante Wachstumsrate von 3.0˚

A/Zyklus bestimmt.

Untersucht wurden die chemischen und elektrischen Eigenschaften der Grenzfl¨

ache mittels

R¨

ontgen-, Ultraviolett- und und Synchrotron-Photoelektronenspektroskopie am BESSY

II. Eine Analyse des Auger Parameters w¨

ahrend des initiellen Wachstums von ALD-ZnO

auf CuInSe2(112) und der Vergleich mit Referenzproben deutet auf die Bildung einer

intrinsischen ZnIn2Se4Grenzschicht hin, welche eine Dicke von nur einer Monolage auf-

weist. Die Kombination aus XPS und UPS erlaubt die Untersuchung der elektronischen

Eigenschaften der CuInSe2|ZnIn2Se4|ZnO Heterostruktur und liefert ein detailliertes Bild

der elektronischen Bandanpassung an den jeweiligen Grenzfl¨

achen.

Zudem wird in dieser Arbeit ein neuer wasserfreier ALD Prozess zur ZnO Abscheidung

mittels Diethylzink (DEZn) und molekularem Sauerstoff (O2) pr¨

asentiert. ZnO zeigt ¨

ub-

licherweise eine starke n-Dotierung, hervorgerufen durch intrinsische Punktdefekte, sowie

von Wasserstoff-Zwischengitteratomen. Um den Wasserstoff-Anteil im ZnO zu reduzieren

wurde ein alternativer Sauerstoff-Lieferant zum H2O benutzt und der erfolgreiche ALD

Prozess gezeigt. Das ALD Fenster der wasserfreien ALD ist leicht zu niederigeren Tempe-

raturen verschoben (185 −210 ◦C) und zeigt eine Wachstumsrate von 5.0˚

A/Zyklus. Der

Anstieg deutet auf einen grundlegend unterschiedlichen Reaktionsprozess hin, der nicht

durch sterische Hinderung der Ethyl-Liganden limitiert wird. Im Vergleich beider Metho-

den zeigt sich, dass beide ZnO Filme jeweils eine zus¨

atzliche Sauerstoff-Komponenten auf-

weisen. Im H2O-ALD Prozess wird diese durch eine Hydroxid-Oberfl¨

achenkomponente er-

zeugt, wohingegen der Ursprung beim wasserfreien ALD Prozess ein O2−

2dumbbell Defekt

ist. Zerst¨

orungsfreie Photoemissions-Tiefenprofilierung zeigen keine weiteren wesentlichen

Unterschiede der chemischen und elektronischen Eigenschaften beider ZnO Schichten.

III

Eidesstattliche Erkl¨

arung

Ich versichere an Eides statt, dass ich diese Dissertation selbst¨

andig verfasst habe. Die

benutzten Hilfsmittel und Quellen sind in der Arbeit vollst¨

andig angegeben. Die Disser-

tation ist bis auf die gekennzeichneten Teile noch nicht ver¨

offentlicht worden. Ich habe

weder fr¨

uher noch gleichzeitig ein Promotionsverfahren bei einem anderen Fachbereich

bzw. einer anderen Hochschule beantragt.

Berlin, den 28.11.2011

(Eike Janocha)

Contents

1 Motivation 1

1.1 Chalcopyrite Solar Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.1 Chalcopyrite Absorber . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.1.2 BufferLayer............................... 4

1.1.3 ZincOxide................................ 6

1.2 Objectives and Outline of this Thesis . . . . . . . . . . . . . . . . . . . . . 8

2 Methods of Surface Analysis 11

2.1 TheIntegratedSystem............................. 11

2.2 Photoelectron Spectroscopies . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.2.1 X-ray Photoelectron Spectroscopy . . . . . . . . . . . . . . . . . . . 14

2.2.2 Ultraviolet Photoelectron Spectroscopy . . . . . . . . . . . . . . . . 31

2.2.3 Synchrotron Radiation . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.3 Low-Energy Electron Diffraction . . . . . . . . . . . . . . . . . . . . . . . . 36

3 Thin-Film Deposition 39

3.1 General Issues of Thin-Film Deposition . . . . . . . . . . . . . . . . . . . . 39

3.2 Physical Vapor Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.2.1 Thermal Evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.2.2 Molecular Beam Epitaxy . . . . . . . . . . . . . . . . . . . . . . . . 47

3.3 Chemical Vapor Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.4 Metal-Organic Molecular Beam Epitaxy . . . . . . . . . . . . . . . . . . . 52

4 Atomic Layer Deposition of ZnO 55

4.1 Basic Principles of Atomic Layer Deposition . . . . . . . . . . . . . . . . . 56

4.1.1 SurfaceReactions............................ 57

4.1.2 Saturation................................ 60

4.1.3 Temperature Dependence . . . . . . . . . . . . . . . . . . . . . . . 62

VII

Contents

4.1.4 DepositionRate............................. 64

4.2 Reactor Design and Commissioning . . . . . . . . . . . . . . . . . . . . . . 64

4.2.1 UHV-ALDReactor........................... 65

4.2.2 Precursor Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

4.2.3 ALD Valve Response . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.3 Growth Parameters for ZnO-ALD . . . . . . . . . . . . . . . . . . . . . . . 73

4.3.1 Substrate Preparation . . . . . . . . . . . . . . . . . . . . . . . . . 73

4.3.2 Surface Saturation and ALD Window using DEZn and H2O . . . . 76

4.4 Chemical and Structural Characterization of ALD-ZnO films . . . . . . . . 80

4.5 Initial Growth of ALD ZnO on Si(111)-H . . . . . . . . . . . . . . . . . . . 83

4.6 Summary .................................... 92

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells 95

5.1 Epitaxial CuInSe2(112) Absorber Substrates . . . . . . . . . . . . . . . . . 96

5.2 ALD-ZnO on CuInSe2(112) Absorber Films . . . . . . . . . . . . . . . . . 99

5.2.1 Initial Growth Characterization . . . . . . . . . . . . . . . . . . . . 101

5.2.2 Interface Formation: Intrinsic Buffer Layer . . . . . . . . . . . . . . 109

5.3 Band Alignment of CuInSe2(112) and ALD-ZnO . . . . . . . . . . . . . . 114

5.4 Summary ....................................120

6 Water-free ZnO ALD 123

6.1 Growth Parameters for ZnO-ALD using O2precursor . . . . . . . . . . . . 124

6.2 Initial Growth of O2-ALD ZnO on Si(111)-H . . . . . . . . . . . . . . . . . 126

6.3 H2O-ALD vs. O2-ALDZnO ..........................131

6.3.1 Electronic Structure and Reaction Mechanism . . . . . . . . . . . . 133

6.3.2 Annealing Behavior of the Oxygen Component . . . . . . . . . . . . 139

6.3.3 Non-destructive Photoemission Depth-Profiling . . . . . . . . . . . 143

6.4 Summary ....................................154

7 Conclusion and Outlook 155

VIII

1 Motivation

During the last decades, the global energy consumption increased dramatically and there

is no indication that this trend reverses in the near future. Especially, the emerging

markets in newly industrialized countries like China or India, will double the demand

for energy in the next 50 years. In 2010, about 81 % of the total energy is produced by

burning fossil fuels such as coal or natural gas[1]. Since these resources are very limited,

renewable energies are the only alternative for future energy production. Today, most of

the power generated by renewable energies comes from biomass and wind power. Solar

energy is one of the most promising candidates for future power production. Neglecting

the difficulty of efficient storage, solar energy is able to cover the global demand for energy

by itself. The high growth rates of the photovoltaic market of 30 −40 % per year[2] show

that the investment in solar energy is of increasing interest, which is also supported by

governmental subsidies in some countries such as Germany.

Nowadays, about 80 % of the photovoltaic energy is produced by solar cells based on

crystalline or polycrystalline silicon. The indirect band gap of silicon and the resulting

absorber layer thickness of at least 100 µm is one of the major disadvantages of these first-

generation devices. To reduce the material consumption, thin-film solar cell technologies

were developed. Besides Si-based thin-film solar cells, this second solar cell generation

is based on direct band gap absorber materials like CdTe, GaAs or CIGS chalcopyrites,

reducing the required amount of absorber material drastically by a factor of 100. To mo-

tivate this thesis, basic knowledge about chalcopyrite-type thin-film solar cells is required.

Hence, the following section will give an introduction of its working principle, the different

materials used and how they are typically deposited in current production lines. After a

short discussion of the most important materials properties, the aim of this thesis will be

motivated and a short outline of this thesis will be presented in the end of this chapter.

1 Motivation

Figure 1.1: Current design of a thin-film chalcopyrite solar cell. Please note, that the film

thicknesses of the buffer layer and the intrinsic zinc oxide (i-ZnO) are not to

scale.

1.1 Chalcopyrite Solar Cells

Chalcopyrite absorber layers used for photovoltaics are based on I-III-VI2CuInSe2com-

pound materials. The current conversion efficiency world-record is 20.3 % achieved at

the Zentrum f¨

ur Sonnenenergie- und Wasserstoff-Forschung Baden-W¨

urttemberg (ZSW)

for laboratory-scale devices[3]. This is the highest efficiency not only for chalcopyrite-

based thin-film solar cells, but the record of polycrystalline thin-film solar cells at all.

The design of such a commercially available chalcopyrite solar cell is shown in Fig. 1.1.

The key component of the chalcopyrite solar cell is its absorber layer, that is usually

deposited by a co-evaporation process of the individual elements on a 1 µm thick molyb-

denum back contact. This electrically conducting back contact is usually deposited onto

a sodium-containing soda-lime glass substrate by sputter deposition. To separate the

charge carriers created in the absorber layer by incident light, a junction has to be formed

with the p-type chalcopyrite material. Today, this is done by a thin layer of n-type CdS

deposited by a chemical bath deposition (CBD) process. It is this heterojunction, mainly

determining the overall efficiency of the solar cell. On top of the buffer layer, a transparent

front contact is deposited by sputter deposition. Usually, a combination of a thin intrinsic

zinc oxide (i-ZnO) and heavily n-doped ZnO:Al is used and shows a beneficial effect on

1.1 Chalcopyrite Solar Cells

the device efficiency. Especially the influence of the i-ZnO is still under discussion[2].

In the following, the absorber layer, the buffer layer and the transparent front contact

materials are discussed in detail, since they play a crucial role for this thesis, while the

substrate, the Mo back contact and the role of Na for the device efficiency are not treated

in this work. For more detail it is referred to several reviews that give additional infor-

mation about the role of the substrate and the back contact[2;4–6].

1.1.1 Chalcopyrite Absorber

As mentioned above, chalcopyrite absorber layers presently used in thin-film solar cells

are usually based on CuInSe2. It crystallizes in a tetragonal chalcopyrite structure (space

group I42d) having lattice lattice parameters of a= 5.77 ˚

A and c= 11.54 ˚

A according to

Shay and Wernick[7]. The chalcopyrite lattice can be derived from the cubic zinc blende

structure of binary II-VI compound semiconductors like ZnSe by occupying the zinc lattice

sites alternately by copper and indium atoms, respectively[6]. Both crystal structures,the

one of ZnSe (cubic zinc blende) and CuInSe2(tetragonal chalcopyrite lattice), are illus-

trated in Fig. 1.2, respectively.

CuInSe2absorbers have very attractive properties to convert sunlight efficiently into

electrical energy. Its direct band gap of Eg= 1.04 eV[6] results in a high optical adsorp-

tion coefficient (≈105cm−1)[2] and hence, the thickness of the absorption layer can be

reduced significantly with respect to indirect band gap materials such as silicon. The band

gap energy of the semiconducting absorber material has a strong influence on the device

Figure 1.2: (left) zinc blende crystal structure of a II-VI compound such as ZnSe; (right)

chalcopyrite unit cell of a CuInSe2I-III-VI2compound

1 Motivation

efficiencies. There are basically two parameters that influence the solar cell’s efficiency:

on the one hand, the amount of photons absorbed in the active layer and on the other

hand the losses in the device that lower the electrical energy created per photon[4]. While

a small band gap would increase the amount of photons absorbed, at the same time, the

electrical losses would increase, where a larger band gap would be advantageous. There

is an optimum band gap that represents a compromise between the absorption efficiency

and electrical losses at Eg= 1.5 eV[4]. Hence, CuInSe2with its band gap of 1.04 eV is

not an ideal absorber material. Fortunately, the band gap of CuInSe2can be increased by

substituting parts of the indium by gallium. Depending on the Ga content, the band gap

can be varied continuously in a range between 1.04 −1.68 eV. In practice, Cu(In,Ga)Se2

absorbers are produced with a [Ga]/[In+Ga] ratio of 20 −30 %, resulting in a band gap

of approximately 1.20 −1.25 eV[4] to achieve optimum device characteristics.

Besides the band gap engineering by changing the gallium concentration, the doping of

the chalcopyrite plays a crucial role for the solar cell efficiency. Doping in chalcopyrites

is controlled by intrinsic point defects that are likely to form due to their low defect for-

mation energies during growth of the material[8]. High-efficiency photovoltaic cells have

p-type absorber layers, that are achieved by growing the material Cu-poor. There, copper

vacancies [VCu] are the dominant acceptors, resulting in p-type doping of the absorber.

The existence of an CuIn3Se5ordered defect compound at the surface of the chalcopy-

rite and its influence on the junction formation has been heavily discussed in the past.

Much more information about the ordered defect compound is presented in the thesis of

Hofmann[9] and is not discussed here in detail.

1.1.2 Buffer Layer

To form a junction with the p-type absorber layer, in today’s production lines n-type CdS

buffer layers are deposited onto the chalcopyrite absorber. Commonly, this is done by a

chemical bath deposition (CBD) process, where the absorber is put into a chemical solu-

tion of Cd ions and thiourea[6]. Besides a very homogeneous junction formation across the

whole surface area, the CBD CdS deposition process has some additional advantages, as

it protects the absorber surface from damage of the subsequent ZnO sputter deposition.

Additionally, the treatment in the chemical bath will also remove natural oxides that can

form at the chalcopyrite’s surface during the process.

Nevertheless, the usage of cadmium sulfide and the chemical bath deposition process itself

1.1 Chalcopyrite Solar Cells

have some disadvantages and both are subject of intensive research activities in the last

years. One drawback of the chemical bath deposition process is its incompatibility with

the vacuum deposition processes in the production line. Currently, the devices have to be

removed from the vacuum after absorber evaporation, to deposit the CdS in the chemical

bath. On the one hand, breaking the in-line production process is economically not very

advantageous. On the other hand, breaking the vacuum can lead to oxidation of the

surfaces and can influence the device characteristics. Hence, having a vacuum compatible

process for the buffer layer deposition would be beneficial for the solar cell production.

The second disadvantage in today’s chalcopyrite solar cell production is the usage of cad-

mium. CdS is known for its high toxicity and it is known for being responsible for causing

cancer. Hence, additional safety regulations are required for the employees[6]. The envi-

ronmental risks of cadmium also result in legal regulations for selling cadmium-containing

devices in some countries[10], for instance Germany as one of the biggest markets for pho-

tovoltaic devices today. Besides those legal issues, substituting the CdS by a different

material is also of interest from the technological point of view. To avoid the absorption

of incident light in the window and buffer layer, materials with wide band gaps are ben-

eficial for the device performance. With a band gap of Eg= 2.4 eV, light in the spectral

region between 350 −550 nm will be absorbed by the CdS[10]. Hence, an alternative buffer

material with a band gap energy exceeding the one of CdS could increase the solar cell

efficiency.

In recent years, several new materials and deposition methods have been investigated

in terms of their benefits in chalcopyrite thin-film solar cell production. Even though

the best efficiencies are still achieved with CBD-CdS buffers, promising results are re-

ported using alternative buffer layers that are Zn- or In-based materials. Among those,

ZnS with a band gap energy of 3.8 eV[11] has been studied extensively[10], just like In2S3

(Eg= 3.0 eV[12]) and ZnSe (Eg= 2.7 eV[11]). Various vacuum compatible deposition

techniques were applied, for instance sputter deposition, co-evaporation, ion layer gas re-

action (ILGAR) and atomic layer deposition (ALD). The latter often shows best results

among the alternative deposition techniques, most likely due to its excellent interface for-

mation with the subjacent chalcopyrite[10;13–15].

It has also been found, that the efficiencies of devices with sulfide-based buffer layers

can be increased, when the material additionally contains some oxides or hydroxides[5].

Studies of the interface formation of Zn(O,S) buffer layers deposited by ALD on poly-

crystalline Cu(In,Ga)Se2absorber layers have been carried out by Yousfi et al.[16] and

Platzer-Bj¨

orkman et al.[15], showing the feasibility of atomic layer deposition for the solar

1 Motivation

cell production process. Also direct ZnO growth on the chalcopyrite absorber layer has

been investigated by those groups using atomic layer deposition[14;17]. As expected, the

resulting efficiencies of those buffer-free solar cells did not reach the world-record cells of

20.3 %. But 14.6 % achieved with ALD-ZnO is close to the 15.0 % record for buffer-free

CIGS|ZnO structures grown by sputtering[18] or the ILGAR method[10] and emphasizes

the potential of atomic layer deposition to substitute the chemical bath deposition process

in the near future.

1.1.3 Zinc Oxide

The front contact of a chalcopyrite solar cell has to fulfill two requirements: on the one

hand, it has to be sufficiently conductive as it acts as the counter electrode to the Mo

back contact. On the other hand, it has to be as transparent to the incident sunlight as

possible to avoid undesired absorption in the window layer. There are not many materials

that fulfill both of those requirements. Often, tin-doped indium oxide (ITO, also known

as indium tin oxide) is used for this purpose. The limited resources are one of the indium

is one of the disadvantages of this material, increasing the costs of the devices. In today’s

chalcopyrite solar cell production, zinc oxide (ZnO) is used as transparent conductive

oxide for the front contact, showing a wide direct band gap of Eg= 3.36 −3.44 eV[19;20]

and hence being a promising candidate for the transparent front electrode in solar cells.

Zinc oxide is a binary II-VI compound semiconductor that usually crystallizes in a hexag-

onal wurtzite structure (space group P63mc), where each oxygen anion is surrounded by

four zinc cations[20]. The crystal structure of hexagonal wurtzite ZnO is illustrated in

Fig. 1.3 in side-view (top) and from the top (bottom). ZnO in (0001) orientation can

have two different surface terminations, zinc or oxygen, respectively. If zinc is the surface-

terminating element, one speaks of (0001) ZnO, while an oxygen-terminated surface is

defined as (0001) ZnO. According to ¨

Ozg¨

ur et al.[20], the lattice parameter of ZnO are

reported being a≈3.25 ˚

A and c≈5.20 ˚

A, respectively. Small deviations in the experi-

mentally determined lattice constants are caused by point defects present in the crystal.

Those point defects, like interstitial zinc atoms [Zni] or oxygen vacancies [VO] have very

low formation energies and are also discussed being responsible for the natural n-type

doping of ZnO crystals[20]. For use as front contact, this n-type conductivity is addition-

ally increased by heavily doping the ZnO by aluminum as a group III element.

1.1 Chalcopyrite Solar Cells

Figure 1.3: Illustration of the crystal structure of wurtzite zinc oxide in two different

viewing directions.

1 Motivation

While additional n-type doping is easily achieved, the same is not true for p-type doping of

zinc oxide. Over the last years, extensive research was done trying to find an effective and

reproducible method to achieve p-type conductivity in ZnO. Most promising dopant can-

didates discussed in the scientific community are nitrogen, phosphorus or arsenic atoms,

substituting the oxygen in the ZnO lattice. High quality p-type ZnO would result in

various new applications, for instance in the field of optoelectronic devices[20]. Difficulties

in p-type doping of ZnO and other wide band gap semiconductors (like GaN, ZnS and

ZnSe) are caused by several factors. Compensation of the p-type doping effects by the

native point defects mentioned above ([Zni], [VO]) is one of them.

Another crucial role might be the presence of hydrogen in the zinc oxide crystal. While

hydrogen is known for being amphoteric in most semiconducting materials (i.e. it can

act as both, donor or acceptor[19]), this is not the case in ZnO. Here, interstitial hydro-

gen always acts as a donor, resulting in additional compensation of potential acceptor

dopants[19;20].

1.2 Objectives and Outline of this Thesis

To understand the high efficiencies achieved by direct ZnO deposition on chalcopyrite

absorbers, earlier work of our group investigated the direct ZnO deposition by metal-

organic molecular beam epitaxy (MOMBE) on epitaxial CuInS2[21] and CuInSe2[22]. Since

polycrystalline chalcopyrite absorbers tend to grow in a preferred (112) orientation[23],

both absorber systems were grown in this technologically important (112) orientation.

The epitaxial nature of the absorber substrate allows detailed investigation of the interface

formation and the electronic band alignment of the two materials. Those studies indicated

no direct ZnO deposition on the chalcopyrite, but the formation of an intrinsic buffer layer

consisting of ZnS in case of CuInS2and ZnSe for CuInSe2absorbers. The electronic band

alignment determined from photoelectron spectroscopies indicated beneficial properties of

the CuInX2- ZnX - ZnO interface (X=S,Se) for photovoltaic devices. To investigate the

interface formation between epitaxial CuInSe2and ZnO grown by atomic layer deposition

is one of the objectives of this thesis. Besides an investigation of the chemical interface

formation, the electronic band alignment can be determined by in situ photoelectron

spectroscopies. This will result in a detailed picture of the processes at the interface,

responsible for the efficiencies achieved for the buffer-free chalcopyrite solar cells.

1.2 Objectives and Outline of this Thesis

As mentioned above, hydrogen plays a crucial role for the electronic properties zinc ox-

ide. Hydrogen incorporation into the zinc oxide crystal is very difficult to prevent, as it

is present in virtually every deposition technique available for ZnO growth. In atomic

layer deposition for instance, H2O is usually used as the oxygen precursor and therefore

provides a possible hydrogen source in the grown zinc oxide layer. To reduce the amount

of hydrogen in the ZnO deposition chamber, a water-free ALD process would be advan-

tageous, using an alternative to H2O as oxygen precursor. There are only few oxidizing

agents, which demonstrated successful atomic layer deposition of ZnO. While in the past

ZnCl2and O2were used as precursors for ZnO atomic layer deposition, nowadays, metal-

organic zinc precursors such as diethylzinc (Zn(C2H5)2) are preferred. Even though there

are few studies of alternative oxygen precursors for the ZnO atomic layer deposition using

diethylzinc, most of them either use reactive ozone (O3) or a plasma-enhanced deposition

technique. Due to their high reactivities, both processes can have negative influences on

the electronic and optical properties of the zinc oxide films deposited. In this study, the

first successful atomic layer deposition of diethylzinc and molecular oxygen (O2) is pre-

sented, that results in ZnO films of comparable chemical and electronic qualities to the

standard water-based ALD process.

This thesis is structured as follows:

After this motivation chapter, an introduction to the analytical techniques used in this

thesis is given. Afterwards, several deposition techniques relevant for this work are shortly

presented in Chp. 3. One objective required for the success of this thesis was the design,

assembly and initial operation of an UHV-compatible atomic layer deposition system for

ZnO growth. The results are presented in Chp. 4, followed by the second goal of this thesis:

the investigation of the CuInSe2|ZnO interface in Chp. 5. Finally, in Chp. 6 the results of

water-free ZnO atomic layer deposition using a precursor combination of diethylzinc and

O2are presented.

2 Methods of Surface Analysis

In the following chapter, some basic information about the methods of surface analysis

used in this thesis are presented. It is mainly focused on the sample characterization by

photoelectron spectroscopies, but also gives a short introduction about other techniques,

such as low-energy electron diffraction (LEED).

2.1 The Integrated System

All surface analysis techniques used in this thesis are sensitive to the utmost atomic layers

of the material under investigation. Therefore, any surface modifications in between the

sample preparation (for instance a deposition step) and its characterization have to be

avoided. This is achieved by carrying out all experiments under ultra-high vacuum (UHV)

conditions in a combined deposition and analysis system called the Integrated System

(IS). An illustration of this complex system is shown in Fig. 2.1. The base pressure in this

system is usually below 5 ·10−10 mbar, except for some deposition chambers, where source

gases cause higher pressures. Several types of vacuum pumps (rotary vane pumps, turbo

molecular pumps, ion getter pumps, sublimation pumps) are necessary to create such low

pressures. In comparison with high vacuum conditions, where only pressures of about

10−7mbar have to be reached, this is not only much more cost-intensive, but also very

time consuming. For instance, to remove water from the walls of the vacuum system after

it has been vented and exposed to air, it has to be baked-out for several hours up to days

at about 130 ◦C, until UHV conditions are achieved. Nevertheless, UHV conditions are of

utmost importance to maintain ultra-clean and unmodified surfaces after preparation of

the samples. Several deposition techniques in the Integrated System allow to investigate

2 Methods of Surface Analysis

Figure 2.1: Schematic drawing of the Integrated System, an ultra-high vacuum system

combining several deposition chambers (red) and surface analysis systems

(blue). Samples can be transferred in the UHV by transport systems (green),

allowing in situ analysis without any sample modification between deposition

and analysis.

the interface formation and properties of different material combinations. Those are in

particular: CuInSe2molecular beam epitaxy (MBE), CuInS2gas-source MBE (GS-MBE)

and ZnSe thermal evaporation. In addition, zinc oxide can be deposited via atomic layer

deposition (ALD) and metal-organic molecular beam epitaxy (MOMBE). Since most of

the techniques were used in this work, they are treated in detail later in a separate chapter.

In Fig. 2.1, all deposition chambers are colored in red.

The samples are mounted on small transferable sample holders, that allow in situ surface

characterization by transferring the specimen under UHV conditions from the deposition

chambers into several analysis systems. The disadvantage of those sample holders is the

limitation of the specimen sizes to 0.8×0.8 cm2at most. Samples are usually introduced

into the IS by one of two load locks.

Various surface analysis techniques are available, labeled in blue in Fig. 2.1. The main

components are two X-ray photoelectron spectroscopy (XPS) systems, one equipped with

a monochromized Al KαX-ray source, the other one with a Mg/Al dual X-ray source.

Both chambers are equipped with helium discharge lamps, allowing the valence band

investigation with ultraviolet photoelectron spectroscopy (UPS). Additionally to the pho-

toemission spectroscopies, the Integrated System allows to investigate the surface struc-

2.2 Photoelectron Spectroscopies

ture of the samples by low-energy electron diffraction (LEED), as well as creating images

of the surfaces by scanning tunneling microscopy (STM).

Furthermore, the Integrated System allows to anneal the samples either under UHV con-

ditions, in oxygen and hydrogen atmospheres or to sputter clean samples by bombardment

with high energetic argon ions. This is usually very useful to prepare clean reference ma-

terials.

One useful feature of the Integrated System not illustrated in Fig. 2.1 is a battery-powered

transfer box. It allows to transport a sample from the laboratory to an analysis system at

the BESSY II synchtrotron radiation facility under UHV conditions. More information

about the BESSY II and the application of synchtrotron light in this work is given later

in Sec. 2.2.3.

2.2 Photoelectron Spectroscopies

When Albert Einstein explained the photoelectric effect in 1905[24], he provided the basis

for one of todays most powerful analytical techniques to investigate electronic properties

of materials, namely the photoelectron spectroscopy (PES). Using Max Planck’s concept

that energy only exists in discrete levels, he explained the energy transfer of a photon with

energy E=h·νby exciting an electron which then might escape the atom. Both were

awarded the Nobel Prize in Physics, Planck in 1918 and Einstein in 1921. Depending on

the photon excitation energy one distinguishes between different types of photoelectron

spectroscopies:

- ultraviolet photoelectron spectroscopy (UPS) with photon energies .120 eV

- X-ray photoelectron spectroscopy (XPS) with excitation energies of ≈1 keV

- hard X-ray photoelectron spectroscopy (HAXPES) with energies &5 keV

In the this chapter, the basic principles of X-ray and ultraviolet photoelectron spectroscopy

will be shortly discussed. For those readers who want to gain deeper insight into this topic

there are lots of textbooks available dealing with photoelectron spectroscopy[25–28].

2 Methods of Surface Analysis

2.2.1 X-ray Photoelectron Spectroscopy

Using the concept of the photoelectric effect, Kai Siegbahn and his co-workers developed

the instrumentation to investigate chemical properties of materials surfaces. In addition,

he developed the theory of XPS and established the term electron spectroscopy for chemical

analysis (ESCA). For his contributions to the field of surface science, also Siegbahn was

awarded the Nobel Prize in Physics in 1981. Since photoelectron spectroscopies can

provide much more information about materials than its chemical properties, as shown

later in this section, nowadays the term X-ray photoelectron spectroscopy (XPS) became

commonly used.

In XPS, a sample is irradiated by X-ray photons of a specific known energy. The pho-

tons can transfer their entire energy to core-electrons in the material, which then are

able to escape the atom. Those photoelectrons are energetically separated by an electron

spectrometer and counted. It is the kinetic energy of those photoelectrons that contain

information about the chemical environment of the emitting atom. The individual pro-

cesses during photoemission spectroscopy, beginning with creating a characteristic X-ray

photon and resulting in photoemission spectra with several spectral features, are discussed

now in detail.

Figure 2.3 shows the configuration of an XPS system as it is used in this work. First,

X-rays have to be created at the X-ray source by accelerating a high-energy electron

beam in the range of 10 −15 keV onto an anode material to create holes in its core

levels. Electrons from outer shells of the atom will fill up this hole, emitting characteristic

radiation of specific energy at the same time. In practice, X-ray source anode materials

Figure 2.2: Simplified energy scheme of the photoelectron creation by energy transfer of

incident X-rays with an energy hν.

2.2 Photoelectron Spectroscopies

Figure 2.3: Schematic illustration of the monochromized X-ray photoelectron spec-

troscopy analysis system mainly used in this work. Explanation of the different

parts is given in the text.

2 Methods of Surface Analysis

are usually either magnesium or aluminum creating characteristic X-rays of 1253.6 eV

(Mg Kα) and 1486.6 eV (Al Kα), respectively[26], where 1 eV = 1.6·10−19 J. Since most

of the energy of the incident electrons is released in form of heat, anodes have to be water

cooled to prevent thermal degradation.

In Fig. 2.4 a spectrum of characteristic Al Kαradiation is shown. The Kαradiation is

quite complex and is composed of two different Kαlines where the Kα1line corresponds

to the 2p3/2→1s transition and the Kα2line arises from a 2p1/2→1s transition. Their

intensity difference originates from the different number of electrons in the 2p1/2and 2p3/2

energy states. In addition, a couple of satellite lines are present which are not shown in

Fig. 2.4.

To increase the energy resolution of the X-radiation, a monochromator is used to separate

the Al Kα1line from the Al Kα2radiation and all satellite lines. To do so, a quartz

crystal is used to diffract the X-rays from the source and to focus it onto a sample.

Aluminum anodes are the most commonly monochromized X-ray sources. This is due to

the fact that the Kα1wavelength of 0.83 nm and the quartz lattice spacing of the (1010)

planes (0.425 nm) fulfill the Bragg relationship nλ = 2dsin θat an angle of 78◦[26] so

that only the Al Kα1line is diffracted. As illustrated in Fig. 2.3, the X-ray source, the

monochromizing crystal and the sample have to be positioned on a spherical surface called

the Rowland sphere, to satisfy the Bragg condition. Using a monochromator significantly

Figure 2.4: Energy distribution of characteristic Al Kαradiation. The illustration is taken

from Alford et al.[28]

2.2 Photoelectron Spectroscopies

increases the quality of the photoemission spectrum due to removing unwanted signal

contributions from other spectral lines as the Al Kα1. Even though it lowers the X-ray

intensity reaching the sample and having higher costs, the benefits in spectral quality

predominate the drawbacks.

Once the X-rays reach an atom of the sample surface, three different interactions may

occur[27]:

1. the photon may pass through the atom without any interactions

2. the photon is scattered by an electron and loses some energy

3. the photon transfers its entire energy to an electron, which then is ejected as a

photoelectron

It is the third event X-ray photoelectron spectroscopy is based on. To gain insight on

the information achieved by photoelectron spectroscopies, we have to take a look at the

interaction of photons and the electrons in the material. One has to differentiate between

initial state and final state effects. In its initial state, before the incident photon interacts

with the atom, the negatively charged electrons are bound to the atom by the positively

charged nucleus. The closer an electron is located at the nucleus, the higher its binding

energy, i.e. the stronger it is bound to the nucleus. Hence, the atomic number (the number

of protons in the nucleus) influences the binding energies of the electrons. This binding

energy can relatively easy be determined using photoelectron spectroscopies. Hence, it is

providing information about the elements present in the sample.

The binding energy (BE) and the measured kinetic energy (KE) of the emitted photoelec-

trons have first been related to each other by Einstein in his work about the photoelectric

effect[24] according to Eq. 2.1:

Ebin =hν −Ekin (2.1)

The excitation energy hν of the incident X-rays is a known parameter determined by the

X-ray source used. In solid materials there is an additional amount of energy necessary to

remove the electron out of the crystal. This additional energy is called the work function

2 Methods of Surface Analysis

Figure 2.5: (left) energetic situation of an electron in the sample; (right) work function of

the spectrometer, if sample and spectrometer are in electrical contact

(ϕ), which has to be considered in the Einstein equation:

Ebin =hν −Ekin −ϕ(2.2)

The work function is defined as the energy required by an electron at the Fermi energy

to escape the crystal, i.e. to reach the vacuum energy Evac:ϕ=EF−Evac. A schematic

representation of the processes in a photoemission process is shown in Fig. 2.5 on the left

hand side. As shown on the right, one also has to take into account the work function of

the spectrometer.

Because sample and spectrometer are in electrical contact, both are also in a thermo-

dynamical equilibrium. Hence, their Fermi levels are equal and the energy an electron

has to overcome to be ejected from the crystal reduces by ∆ϕ=ϕ−ϕspec, where ϕspec

corresponds to the work function of the spectrometer. Including the work function of the

spectrometer in Eq. 2.2 results in:

2.2 Photoelectron Spectroscopies

Ebin =hν −Ekin −ϕ+ (ϕ−ϕspec) = hν −Ekin −ϕspec (2.3)

ϕspec can be determined during calibration of the spectrometer using clean Cu, Ag and

Au metal samples. Once calibrated, it can be assumed to stay constant as long as the

spectrometer is maintained under ultra-high vacuum conditions.

It is not only the atomic number that is responsible for the binding energy of the electrons,

but also its chemical environment. If an atom forms chemical bonds to other atoms, their

valence electrons, responsible for chemical bonding, will redistribute depending on the

electronegativities of the bonding partners. This changes the electrostatic potential felt

by the core-electrons, which in turn can result in a change in the binding energies of

the core-electrons. This change is called chemical shift and contains information on the

bonding situation of elements in a sample.

Figure 2.6: X-ray photoelectron survey spectrum of Ag. The major photoemission and

Auger transition peaks are labeled. The small spectral features around

371.5 eV and 377.5 eV originate from plasmon losses of the photoelectrons.

The spectrum has been recorded with monochromized Al Kαradiation with

hν = 1486.6 eV

2 Methods of Surface Analysis

A survey spectrum of a metallic Ag sample is presented in Fig. 2.6 to exemplify a typical

photoemission spectrum. The X-ray source was a monochromized Al Kαanode with

hν = 1486.6 eV. Survey spectra like this are useful to get an overview about the elements

present in a sample. They are recorded over a broad range of binding energies with a

lower energy step size of typically 0.5 eV to reduce recording times. The most prominent

feature in Fig. 2.6 is originating from Ag 3d electrons at binding energies of about 368 eV

and 374 eV, respectively. This is shown in detail in the inset recorded with an smaller

energy step size of only 0.05 eV. Another initial state effect is shown in this inset is the

spin-orbit splitting of two electron states of same energy but different spin. The magnetic

interaction between the electron’s spin (up or down) and its orbital angular momentum

leads to the splitting into two components. In case of d-orbitals as in the example of

Ag 3d they split into 3d3/2and 3d5/2with a difference in binding energy of ∆BE = 6 eV

in this particular case. The intensity ratio between those two peaks is d3/2: d5/2= 2 : 3.

Other less prominent photoemission lines in Fig. 2.6 belong to electrons of the 3p state

around 600 eV, that split into p1/2and p3/2with an intensity ratio of 1 : 2. In case

of Ag, the core-electrons of deeper energy levels than 3s are not visible using Al Kα

radiation since their binding energies (e.g. 3352.6 eV for Ag 2p3/2electrons[29]) exceed

the excitation energy of 1486 eV. The survey spectrum also provides information about

the contamination of the sample surface. In case of the Ag example, the surface was

sputter cleaned by Ar+ion bombardment and only small traces of carbon are present at

the surface indicated by the C 1s photoemission peak at 284 eV.

In X-ray photoelectron spectroscopy, there is another important feature that contributes

to the photoemission spectra. Those features arise from X-ray excited Auger electrons,

as illustrated in Fig. 2.7 in a simplified form.

On the left hand side of Fig. 2.7, the process of creating the core-hole by ejection of a

photoelectron is demonstrated again. On the right, this particular hole is refilled by an

3d outer shell electron to reduce the total energy of the excited atom. For reasons of

energy conservation, the surplus energy of this process has to be emitted either in form

of an X-ray photon or by ejecting a so-called Auger electron. Its kinetic energy depends

on the energy level of the emitted photoelectron, the initial state of the electron that fills

up the core-hole and on the state the Auger electron has been prior to being emitted

to the vacuum. Hence, the kinetic energy is independent of the excitation energy of the

incident photons. For historical reasons, for Auger electrons are labeled in X-ray notation,

whereas for photoelectrons the spectroscopic notation is used. In spectroscopic notation,

2.2 Photoelectron Spectroscopies

Figure 2.7: Schematic illustrations of the excitation of an photoelectron by absorption of

an X-ray (left) and the emission of an Auger electron created by a relaxation

process (right)

the electron states are a combination of the principal quantum number n= 1,2,3,4, the

angular momentum l= s, p, d, f and the total angular momentum quantum number

j=l+s, where sis the spin of the electron, i.e. s= 1/2 or (−1/2). In X-ray notation,

nis expressed by the capital letter K, L, M, N followed by suffixes 1, 2, 3, 4, ... which

represent the various combinations of land j. Table 2.1 summarizes the most important

electron states and their particular notation.

The combination of the three energy levels that determine its kinetic energy is used to

Spectroscopic X-ray notation

1 s K1

2 s L1

2 p1/2L2

2 p3/2L3

3 s M1

3 p1/2M2

3 p3/2M3

3 d3/2M4

3 d5/2M5

Table 2.1: Spectroscopic notation and X-ray notation for photo- and Auger electrons

2 Methods of Surface Analysis

label an Auger electron. In Fig. 2.7 the photoelectron was ejected from the 2p3/2(i.e. L3)

state and filled up by an electron of the 3d3/2or 3d5/2(M4or M5) shell. The Auger

electron is released from the same energetic state and therefore the Auger electron is

labeled L3M4,5M4,5. This example has been chosen since it is the most prominent Auger

transition in case of zinc and of great importance for this work as shown later.

In many cases, additional analysis of the Auger lines in a photoemission spectrum provides

more chemical information about the sample than just measuring the photoelectrons. In

fact, the shift in kinetic energy of the Auger electrons might exceed the shift in binding

energy of the photoelectrons significantly[30]. In general, this occurs for materials that

are electrically conducting and when the initial vacancy is in one of the inner electron

shells. In case of zinc compounds Wagner et al.[30] showed that the Zn photoelectrons

show almost no shift in binding energy, while the corresponding Auger lines can shift

several electron volts. This deviation can be explained by differences in the ionized states:

if the atom emits a photoelectron, its final state remains singly ionized. This affects the

outer shell electrons that can be emitted as Auger electrons. Therefore, electrons emitted

from an ionized atom in form of an Auger electron will be more sensitive to the chemical

environment than photoelectrons emitted from atoms in their ground state[31]. Wagner[30]

combined the photoelectron binding energies and the kinetic energy of the corresponding

Auger electrons in his concept of the so-called Auger parameter αas an empirical method

to determine the chemical state of an atom in a sample. The Auger parameter is defined

as:

α=Ebin(photoelectron) + Ekin(Auger electron) (2.4)

His concept is based on the idea that the energy difference between photoelectron and

Auger electron of an element is fixed for particular compounds. In addition, the combina-

tion of kinetic energy of the Auger electron and the binding energy of the photoelectron

cancel out effects due to surface dipoles or band bending and make work function cor-

rections unnecessary. It has also the advantage of being independent on the excitation

energy, which allows for comparing measured data with the results of other groups inde-

pendent on the X-ray source used. There are two database of experimentally determined

Auger parameter available in the internet worth being mentioned: one the one hand an

extensive database of the National Institute for Standards and Technology (NIST)[29]

and on the other hand one database provided by the UK Surface Analysis Forum[32]. A

2.2 Photoelectron Spectroscopies

comprehensive review of the Auger parameter concept can be found elsewhere[30;33].

An unwanted but important spectral feature to be aware of are X-ray satellite features.

Those features arise due to the differences in excitation energy of the different X-ray

lines in case of not using a monochromator (c.f Fig. 2.4). Fortunately, those features can

nowadays easily be removed during data evaluation by aid of proper software routines.

Additional final state effects that can occur in the photoemission spectra like plasmon

energy loss features, multiplet splitting, shake-up and shake-off peaks will not be discussed

here but are treated in various text books[26–28].

Photoelectron spectroscopies are very surface specific techniques. Even though the X-rays

with energies usually used in XPS can penetrate about 1000 nm into the sample, the

distance an electron can travel through matter without losing energy (and hence their

information content) is much smaller. The distance an electron can travel without un-

dergoing inelastic collisions is called the inelastic mean free path (IMFP) of an electron.

According to Seah and Dench[34] the IMFP λfor inorganic compounds can be calculated

by Eq. 2.5:

λ=2170

kin

+ 0.72paEkin·a(2.5)

ais the monolayer (ML) thickness of the material the electron travels through, expressed

in nanometers. To determine the monolayer thickness for ZnO, we assume a crystal growth

without a preferred growth direction. One can then calculate the monolayer thickness via

the volume of the unit cell Vand its number of atoms n:

a3=V

n=(~a ×~

b)·~c

n, (2.6)

where the volume can be calculated forming the triple product over the basis vectors of

the unit cell (~

b=~a). Solving Eq. 2.6 results in a ZnO monolayer thickness of:

a=3

r(~a ×~a)·~c

n=3

r0.3252·cos 30◦·0.52

4= 0.228 nm (2.7)

2 Methods of Surface Analysis

Figure 2.8: Calculated inelastic mean free path (IMFP) of photoelectrons in zinc oxide

depending on their kinetic energy.

In case of ZnO, Fig.2.8 shows the IMFP calculated by Eq. 2.5 plotted against the kinetic

energy of the excited photoelectrons. Even if the absolute values of the electron IMFP

are just a rough estimate, it gives an idea about the surface sensitivity of the technique.

Since the IMFP is a statistical value it does not mean that electrons excited deeper in

the material than the IMFP do not contribute to the measured photoemission signal.

95 % of all photoelectrons escape from the so-called sampling or information depth of

3λwhich results in a surface sensitivity of roughly 10 nm in photoelectron spectroscopy

experiments. As mentioned above, the calculated IMFP is just a rough estimate. The

actual values are difficult to determine and not only depend on the kinetic energy of

the electrons but also on the density of the material the electrons have to penetrate, its

composition and crystal structure.

Those electrons that are able to escape the crystal without losing energy contribute to

the characteristic photoemission peak containing chemical information. In addition, there

are photoelectrons that lost some energy by inelastic scattering processes but still have

enough energy to escape the work function of the surface. Those photoelectrons contribute

to a cumulative background signal which increases after each photoemission line in the

spectrum also visible in Fig. 2.6. This background signal increases continuously towards

2.2 Photoelectron Spectroscopies

lower kinetic energy and has to be considered during quantification as shown later in this

section.

Once an electron has been ejected from the sample, its kinetic energy has to be determined

using an electron spectrometer. As illustrated in Fig. 2.3 those spectrometers are com-

posed of three components: electron optics to retard the incident electrons, an electron

analyzer to separate the electrons with respect to their kinetic energy, and a detector to

count the electrons.

Nowadays, the most common electron analyzers are energy dispersive concentric hemi-

spherical analyzers (CHA) also illustrated in Fig. 2.3. A CHA consists of two hemispheres

in between an electric field is applied such that the outer hemisphere is more negative

and the inner hemisphere more positive with respect to the potential at the center line[27].

This potential at the center of the analyzer is called the pass energy Epass.

Electrons that enter the entrance slit of the analyzer with Ekin ≈Epass are deflected by

the electric field, pass through the analyzer and are counted by a multichannel detector

behind the exit slit. Electrons that enter the entrance slit with kinetic energies much

higher than the pass energy will hit the outer hemisphere, those with much lower kinetic

energies will be attracted by the inner hemisphere and will also not contribute to the

measured signal.

Quantification

The quantification of the measured raw data usually includes two steps: the removal of

the unwanted inelastically scattered secondary electron background and a peak fitting

routine to determine the position and intensity of the photoemission peak. In this work,

background subtraction is done by subtraction of a Shirley function[35] from the measured

spectra. The Shirley background is illustrated by the blue shaded area in the left part of

Fig. 2.9.

After removing the secondary electron background, a software-controlled peak fitting rou-

tine was applied to the data. It simulates the intensity distribution of the photoemission

signal by a Voigt profile, which is a convolution of a Lorentzian and a Gaussian profile.

Integration of the line profile then results in the particular peak intensity, as shown in

Fig. 2.9 (b). To get comparable peak positions, it is important to calibrate the position

of the Fermi energy and the energy scale of the spectrometer from time to time. Both is

done by defined reference materials. Using clean metal surfaces, the position of the Fermi

2 Methods of Surface Analysis

Figure 2.9: (a) inelastically scattered secondary electron background removal by applying

a Shirley function (blue); (b) peak fitting by a Voigt profile. The integrated

peak area (red) gives the peak intensity.

edge defines the energetic reference point for the core-level spectra. In addition, clean

copper, silver and gold samples are used to calibrate the spectrometer for slow electrons

(Cu), electrons of medium energy (Ag) and fast electrons (Au), respectively.

Many factors have influences on the measured peak intensity. According to Wagner et al.

the detected photoemission intensity is given by Eq. 2.8[36]:

I=n·f·σ·Θ·y·A·T·λ(2.8)

The contributing parameters are in particular:

I = photoemission intensity of a specific element per time unit [s−1]; n = particle density of

the element [cm−3]; f = photon flux of the X-ray source [cm−2s−1]; σ= photoionization

cross section of the particular element; Θ = angular efficiency of the spectrometer; y

= photoelectron yield; A = surface area under investigation [cm2]; T = transmission

function of the spectrometer; λ= inelastic mean free path of the photoelectrons in a

specific material.

2.2 Photoelectron Spectroscopies

Zn 2p3/2O 1s Cu 2p3/2In 3d5/2Se 3d

Al Kα23.93 3.1 20.28 13.89 2.22

Mg Kα3.726 0.711 5.321 4.359 0.853

Table 2.2: Collection of the atomic sensitivity factors (ASFs) used in this work for com-

positional analysis in both XPS systems of the Integrated System.

While some of those parameters such as the illuminated area A, the angular efficiency

Θ and the photon flux f can be assumed to be equal for different elements under in-

vestigation, some of them are material specific. In particular, this is of importance in

compositional analysis, when comparing the individual photoemission intensities among

each other. To simplify the compositional analysis, there are atomic sensitivity factors

(ASFs) Sxprovided by the spectrometer manufacturers for each element and their specific

electron transitions. The atomic sensitivity factors used in this thesis are given in Tab. 2.2

for both XPS systems, respectively. They are given for a fixed angle between X-ray source

and electron analyzer of 54.7◦. There are differences in magnitude between the ASFs for

Al Kαand Mg Kαradiation. They occur due to different normalization of the factors:

while the data for Al Kαare normalized to the carbon 1s photoemission line, the ones

for Mg Kαare normalized to F 1s. With those sensitivity factors, in compositional XPS

analysis Eq. 2.8 reduces to:

I=n·Sx(2.9)

Hence, for zinc oxide the composition is given by:

[Zn]

[O] =nZn

=IZn :SZn

IO:SO

(2.10)

Another useful feature of the photoemission intensity used in this thesis is the possibility

of directly calculating the thickness of an overlayer by the attenuation of the substrate’s

photoemission signal(s). The only requirement is, that the overlayer has to be thin enough

to allow photoelectrons of the substrate to travel-through without any inelastic scattering,

i.e. its inelastic mean free path λ.

To determine its attenuation, one has to characterize the substrate prior to deposition of

the overlayer by means of X-ray photoelectron spectroscopy. An easy way to calculate

the film thickness after deposition of the overlayer is given by Seah and Dench[34]:

2 Methods of Surface Analysis

Figure 2.10: Schematic drawing of the overlayer thickness determination by attenuation of

the substrate’s photoemission attenuation and simultaneous increase of the

overlayer photoemission signal according to Jablonski and Zemek[37].

I=I0·exp −d

λ⇒doverlayer =−λ·ln I

I0(2.11)

where d= thickness of the overlayer; λ= inelastic mean free math of the particular

photoelectron in a specific material; I= measured intensity of the substrate; I0= intensity

of the substrate before overlayer deposition.

In this work, a more precise way to calculate the overlayer thickness reported by Jablonski

and Zemek has been used[37]. Additionally to the attenuation of the substrate, they take

into account the increase of the overlayer signal. In Eq. 2.12 the overlayer thickness

calculation is presented for the case of ZnO deposition on a silicon substrate:

dZnO =λ(Si 2p3/2)·cos α·I(Zn) ·I0(Si)

I(Si) ·I∞(Zn) + 1(2.12)

There, αis the emission angle of the photoelectrons with respect to the surface normal.

For normal emission (i.e. 90◦with respect to the surface or 0◦with respect to the surface

normal), the cos αterm has no influence on the measurement, since cos(0◦) = 1.

To be able to calculate the overlayer thickness according to Eq. 2.12, one has to deter-

mine the intensity of the overlayer signal much thicker than the information depth of the

substrate’s signal I∞(Zn). This has to be done just once for all subsequent measurements

if the analysis system remains unchanged. A schematic drawing of the overlayer thickness

determination with X-ray photoelectron spectroscopy is shown in Fig. 2.10.

2.2 Photoelectron Spectroscopies

In almost all XPS experiments the angle between sample surface and the electron optics

remains constant at 90◦with respect to the surface normal. Even though photoelectron

spectroscopy is a very surface sensitive technique, in some cases it can be of advantage

to further decrease the information depth of the photoelectrons. There are basically two

ways to obtain increased surface sensitivity. One possibility is to reduce the inelastic

mean free path of the photoelectrons by lowering the kinetic energy of the photoelectrons

according to Fig. 2.8. Since the kinetic energy is directly connected to the excitation

energy of the photon source (cf. Eq. 2.1) and each X-ray source has only one specific

photon energy, it is usually not possible to reduce the IMFP by lowering the electron’s

kinetic energy in the laboratory. As shown later in this chapter, it is possible to vary

the excitation energy using synchrotron radiation. Since those facilities are usually not

available and measurements are very cost-intensive, there is a second way to reduce the

information depth of the photoelectrons in the laboratory by reducing the exit angle of

the photoelectrons with respect to the surface normal. Fig. 2.11 illustrates those angle-

dependent XPS experiments. In the following, the exit angle is always denoted with

respect to the surface, i.e. normal emission corresponds to 90◦.

For normal photoelectron emission, the information depth (≈3λ) is identical to the

effective information depth of the photoelectrons. If one changes the angle between the

Figure 2.11: Illustration of angle-resolved XPS measurements to reduce the effective in-

formation depth of the photoelectrons. This increases the surface sensitivity

of the measurements without changing the excitation energy of the X-ray

photons.

2 Methods of Surface Analysis

surface and the electron optics, the path the electrons have to travel to escape the sample

increases. Since λremains constant, the effective information depth decreases and the

measurements become more surface sensitive. This is illustrated at the right-hand side of

Fig. 2.11. This allows to create non-destructive depth-profiling for instance of the samples’

composition and is applied later in this work.

Accuracy of XPS measurements

Estimating the absolute error of photoemission spectroscopy experiments (in terms of the

photoemission peak intensity and the accuracy in binding energy determination) is not

trivial, since there are some parameters which might cause large inaccuracies. For instance,

when calculating the thickness of an ZnO overlayer by attenuation of the substrate’s pho-

toemission signal (cf. Eq. 2.12), inaccuracies contributing from the peak fitting procedure

(<5 %) can usually be neglected, since the significant error arises from large uncertainties

of the electrons inelastic mean free path (≈40 %[37]). Fortunately, the relative error, com-

paring the results of the experiments performed in the Integrated System under the same

conditions among each other, is much smaller. These relative errors usually contribute

mainly from data evaluation like the background removal procedure or peak fitting. Here,

the inaccuracies depend on the signal-to-noise ratio of the measured signal, which in turn

improves with increasing the overlayer thickness, as demonstrated in Fig. 2.12 for a very

thin ZnO layer (left) and a film thickness in the range of the information depth (right).

Figure 2.12: (left) Zn 2p3/2peak fit of a very thin ZnO layer, resulting in an inaccuracy

of σ≈5 % (right) same peak fitting of a thick ZnO film results in a smaller

peak fitting error of only σ≈1.5 %

2.2 Photoelectron Spectroscopies

In case of a thin ZnO overlayer, the error of the data evaluation σwas determined being

σ≈5 %, represented by the difference spectrum between the measured and the fitted

data (green). With increasing film thickness, the signal-to-noise ratio decreases for the

elements present in the overlayer, resulting in an relative error of only about 1.5 %. At

the same time, the signal-to-noise ratio of the substrate’s photoemission signals increases

due to their decreasing peak intensity. Since the overlayer thickness is calculated from

the attenuation of these signals, the relative error of film thickness determination also

depends on the overlayer thickness as discussed just recently, being 5 % at most.

In angle-dependent XPS experiments, an additional error results from inaccuracies in ad-

justing the emission angle by tilting the sample. Assuming an error of 1◦at most, the

cosα dependency of the peak intensity given in Eq. 2.12 results in an additional error of

2 % for small emission angles. For normal emission, the error is only about 0.2 % per

degree deviation from the surface normal and hence can be neglected. Besides an error

in peak intensity determination, one also has to consider inaccuracies in determination

of the peak’s position. Since the spectrometer was calibrated by well-defined reference

samples, the absolute error is very small, usually assumed being below 0.1 eV.

2.2.2 Ultraviolet Photoelectron Spectroscopy

While X-ray photoelectron spectroscopy contains information about the core-level elec-

trons, ultraviolet photoelectron spectroscopy (UPS) allows analysis of the valence electron

states that form interatomic bonds between the atoms. Instead of using high energy X-

rays to excite the photoelectrons, UPS uses ultraviolet radiation of a helium discharge

lamp, resulting in an excitation energy of hν = 21.22 eV for the He I transition. This

low excitation energy is not able to excite core-level electrons but can emit loosely bound

photoelectrons of the valence band. The helium discharge lamp does not only provide very

intense photons of the He I discharge, but also He IIαof hν = 40.8 eV and He IIβwith

hν = 48.0 eV energy. As in X-ray photoelectron spectroscopy, satellite peak contribute to

the measured spectra, that have to be removed for quantitative analysis. The line width

of the He I line of only 3 meV results in very high energy resolution of the valence band

spectra, usually limited by the energy resolution of the electron analyzer[28].

A typical valence band spectrum of a semiconductor recorded by He I irradiation is

shown in Fig. 2.13 (a). One can directly determine the work function of the material

by the position of the secondary electron cutoff at the high binding energy side of the

2 Methods of Surface Analysis

Figure 2.13: (a) Ultraviolet photoelectron spectrum containing the valence band structure

of a semiconductor; (b) secondary electron cutoff; (c) valence band maximum

spectrum. An enhancement of the secondary electron cutoff is shown in Fig. 2.13 (b).

The work function φis simply calculated by subtraction of the secondary electron cutoff

energy from the excitation energy of the He I discharge (hν = 21.22 eV):

φ=hν −Esec (2.13)

At the low binding energy side of the valence band spectrum, the valence band maximum

(VBM) can be determined. The easiest way is to do a linear extrapolation of the valence

band edge. This is illustrated in Fig. 2.13 (c) by the blue lines. The point of intersection

corresponds to the valence band maximum of the semiconductor material. Knowledge of

the VBM and the secondary electron cutoff Esec directly results in the ionization energy

Eion of the material:

Eion =hν −Esec +EVBM (2.14)

2.2 Photoelectron Spectroscopies

In addition, knowledge of the band gap width Egbetween the conduction band minimum

and the valence band maximum of the semiconductor results in the electron affinity of

the sample. The band gap energy is known from literature values. One has to be aware

that those values are usually determined for bulk crystals and the actual band gap for

ultra-thin films can differ from those values. Nevertheless, the electron affinity can be

calculated according to Eq. 2.15:

χ=φ+EVBM −Eg(2.15)

The combination of UPS and XPS is a very powerful method to investigate the heteroin-

terface of two semiconducting materials. Especially the in situ deposition and analysis of

semiconductor A on semiconductor B is of interest in this work and can help to investigate

the electronic band alignment of the heterocontact. This contact is of utmost importance

for the efficiency of electronic devices, such as for instance solar cells and therefore detailed

knowledge of the band alignment is required. More details about the band alignment is

given later in Chp. 5.

2.2.3 Synchrotron Radiation

In contrast to laboratory X-ray sources, which are limited to one specific photon en-

ergy due to the defined electron transitions in the anode material, synchrotron radiation

provides tunable radiation in a wide energy range. Unfortunately, synchtrotron radiation

cannot be created in the laboratory but requires large electron acceleration facilities, what

limits their every day use in science and technology.

The experiments presented in this thesis were all performed at the BESSY-II synchrotron

radiation facility in Berlin, where a cooperative research group of the University of

Cottbus, the University of Darmstadt and the Helmholtz-Center Berlin operate the

U49/2-PGM2 beam line. The close vicinity of the synchtrotron radiation facility allows

to transport the samples from the laboratory to the BESSY II beam line under ultra-high

vacuum conditions.

Fig. 2.14 illustrates the principle of the BESSY II synchrotron radiation facility. Bunches

of electrons are injected into the inner synchrotron ring, where they are accelerated to

energies of 1.7 GeV, resulting in velocities close to the speed of light. After the electrons

have been accelerated, they are injected into the electron storage ring, with a circumference

2 Methods of Surface Analysis

Figure 2.14: Schematic of the BESSY synchtrotron radiation facility: (1) electron accel-

erator; (2) electron storage ring; (3) bending magnets; (4) undulator; (5)

collimating mirror; (6) monochromator; (7) toroidal mirror; (8) sample

2.2 Photoelectron Spectroscopies

of 240 m. There, bending magnets and undulators force the electrons to circulate around

the storage ring for a long period of time. Forcing a charged particle into a circular orbit,

they are accelerated radially and emit a continuum of light, the so-called synchtrotron

radiation. For the U49/2-PGM2 beam line, this synchtrotron radiation is created and

inserted into the beam line by an undulator, as illustrated in Fig. 2.14. An undulator is

composed of a linear periodic structure of dipole magnets. Their static magnetic fields

force the electrons into a sinusoidal oscillation, thus emitting intense synchtrotron light.

This light enters the beam line as a cone with a small divergence angle. After entering

the beam line, a cylindrical mirror is used to collimate the beam. This parallel beam

of light them enters the monochromator, where the desired photon energy is chosen.

This particular beam line is able to provide energies in a range between 86 −1890 eV.

After monochromization, the light beam is focused onto the sample, resulting in a small

illuminated area of high intensity. Synchrotron radiation photoelectron spectroscopy (SR-

PES) offers very high spectral resolution and excellent signal-to-noise ratios. But it is

its tunable radiation source that allow some unique experiments and makes synchrotron

radiation a powerful tool in physics and materials science.

In X-ray photoelectron spectroscopy, photons of constant energy create photoelectrons

of different kinetic energies. As discussed earlier, the distance an electron can travel

in a crystal without undergoing inelastic scattering events (i.e. the inelastic mean free

path) depends on the kinetic energy of the photoelectrons. Hence, the photoelectrons

contributing the spectra can have their origin in different depths of the material, possibly

resulting in significant errors during data evaluation. The tunable excitation energy of the

photons allow to excite photoelectrons with specific kinetic energy according to Eq. 2.1.

According to Fig. 2.8, the information depth of the SR-PES experiments can be chosen

either very surface sensitive (for low kinetic energies) or electrons can be created deeper in

the material. Recording a range of different kinetic energies will result in non-destructive

depth profiling of the surface under investigation.

Quantification

Quantification of the SR-PES data is much more time consuming than in case of X-ray

photoelectron spectroscopy. In contrast to XPS experiments, where the flux of incident

photons on the surface can be assumed being constant, this is not the case for the syn-

chrotron radiation. Here, the photon flux depends on several factors, such as the selected

photon energy, the amount of photons rejected by the exit slit, and the electron beam

2 Methods of Surface Analysis

current in the electron storage ring. A proper way to normalize the spectra to allow com-

parability between the different measurements is to measure the intensity of the incident

photons with a GaAs photodiode. The spectra are normalized according to Eq. 2.16:

Inorm =I0·X·Ydiode

ring ·σ(2.16)

To normalize the measured photoemission spectra of intensity I0, the diode current Idiode

and the current in the storage ring Iref

ring have to be determined for the applied excitation

energy / exit slit combination. This results in a factor X=Iref

ring/Idiode to normalize

the spectra with respect to the ring current I0

ring when a particular spectrum has been

recorded. The yield of the GaAs diode Ydiode is also energy-dependent and has to be

determined. Its characteristic curve is well-known for the photon energy range used in

this work. σin Eq. 2.16 describes the ionization cross-section of a particular chemical

element. This is also energy-dependent and can be obtained from reference works, such

as the work of Yeh and Lindau[38].

Since there are several factors contributing to the normalization of the measured data,

the error in data evaluation is expected to be increased with respect to the laboratory

XPS work. Due to the very good signal-to-noise ratio, the error of data evaluation of I0

can be expected being below 1.5 %. The ring current of the storage ring does not change

significantly during the particular measurements. The absolute error is approximated

being below 0.5 % for Iref

ring and I0

ring, respectively. Additionally assuming the error of

Ydiode,Idiode, and σeach with 1 % at most, the total error of Inorm can be estimated

estimated being below 6 %. As in case of XPS, the energy scale of the spectrometer was

calibrated by standard reference samples, resulting in an absolute error below 0.1 eV of

the peak’s energy position.

2.3 Low-Energy Electron Diffraction

Applying low-energy electron diffraction (LEED) is a very useful tool to characterize

crystalline surface structures. Even though the focus of this work is on photoelectron

spectroscopy, LEED has been used to verify the quality of the substrate surfaces for the

subsequent ZnO deposition experiments.

The basic principle of electron diffraction in a crystal is illustrated in Fig. 2.15. A colli-

mated beam of electrons, typical kinetic energies are in a range between 20 −200 eV, is

2.3 Low-Energy Electron Diffraction

Figure 2.15: Principle of electron diffraction at a crystal surface.

directed on a crystal surface. The electrons are created in a hot cathode and accelerated

to the substrate by a couple of electrodes. The electrons interact with the crystal an

electrons can be elastically scattered. In some cases, the path difference of the scattered

electrons match the Bragg condition nλ = 2dhlk ·sinβ, resulting in constructive interfer-

ence of the scattered electrons. In reciprocal lattice, the Bragg condition can be expressed

as G=kj−k0, where Gcorresponds to the reciprocal lattice vector, while k0and kj

represent the wave vectors of the incident and diffracted beams, respectively.

Fig. 2.16 shows a combination of the reciprocal lattice and the spot pattern observed for

LEED experiments. The electrons that fulfill Bragg’s law, i.e. where the Ewald sphere

intersect a reciprocal lattice point, appear as bright diffraction spots on the luminescent

screen. Hence, the diffraction pattern provides a direct reproduction of the reciprocal

lattice. For detailed information about the technique and its information content it is

referred to various text books[39–41]. In this work, LEED is only used to check the crys-

talline quality of the prepared substrates. In addition, the presence of a secondary electron

background can give an indication about the cleanliness of the prepared surface.

2 Methods of Surface Analysis

Figure 2.16: Illustration of a LEED diffraction pattern formation by constructive interfer-

ence of diffracted electron fulfilling Bragg’s law.

3 Thin-Film Deposition

Nowadays thin-films are used in a wide field of applications as for instance in thin-film

chalcopyrite solar cells as introduced in Sec. 1.1. Producing thin-films from the vapor

phase is of great importance and the processes occurring when atoms or molecules impinge

on a materials surface are crucial for the quality and properties of the films produced. In

the following chapter, first some general remarks about the atomic events of initial layer

growth are presented. Afterwards examples of several important vapor phase deposition

techniques are given and their particular advantages and disadvantages are discussed.

3.1 General Issues of Thin-Film Deposition

There are a couple of different processes in thin-film growth as the adsorption of the gas

phase atoms and molecules on the surface to be coated, surface diffusion of the adsorbed

species, re-evaporation processes and the final film formation. The structural properties

of the formed films depend on both, the conditions in the deposition system as well as the

combination of the substrate and film materials. The conditions in the deposition system

will vary for the different techniques and hence their influence on the film structure will

be discussed in the particular sections.

First we will take a look at the atomic processes when a molecule approaches a surface.

To be deposited on the substrate it has to be bound to the surface, a process known as

adsorption. Usually one distinguishes two types of adsorption mechanisms and that is

physisorption and chemisorption, respectively. Figure 3.1 illustrates the differences be-

tween both adsorption mechanisms in an energetic manner. If a molecule reaches the

surface from the vapor phase it will feel an attractive force which accelerates it towards

the surface. Even it is, from a thermodynamic point of view, energetically favorable for

3 Thin-Film Deposition

Figure 3.1: Energetic conditions of atoms approaching a surface where they either are

physisorped by weak van-der-Waals forces or strongly bound when chemi-

cal bonds are formed between surface and adsorbed atom in a chemisorption

process.

the arriving molecule and the substrate’s atoms to react and form a chemical bond, the

interatomic bonds in the arriving molecule might be too strong to reorganize sponta-

neously. In this case, the molecule will remain loosely bound at the surface bound by

electrostatic van-der-Waals forces. In Fig. 3.1 this process is illustrated in terms of a

Lennard-Jones potential which is the result of attractive and repulsive forces affecting

the incident molecules. Mathematically, the Lennard-Jones potential can be expressed as

shown in Eq. (3.1):

E=rm

r12 −2rm

r6(3.1)

As shown in Fig. 3.1 the Lennard-Jones potential will form a well of depth which defines

an energetically favorable position rmfor the molecule to remain at. To form a stable

chemical bond to the surface the physisorbed atom in the potential well of the blue curve

first requires some extra energy to break some intramolecular bonds. This is illustrated

by the energy barrier between the potential well of the blue curve and the intersection

point with the red curve. The additional energy might be provided by thermal heat which

allows the atom to overcome the energetic barrier. Chemical bonds are formed and the

adsorbed atom is strongly bound, i.e. it is chemisorbed at the substrate’s surface. On the

3.1 General Issues of Thin-Film Deposition

other hand, if the energy required to change from the physisorped state to the chemisorbed

state is higher than the binding energy between molecule and surface, it is more likely for

the adsorbed molecule to desorb from the surface.

Once a molecule is adsorbed on the surface it might move to a neighboring surface site

which is not occupied by other molecules adsorbed from the vapor phase. Depending

on the surface coverage (and therefore on the deposition rate of the material on the

surface) surface diffusion might be more or less favorable. Generally speaking, the lower

the deposition rate, the higher the surface diffusivity. This, in turn, results in improved

structural quality of the deposited films.

When a thin-film of material B is deposited onto a substrate A the film is desired to

grow with uniform thickness distribution across the whole surface area. In practice, the

morphology of the deposited film depends on a couple of properties of both the substrate

as well as the material to be deposited.

Much of the thin-film structure achieved during deposition from the vapor phase can be

attributed to surface energies of the substrate γSA, the film material γSB and the energy

of their interface γAB. The surface energy is a quantity that expresses the additional

amount of energy of a surface compared to the energy of the bulk crystal. At the surface

the bonding conditions of the particular atoms differ from that in the bulk because there

is no partner to form bonds with which results in dangling bonds. By reconstruction,

i.e. reorganization of the surface atoms, the number of dangling bonds can be reduced

and the surface energy is lowered. An illustration of the different surface energies for a

material B deposited on a substrate A is given in Fig. 3.2.

Figure 3.2: Contact angle Θ in case of deposition of a material B onto a substrate material

A. γSA denotes the surface energy of the substrate, γSB the one of the deposit

B and γAB the interface energy between A and B.

3 Thin-Film Deposition

It is common to express the surface energies in terms of the contact angle Θ. Mathemat-

ically, this can be described by Young’s Equation given in Eq. (3.2).

γSA =γSB +γAB ·cos Θ (3.2)

Desired is a contact angle of cos Θ = 0◦where the material does not form a droplet but

spreads over the whole surface area and forms a thin-film. This is the case when the

film deposited onto material A lowers its surface energy by saturating the dangling bonds

without further reconstruction. On the other hand, if the surface energy of material A

is increased by depositing film B the film will tend to minimize the covered area of the

substrate. This will lead to the formation of islands of material B. A third mode of

thin-film growth is the so-called Stranski-Krastanov or layer-plus-island growth mode. In

this process first some monolayers are deposited in a layer-by-layer mode and beyond a

critical thickness, depending on lattice strain and the chemical potential, the film continues

growing in an island growth mode. All three possibilities, layer-by-layer (or Frank-van

der Merwe) growth, island (or Volmer-Weber growth), and Stranski-Krastanov growth

are illustrated in Fig. 3.3.

Figure 3.3: Three different growth modes observed during thin-film growth: (top) layer-

by-layer growth, (middle) island growth, (bottom) Stranski-Krastanov growth

3.1 General Issues of Thin-Film Deposition

Figure 3.4: Photoemission signal characteristics for the three different primary thin-film

growth modes. The attenuation of the substrate photoelectrons A and the

increase of the film material B can be measured in situ during the initial

growth.

Surface energy is not the only factor influencing the growth mode. Lattice mismatch is

another important one. If the lattice constants between substrate material and deposit

do not fit a layer-by-layer growth is not very likely. Therefore, even if a deposit would

lower the surface energy of the substrate there may be an energetic favor of growing in

an island growth mode.

Which type of growth mode occurs in the initial growth regime of the thin-film can be

measured by photoelectron spectroscopy. As long as the deposited film is thin enough

for photoelectrons of the substrate to escape the growth mode can be determined by the

attenuation of the substrate signal and the increase of the film signal as shown in Fig. 3.4.

Another point of interest in crystalline thin-film deposition is the orientation of the

growing film. In case of island growth, polycrystalline films are formed whose grains usu-

ally try to reduce the density of dangling bonds by growing in a preferred orientation to

the substrate[42]. The distribution of crystallographic orientations is called the texture

of the film and can be measured by X-ray diffraction techniques. Epitaxy is the most

perfect way of growing a material B onto a material A. It results in a single-crystalline

film which has taken on the lattice structure and crystal orientation of the subjacent

substrate. One distinguishes between two types of epitaxial growth, homoepitaxy and

heteroepitaxy, respectively. In homoepitaxy the substrate and the deposited film consist

3 Thin-Film Deposition

of the same material and therefore their structural properties match perfectly. In contrast,

in heteroepitaxy the deposit and the substrate material differ from each other. Therefore,

their lattices will usually not match perfectly which leads to lattice strains in the begin-

ning of the overlayer growth. This can influence the electronic properties of the interface

as will be shown later in this work. For further details of thin-film growth the reader is

referred to the books of Ohring[43] and Rockett[42].

Generally, one can divide the wide field of thin-film deposition techniques into the two ma-

jor classes of physical vapor deposition (PVD) and chemical vapor deposition (CVD). The

following two sections will give an overview of their basic properties, their differences and

their particular advantages and disadvantages with respect to the final film properties.

3.2 Physical Vapor Deposition

From an historic point of view physical vapor deposition is one of the oldest thin-film

deposition techniques. The most widely used methods are thermal evaporation of a hot

source material and sputter deposition where the material to be deposited is removed

from a solid target by impact of high energy gaseous ions, usually Ar+. The momentum

transfer during ion impact evaporates the target material which is then deposited onto

the substrate. The use of an ionized sputter gas results in a higher background pressure

during film deposition. In contrast, thermal evaporation is operated under high vacuum

conditions (p= 10−6mbar) and hence it shows beneficial results in terms of film purity

and the deposition rate is increased with respect to sputter deposition. Since sputter

deposition has not been used in this thesis it will not be treated here in more detail.

For further reading there are various textbooks (e.g.[43;44]) giving extensive reviews of

sputter deposition techniques. Instead, we take a closer look at thermal evaporation and

a techniques that is closely related and of utmost importance, namely molecular beam

epitaxy (MBE).

3.2.1 Thermal Evaporation

In thermal evaporation processes a crucible containing the source material to be deposited

as a thin-film is heated resistively. This heating process converts the condensed phase of

the deposit into its vapor phase. If those atoms reach a substrate in the deposition

3.2 Physical Vapor Deposition

chamber they condense and form a thin film as described in Sec. 3.1. Hence, it is obvious

that the key parameter determining the deposition rate of the film is the temperature

of the evaporation source. A relation between temperature Tand vapor pressure pof a

material is given by the Clausius-Clapeyron equation[44]:

dT =L

T(νg−νc)(3.3)

where Lis the amount of energy needed to evaporate 1 mol of material, νgits molar

volume in the gas phase, and νcthe molar volume in its condensed state. Assuming that

Lis independent on temperature, the vapor can be described with the ideal gas law and

assuming that the volume of an amount of material is much larger in its vapor phase than

in its solid state, i.e. νgνc, Eq. (3.3) can be written as:

log10 p=−A

T+B(3.4)

where Aand Bare constants that can be experimentally determined. In case of zinc selenide

(ZnSe) evaporation one can find reported values[45] for A= 6250±187.5 and B= 3.762 ±0.41.

Figure 3.5 plots the calculated vapor pressure of ZnSe over the temperature range inter-

esting for ZnSe evaporation.

Figure 3.5: Calculated ZnSe vapor pressure during evaporation at different source temper-

atures. The vapor pressure was calculated by the Clausius-Clapeyron equation

using A= 6250 and B= 3.762

3 Thin-Film Deposition

While metal sources vaporize mostly in the form of single atoms the same is usually

not valid evaporating compounds consisting of two elements Aand B. There are only

few compounds which evaporate congruently, i.e. in a single vapor phase AB(g) without

dissociation. Examples are fluorides like CaF2or some simple oxides like SiO[44] due to

their strong interatomic bonds. Most other compounds decompose during heating because

of differences in the vapor pressures of their elemental components, i.e. they vaporize non-

congruently. This can lead to a change in composition in the deposited film with respect

to the stoichiometric source material. Usually those films tend to be metal-rich. A typical

reaction for chalcogenides like ZnSe where the compound dissociates into two vapor phases

can be expressed as:

AB(s)→A(g) + B(g)

In compounds whose elements have significant differences in their vapor pressures like in

nitrides or carbides one component remains solid while the other evaporates:

AB(s)→A(s) + B(g)

Both cases, congruent and noncongruent evaporation, are schematically depicted in Fig. 3.6.

Evaporating the source material by resistive heating of the crucible is the most commonly

used thermal evaporation technique. One important requirement of every evaporation

source is that there is only an insignificant amount of vapor created by the crucible ma-

terial itself. Otherwise the purity of the formed thin-film cannot be guaranteed. Typical

Figure 3.6: Schematic illustration of different types of thermal evaporation of compounds:

(a) congruent evaporation where the compound evaporates without dissocia-

tion and (b) noncongruent evaporation where the two atomic species evaporate

at different temperatures

3.2 Physical Vapor Deposition

materials for evaporation sources are tungsten, tantalum or ceramics with high melting

points and low vapor pressures. Further origin for possible film contaminations are impu-

rities in the source material to be deposited and residual gases in the deposition chamber.

The former can be reduced to a minimum by outgassing of the sources, the latter is

prevented by high vacuum conditions in the evaporation system.

A somehow special case of thermal evaporation is the important and powerful technique

of molecular beam epitaxy (MBE) which will be treated in the following section.

3.2.2 Molecular Beam Epitaxy

As mentioned before molecular beam epitaxy (MBE) can be seen as a more sophisticated

type of thermal evaporation. The major difference between those two techniques is the

vacuum level where the film deposition is operated at and the use of advanced effusion

cells, so-called Knudsen cells. A schematic picture of a MBE deposition chamber is given

in Fig. 3.7.

Figure 3.7: Schematic drawing of a molecular beam epitaxy (MBE) system, taken from

Capper et al.[46]

3 Thin-Film Deposition

While in thermal evaporation systems high vacuum conditions with pressures in the

10−6mbar region are sufficient to produce high quality films, the same is not true for

MBE systems. The reason for this is the reduced deposition rate in molecular beam epi-

taxy. Those are typically below 1000 nm/h to produce highest crystalline qualities atomic

layer by atomic layer. To prevent incorporation of impurity atoms at such low growth rates

ultra-high vacuum conditions with background pressures below 10−9mbar are required.

As already mentioned in Sec. 2.1 achieving UHV conditions is time consuming and the

equipment is much more expensive. But since the advantages of molecular beam epitaxy

clearly exceed those disadvantages, MBE has become one of the most important thin-film

deposition techniques in the last decades. In addition, using UHV conditions makes MBE

compatible with in situ analytical techniques like reflective high energy electron diffrac-

tion (RHEED) to control the monolayer growth during deposition or direct connection to

other UHV analytical methods like photoelectron spectroscopies (cf. Sec. 2.1).

The requirement of precisely controlling the beam fluxes of each evaporated material led

to an improvement of effusion cells. Knudsen did extensive research on molecular gas flow

in the early 20th century[47]. He invented an evaporation source fulfilling the requirements

of precisely controllable effusion cells with a uniform flux in forward direction. Their main

difference to common effusion cells is a small opening from which the evaporant escapes.

As shown in Fig. 3.7 shutters allow to stop the growth process without switching off the

effusion cells itself.

One disadvantage of using Knudsen cells or any other type of solid evaporation sources

in MBE systems is the problem of refilling them. Since they are located in the vacuum

chamber a refilling requires breaking the vacuum. Since establishing UHV conditions is

fairly time consuming there are some popular modifications of the MBE process using

volatile gas sources. Those gas-source MBE (GSMBE) or metal-organic MBE (MOMBE)

systems are explained later in detail in Sec. 3.4. First, it is necessary to take a closer look

at chemical vapor deposition (CVD) processes since using gas sources films are deposited

by chemical reactions instead of physical deposition.

3.3 Chemical Vapor Deposition

As mentioned above, external sources containing gaseous reactants that flow into the

deposition chamber can be refilled easily and timesaving compared to thermal evaporation

or MBE effusion cells. While evaporation and film formation in PVD techniques are

3.3 Chemical Vapor Deposition

physical processes, in chemical vapor deposition (CVD) volatile gases are introduced in

the deposition chamber to react there and form a solid film at a heated substrate surface.

CVD has many advantages over PVD techniques. Surfaces with complex topographies

can be coated without any shadowing effects since there is no requirement for line-of-

sight geometry as in evaporation[44]. The deposition itself is generally more conformal

compared to PVD, i.e. it covers rough surfaces in a much more uniform manner. It is

able to produce high quality epitaxial films as well as amorphous layers depending on the

deposition parameters and there is usually no need for very high vacuum conditions using

CVD, even though this can have advantages as will be shown later in this chapter.

Chemical vapor deposition systems consist of three main parts: the gas inlet, the reaction

chamber and an exhaust for maintaining a constant gas flow and removal of reaction by-

products. Depending on the condition of aggregation precursors are dispensed into the

reaction chamber in different ways. If the vapor is in its gaseous phase and its vapor

pressure sufficiently high it can be fed directly into the system. For precursors that

are in a liquid or solid state at ambient conditions an introduction into the deposition

chamber is done by aid of evaporators or sublimators. Usually transport gases like H2are

used to transport the precursor into the chamber. Especially H2as carrier gas can have

advantageous properties on the deposited films in terms of reduction of contaminants[48].

The reaction chamber containing the substrate to be deposited can either be a hot-wall

reactor where both substrate and reactor walls are heated or it can be a cold-wall reactor

where only the substrate is kept at elevated temperatures. Cold-wall reactors do have the

advantage over hot-wall reactors that no reactions occur at the reactor walls which form

particles that can fall on the substrate and contaminate the surface[49].

In CVD processes the precursor gases flow through the reaction chamber and the substrate

is placed into this gas stream. Depending on the gas velocity there are two different flow

regimes: a preferred laminar flow where the gas molecules flow parallel to each other and

a turbulent flow regime. Turbulent flows are unfavorable and to prevent since resulting

swirls can lead to a nonuniform defect-rich thin-film. The dimensionless Reynold’s Number

Re is used to express the type of flow regime. For Re < 1100 the flow will have a laminar

character while there turbulent flow regime starts at Re > 2100. In between there is a

mixed flow type.

From theory of fluid mechanics it is known that near the sample surface a boundary layer

is formed in which the gas flow velocity is reduced down to zero directly at the surface

and vapor concentrations differ from those in the gas stream. Both, precursor molecules

3 Thin-Film Deposition

and gaseous reaction by-products have to pass through this boundary layer by chemical

diffusion processes before they can be deposited onto the substrate as shown later.

Depending on the material to be deposited there are several different general classes of

reaction types used in CVD processes. Those include amongst others:

•pyrolysis: thermal decomposition of precursor molecules on a hot substrate surface,

e.g. Si or metal deposition

•reduction: reduction by hydrogen gas as reducing agent, e.g. Si or metal deposition

•oxidation: precursor molecule is oxidized by O2as oxidizing agent to form nonvolatile

oxide, e.g. SiO2deposition

•compound formation: formation of compound by reaction of two precursors each

containing one of the compound elements, e.g. carbide or nitride formation for

wear-resistant coatings

The reactions used in this thesis are all of the compound formation type which will be

discussed later in Sec. 3.4 in more detail. The other reaction types are will not be treated

here since they have not been used. For further details various textbooks about chemical

vapor deposition are available[43;44;49] for the interested reader.

Several factors do have influences on the growth kinetics in CVD. Figure 3.8 schematically

depicts the different reaction steps that are necessary to produce a thin-film of the desired

compound material: In order for a film to grow, precursor molecules from the introduced

vapor have in a first step (a) to travel towards the substrate. Due to the boundary layer

formation step (b) involves diffusion through this layer (indicated by a yellow background

in Fig.3.8) towards the substrate surface. Reaching it, the molecules might be adsorbed

(cf. Sec. 3.1) if they find a surface site which is not already occupied by other precursor

molecules in step (c). Step (d) involves the chemical reaction of the adsorbed precursor

molecule. There are two possible reaction partners to form a compound with, either one

other absorbed molecule on the substrate surface or one which already diffused through

the boundary layer but has not been adsorbed yet. Since chemical reactions require

the breaking of intramolecular bonds those reactions are likely to occur just at the heated

substrate where the required activation energy is supplied in the form of heat and catalytic

activity. At the same time chemical reactions between the precursor molecules in the main

3.3 Chemical Vapor Deposition

Figure 3.8: Primary atomic processes during chemical vapor deposition: (a) transport

to the substrate, (b) diffusion through boundary layer, (c) adsorption, (d)

chemical reaction, (e) deposition directly from the gas phase, (f) nucleation

and film formation, (g) diffusion of reaction by-products through boundary

layer, (h) incorporation of by-products into gas stream

vapor stream are unlikely to happen. This is not necessarily true for precursor molecules

that are in the boundary layer at the closer vicinity to the heated surface. Those reactants

may gain enough thermal energy to react with each other and deposit undesired weakly

bound material onto the surface as illustrated in step (e). In step (f) the atoms have

nucleated and the film is started to form. Depending on the substrate temperature and the

deposition rate, layers with different qualities are possible to grow. If the deposition rate is

low, less nuclei are formed on the surface and the film will show good structural qualities.

At higher deposition rates, the atoms tend to cluster which results in polycrystalline or

even amorphous films. In addition, steps (g) and (h) in Fig. 3.8 show the removal of

gaseous reaction by-products from the surface by diffusion through the boundary layer

and incorporation into the precursor gas stream.

To improve the film quality, several modifications of the standard CVD process have been

developed. Low-pressure CVD (LPCVD) distinguishes from conventional CVD by reduc-

ing the pressure from atmospheric to pressures around 1 mbar[43]. Advantages are higher

deposition rates, improved film thickness uniformities and less defects in the deposited

3 Thin-Film Deposition

films. Another modification is the plasma-enhanced CVD (PECVD) where the pressures

are reduced as in case of LPCVD but in addition a glow discharge is initiated in the cham-

ber. In this plasma the precursor molecules will decompose and therefore the substrate

temperatures can be significantly lowered. In case of metal-organic CVD (MOCVD) as

a third variant of CVD, no changes in the setup or pressure region are carried out. In

contrast to conventional CVD at least one of the precursors has to be a metal-organic

volatile compound as the name already implies. Usually, H2or N2are used as carrier

gas for the precursor transport. MOCVD has become an important deposition method

because of its ability to grow high quality epitaxial films especially for the production of

III-V semiconductors. Hence, the notation metal-organic vapor phase epitaxy (MOVPE)

is also quite common in use.

3.4 Metal-Organic Molecular Beam Epitaxy

Actually related to chemical vapor deposition but being treated in a separate section due

to its hybrid character between metal-organic chemical vapor deposition and molecular

beam epitaxy is metal-organic molecular beam epitaxy (MOMBE). In this technique the

particular advantages of metal-organic precursors are combined with ultra-high vacuum

conditions as used in MBE. The metal-organic vapor introduced into the chamber from

external reservoirs is directed towards the substrate through small tubes as molecular

beams. Due to the UHV conditions gas phase reactions can virtually be excluded and

growth reactions are occurring almost exclusively at the substrate’s surface resulting in

very uniform films. MOMBE first was used to produce high quality GaAs films from

trimethylgallium (TMGa) and cracked arsine to avoid morphological defects in 1985[50].

In literature, MOMBE sometimes is also referred to as chemical beam epitaxy (CBE).

Besides the UHV conditions, the choice of the metal-organic precursor determine the

resulting film properties significantly in MOMBE. There are several important character-

istic features a precursor has to fulfill to be suitable to use in MOMBE growth processes.

On the one hand they have to have high vapor pressures to achieve high transport rates.

Usually this is the case for metal-organic compounds and therefore a carrier gas as in

case of CVD is not necessary. Another requirement for precursor is the stability of the

molecules at ambient temperatures. Otherwise decomposition reactions may occur and

the molecule fragments may contaminate the growing layer. To prevent carbon or oxygen

3.4 Metal-Organic Molecular Beam Epitaxy

incorporation into the growing films the reaction on the surface has to remove the organic

ligands of the metal atoms and the reaction products have to be volatile to be effectively

removed from the substrate.

An example is the MOMBE deposition of zinc oxide[21;51]. A precursor selection of di-

ethylzinc Zn(C2H5)2and ultrapure water H2O can result in very high quality epitaxial

films on appropriate substrates, for instance Al2O3(1102) or SiC (0001)[51]. The base

pressure of the MOMBE system included in the Integrated System (cf. 2.1) usually is

in the 10−10 mbar range. For deposition of ZnO films diethylzinc (DEZn) with a partial

pressure of pDEZn = 2 ·10−6mbar and water with a pressure of pH2O= 8 ·10−6mbar

are introduced to the deposition chamber by two leak valves. The temperature of the

substrate is kept at a temperature of T= 450 ◦C. On the surface, the ethyl ligands of

the diethylzinc molecule react with the hydrogen of the water and form gaseous ethane

C2H6. The remaining hydroxide is bound to the zinc atom, forming an Zn(OH)2inter-

mediate state, until it reacts with another diethylzinc molecule, finally forming a ZnO

compound.

4 Atomic Layer Deposition of ZnO

In comparison to those growth techniques presented in Chp. 3, atomic layer deposi-

tion (ALD) differs significantly in its growth mechanism. ALD belongs to the field of

CVD techniques (cf. Sec. 3.3) but in contrast to classical CVD the precursor gases are

not supplied simultaneously. Instead, two separated surface reactions are forming the com-

pound to be deposited. It is its characteristic self-limited growth mode that allows atomic

layer deposition a precise thickness control of the deposited materials in the sub-monolayer

regime.

ALD has already been invented in the 1970s by Suntola et al[52]. Its been developed for

depositing thin dielectric zinc sulfide (ZnS) layers for use in thin-film electroluminescent

displays. Even though ALD has already been invented nearly half a century ago, it is

becoming more and more important just recently because the need for ultra-thin films

of high uniformity rises. For instance, the International Technology Roadmap of Semi-

conductors (ITRS) included atomic layer deposition for the production of high dielectric

constant gate oxides and as Cu diffusion barriers just recently. In chalcopyrite thin-film

solar cell technology, ALD is more and more applied for buffer layer deposition. Further

details about buffer layers in chalcopyrite solar cells are given later in Chp. 5.

Since in the beginning of ALD the deposited films were mostly of single-crystalline epi-

taxial nature, the acronym ALE (atomic layer epitaxy) was introduced. In this work,

independently on producing single- or polycrystalline films, the term ALD is used.

This chapter is structured as follows: first, the ALD process of ZnO deposition is treated

in detail. Afterwards, the assembly of the UHV-ALD reactor and the process of commis-

sioning is presented. In Sec. 4.3 the basic growth parameter for zinc oxide ALD using

diethylzinc (DEZn) and water (H2O) as precursor gases are determined and the initial

growth of ALD-ZnO on hydrogen-terminated silicon is investigated in Sec. 4.5.

4 Atomic Layer Deposition of ZnO

4.1 Basic Principles of Atomic Layer Deposition

As mentioned above, the characteristic feature of ALD is the sequential supply of reaction

gases, delivering one of the components of the compound to be formed. A common ALD

reaction sequence consists of four stages. First, a gaseous reactant A is introduced into

the deposition chamber onto a heated substrate until the sample’s surface is completely

saturated. This can happen either by the precursor gas or one of its reaction products

in case of a chemical reaction occurring at the surface which is usually the case. After

saturation, the excess precursor gases and gaseous reaction by-products are flushed away

by a non-reactive purging gas. Commonly, this is argon (Ar) as non-reactive noble gas.

Afterwards, the second precursor gas (reactant B) is fed into the chamber and reacts with

reactant A at the sample surface until the reaction product saturate the surface. Again,

reaction products and surplus precursor gases are removed by an Ar purge. The sequence

of these 4 steps is called one ALD cycle. Since all reactions in ALD occur at the sample’s

surface without any gas phase reactions, only up to one monolayer (ML) of the compound

to be formed can be deposited during one cycle. This is schematically shown in Fig. 4.1

for the atomic layer deposition of a compound AB.

This unique reaction mechanism is the reason for atomic layer deposition offering a couple

of advantages over other deposition techniques:

Figure 4.1: Reaction scheme of AB compound formation during atomic layer deposition.

In contrast to other deposition techniques, the reactants which deliver the A

and B components are not fed in at the same time but sequentially one after

another. It is self-limited by saturation of the available reactive sites and the

compound forms in two separated half-reactions.

4.1 Basic Principles of Atomic Layer Deposition

deposition rate: in ALD the deposition rate is not depending on the precursor flux but

on the number of cycles due to saturating the limited number of reactive surface sites.

Hence, its self-limiting growth mode provides thickness control in the range of an atomic

layer.

film conformity: another feature of ALD is its superior conformity on surface structures

with high aspect ratios. No other deposition technique is able to coat trenches or steps

as uniform as ALD does. In addition, the films are very continuous and pinhole-free over

large areas and result in very sharp interfaces.

substrate temperature: right from the start, ALD was used to grow high-quality epi-

taxial films. In contrast to other techniques, the substrate temperatures are significantly

lower because diffusion processes are less important for saturating the reaction sites. Those

lower substrate temperatures also enable atomic layer deposition on substrates that would

usually decompose at elevated temperatures as for instance most polymers do.

4.1.1 Surface Reactions

The surface reactions in atomic layer deposition differ significantly from other deposition

techniques because of the sequential supply of the reactive precursor gases. This section

will provide a detailed view on the reaction mechanisms during zinc oxide deposition using

organometallic diethylzinc (Zn(C2H5)2) as metal-precursor and water (H2O) as oxidizing

agent. Diethylzinc is highly reactive with water. Supplied simultaneously, both ethyl

ligands will react with the two hydrogen atoms of the water molecule and form gaseous

ethane (C2H6). At the same time, the remaining Zn and O atoms will form a ZnO

compound:

Zn(C2H5)2+ H2O→ZnO + 2 C2H6(g) (4.1)

This reaction is highly exothermic with a formation enthalpy of ∆H=−70 kcal[53].

Since atomic layer deposition does not supply both reactants at the same time, this re-

action splits into two so-called half-reactions. The first half-reaction in an ZnO ALD

process is shown in Fig. 4.2 where a substrate surface terminated by hydroxides (-OH) is

assumed. This is not necessarily true for the first deposition cycles but a proper assump-

tion if the first complete monolayer of ZnO has formed as shown later. After introducing

4 Atomic Layer Deposition of ZnO

Figure 4.2: Reaction mechanism of the first ALD half-reaction using diethylzinc and water

to form ZnO. (a) DEZn exposure; (b) ligand exchange; (c) Ar purge; (d)

monoethylzinc saturated surface

the diethylzinc into the deposition chamber (a), the chemisorption of the molecule takes

place by a ligand exchange mechanism: one of the ethyl ligands reacts with one of the the

hydrogen atoms of the hydroxides located at the substrate’s surface. After reacting, one

ethane molecule is released as gaseous reaction by-product to the vacuum. The remaining

monoethylzinc (MEZn) is bound to the surface at the left over oxygen atom. This reac-

tion occurs until all reactive sites are saturated as shown in Fig. 4.2 (b). After complete

saturation, the surplus reactant gases and reaction by-products are removed by flushing

the chamber with non-reactive argon (c) to prevent further unwanted gas-phase reactions.

One ends up with a surface completely saturated by monoethylzinc molecules as shown

in Fig. 4.2 (d). One can express this half-reaction by Eq. 4.2, where kdenotes the surface

of the substrate:

k − OH + Zn(C2H5)2→ k − O−Zn(C2H5)+C2H6(g) (4.2)

4.1 Basic Principles of Atomic Layer Deposition

Figure 4.3: Reaction mechanism of the second ALD half-reaction using diethylzinc and

water to form ZnO. (e) H2O exposure; (f) ligand exchange; (g) Ar purge; (h)

hydroxide saturated ZnO monolayer

The second half-reaction is called the oxidizing half-reaction and is shown in Fig. 4.3. First,

the oxygen precursor is introduced into the chamber (Fig. 4.3 (e)). In case of ZnO atomic

layer deposition usually ultra-pure water is used as reactant. Again, chemisorption takes

place via a ligand exchange: the water reacts with the MEZn at the substrate’s surface,

formed during the first half-reaction. One of the hydrogen atoms removes the one ethyl

ligand left, again forming ethane. The remaining hydroxide forms a chemical bond with

the zinc atom on the surface as shown in Fig. 4.3 (f). Afterwards, the excess precursor

gases and the ethane is removed by argon flushing (g) just like in the zinc half-reaction.

Describing this half-reaction in form of a chemical reaction equation results in:

k − O−Zn(C2H5)+H2O→ k − O−Zn −OH + C2H6(g) (4.3)

4 Atomic Layer Deposition of ZnO

Both half-reactions are irreversible chemical reactions. This is one of the important pre-

cursor requirements to achieve self-limiting conditions necessary for ALD. After the purg-

ing step, one ends up with one ZnO monolayer covered by one hydroxide monolayer

(Fig. 4.3 (h)). Hence, the assumption made above, having an hydroxylated surface be-

fore starting the first half-reaction, can considered being true after the initial complete

ZnO monolayer was formed. What happens during the initial growth of zinc oxide is

also investigated in this work and shown later in Sec. 4.5. Usually, the deposition of one

monolayer is not achieved after one ALD cycle but requires several deposition steps. This

is treated in detail in the following section. Since its self-limiting growth is one of the most

important features of ALD, we will take a closer look at the surface saturation behavior

during precursor exposure.

4.1.2 Saturation

The self-limiting characteristic of ALD is most important to control the thickness of the

deposited films. Since chemical reactions only occur between the precursor molecules and

reactive surface sites and not among the precursor molecules themselves, there is only a

finite number of surface sites available to react as shown schematically in Fig. 4.1. If all

reactive sites are occupied by reaction products, no further molecules can be adsorbed

and the surface is saturated. Hence, only one monolayer of the adsorbate is accepted

by the surface. The number of adsorbed precursor molecules does not only depend on

the number of reactive sites but also on the size of the chemisorbed molecule. Fig. 4.2

(d) shows how the size of the adsorbed precursor ethyl ligands cause steric hindrance so

that not all reactive sites on the surface can be saturated in one ALD cycle. Even if in

literature often it is stated that one ALD cycle deposits one monolayer of material, it is

obvious that those steric hindrance effects effectively limit the maximum deposition of

material per ALD cycle to some value below one complete monolayer.

One requirement for complete saturation in atomic layer deposition is a sufficiently long

exposure time of the surface by the particular precursor gases. Fig. 4.4 plots the surface

coverage against the exposure time. In the beginning, the surface coverage increases

rapidly, until a maximum is reached. How long it takes until the complete surface is

saturated depends on the pressure of the precursor gas and the adsorption rate[55]. Ideally,

the reaction is as fast as possible to minimize the exposure times to keep the contamination

4.1 Basic Principles of Atomic Layer Deposition

Figure 4.4: Surface coverage vs. duration of precursor exposure according to Ritala and

Leskel¨

a[54]. If sufficient exposure times are chosen the surface is completely

saturated. The maximum saturation depends on both, reactive surface sites

and steric hindrance effects of the precursor molecules.

level in the deposition chamber as low as possible. If the exposure times are chosen too

short, thickness variations can occur in the films due to the incomplete surface coverage.

During Ar purge, the surface coverage should remain constant, which is usually true since

the chemisorbed surface species are sufficiently stable. If the Ar purging times are not long

enough for sufficiently removing all residual precursor gases and reaction products, gas-

phase reactions can occur during exposure of the next reactant. Those gas-phase reactions

would lead to an unwanted CVD-like growth which results in growth rates larger than

the self-limited ALD growth. Therefore, it is important to carefully pay attention to the

particular exposure and purging times when start running an ALD reactor. One ends up

with a ALD sequence scheme as shown at the top of Fig. 4.5.

In addition to the exposure and the purging steps, there is an additional pumping stage

after each Ar flush. This step is not present in standard ALD reactors that operate under

ambient pressures. In contrast to those reactors, the one designed and assembled in this

work is an UHV-compatible ALD reactor as shown later in Sec. 4.2. Since during exposure

and purging the pressure rises significantly in the chamber, a pumping step was added to

decrease the total pressure of the system at least down to high vacuum conditions after

each half-reaction. The pressure evolution in the deposition chamber is shown in the lower

part of Fig. 4.5.

4 Atomic Layer Deposition of ZnO

Figure 4.5: (top) ALD pulse scheme of the UHV-ALD reactor assembled for this work;

(bottom) pressure evolution during one complete ALD cycle.

4.1.3 Temperature Dependence

Since all reactions in an ALD process are surface-controlled, there are not many pa-

rameters having major influences on the deposition. Besides the exposure and purging

times discussed above the most important parameter is the substrate’s temperature dur-

ing growth. The additional energy that is supplied in form of heat is necessary to initiate

the ligand exchange to chemisorb the precursor molecule at the surface[56]. Hence, atomic

layer deposition process is also referred to as thermal ALD [53]. The temperature regime

where the ALD process fulfills the requirement of self-terminating chemical reactions is

termed the ALD window [55]. In this temperature range, a constant amount of material is

deposited during one ALD cycle, the so-called growth-per-cycle or GPC, due to complete

saturation of all surface sites. Usually, the ALD window spreads over a wider temper-

ature range since minor temperature changes do not have a significant influence on the

adsorption behavior of the precursor molecules.

Fig. 4.6 shows all typical temperature dependencies of the GPC on temperature. In

the middle, the ALD window regime of constant self-limited ALD growth is present.

The center of this temperature range is the preferred substrate temperature since small

4.1 Basic Principles of Atomic Layer Deposition

Figure 4.6: Temperature dependence of the growth-per-cycle (GPC) according to Sun-

tola[56]. The temperature region where the characteristic self-limited growth

keeps the GPC constant is referred to as ALD window.

temperature variations do not affect the deposition rate.

If the substrate’s temperature is too low, the energy supplied to the reactants is smaller

than the activation energy to have a catalytic effect and therefore no chemical reaction

occurs between surface and precursor molecule. If the temperature exceeds the activation

energy, the reactions might still be too slow to saturate the surface completely within

the exposition time. The growth-per-cycle decreases with respect to the ALD window

as indicated by branch (1) in Fig. 4.6. This behavior is most commonly investigated in

thermal ALD processes. Nevertheless, a lower substrate temperature can also show the

effect of an increase in growth rate as indicated by the dashed line of branch (2). In

this case, the lower temperature leads to condensation of precursor molecules or reaction

products on top of the surface. This results in lower quality films since the deposition

process does not base on the characteristic self-limiting ALD properties.

On the high temperature side, when temperature exceeds the ideal ALD window regime,

the growth rate usually decreases (3). In this case, the adsorbed precursor molecules

desorb, leaving behind unsaturated surface sites where no reaction partner is present

during the subsequent precursor exposition (cf. Fig. 4.1). In contrast, the GPC might also

increase at elevated temperatures as indicated by branch (4) in Fig. 4.6. This happens

in case of an undesired precursor decomposition which then condenses at the formed

monolayer[56].

The ideal growth-per-cycle dependence on substrate temperature is indicated by the solid

lines in Fig. 4.6.

4 Atomic Layer Deposition of ZnO

4.1.4 Deposition Rate

A theoretical model developed by Puurunen[57] relates the amount of material deposited

in one ALD cycle under ideal growth conditions, i.e. complete surface saturation in the

ALD window temperature range, on either size effects of the adsorbed precursor molecules

(steric hindrance of the ligands) or the number of reactive surface sites. Considering the

molecule size, in case of ZnO deposition the number of reactive hydroxide surface sites can

be considered much higher than the number of adsorbed monoethylzinc molecules. Hence,

the growth-per-cycle is mainly limited by steric hindrance effects of the ethyl ligands.

Since the original substrate surface and the deposited film usually show different chemical

compositions, the GPC is also expected to vary during initial atomic layer deposition.

The formation of the first monolayer can take several ALD cycles resulting in a reduced

growth-per-cycle. After the initial monolayer completed, the growth rate settles to the

constant value determinable from the ALD window in Fig. 4.6.

After introducing of the atomic layer deposition process, the following section will concen-

trate on the setup of the ZnO ALD reactor and the determination of its ideal deposition

parameters.

4.2 Reactor Design and Commissioning

Since its invention, a couple of different ALD reactor types have been developed. The most

commonly used reactor type is the so-called traveling-wave reactor where the precursor

is transported towards the sample by a continuous flowing inert carrier gas. The first

commercially available ALD reactor was the F-120 by Microchemicals Ltd. Nowadays,

further companies like Cambridge NanoTech or Picosun (founded by Tuomo Suntola, the

inventor of this technique), offer ALD reactors.

As in CVD reactors, distinctions are made between hot and cold wall ALD reactors. Since

their advantages and disadvantages are identically to those in CVD reactors, they are not

repeated here. More details are found either in Sec. 3.3 or in the reviews of Ritala and

Leskel¨

a[54] or George[53].

In addition to thermal ALD reactors, there are also plasma-enhanced ALD (PE-ALD)

reactors in use. As in plasma-enhanced CVD, those reactors are using a plasma to induce

chemical reactions that would not be possible just using thermal energy[53]. An advantage

4.2 Reactor Design and Commissioning

of PE-ALD is the possibility of reducing the substrate temperature without negative

influences on the film quality. This might be important using substrates that decompose

at higher temperatures like polymer substrates. On the other hand, the self-limiting

reactions are best fulfilled by thermal ALD, since in PE-ALD reactant decomposition can

cause problems with conformity[55].

In this work, a thermal ALD reactor for zinc oxide deposition has been designed. To

allow in situ growth and analysis of the grown films, the reactor is compatible to ultra-

high vacuum conditions and attached to the Integrated System (see Sec. 2.1).

4.2.1 UHV-ALD Reactor

Figure 4.7 shows schematically the setup of the UHV-compatible cold-wall ZnO-ALD

reactor build up in this work. Its key components are pneumatic Swagelok Atomic Layer

Deposition Diaphragm Valves which are additionally equiped with solenoid pilot valve

assemblies. These valves allow to achieve high-speed valve opening and closing times of

up to 5 ms. Each precursor and purging gas requires one ALD valve to control its intake

into the deposition chamber. Hence, 3 ALD valves are the minimum requirement for the

deposition of a compound formed by the reaction of two precursor gases separated by one

purging gas.

The ALD valves run in normally closed operation, i.e. they just open if they get a sig-

nal supplied by an Advantech USB-4761 8 channel relay control unit. This control unit

translates the commands of a software that has been written to actuate the ALD valves

into opening signals.

As mentioned above, this ALD reactor is attached to the Integrated System and oper-

ated under ultra-high vacuum (UHV) conditions to allow in situ growth and analysis

experiments. Hence, a turbo molecular pump and a rotary vane pump are necessary to

achieve those low pressures. The base pressure in this ALD system is typically in the

range of 5 ·10−9mbar. Not reaching lower pressures is mainly caused by residues of the

water used as oxygen precursor. Not in operation, the deposition chamber pressure is

determined by an ionization gauge. Running the atomic layer deposition process, the

pressure can shortly rise into the millibar region, during the purging steps. Since such

pressure fluctuations would destroy the hot filament of the ionization gauge, the pressure

during operation is measured by the Pirani gauge of a Balzers full range vacuum gauge.

4 Atomic Layer Deposition of ZnO

Figure 4.7: Sketch of the UHV-compatible ZnO-ALD reactor build up for this work. The

key components are discussed in the text.

One advantage of those Balzers pressure gauges is the possibility of constantly logging

the pressure in the deposition chamber. This can be useful to control process parameters

like valve opening times. The lower pressure limit of the Pirani is about 2 ·10−2mbar

and therefore still sufficient to measure the pressures in the deposition chamber during

precursor exposure.

One crucial part of the ALD reactor is the sample heater. In the beginning, a coiled

tungsten wire was used to heat the sample. This attempt was not very successful because

either the shock pressure during purge or deposited ZnO on the tungsten wire led to fail-

ure of the heater after a couple of weeks of operation. This problem was solved by using

a Boralectric heating element, which are build up of a graphite conductor surrounded by

a boron nitride ceramic. Those heating elements have high thermal shock resistance and

are inert to corrosive gases. Figure 4.8 shows the power input characteristics for different

temperatures in a regime interesting for atomic layer deposition. The substrate tempera-

ture is regulated automatically using a Eurotherm 815s PID controller. The temperature

is raised by 15◦C/min until the final temperature is reached to prevent thermal stress in

4.2 Reactor Design and Commissioning

Figure 4.8: Power consumption characteristics of the boron nitride / graphite Boralec-

tric heater in a temperature regime relevant for successful ZnO-ALD. In the

required temperature regime between 100 −300 ◦C the power consumption

shows a linear behavior.

the samples. The same applies for cooling down the samples after the zinc oxide has been

deposited.

4.2.2 Precursor Materials

As described in Sec. 4.1, two precursor gases and one purging gas is required for successful

ALD of zinc oxide. Argon is chosen as purging gas because of its inertness and its availabil-

ity at the Integrated System. The argon pressure in the gas supply is about 1.5−2 bar.

Setting up lower pressures could lead to diffusion of impurities from the environment into

the gas supply system. Because such high pressures would lead to a shut-down or even a

damaging of the turbo molecular pump, a short purging time of 25 ms was chosen. While

flushing the chamber, the pressure reaches values of around 1 mbar but decreases very

quickly after the purge.

Organometallic diethylzinc (DEZn) supplied by Sigma-Aldrich is chosen as the zinc pre-

cursor for the ZnO deposition. The ALD sources consist of a custom made two-branched

glass tube equipped with a CF-flange and an all metal angle valve for UHV-compatibility.

Diethylzinc is liquid at room temperature and very reactive with oxygen. Therefore, filling

the source had to be carried out in a glove box under argon atmosphere.

The standard oxygen precursor for ZnO ALD is water. A source identical in construction

with the DEZn source is filled with ultra-pure water of highest purity. Additionally, pure

4 Atomic Layer Deposition of ZnO

Figure 4.9: Organometallic diethylzinc (Zn(C2H5)2) molecule being composed of one zinc

atom and two ethyl (C2H5) ligands.

O2can be used as oxygen precursor as shown later in Chp. 6. The oxygen is provided by

a 1 l O2lecture bottle (99.998 vol %) controlled by an additional ALD valve.

To control the vapor pressure of the liquid precursors DEZn and water and to prevent

thermal degradation of the diethylzinc, both sources can be cooled independent of each

other using Peltier elements. Therefore, each source is equipped with a thermocouple and

Eurotherm 815s PID controllers to keep the temperature of each source constant. One

further advantage of the cooled DEZn source is the possibility of additional cleaning of

the liquid precursor by sublimation from the uncooled branch of the source into the cooled

one to minimize impurities.

Figure 4.10: Mass spectrum of diethylzinc fed into the atomic layer deposition chamber.

Besides the dominant peaks resulting from fragments of the molecule, no

contaminations of the DEZn are observable.

4.2 Reactor Design and Commissioning

Fig. 4.10 shows a mass spectrum of room temperature DEZn during introduction into

the deposition chamber. The spectrum has been recorded with a quadrupole mass spec-

trometer. According to Kuniya et al. the peaks arising around 122 m/e correspond to

DEZn[58]. Additional peak regions arise due to decomposition of the diethylzinc in the

mass spectrometer ion source into Zn(C2H5)+(93 m/e) and Zn (64 m/e) mainly by losing

their ethyl ligands (29 m/e). Even though gaining information from the mass spectra

about the DEZn is not easy it gives an idea about the purity in the system. For instance,

the amount of water (18 m/e) or hydroxides (17 m/e) is very low. Other contaminants

are not identified in the mass spectrum.

To prevent decomposition of the precursor, the sources are cooled as mentioned above. At

the same time the vapor pressure has to be sufficiently high to achieve complete saturation

on the surface in an adequate time scale. To get an idea about the temperature dependence

of the vapor pressure, it has been calculated using the Antoine equation:

log p=A+B

(T[K] + C) [Pa] (4.4)

A,Band Care parameter that can be experimentally determined[59], while Tis the

absolute temperature. An excellent online database is the NIST Chemistry WebBook[60]

providing a huge selection of data for different materials. For diethylzinc, Stull determined

the following values[61] being valid in the temperature regime of interest:

A= 4.41445, B= 1571.638, C=−34.978 between T= 250.7−391.0 K

Bridgeman and Aldrich determined the Antoine equation parameters for water[62]. There,

the parameters vary in the range of interest what has to be taken into account. Those

parameters are:

A= 5.40221, B= 1838.675, C=−31.737 between T= 273 −303 K

A= 5.20389, B= 1733.926, C=−39.485 between T= 304 −333 K

The vapor pressures have been calculated in a temperature regime between −20◦C and

60◦C for diethylzinc and water, respectively. The results are plotted in Fig. 4.11. Both

curves intersect at 8◦C having a vapor pressure of 10.72 mbar. Hence, 8◦C has been chosen

to be the optimal temperature the sources are kept at.

4 Atomic Layer Deposition of ZnO

Figure 4.11: Vapor pressures of diethylzinc and water calculated by the Antoin equation

(Eq. 4.4). At 8 ◦C both vapor pressures equal 10.72 mbar.

There are several zinc precursor materials available to deposit zinc oxide films by ALD

or CVD methods. Amongst others, ZnO was successfully produced using ZnCl2[63],

organometallic zinc acetate (Zn(CH3COO)2)[64] or dimethylzinc (DMZn, Zn(CH3)2)[65].

Ye et al. compared the structural properties of MOCVD deposited ZnO films using both,

diethylzinc and dimethylzinc as zinc precursor, respectively[65]. Both films showed pref-

erential growth orientation along the c-axis as X-ray diffraction experiments revealed. In

contrast to DEZn-grown films, the XRD peaks of the films deposited with DMZn exhibit a

larger peak widths. This indicates an improved structural quality of films deposited with

DEZn. In addition, Raman spectroscopy of DEZn-grown ZnO do not show any phonon

modes besides those of ZnO. In contrast, DMZn-grown ZnO films show additional phonon

modes assigned to carbon and hydrogen content in the zinc oxide. This clearly indicates

a much higher chemical and structural film quality using diethylzinc as zinc precursor.

This, and the matter of fact that DEZn was already successfully used for MOMBE ZnO

deposition in our group[21;51], made the decision for diethylzinc as zinc precursor in this

ALD reactor.

4.2.3 ALD Valve Response

In ALD, surface saturation is achieved very fast and hence, the ALD valves have to have

reliable response times in the millisecond range. In order to check the valve response on

the opening signal of the control unit, each ALD valve is additionally equipped with an

4.2 Reactor Design and Commissioning

Figure 4.12: (left) Actuator-position sensor response signal for a valve opening signal of

10 ms; (right) valve specific delay of response time for the DEZn valve.

electronic actuator-position sensor, transmitting a response signal to a data acquisition

unit. In addition to the opening and closing times the manufacturer specified with 5 ms

respectively, a significant delay in valve response has been found. On the left hand side

of Fig. 4.12 the response signal of an ALD valve is presented. The control unit delivered

a 10 ms opening signal whereas the valve actuator-position sensor gave notice of a valve

opening of 20 ms.

For deeper insight, the valve response times have been determined with respect to the

input signal of the control unit. As shown at the right part of Fig. 4.12, at very short

exposure times, the discrepancy between both signals rises until a maximum error of

10 ms is reached. For exposure times longer than 10 ms, the error remains constant.

Hence, 11 ms has been chosen the minimum exposure time for this particular valve, since

the error can easily be corrected by the ALD control software.

This response characteristic has been determined for each of the ALD valves, resulting in

minimum exposure times of 10 ms for the DEZn valve as just shown, 12 ms for the H2O

and O2valves, respectively and 13 ms for the Ar valve.

Fig. 4.13 shows the pressure evaluation during one typical ALD cycle as shown later

in Sec.4.3. The two prominent features in the upper part arise from the argon purge.

The maximum pressure usually reaches values of 1 −2 mbar and it takes about 5 s until

the Pirani detection limit of 2 ·10−2mbar is reached. Due to this pressure impulse, an

additional pumping step of 20 s has been implemented after the purging step for two

reasons: on the one hand, at least high-vacuum conditions are desired during precursor

exposure to prevent incorporation of impurities into the deposited film. On the other hand,

4 Atomic Layer Deposition of ZnO

Figure 4.13: (top) Pressure evolution in the reactor chamber during one ALD cycle; (a)

400 ms DEZn exposure (red) followed by 25 ms Ar purge (green); (b) 200 ms

H2O exposure (blue) followed by 25 ms Ar purge.

the turbo molecular pump would shut-down due to those high pressures if no additional

pumping step is performed.

Taking a closer look, it is even possible to check the pressure rising in the chamber during

precursor exposure as shown in Fig. 4.13 (a) for 400 ms DEZn and 200 ms H2O followed

by 25 ms Ar flush. The length of the valve control signal for DEZn exposure is indicated

by the red background in (a), whereas the signal for water exposure is indicated by a

blue background. Especially in case of DEZn one can nicely see the pressure rising until

a constant value of about 1.3·10−1mbar is reached.

Integrating the area of the pressure peaks can help controlling the reproducibility of the

ALD reactor: a constant precursor intake at each ALD cycle is required for saturation

of the sample surface and at the same time one can easily check whether the precursor

sources have to be refilled.

4.3 Growth Parameters for ZnO-ALD

After the ALD reactor has successfully started operation, the growth parameters for re-

producible ZnO atomic layer deposition have been determined as shown in this section.

Furthermore, the chemical and structural properties of the deposited ZnO films are char-

acterized.

As already mentioned in the Sec. 4.1, the basic process parameters for ideal self-limiting

atomic layer deposition are the exposure times of the precursors to completely saturate

the surface and the substrate’s temperature. To determine those parameters for ALD of

ZnO using diethylzinc and water as precursor materials, well-defined substrates have to

be used to prevent substrate-induced effects on the growth rate of the ZnO.

4.3.1 Substrate Preparation

Polished silicon in a (111)-orientation was chosen to be used as substrate material because

clean planar surfaces can be achieved relatively easily. To remove the thin natural oxide

layer on the silicon surface, the substrates are prepared in a wet-chemical process directly

before introduction into the Integrated System. Since the symmetry of the bulk crystal

is broken at the surface, the bonding situation of the surface atoms differs from those in

the bulk. This results in dangling bonds of the atoms located at the surface illustrated

by the red atoms in Fig. 4.14 (a). In addition, reconstruction processes of the atoms

occur at the surface to minimize the energetic situation of the system. The wet-chemical

preparation on the one hand removes the surface oxides but at the same time saturates

the dangling bonds by hydrogen atoms. This results in a stable, hydrogen-terminated

Si(111)-H substrate surface, preventing an immediate re-oxidation of the silicon during

transport to the Integrated System. Fig. 4.14 (b) presents the LEED pattern of one of

the prepared Si(111)-H substrates, showing an unreconstructed 1x1 surface structure.

The wet-chemical sample preparation process starts with degreasing the silicon succes-

sively in an acetone and ethanol ultrasonic bath at 40 ◦C. Afterwards, possible solvent

residues are removed by rinsing with ultrapure water. To obtain clean surfaces, a thick

oxide layer is formed containing impurity atoms located at the silicon surface. This is

achieved by treating the substrates in a boiling solution of 50 % sulfuric acid (H2SO4)

and 50 % hydrogen peroxide (H2O2) for 10 min at 80 ◦C. After rinsing in ultrapure water,

4 Atomic Layer Deposition of ZnO

(a)

(b)

Figure 4.14: (a) crystal structure of Si(111) 1x1, atoms with unsaturated dangling bonds

are shown in red; (b) LEED pattern of a hydrogen-terminated Si(111) 1x1

with hν= 61.6 eV

the oxide layer (and all impurities within) is removed in an ammonium fluoride (NH4F)

bath (40 %) at room temperature. At the same time the NH4F removes the silicon ox-

ide, it terminates the dangling bonds at the substrate’s surface with hydrogen resulting

in a clean Si(111)-H passivated surface. This hydrogen termination is sufficiently stable

to prevent an oxidation of the surface during dry blowing with nitrogen, mounting the

substrate on an sample holder and transferring it into the UHV system. Further details

of the wet-chemical preparation process are described elsewhere[66].

After transfer into the Integrated System, the substrate is characterized by means of

photoelectron spectroscopy for two reasons: first to make sure the cleanliness of the

prepared surface and to determine the intensity of the Si 2p photoemission peak prior to

4.3 Growth Parameters for ZnO-ALD

Figure 4.15: X-ray photoemission survey spectrum of Si(111)-H and the corresponding

detail spectra of: (a) F 1s, showing small amounts of fluorine due to NH4F

treatment; (b) O 1s; (c) C 1s; (d) Si 2p; (e) valence band

atomic layer deposition. This peak is used to calculate the ZnO thickness after ALD by

attenuation of the overlayer as described earlier in Sec. 2.2.1. The Si 2p spectrum shown

in Fig. 4.15 (e) also demonstrates the spin-orbit splitting, having an energy difference of

about 0.8 eV.

Figure 4.15 shows the survey spectrum of a Si(111)-H substrate prepared by the wet-

chemical process described above. The insets show detail spectra of possible surface

contaminants: fluorine (a), oxygen (b), and carbon (c). There are only very small amounts

of oxygen and carbon present at the silicon surface. Hence, we neglect any influences of

those traces on the initial growth of the ZnO.

As shown in Fig. 4.15 (a), a small amount of fluorine is present at the surface originating

4 Atomic Layer Deposition of ZnO

from the ammonium fluoride treatment. Annealing the sample to 150 ◦C leads to complete

removal of the fluorine.

4.3.2 Surface Saturation and ALD Window using DEZn and H2O

First, the necessary exposure times of the two precursor materials diethylzinc and water

had to be determined. This is crucial to achieve complete surface saturation and hence

self-limited atomic deposition as mentioned earlier. The easiest method to make sure that

the precursor exposure length is sufficient to saturate the surface is to determine the ZnO

deposition per ALD cycle. This growth-per-cycle (GPC) will saturate at a maximum

when the substrate is entirely covered with one monolayer of the precursor. To exclude

effects of the second precursor on the GPC, an exposure time is chosen that definitely

saturates the whole substrate.

The growth-per-cycle does not only depend on the amount of material that is provided

during an ALD cycle but can also be affected by the number of reactive surface sites

of the topmost atomic layer[57]. Since number and type of the reactive surface sites will

differ between the pure Si(111)-H surface and ZnO, the GPC is expected to change during

initial growth of the first completed ZnO monolayer. To avoid those surface effects, a very

Figure 4.16: (a) GPC dependence on initial ZnO monolayer formation; (b) Normalized

peak intensities plotted against ALD cycle. During the first approx. 10 ALD

cycles the first ZnO monolayer is formed. Afterwards ideal layer-by-layer

growth is achieved (cf. Fig 3.4).

4.3 Growth Parameters for ZnO-ALD

thin layer of about 1 nm ZnO has been deposited prior to the actual experiment. This

excludes any nucleation effects on the growth rate. As can be seen from Fig. 4.16 (a) it

takes about 10 ALD cycles to complete the first ZnO monolayer. Comparing Fig. 4.16 (b)

and Fig. 3.4 indicates that after the first monolayer has formed, an ideal layer-by-layer

growth is achieved as expected in atomic layer deposition. The thickness of the initial

ZnO layer still allows Si 2p photoelectrons of the substrate to travel through the overlayer

without much attenuation. After characterization with XPS, five additional ALD cycles

were grown and the film thickness has been determined by the attenuation of the Si 2p

photoelectrons (cf. Sec.2.2.1). The GPC is simply the gain in film thickness divided by

the number of ALD cycles deposited. Photoemission spectra of the involved elements

are presented in Fig. 4.17. The left spectra show the Si 2p photoemission lines being

attenuated by the ZnO overlayer. The spectra in the middle and at the right hand side

show the rise of the Zn 2p3/2and the O 1s photoemission, respectively. The O 1s peak

consists of two separate photoemission peaks chemically shifted by approximately 1.5 eV.

Detailed analysis of this second component will be the focus of Sec. 6.3.

The results of the surface saturation experiments are shown in Fig. 4.18. In case of DEZn

saturation, a H2O exposure time of 200 ms was selected to be sufficient to saturate all

monoethylzinc surface adsorbates. All experiments were carried out at 210 ◦C.

An exposure time of diethylzinc of 400 ms has been determined to be sufficient to achieve

Figure 4.17: X-ray photoemission spectra of Si 2p, Zn 2p3/2and O 1s. All spectra are

excited by Al Kαradiation of 1486.7 eV. The attenuation of the silicon

signal contains information about the ZnO overlayer thickness.

4 Atomic Layer Deposition of ZnO

Figure 4.18: Growth-per-Cycle (GPC) dependence over precursor exposure of diethylzinc

and water. At exposure times of 400 ms in case of DEZn and 50 ms in

case of H2O, the surface is completely saturated, indicated by the constant

deposition rate of about 3.0˚

A/cycle.

total saturation of the substrate’s surface. The amount of ZnO deposited per ALD cycle

approaches about 3.0˚

A/cycle. The same value is reached in case of keeping the DEZn ex-

posure time constant at 400 ms and varying the H2O exposure time. 50 ms H2O exposure

are already sufficient to completely react with all available surface sites. To guarantee

reproducibility of the atomic layer deposition process, exposure times of 400 ms for DEZn

and 200 ms in case of H2O were chosen for all upcoming ZnO ALD cycles.

The second crucial parameter of ZnO atomic layer deposition is the temperature of the

substrate. To determine the temperature range of ideal self-limiting ALD growth, a thin

ZnO layer was deposited onto the Si(111)-H for the same reasons as in case of the satu-

ration experiments. Again, five ZnO-ALD cycles were deposited afterwards and the film

thickness was determined by means of XPS. These experiments have been carried out

for different well-defined substrate temperatures to determine the growth-per-cycle de-

pendence on deposition temperature. In Fig. 4.19 the GPC is plotted against deposition

temperature. As in theory (see Fig. 4.6) three temperature regimes can be classified in

Fig. 4.19. At low temperatures, low reaction kinetics, i.e. slow reactions of the precursor

molecules with the surface species, result in a decreases GPC. With increasing temper-

4.3 Growth Parameters for ZnO-ALD

Figure 4.19: Growth-per-cycle dependence on deposition temperature. The temperature

regime showing constant deposition rate and hence temperature-independent

self-limiting growth is referred to as the ALD window.

ature, the GPC increases until it reaches a value of 3.0˚

A/cycle. This point defines the

beginning of the ALD window, the regime of temperature independent self-limited ALD.

The ALD window using the precursor combination DEZn and water is located between

200 ◦C and 225 ◦C. Hence, 210 ◦C is determined as the optimal deposition temperature

of this ALD reactor. At temperatures higher than 225 ◦C, desorption of the precursors

molecules adsorbed on the surface increases, which again results in a decrease in growth

rate.

To verify this growth rate, a ZnO film of 200 ALD cycles has been deposited and a

cross-section scanning electron microscopy (SEM) image has been taken. Analysis of this

image resulted in a film thickness of approximately 42.4 nm. In combination with the

amount of deposition cycles, a growth rate of 2.1˚

A/cycle has been determined. Both

growth rates match to literature values of 1.9−2.8˚

A/cycle previously reported by other

groups[14;67–71] for the ZnO atomic layer deposition using DEZn and H2O as reactants.

One has to keep in mind that both methods offer sources of error, namely imprecise film

thickness determination from the SEM images and errors during data processing and

IMFP calculation in case of PES. From now on, the value determined by photoelectron

spectroscopy is used to express the film thickness, since it can be determined in situ and

without big effort.

4 Atomic Layer Deposition of ZnO

Assuming the zinc oxide growing in a preferred orientation along its c-axis (shown later

in Sec. 4.4) results in a monolayer thickness for ZnO of 5.2˚

A/cycle. Hence, a deposition

rate of 3.0˚

A/cycle corresponds to a growth of ≈60 % of an entire monolayer, limited by

steric hindrance effects caused by the ethyl ligands of the DEZn. Thus, one can assume

a complete ZnO monolayer to be formed each two ALD cycles.

An additional information about the ALD process can indirectly be achieved from the

ALD window in Fig. 4.19: due to the existence of the self-limited growth character it is

proofed that the argon purging step is sufficient to remove all residual precursor gases

from the reactor and therefore prevent CVD-like deposition from the gas phase during

the ALD process.

4.4 Chemical and Structural Characterization of

ALD-ZnO films

After determination of the optimal growth parameter for self-limited atomic layer depo-

sition of ZnO, chemical and structural film properties of the deposited layers are inves-

tigated. Therefore, a ZnO layer of 50 ALD cycles and an estimated film thickness of

dZnO = 15 nm is deposited onto Si(111)-H. Since this thickness is much larger than the

Figure 4.20: Photoemission spectra of an ALD-ZnO layer much thicker than the escape

depth of the photoelectrons. All spectra are recorded using Al-Kαradiation

of 1486.7 eV. Quantification of the spectra results in a [Zn]:[O] ratio of 0.97

and an zinc Auger parameter of 2009.68 eV.

4.4 Chemical and Structural Characterization of ALD-ZnO films

escape depth of the photoelectrons, possible effects of the Si(111)-H/ZnO interface can

be ruled out. Fig. 4.20 plots the photoemission spectra of the Zn 2p3/2and O 1s photo-

electrons and the Zn LMM Auger electron peak. After background removal, peak fitting

results reveal a [Zn]:[O] ratio of 0.97, i.e. the ZnO layer is close to the stoichiometric

composition of 1.00. At this point it is reminded that for compositional analysis the sum

of both O 1s components is taken into account.

Furthermore, the Auger parameter has been calculated to investigate the chemical envi-

ronment of the zinc component. Peak fitting results in a Zn 2p3/2photoelectron binding

energy of BE = 1021.71 eV. In combination with the kinetic energy of the Zn L3M4,5M4,5

Auger electrons of KE = 987.97 eV the Zn Auger parameter results in:

AP (Zn) = BE (Zn 2p3/2) + KE (Zn L3M4,5M4,5) (4.5)

= 1021.71 eV + 987.97 eV

= 2009.68 eV

This Auger parameter is in excellent agreement with the zinc Auger parameter of ZnO

observed previously for the MOMBE ZnO deposition[21]. Combination of the Auger pa-

rameter and the results of the composition analysis clearly reveal that the ALD reactor

has the ability to deposit ZnO layers of high quality.

Figure 4.21: Grazing-incidence X-ray diffraction data of an approx. 50 nm thick ALD-

ZnO layer. In comparison with XRD data of ZnO powders[72], a preferred

orientation along the c-axis is assumed[73].

4 Atomic Layer Deposition of ZnO

Since this work is focused mainly on the electrical and chemical properties of ZnO grown

by atomic layer deposition, only brief analysis of the structural properties is presented in

this work. To characterize the structural quality of the ALD-grown ZnO films, grazing-

incidence X-ray diffraction (GI-XRD) experiments have been carried out. Using GI-

XRD, the thickness of the film under investigation can be significantly smaller than in

conventional X-ray diffraction experiments. Anyhow, the thickness should exceed 50 nm

to get an satisfactory signal-to-noise ratio. Hence, a ZnO film of 200 ALD cycles, resulting

in roughly 60 nm ZnO film thickness, has been prepared again using Si(111)-H as substrate

material. The results of the GI-XRD using Cu-Kαradiation are shown in Fig. 4.21. The

most prominent feature in Fig. 4.21 can be assigned to a (002) reflex[73]. Most references in

XRD databases[72] usually provide structural information of statistically oriented powders

without any preferred growth orientation. Nevertheless, the line positions of those powders

match to those of the ALD-ZnO film presented in Fig. 4.21. This verifies the ZnO is

growing in wurtzite structure, as expected. The intensity distribution of the Bragg peaks

differs from those in the powders, giving an indication that the films grow along a preferred

orientation. Matsubara et al. present XRD measurements of epitaxially deposited ZnO

films on LiNbO3substrates. Their XRD measurements only result in one single Bragg

peak related to a (002) reflection. Since the (002) peak of the ALD-ZnO films is the

most prominent one, we assume those layers to grow in a preferred orientation along the

c-axis.

(a) (b)

Figure 4.22: (a) Low-energy electron diffraction (LEED) pattern of the clean Si(111)-

H 1x1 substrate; (b) LEED spots disappear after deposition of ALD-ZnO

since the lattice mismatch of the two materials avoids highly-ordered growth

of the ZnO layer.

4.5 Initial Growth of ALD ZnO on Si(111)-H

Furthermore, ALD-ZnO films have been investigated in situ by low-energy electron diffrac-

tion (LEED). Prior to ZnO deposition, the Si(111)-H substrate has been characterized to

validate its structural quality. Fig. 4.22 (a) verifies the Si(111)-H substrate having a 1x1

surface orientation[74] without showing any superstructure spots. After 50 ALD cycles,

the LEED pattern of Fig. 4.22 (b) does not show any reflexes. Instead, the background

intensity has risen. Hence, even though the films grow in a preferred orientation along the

c-axis as resulted from the GI-XRD measurements, the films do not show a long-range

order resulting in LEED spots. This is also expected, since the lattice constants of (111)-

oriented Si (a= 3.84 ˚

A[74]) and ZnO along its c-axis (a= 3.25 ˚

A[20]) differ significantly,

resulting in lattice mismatch of 15.4%

4.5 Initial Growth of ALD ZnO on Si(111)-H

In general, films produced by atomic layer deposition are usually in the nanometer range.

Hence, the interface between substrate and ZnO deposit cannot be neglected anymore.

The thinner the deposited film, the more chemical or structural changes at the interface

influence the electronic properties of the material and, depending on its application, the

efficiency of the device. Therefore, the focus of the following section is on the initial

ALD-ZnO growth on Si(111)-H substrates and the properties of its interface.

To investigate the initial growth on silicon, monolayers of zinc oxide are deposited stepwise

onto the substrate. As determined in Sec. 4.3, it takes about two ALD cycles until one

complete ZnO monolayer is formed. After deposition of each monolayer, the sample is

characterized by means of photoelectron spectroscopies. These steps are repeated until the

deposited layer thickness exceeds the escape depth of the silicon photoelectrons. Fig. 4.23

plots the photoelectron spectra of the involved elements oxygen, zinc and silicon after

each deposition step.

The spectra in the front of Fig. 4.23 belong to the pure Si(111)-H substrate. The Si 2p

peaks have a FWHM of about 0.5 eV and there is no indication of a surface oxide. The

intensity of the Si 2p photoelectron peak is determined for calculation of the ZnO layer

thickness as described above. In Sec. 4.3.2 it has already been mentioned that it takes

several ALD cycles until the initial growth of ZnO on Si(111)-H starts. Therefore, in the

first step 5 ALD cycles were deposited, resulting in a calculated film thickness of 0.96 ˚

4 Atomic Layer Deposition of ZnO

Figure 4.23: Photoemission spectra of (left) O 1s; (middle) Zn 2p3/2; (right) Si 2p during

monolayer-wise atomic layer deposition of ZnO. While the substrate signal

constantly decreases during deposition, both film elements rise until they

reach a maximum. No oxidation of the silicon is observed during deposi-

tion of ZnO with ALD. All spectra are recorded using Al Kαradiation and

background signals were removed.

corresponding to an initial surface coverage of the deposit of about 0.33 %. In step two, 4

additional ALD cycles were deposited to complete this first monolayer. Afterwards, two

ALD cycles were sufficient to grow ZnO in monolayer steps. The spectra in Fig. 4.23 clearly

show the attenuation of the substrate’s Si 2p photoemission signal until it completely

disappears after 27 x ALD. The corresponding film thickness after this deposition step

is about 45 ˚

A. At the same time, the photoemission peaks of both film elements zinc

and oxygen rise until they reach a maximum. This maximum is achieved after about

23 x ALD while there is still a small silicon peak visible. This might occur due to the

differences in the inelastic mean free paths (IMFP) of the contributing elements. While

the Al Kαexcited Si 2p photoelectrons exhibit high kinetic energies of 1388 eV resulting

in an IMFP of about 30 ˚

A, the kinetic energies of the O 1s and Zn 2p3/2electrons are

significantly lower. Their IMFPs amount to 24.2˚

A and 16.9˚

A, respectively, resulting in

a slightly reduced information depth in comparison with Si 2p.

The resulting peak intensities of the spectra plotted in Fig. 4.23 are determined and

normalized to 100 % with respect to the most intense peak of each element as shown in

Fig. 4.24. To illustrate the peak intensity behavior during initial growth, the results are

plotted against the number of atomic layer deposition cycles in Fig. 4.24 (a) and the ZnO

4.5 Initial Growth of ALD ZnO on Si(111)-H

Figure 4.24: Photoemission peak intensity development of initial ZnO atomic layer de-

position. Peak intensities for each element are normalized to 100 % of the

maximum intensity. Peak intensities are plotted against deposition cycles (a)

and ZnO overlayer thickness (b), respectively.

overlayer thickness in Fig. 4.24 (b). Both, initial ZnO monolayer formation and saturation

of the zinc and oxygen signals at a maximum are clearly noticeable. The attenuation of

the silicon substrate signal again matches to the expected layer-by-layer growth mode

after the initial ZnO monolayer was formed.

Figure 4.25 (a) presents the regional spectrum of the Zn 2p3/2photoemission line after

the last deposition step. The black circles represent the measured data after removing the

secondary electron background by a Shirley routine. Fitting the peak by a Voigt function,

a convolution of a Gaussian and Lorentzian profile, ends up in the peak fit shown as a solid

red line. The resulting error is illustrated by the green line. The zinc spectrum shows no

indication for a second chemical component in the ZnO layer. In addition, the Zn 2p3/2

photoemission contains information about the crystalline quality during initial growth.

Therefore, the peak full width at half maximum (FWHM) of the Zn 2p3/2photoemission

line was determined after each deposition step and plotted as a function of ZnO thickness

in Fig. 4.25 (b).

The peak width can be taken as an indication of the order in the crystal. The broader

the photoemission peak, the more disordered the film[75]. Fig. 4.25 (b) shows a signifi-

cant decrease from an initial value of 1.72 eV after the first deposition step down to a

constant value of 1.53 eV. This indicates an increased amount of disorder in the film at

4 Atomic Layer Deposition of ZnO

Figure 4.25: (a) Peak fitting (red line) of the Zn 2p3/2photoemission line (black circles)

revealing one single zinc component in the ZnO film; (b) FWHM dependency

on deposited ZnO film thickness. The peak width is giving an indication of

the disorder at particular film thicknesses.

the ZnO-Si(111) interface, while the chemical environment becomes more homogeneous

with increasing film thickness. This can cause considerable effects on device efficiencies

as the interface disorder may induce increased charge carrier scattering lowering their

mobilities[75].

While the Zn 2p3/2photoelectrons do not show significant changes during initial growth,

the opposite is true in case of the oxygen component. As mentioned earlier, the O 1s

does not only show one but two chemical components as readily identifiable in Fig. 4.23.

To clarify its extensive change during initial growth, the O 1s photoemission spectra at

different deposition stages are plotted in Fig. 4.26. Three different deposition steps are

chosen to illustrate the O 1s behavior with increasing film thickness: a very thin, not

even completed film corresponding to the first deposition step is shown in Fig. 4.26 (a).

Fig. 4.26 (b) represents an intermediate thickness regime of about 4 ˚

A, while the spectrum

shown in Fig. 4.26 (c) belongs to a film thickness of 45.45 ˚

A, where the initial ZnO growth

already finished.

To identify the two different chemical components present in the crystal, the determined

binding energies are compared to reference work by other groups. Dupin et al. published

a study on various metal oxides, amongst others zinc oxide[76]. Almost all metal oxides

show two oxygen components with the main component being O2−ions bound to the

4.5 Initial Growth of ALD ZnO on Si(111)-H

Figure 4.26: Evolution of the O 1s photoemission during initial growth. Two chemical

components are observed, that are related to [O2−] anions forming ZnO (dark

blue) and an [O−OH] component usually attributed being located at the sur-

face of the ZnO (light blue).

metal ion. In case of ZnO, those O2−ions that are bound to the zinc atoms have binding

energies around 530.5 eV.

The second oxygen component is usually attributed to hydroxide oxygen (-OH)[76] mainly

located at the surface of the ZnO film. Their O 1s electrons exhibit a chemical shift of

∆BE ≈1.5 eV. This especially fits for ZnO-ALD using H2O as oxygen precursor, where

the reaction model predicts a hydroxide terminated surface after each deposition cycle

(cf. Eq. 4.3).

Hence, all three O 1s spectra were fitted with two oxygen components, as labeled in

Fig. 4.26 (c). A third component related to water molecules weakly adsorbed at the sur-

face around binding energies of 533.5 eV[77] is not observed. During formation of the first

ZnO monolayer, both oxygen components show equal intensities. There is no evidence

for an oxidation of the silicon. On the one hand, the formation of SiO2would result in

a photoemission peak at 103.5 eV[78] and on the other hand, a third oxygen component

would be observed in Fig. 4.26 (a) in an energy range between 532.5−533.2 eV[79]. Fur-

ther deposition would decrease an O 1s signal originating from SiO2in Fig. 4.23 (b) until

it vanishes completely for thick ZnO films in Fig. 4.26 (c). This is not observed and in

4 Atomic Layer Deposition of ZnO

combination with the absence of an Si 2p photoemission signal at 103.5 eV, an oxidization

of the Si(111)-H substrate can be ruled out.

Peak fitting results show no significant change in peak width ([O2−] = 1.07 eV and

[O−OH] = 1.97 eV) or an increased chemical shift between those two oxygen components

(∆BE = 1.25 eV) during initial growth. Instead, Fig. 4.23 (b) and (c) show a continuous

increase of the [O2−] component with respect to the [O−OH]. Plotting the [O−OH] : [O2−]

development over the corresponding film thickness shows a constant decrease of the in-

tensity ratio. During deposition of the first ALD cycles both components are present to

same amounts. After initial monolayer formation the intensity of the [O−OH] decreases

until it settles at 40 % of the [O2−] intensity, as shown in Fig. 4.27 (a).

Deeper insight into the initial atomic layer deposition of ZnO on Si(111)-H is obtained by

composition analysis after each deposition step. As already mentioned in Sec. 2.2.1, the

total O 1s intensity is used for determination of the film composition. Hence, both oxygen

components contribute to the calculated [Zn]:[O] ratio. The determined [Zn]:[O] ratio is

illustrated in Fig. 4.27 (b) plotted against the calculated film thickness. Ideal stoichiome-

try of [Zn]:[O] = 1.00 is indicated by the dotted line. During initial monolayer formation,

the film is extremely oxygen-rich with a calculated composition of [Zn] : [O] ≈0.5. After

completing the first monolayer, a sudden jump is observed in the [Zn]:[O] ratio. With in-

creasing film thickness, the excess amount of oxygen incorporated into the ZnO decreases,

Figure 4.27: (a) Intensity ratio of the two oxygen components O2−and O−OH plotted

against ZnO thickness; (b) Thickness dependency of the composition. The

interface is highly oxygen-rich, even though no silicon oxide formation is

observed.

4.5 Initial Growth of ALD ZnO on Si(111)-H

ending up in a composition that saturates at [Zn] : [O] = 0.91. The films deposited in

monolayer steps show a slight zinc deficiency, while those films deposited in one single

step are close to stoichiometry as shown earlier. Since all deposition parameters were kept

unchanged, there are only two possible explanations. Both are related to a removal of

surface hydroxides, either by the raise in substrate temperature prior to the next atomic

layer deposition step or by irradiation damage during surface analysis.

Additional information about the local chemical environment during initial deposition

is obtained by analysis of the zinc Auger parameter. The Zn Auger electron spectra

corresponding to each deposition step are presented in Fig. 4.28 (a). To determine the

kinetic energy of the Zn L3M45M45 peak, a single Gaussian peak has been fitted. Combi-

nation of the kinetic energy of the Auger electrons and the binding energy of the Zn 2p3/2

photoelectrons as given in Eq. 2.4 results in its modified Auger parameter. Fig. 4.28 (b)

illustrates its thickness dependency during initial growth on Si(111)-H. Comparison of the

determined Auger parameter with databases and values earlier reported by other groups

yield information about the local chemical environment of the zinc after each ZnO atomic

layer deposition step.

After five ALD cycles, i.e. at stages where the first ZnO monolayer is not yet completed,

the Auger parameter amounts 2010.03 eV. The corresponding binding and kinetic energies

Figure 4.28: (a) Zn L3M45M45 Auger electron spectra during initial ZnO deposition; (b)

Auger parameter development with increasing film thickness.

4 Atomic Layer Deposition of ZnO

Compound Auger Parameter BE (Zn 2p3/2) KE (Zn L3M45M45)

ZnO 2009.5−2011.0 eV 1021.2−1022.5 eV 987.7−988.9 eV

Zn(OH)22009.2 eV 1022.7 eV 986.5 eV

Zn 2013.4−2014.4 eV 1020.8−1022.1 eV 991.8−992.5 eV

Table 4.1: Reported Zn Auger parameter[29] for different compounds potentially present

during initial ZnO atomic layer deposition on Si(111)-H substrates.

are BE (Zn 2p3/2) = 1022.23 eV and KE (Zn LMM)= 987.80 eV, respectively. After

completing the first monolayer, a sudden decrease to 2009.64 eV (BE = 1022.06 eV; KE =

987.58 eV) is observed, followed by a slight increase up to a film thickness of about 25 ˚

Additional ZnO deposition does not affect the Auger parameter, remaining constant at

2009.75 eV (BE = 1021.83 eV; KE = 987.92 eV). Even if the results of the composition

analysis suggest a formation of zinc hydroxide (Zn(OH)2) at very early stages of the

ZnO deposition, Auger parameter analysis disagrees. Table 4.1 summarizes all reported

Zn Auger parameter values for zinc compounds possibly present at the ZnO-Si(111)-H

interface. Even though there is only one reported Auger parameter available for Zn(OH)2,

there is a significant difference to the measured value of ∆AP = 0.8 eV. The measured

zinc Auger parameter rather speaks for the formation of oxygen-rich zinc oxide after

completing the first deposited monolayer. The Auger parameter also indicates this initially

formed monolayer to be ZnO. With increasing film thickness, the amount of surplus oxygen

decreases as composition analysis pointed out. This is also observed in the development

of the Auger parameter. Comparing the literature values for ZnO and Zn(OH)2reveal a

decrease in Auger parameter with increasing amount of oxygen. Hence, the increase in

measured Auger parameter with increasing film thickness indicates a decrease of oxygen

the thicker the deposited ZnO films. This is in total agreement with the rising [Zn]:[O]

ratio in this thickness region.

All individual results presented in this section help to develop a phenomenological model

of the initial atomic layer deposition of ZnO on Si(111)-H. An attempt to illustrate the

initial ZnO monolayer formation on Si(111)-H is made in Fig. 4.29. Before deposition of

the first ALD cycle, the substrate’s surface is assumed to be entirely hydrogen-terminated.

However, it is obvious that this hydrogen-termination has to be removed to allow the for-

mation of a stable bond between silicon and zinc oxide. Hence, either the diethylzinc or

the water has to remove the hydrogen. For reason shown in a moment, we assume this to

occur during the water exposure. Hence, as shown in the top left part of Fig. 4.29, the

original hydrogen-terminated surface remains unchanged during the first DEZn exposure.

4.5 Initial Growth of ALD ZnO on Si(111)-H

Figure 4.29: Illustration of the developed initial growth model of ZnO on Si(111)-H for

the first two ALD cycle half-reactions.

There is no reactive site to bond with and all DEZn is removed by the subsequent argon

purge. Afterwards, the first water pulse is introduced into the deposition chamber. How-

ever, few water molecules remove the hydrogen-termination of the silicon surface. This

would end up in a hydroxylation of some silicon surface sites as shown at the top right

part of Fig. 4.29.

At this stage, an investigation of the sample by means of photoelectron spectroscopy

would not show big differences to the hydrogen-terminated silicon substrate. Only small

amounts of oxygen might be observed if not below the detection limit of XPS.

The second ALD cycle again exposes the surface with DEZn which now finds some reac-

tive sites in terms of surface hydroxides. As assumed in the standard model of ZnO ALD

(cf. Sec. 4.1.1), one ethyl ligand will undergo a bond with the hydrogen atom forming

gaseous ethane. After this half-reaction has finished, part of the surface is covered by

monoethylzinc (ZnEt), while the rest is still hydrogen-terminated. This is illustrated at

the bottom left of Fig. 4.29.

While the second ALD cycle is completed by exposing the substrate to water (bottom

right), two different reactions do occur. At the one hand, the water removes the remain-

ing ethyl ligand of the monoethylzinc by exchanging one hydrogen atom. On the other

hand, the water will further remove hydrogen from the original Si(111)-H surface. Only

for reason of illustration, a complete removal of the hydrogen-termination is assumed in

the bottom right part of Fig. 4.29. After this second cycle, the entire surface is terminated

4 Atomic Layer Deposition of ZnO

by hydroxides, as the standard model of ZnO atomic layer deposition assumes.

X-ray photoelectron spectroscopy after this cycle would show a completely different pic-

ture as after the first ALD cycle. On the one hand, the adsorbed zinc atoms would produce

a photoemission peak. On the other hand, two oxygen components would be observed:

one originating from the oxygen atoms bound to silicon and zinc and one originating from

the hydroxides, the latter more intense than the first. Further atomic layer deposition

would lead to more incorporation of [O2−] forming ZnO, while the amount of hydroxides

will remain almost constant. This might be true if it mainly terminates the surface and

only small amounts are incorporated into the crystal. This would lead to a constant

decrease of the [O−OH]:[O2−] ratio with increasing film thickness, until it saturates at a

constant level. This is exactly what is observed in Fig. 4.27 (a).

Additionally, the model presented in Fig. 4.29 results in a very oxygen-rich initial film,

observing a composition of [Zn] : [O] 1.00 during formation of the initial monolayer.

Increasing film thickness reduces the effect of the surplus oxygen at the silicon surface,

resulting in an increase of the [Zn]:[O] ratio as observed in Fig. 4.27 (b).

As mentioned above, it is assumed that the hydrogen-termination of the silicon is removed

during the water exposure and not affected by the DEZn. If this would not be the case

and the DEZn would remove the H-termination to form stable bonds to the silicon, the

[Zn]:[O] analysis would result in a stoichiometric or even zinc-rich initial monolayer. Since

this is not observed, the removal of hydrogen is assigned to the H2O, possibly by formation

of H2reaction products.

4.6 Summary

In this chapter, the characteristic properties of atomic layer deposition were introduced,

followed by a detailed description of the design, assembly and commissioning of an UHV-

compatible ALD reactor. In this work, ultra-high vacuum conditions are necessary for the

in situ deposition and analysis experiments, without any modification of the material’s

surfaces under investigation.

Furthermore, the successful atomic layer deposition of zinc oxide has been demonstrated,

using the widely-used precursor combination of diethylzinc and water as reactants. The

existence of the so-called ALD window, i.e. a temperature-regime where the deposition

rate remains constant, indicates an self-limited atomic layer deposition process. In addi-

tion, photoemission studies of the stepwise deposition of ZnO monolayers on hydrogen-

4.6 Summary

terminated silicon show ideal layer-by-layer growth of the material.

After investigation of the film properties that reveal the high quality of the ZnO grown

by the ALD system, the initial growth of ZnO on Si(111)-H has been determined by in

situ photoelectron spectroscopies. Analysis of the [Zn]:[O] ratio and the combined Auger

parameter of the zinc component show an oxygen-rich initial growth that turns into sto-

ichiometric ZnO during the growth of the first 3 nm. It is worth mentioning that no

oxidation of the silicon substrate has been observed during the photoemission studies.

All things considered, we showed that we are able to deposit high-quality ZnO films with

the accuracy of an atomic monolayer. Hence, the ALD reactor designed for this work is

suitable for further experiments on the interface formation of ZnO and epitaxial CuInSe2

(112), presented in the following chapter.

5 ALD-ZnO as Cd-free Buffer Layer for

Chalcopyrite Solar Cells

One of the main interests in chalcopyrite solar cell research is to substitute the CdS buffer

layer by a cadmium-free alternative for junction formation with the p-type chalcopyrite

absorber. As extensively discussed in the introductory chapter (see Sec. 1.1.2), one of the

main motivations is the toxicity of Cd, which might lead to legal regulations in different

countries, such as Germany[5]. On the other hand, it is technologically advantageous to

replace the chemical bath deposition (CBD) by a vacuum deposition process, resulting in

a continuous in-line vacuum production of the solar cell device.

Various materials are investigated as alternative buffer layer candidates. Most promising

results are achieved by In2S3and Zn-based materials such as ZnS, ZnSe, or ZnO[5;10].

Often even higher efficiencies are achieved if the buffers additionally contain some oxygen,

most often in terms of hydroxides. These materials were deposited by several deposition

techniques besides the standard chemical bath deposition, as for instance chemical vapor

deposition methods, sputtering, or atomic layer deposition. An additional advantage to

the cadmium issue is their increased band gap with respect to CdS. This reduces the

absorption of photons in the spectral region between 350 −550 nm which can result in an

increase of the solar cell’s efficiency. Also ZnO deposited directly on chalcopyrite absorbers

showed some good results in terms of device efficiency[5;10]. Especially the role of oxygen

in the buffer layer material is not well understood, yet.

While all of those studies investigated the alternative buffer materials in terms of the

efficiency of the fabricated devices, our approach is focused on the chemical and electronic

properties of the interface between the chalcopyrite absorber and the buffer material. To

understand the fundamental properties of this contact, all our experiments are carried

out on well-defined single crystalline chalcopyrite absorbers, grown in the technologically

relevant (112) orientation. The (112) orientation was chosen, since it is the preferred

growth orientation of polycrystalline chalcopyrites used in solar cell production.

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

In an earlier study of ZnO deposited by MOMBE (cf. Sec. 3.4) on CuInS2(112), it was

found evidence for an intrinsically formed ZnS layer at the interface of both materials[21].

This layer might improve the alignment of the electronic bands at the interface. Similar

experiments are part of the thesis of Hofmann, where the interface of MOMBE-ZnO and

CuInSe2(112) is investigated[9]. As observed in the case of CuInS2, the chalcopyrite anion

forms an intrinsic buffer layer with the zinc atoms, forming an intrinsic ZnSe layer at the

interface of the CuInSe2and the ZnO[22].

In this section, the interface formation of CuInSe2(112) and ZnO deposited by atomic

layer deposition is investigated. This is motivated mainly by two reasons: first, the self-

limited ALD growth mode allows controlled deposition of the ZnO in monolayer steps.

In combination with in situ photoelectron spectroscopy analysis, it is the best method to

determine the band alignment and to investigate the formation of intrinsic interface layers.

Secondly, the significantly lower substrate temperature of the atomic layer deposition

process allows to investigate the influence of the substrate temperature on the interface

layer formation. While during MOMBE ZnO deposition the sample’s temperature usually

is 400 ◦C, the ideal substrate temperature during ZnO-ALD was determined being 210 ◦C,

as shown in the previous chapter. The influence of this strongly decreased substrate

temperature on the initial growth of the ZnO on CuInSe2(112) is one of the key questions

of this chapter.

5.1 Epitaxial CuInSe2(112) Absorber Substrates

As mentioned above, highly-ordered epitaxial CuInSe2absorber films are used to investi-

gate the fundamental properties of the CuInSe2|ZnO interface. All films are prepared

by molecular beam epitaxy (see Sec. 3.2.2) in a deposition chamber attached to the Inte-

grated System. This allows real in situ deposition and analysis, i.e. all experiments are

carried out under ultra-high vacuum conditions without any environmental influences on

the samples.

To grow epitaxial films, proper substrates with similar lattice constants are required. In

case of CuInSe2in (112) orientation, GaAs (111)A (i.e. Ga-terminated) substrates are

used. Before they are introduced into the Integrated System, the natural oxide is removed

and the surface sulfur-terminated to prevent any contamination during transfer into the

vacuum system. The procedure of GaAs preparation is described in detail elsewhere[80].

The temperature of the evaporation sources is controlled by PID controllers and set to

5.1 Epitaxial CuInSe2(112) Absorber Substrates

Figure 5.1: Photoemission survey spectrum of an CuInSe2(112) absorber layer epitaxially

grown on GaAs (111)A. Some of the important photoemission lines of this work

are labeled in red (photoelectrons) or blue (Auger electrons), respectively.

TCu = 1040 ◦C, TIn = 900 ◦C and TSe = 210 ◦C, respectively. For CuInSe2deposition,

the substrate temperature is raised to 525 ◦C under selenium atmosphere. This leads to

a replacement of the sulfur surface-termination by a selenium-termination. At such high

temperatures, excess Se does not stick to the surface and reevaporates immediately. After

the deposition temperature is reached, the shutters of the Cu and In sources are opened si-

multaneously and the CuInSe2starts to grow with a rate of approx. 5 nm/min[9]. Typical

absorber layer thicknesses are in the range of 100 nm to achieve relaxed layers by prevent-

ing internal stresses due to interface effects. More details, not only about the growth of

the absorber films, but also their electronic structure and the formation of the Cu-poor

defect compound CuIn3Se5, are investigated in detail in the thesis of Hofmann[9].

After successful deposition of the CuInSe2absorber, the sample is transferred under

UHV conditions to the analysis chamber where it is characterized using photoelectron

spectroscopy. A survey spectrum of one of those absorber layers is presented in Fig. 5.1

using Al Kαradiation. In comparison to Si or ZnO spectra shown in previous chapters

of this work, the ternary chalcopyrite produces a more complex survey spectrum with

various different photoelectron and X-ray induced Auger electron emission peaks. There-

fore, the most important photoelectron peaks of the contributing elements in Fig. 5.1 are

labeled in red. In particular, these are Cu 2p electrons at binding energies around 933 eV,

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

Figure 5.2: Regional XPS spectra of the most dominant photoelectron lines present in the

ternary CuInSe2absorber layer. (a) Cu 2p3/2around 933 eV; (b) In 3d1/2at

452.5 eV and In 3d5/2at 445 eV; (c) Se 3d. All spectra are recorded using

monochromatic Al Kαradiation to excite the photoelectrons.

In 3d at 444 eV, and Se 3d having binding energies of approximately 54 eV. In addition

to the photoelectron lines, there are corresponding Auger emission peaks present in the

survey spectrum. In Fig. 5.1, those lines are indicated by blue labels. The regional pho-

toelectron spectra are presented in Fig. 5.2 (a)-(c) after removal of the secondary electron

background by a Shirley routine.

Analysis of the [Cu]:[In] peak intensities results in a near-stoichiometric CuInSe2surface

composition of 0.98 and an anion to cation ratio of [Se]/[Cu] + [In] = 1.17. A binding

energy of the Cu 2p3/2photoelectrons of 932.78 eV has been determined. In combination

with the kinetic energy of the Cu L3M45M45 Auger electrons (916.65 eV), this results in a

combined Auger parameter of α(Cu) = 1849.43 eV. According to reported literature and

database values, this Auger parameter also indicates a near-stoichiometric composition of

the CuInSe2absorption layer.

One further feature present in the survey spectrum should be discussed shortly. At bind-

ing energies of 1144 eV and 1117 eV there are two very small peaks observed. These can

be identified as Ga 2p1/2and Ga 2p3/2photoelectrons, respectively[79]. If these peaks do,

however, originate from the substrate, there should also be an arsenic signal contributing

the spectrum in terms of two As 2p peaks being visible at 1324 eV and 1359 eV, respec-

tively. Since this is not the case, it has to be assumed that small amounts of the gallium

diffuses into the CuInSe2crystal during the deposition process.

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

Figure 5.3: Low energy electron diffraction (LEED) pattern of one of the CuInSe2(112)

absorber layers used as substrates for the CISe/ALD-ZnO interface investiga-

tion. The pattern indicates a c(4x2) surface reconstruction.

The low energy electron diffraction (LEED) pattern presented in Fig. 5.3 demonstrates the

epitaxial growth of the CuInSe2on GaAs (111)A, resulting in a (112) oriented film. The

dominant spots in the LEED pattern show a hexagonal zinc blende structure. Additional

spots are introduced by the chalcopyrite order, in total resulting in a c(4x2) surface

reconstruction LEED pattern. More on the surface structure of CuInSe2having different

orientations and their dependency on the surface composition is found extensively in the

work of Hofmann[9].

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

After characterization of the CuInSe2absorber substrates, the investigation of the interface

formation is carried out. Therefore, the ZnO is deposited in monolayer steps onto the

chalcopyrite and after each deposition step its composition and electronic properties are

analyzed. These experiments are primary performed in the laboratory using combined

X-ray and ultraviolet photoelectron spectroscopy analysis in the Integrated System. In

addition, the same growth experiments are also carried out at the BESSY synchrotron

radiation facility to benefit from the increased surface sensitivity provided by the tunable

excitation energy and its high energy resolution. The combination of all those experiments

result in a detailed picture of the interface formation. Additionally, the combination of

core level and valence band spectroscopy yields complete information about the band

alignment of the two materials.

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

Figure 5.4: Thickness dependence of the photoemission spectra of the substrate (top), the

overlayer elements, and the valence band spectra (bottom).

100

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

5.2.1 Initial Growth Characterization

The regional spectra evolution with increasing overlayer thickness is plotted in Fig. 5.4

for all contributing elements, i.e. Cu 2p3/2, In 3d5/2, Se 3d, Zn 2p3/2, and O 1s. All of

those spectra were recorded using Mg KαX-rays as excitation radiation in the laboratory.

In addition to the core-level peaks of the contributing elements, the valence band spectra

excited by He I radiation are also presented in Fig. 5.4.

First we will focus on the X-ray photoemission spectra. As expected, the elements of the

substrate are attenuated with increasing film thickness, while the photoemission signals

of the overlayer start increasing. The intensities of the photoemission lines of all five

elements are plotted against the overlayer thickness in Fig. 5.5. Here, the attenuation of

the Cu 2p3/2substrate peak is used to calculate the overlayer film thickness.

Figure 5.5: Normalized photoemission peak intensity of the contributing elements plotted

against the calculated overlayer thickness, determined by the attenuation of

the Cu 2p3/2peak.

The normalized peak intensities show the characteristic behavior of ideal layer-by-layer

growth, as described earlier in Sec. 3.1. Nevertheless, it is already visible in Fig. 5.5 (a)

that the indium and selenium photoemission signals show a slight delay in their atten-

uation behavior with respect to the Cu 2p3/2. This is also illustrated by the image en-

largement shown in Fig. 5.5 (b) for the initial deposition steps. While the Cu 2p3/2is

101

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

Figure 5.6: Photoemission spectra evolution with increasing overlayer thickness using syn-

chrotron radiation to excite the photoelectrons with constant kinetic energy

of KE = 130 eV to achieve higher surface sensitivity and energy resolution.

attenuated from the first deposition step on, the peak intensities of the In 3d5/2and Se 3d

even seem to increase rather than being attenuated.

A magnification of the normalized peak intensities of the ZnO overlayer during initial

growth is plotted in Fig. 5.5 (b). While the zinc photoemission starts raising immediately,

there is no oxygen observed at the first deposition steps. To gain deeper insight on the ini-

tial deposition behavior, same experiments were performed using synchrotron radiation at

the BESSY II. There, the surface sensitivity was increased by tuning the excitation energy,

that all emitted photoelectrons have the same kinetic energy of KE = 130 eV. This has

basically the advantage of equal inelastic mean free paths (IMFP) for all elements, which

is resulting in better comparability of the particular elements, since all photoelectrons

are created within the same depth region of the material under investigation. To avoid

the time consuming UHV transfer using the transfer box, the ALD reactor was directly

attached to the SoLiAS analysis system. This allows much faster in situ analysis of the

deposition experiments. In analogy to the experiments performed in the laboratory, the

ZnO was deposited in monolayer steps onto the chalcopyrite absorber. Fig. 5.6 presents

the regional spectra of the contributing elements after normalization (cf. Sec. 2.2.3) and

removal of the secondary electron background.

102

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

Even though the experiments performed at the synchrotron radiation facility show compa-

rable spectral characteristics, there are some differences discussed in detail in the following.

First, the photoemission peaks of the substrate elements Cu, In and Se show an enhanced

shift in their binding energies after deposition of the first monolayer with respect to the

experiments in the laboratory. Since the magnitude of the shift is similar for all three ele-

ments, we can assign this shift to a band bending at the interface, rather than a chemical

shift in the substrate.

As already indicated by the laboratory experiments, the attenuation of the indium and

selenium show a different attenuation characteristics than the copper signal. This be-

havior has already been observed in experiments performed by Andres et al. for the

MOMBE ZnO deposition on CuInS2(112). A delay in the initial deposition of oxygen

was observed, while at the same time the sulfur photoemission intensity did not decrease

immediately. This led to the conclusion of the formation of a ZnS interfacial layer between

the CuInS2absorber and the zinc oxide[21]. A similar behavior was observed just recently

for MOMBE ZnO deposition on CuInSe2(112) by Hofmann[9;22], where a ZnSe interfacial

layer was identified, having a thickness of about 2 nm.

Initial Growth - Influence on Substrate’s Elements

In contrast to MOMBE deposition, where the Cu and In signals were both attenuated

during the formation of the ZnSe layer, this is not observed in case of atomic layer depo-

sition. In fact, only the Cu 2p3/2is attenuated, while the indium even appears to increase

during initial growth. The enhanced surface sensitivity of the SR-PES experiments can

help identifying the interface formation in more detail. Therefore, the photoemission peak

intensities are normalized to 1 and plotted over the calculated overlayer thickness. The

resulting graph is shown in Fig. 5.7. The most obvious feature is the strong increase in

peak intensity for the In 3d3/2and the Se 3d photoelectrons during the initial deposition

steps. They both reach a maximum in intensity after about 0.4 nm of roughly 150 % (In)

and 170 % (Se), respectively. Even though the inaccuracies in data quantification using

synchrotron radiation is increased (see Sec. 2.2.3), an error of such high magnitude can be

excluded, particularly since the Cu 2p photoelectron line shows ideal attenuation behavior

(blue line).

Further analysis of the substrate’s photo emission during initial ZnO deposition does not

show a significant chemical shift or change in Auger parameter. Hence, it is difficult to

103

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

Figure 5.7: Normalized photoemission intensities during ALD-ZnO deposition investi-

gated with synchrotron radiation at constant kinetic energies of KE = 130 eV.

Due to the increased surface sensitivity, effects at the surface are pronounces

with respect to the XPS investigation in the laboratory. It is observed, that

the intensities of both, the selenium and the indium increase during initial

growth. This is in contrast to the MOMBE deposited ZnO of earlier studies,

where only the Se signal was influenced by the ZnSe formation.

104

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

draw conclusions about the local chemical environment of the elements directly from the

photoemission peaks. Annealing experiments of the ZnO layer after finishing the initial

growth experiments can give some indication of the mobility of the particular elements.

The corresponding photoemission spectra before and after annealing the sample at 400 ◦C

for 30 min are presented in Fig. 5.8.

While there is no indication for an intensity change of the Cu 2p3/2or Se 3d photoelectrons,

the indium signal shows a significant increase in intensity. This suggests that the indium

atoms are mobile and diffuse into the ZnO film. A decrease in ZnO overlayer thickness by

the annealing process can be ruled out, since this would result in a simultaneous increase

of the copper and selenium photoemission peaks as well.

Compared to the XPS results shown in Fig. 5.5, the zinc and oxygen peak development

shows a more linear behavior. It is also observed that the oxygen peak is significantly

reduced with respect to the Zn 2p photoelectrons. But in contrast to the experiments

performed by Andres and Hofmann, an oxygen photoemission peak is arising already in

the early deposition stages. As usual, the O 1s peak shows two components: one at lower

binding energies assigned to [O2−] lattice oxygen and the characteristic surface hydroxide

oxygen peak at the high binding energy side. A selection of the corresponding O 1s

photoemission spectra are presented in Fig. 5.9.

Figure 5.8: Photoemission spectra before (black) and after annealing (red) of a ALD-

ZnO film on CuInSe2in the thickness range of the photoelectron information

depth of the substrate. Annealing for one hour at 400 ◦C shows an increase in

photoemission intensity of the indium signal, indicating an indium diffusion

into the ZnO overlayer.

105

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

Overlayer Growth - Oxygen Component

Compositional analysis of the overlayer in terms of the [Zn]:[O] ratio is presented in

Fig. 5.10 (a). Even though an oxygen peak arises from the very beginning on, the initial

overlayer is extremely zinc-rich with a [Zn]:[O] ratio exceeding 7.6. With increasing film

thickness, the ratio drops rapidly, resulting in stoichiometric zinc oxide after about 5 ˚

corresponding to one ZnO monolayer.

Comparing the two oxygen components presented in Fig. 5.9 show an increased amount of

hydroxides at the surface during initial ZnO-ALD. The ratios of the two particular oxygen

components over the film thickness are plotted in Fig. 5.10 (b). The increased amount

of hydroxides during initial atomic layer deposition has already been observed earlier

in Chp. 4 for the ZnO deposition on Si(111)-H. In contrast, the amount of hydroxides

strongly exceed the amount of lattice oxygen, resulting in a [O−OH]:[O2−] ratio of 1.8,

while for ZnO-ALD on Si(111)-H the [O−OH]:[O2−] ratio for initial growth was almost 1:1.

Hence, in combination with the results of the composition analysis, there is no indication

of ZnO formation from the very first deposition step on. The present oxygen is most likely

Figure 5.9: O 1s spectra evolution during initial ZnO atomic layer deposition on CuInSe2.

While for ZnO MOMBE on CuInSe2there was no oxygen visible during initial

growth, this is not the case for the ALD process.

106

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

Figure 5.10: (a) Composition analysis of the overlayer plotted against its thickness. Even

though there is a weak oxygen signal, the film is strongly zinc-rich before

it becomes stoichiometric for thicker layers. (b) Intensity ratio of the two

oxygen components observed in the O 1s spectra. In contrast to the ZnO

atomic layer deposition on silicon substrates, the amount of hydroxides during

initial growth is increased with respect to the lattice oxygen.

originating from hydroxides attached to the surface after the water pulse of the second

ALD half-reaction.

Overlayer Growth - Zinc Component

The major difference of the ZnO |CuInSe2interface achieved by ALD and MOMBE ob-

served so far is found in the significantly decreased interface layer thickness to 5 ˚

A and the

presence of a considerable amount of indium (cf. Fig. 5.7) besides the zinc and selenium

component in the initially formed monolayer.

A powerful tool to identify the local chemical environment of an element in the compound

under investigation is the combined Auger parameter α. The inset in Fig. 5.11 (a) shows

no observable change in the Auger parameter for Cu and In. For technical reasons, the

experiments had to be performed using Mg Kαradiation and the Auger parameter could

not be determined for selenium. Hence, the Auger parameter of the zinc component is

used trying to identify the interfacial layer between absorber and ALD-ZnO. Its thickness

107

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

dependence is plotted in Fig. 5.11 (a) for the laboratory experiments, while Fig. 5.11 (b)

represents the experiments with more surface sensitive synchrotron radiation. In both

cases, an initial zinc Auger parameter of 2011.8 eV is determined. After 5 ˚

A, a significant

drop by about 1.8 eV is observed. At this film thickness, stoichiometric ZnO is observed

in the compositional analysis and therefore supports the assumption of the formation of

an interface layer having that particular thickness of 5 ˚

A. At higher film thicknesses,

the Auger parameter clearly reveals ZnO formation with an Auger parameter around

2010 eV.

The results so far do not indicate a ZnSe formation similar to the situation in the MOMBE

process. The additional presence of indium during initial growth gives an indication for a

different chemical situation at the interface of CuInSe2and ALD-ZnO. Trying to identify

this boundary layer, possible reference materials were prepared and their Auger parame-

ters were determined using the same analytic setup, respectively. First, a ZnSe film was

deposited by thermal evaporation on CuInSe2in the Integrated System. Afterwards, its

Auger parameter has been determined by XPS using Al Kαradiation. The corresponding

photoemission spectra are shown in Fig. 5.12 (a) and (b). The binding energy of the

Figure 5.11: Auger parameter analysis of the zinc component with increasing film thick-

ness. Part (a) on the left hand side shows the results of the experiments per-

formed in the laboratory using Mg Kαradiation, while in (b) synchtrotron

radiation was used for surface analysis. There is a significant drop in the

Auger parameter observed, indicating the formation of an interface layer be-

tween CuInSe2and ZnO during initial ALD.

108

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

Figure 5.12: Regional photoemission spectra of the Zn 2p3/2photoelectron and Zn LMM

Auger electron lines for a ZnSe film deposited on CuInSe2to determine the

combined Auger parameter as a reference to the one observed for initial ZnO-

ALD on CuInSe2.

Zn 2p3/2photoelectrons amounts to 1022.26 eV, while the Auger electrons have a kinetic

energy of 989.22 eV. This ends up in a combined Auger parameter of:

AP (ZnSe) = BE (Zn 2p3/2) + KE (Zn L3M4,5M4,5) (5.1)

= 1022.26 eV + 989.22 eV

= 2011.48 eV ±0.2 eV

Considering the error of measurement of about 0.2 eV, the zinc Auger parameter deter-

mined for ZnSe is clearly below the one determined for initial ALD-ZnO deposition on

CuInSe2. A ZnSe boundary layer would also not explain the significant indium signal

observed at the interface. Hence, the formation of an intrinsic ZnSe buffer layer can be

ruled out for ALD-ZnO deposition on chalcopyrite absorber materials.

5.2.2 Interface Formation: Intrinsic Buffer Layer

This results in an alternative interpretation of the experimental data, assuming that all el-

ements identified at the boundary layer, i.e. zinc, indium and selenium, form the interface

layer. Combining those materials, the formation of a ZnIn2Se4(ZISe) phase seems to be

109

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

most likely. Interestingly, this material is investigated by some few groups as alternative

to the CdS buffer layer in chalcopyrite solar cells[81–83]. Reports on the crystal structure of

ZnIn2Se4reveal, that it belongs to the defect chalcopyrite family, crystallizing in a tetrag-

onal structure with the space group S2

4. Their lattice parameters were determined being

a= 5.69 ˚

A and c= 11.49 ˚

A[81]. Fig. 5.13 illustrates the crystal structure of ZnIn2Se4and

shows the differences to the CuInSe2chalcopyrite and its crystal structure. There are only

few studies about ZISe and their physical, chemical and electronic properties. While Babu

et al. demonstrated ZISe thin film growth by chemical bath deposition[81] just recently,

Ohtake et al. investigated the performance of Cu(In,Ga)Se2devices with coevaporated

ZISe buffer layers[82]. Their measurements showed very promising results. While the ref-

erence cells using CdS buffer layers showed efficiencies of 15.9 % (1997), the alternative

ZISe buffers reached efficiencies of already 15.1 %. There are only few reported values of

the band gap energy of ZnIn2Se4. Babu et al. reported values between 2.15 −2.64 eV,

depending on the pH-value of the chemical bath. A similar value of 2.22 eV has been

determined by Hendia et al.[84] for evaporated thin-films. Choe reports a considerable

lower value of 1.82 eV[85]. Compared to the band gap energies of CdS (2.42 eV) or ZnO

(3.37 eV), the reduced band gap is not very beneficial for high efficiency solar cells. Un-

der these circumstances it is even more surprising that Ohtake et al. achieved such high

efficiencies. One reason might be the good heterojunction formation, since the crystal

structures of the ZISe defect chalcopyrite and the absorber material are very similar. In

addition, the vacuum coevaporation process also might have advantages over the unfavor-

able chemical bath deposition process in terms of surface contamination.

To give further indications about the formation of ZnIn2Se4at the interface of the

CuInSe2absorber and the ZnO, comparison of the initial zinc Auger parameter shown

in Fig. 5.11 with reported values for ZISe would be helpful. Unfortunately, there are no

experimental Auger parameter data reported for ZnIn2Se4in literature or the common

databases. To have some comparable data, ZnIn2Se4crystals were grown in cooperation

with the group of Prof. Binnewies at the University of Hannover. They are able to grow

ZISe bulk crystals by the chemical vapor transport (CVT) method using iodine as trans-

port agent[87]. Before introducing those crystals into the vacuum system, they were glued

on a sample holder. The thermal instability of the glue does not allow to heat the ZISe

samples after introduction into the UHV system. To remove surface contaminants such as

oxygen and carbons, the sample was sputter cleaned for 60 min using Ar+ions of 500 eV

energy. Afterwards, the ZnIn2Se4crystals were transfered under UHV conditions into

the XPS system and analyzed. No remaining contaminations of oxygen or carbon were

110

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

Figure 5.13: Chalcopyrite crystal structure of ZnSe, CuInSe2and the defect chalcopyrite

structure of ZnIn2Se4according to Margaritondo et al.[86]

observed. At binding energies of 630 eV a very weak additional peak was found, that can

be attributed to iodine of the transport gas, implemented into the crystal during growth.

For determination of the combined Auger parameter, the Zn 2p3/2photoelectron peak and

the corresponding Zn L3M45M45 Auger electron spectrum are measured. The particular

peak positions were determined after secondary electron removal by peak fitting routines

described earlier in Sec. 2.2.1. The corresponding spectra are plotted in Fig. 5.14 (a) for

Zn 2p3/2and in (b) for Zn LMM, respectively.

A binding energy of BE = 1021.99 eV was determined for the Zn 2p3/2photoelectrons. In

combination with the Zn LMM Auger electrons’ kinetic energy of KE = 989.73 eV, this

results in a combined Auger parameter for ZnIn2Se4of:

AP (ZISe) = BE (Zn 2p3/2) + KE (Zn L3M4,5M4,5) (5.2)

= 1021.99 eV + 989.73 eV

= 2011.72 eV ±0.2 eV

Comparing this value with the one of the initially formed layer during ZnO-ALD de-

position (2011.84 eV) can also indicate the formation of ZnIn2Se4at the CuInSe2/ZnO

interface: while the zinc Auger parameter of ZnSe with a value of 2011.48 eV (cf. Eq. 5.2)

is slightly lower than the one measured during initial ZnO deposition of approx. 2011.8 eV,

the Auger parameter of the ZnIn2Se4crystal fits the one determined for ZISe within the

range of the measurement tolerances (±0.2 eV) of the XPS measurements.

111

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

Figure 5.14: Regional photoemission spectra of the Zn 2p3/2photoelectron and Zn LMM

Auger electron lines for a ZnIn2Se4crystal to determine their combined Auger

parameters. This reference helps identifying the intrinsically formed buffer

layer at the interface of CuInSe2and ZnO.

Furthermore, the Auger parameters of the indium and selenium are determined for the

CuInSe2substrate, the ZnSe and ZnIn2Se4references and for CuInSe2after the first ZnO

deposition step, respectively. The resulting combined Auger parameters are summarized

in Tab. 5.1. Especially the selenium Auger parameter of the ZnSe reference differs strongly

from the value determined after initial ZnO deposition. Since the other values show no

significant changes, this might be another indication for the presence of ZISe rather than

ZnSe formed at the CuInSe2|ZnO interface.

In the following, an attempt to explain the boundary layer creation is given. The main

effect being responsible for the interface formation observed during ZnO-ALD is assumed

to be the so-called effect of self-compensation in the chalcopyrite material. This self-

compensation occurs as a reaction of doping the material and zinc is one of the typical

materials known for being an n-type dopant in CuInSe2chalcopyrites by substituting the

Cu sites[88]. Once zinc is deposited onto the CuInSe2surface, for instance in the form of

diethylzinc, the donor atoms create electronic states in the forbidden band gap near the

conduction band minimum (CBM). This results in a shift of the Fermi energy EFtowards

the CBM in the uncompensated chalcopyrite[89]. In CuInSe2, this shift towards the con-

duction band affects the defect formation energies significantly[9;90]. The chalcopyrite will

form copper vacancies [VCu] as a reaction on the Fermi energy shift. The free copper [Cu−]

atoms react with the excess electrons introduced into the material by the zinc dopants

and form neutral [Cu0]. Since there is a gradient in [Cu0] concentration, the atoms will

112

5.2 ALD-ZnO on CuInSe2(112) Absorber Films

Compound AP (Zn) AP (In) AP (Se)

CuInSe2- 852.5 eV 1360.84 eV

ZnSe 2011.48 eV - 1360.45 eV

ZnIn2Se42011.72 eV 852.5 eV 1361.0 eV

CuInSe2+ ZnO 2011.8 eV 852.7 eV 1361.16 eV

Table 5.1: Summary of all Auger parameters determined in this work. CuInSe2, ZnSe

and ZnIn2Se4have been measured as references for the Auger parameter values

determined for the interface layer between CuInSe2and ZnO shown in the last

row.

diffuse deeper into the material, leaving behind [VCu] sites. It is not unlikely, that those

vacancies are (partially) filled by the zinc atoms, resulting in a ZnIn2Se4crystal configu-

ration as illustrated in Fig. 5.13, as also indicated by the Auger parameter analysis. This

effect would be limited to the outermost atomic layers, since further deposition of zinc

on the ZISe buffer layer would not affect the chalcopyrite material underneath as intense

as in case of direct deposition on the CuInSe2. This, in addition with the decreased sub-

strate temperature in the ALD process, would explain the reduced buffer layer thickness

of about 5 ˚

A compared to 20 ˚

A in case of MOMBE ZnO deposition[9].

The reason for the different interface boundary layers achieved by the two different depo-

sition processes, MOMBE and ALD, respectively, has to be explained by the differences in

their process characteristics. As mentioned earlier, one of the main differences is the lower

substrate temperature in case of atomic layer deposition of 210 ◦C with respect to 400 ◦C

for MOMBE ZnO growth. In general, the higher temperature of the MOMBE process is

expected have some kind of influence on the diffusion behavior of the contributing atoms.

This can explain the increased buffer layer thickness of about 2 nm for MOMBE deposited

ZnO with respect to 5 ˚

A for ALD-ZnO.

The second main difference is the sequential deposition mechanism of the ALD process.

This results in significantly changed growth kinetics during initial growth of the ZnO on

the chalcopyrite and is most likely the reason for the ZnIn2Se4formation instead of ZnSe

as in case of MOMBE ZnO deposition.

113

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

5.3 Band Alignment of CuInSe2(112) and ALD-ZnO

The interface between the light absorbing material and the buffer layer forming the

p-n junction has a crucial influence on the overall efficiency of the solar cell device. The

investigation of the electronic band alignment at the junction is therefore of great interest

to decide, whether a buffer layer material is applicable or not. In situ X-ray and ultravi-

olet photoelectron spectroscopies in combination with stepwise deposition of the junction

forming material is able to determine the electronic band alignment of the two materials

in contact. As shown in the previous section, a ZnIn2Se4interface layer is formed and

therefore, one has to consider two interfaces for the experimental band alignment deter-

mination: on the one hand the CuInSe2|ZnIn2Se4interface and on the other hand the

ZnIn2Se4|ZnO interface.

Prior to the deposition of ZnO, the substrate is characterized in detail. As already intro-

duced in Sec. 2.2.2, basic material parameters can be determined by ultraviolet photoelec-

tron spectroscopy. According to Eq. 2.13, the work function is calculated by the position

of the secondary electron cutoff energy Esec and knowledge of the excitation energy, i.e.

hν = 21.22 eV in case of He I excitation. For CuInSe2, this results in a work function ϕ

Figure 5.15: (left) determination of the true valence band maxima of the chalcopyrite by

superposition of a reference spectrum recorded with synchrotron radiation of

hν = 15.5 eV; (right) valence band maximum of ZnO determined by linear

extrapolation.

114

5.3 Band Alignment of CuInSe2(112) and ALD-ZnO

and an ionization energy according to Eq. 2.14 of:

ϕ=hν −Esec (5.3)

= 21.22 eV −16.57 eV

= 4.65 eV

Eion =hν −Esec +EVBM (5.4)

= 21.22 eV −16.57 eV + 0.79 eV

= 5.44 eV

Since the real position of the valence band maximum of CuInSe2cannot be determined

directly by He I radiation (cf. thesis of Hofmann[9]) due to the overlap of different valence

band states, an indirect method is applied as shown in Fig. 5.15 (a). Here, superposition

of a reference spectrum recorded at the BESSY II synchtrotron facility with an excitation

energy of 15 eV is used to determine the correct VBM of the chalcopyrite.

After each atomic layer deposition step, the sample is characterized by means of XPS and

UPS, respectively. To determine possible band bending in the substrate or the overlayer,

the binding energy changes for the contributing elements in the substrate and the overlayer

are plotted against the calculated overlayer thickness in Fig. 5.16. For the substrate

elements, the binding energy changes are calculated with respect to the clean substrate,

while the overlayer element’s energy change is given with respect to the binding energy of

their first appearance in the spectra.

The band bending of the substrate is determined using only the copper and indium

shifts in their particular binding energies, since their Auger parameters did not show

any change of their chemical environment (cf. Fig. 5.11). The significant change of the

selenium binding energy can be caused by a superposition of a possible chemical shift

due to the ZnIn2Se4formation. The initial ZnO deposition results in a band bending

of eV CISe

b= 0.2 eV. Since the ZnIn2Se4thickness is in the range of one monolayer, a

band bending in the ZnIn2Se4overlayer is not expected. This results in a valence band

discontinuity ∆EVB of the CuInSe2|ZnIn2Se4interface of:

∆ECISe/ZnIn2Se4

VB =EZnIn2Se4

VBM −ECISe

VBM −eV CISe

b−eV ZnIn2Se4

b(5.5)

= 1.8 eV −0.79 eV −0.2 eV −0.0 eV

= 0.81 eV

115

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

Figure 5.16: Evolution of the relative binding energies of the substrate (left) and the

overlayer (right) with increasing overlayer thickness.

EZnIn2Se4

VBM was not determined directly from the initial growth experiments, but indirectly

using information gained by PES measurements of a ZnIn2Se4bulk crystal. Since there

is no abrupt interface between CuInSe2and ZnIn2Se4, the VBM is determined indirectly

by the binding energy of the Zn 3d photoemission peak, which is located 9.5 eV below

the valence band maximum of ZnIn2Se4. The determination of the conduction band

discontinuity ∆ECB is not possible for two reasons. On the one hand there is no reliable

band gap energy reported for ZnIn2Se4. The few ones found in literature differ significantly

among each other and are not reliable. On the other hand, even if the band gap energy

is known, a film in the thickness regime of one monolayer might not match the reported

band gap values of bulk ZnIn2Se4crystals.

Instead, the surface dipole between the CuInSe2and the ZnIn2Se4interface layer can be

observed and calculated by the difference of the ionization energies of the two materials

and the valence band offset:

eD CISe/ZISe =ECISe

ion −EZISe

ion + ∆ECISe/ZISe

VB (5.6)

= 5.44 eV −6.48 eV + 0.81 eV

=−0.23 eV

116

5.3 Band Alignment of CuInSe2(112) and ALD-ZnO

Figure 5.17: Band diagram of the CuInSe2- ALD-ZnO system with intrinsically formed

ZnIn2Se4of one monolayer in thickness.

117

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

As expected, the interface dipole is quite small, since both materials contain the same

anions and have similar crystal structures.

To complete the band alignment, same anaylsis is performed for the ZnIn2Se4|ZnO

interface. The valence band maximum of ZnO was determined by linear extrapolation of

the valence band edge, shown in Fig. 5.15 (b). This results in an ionization energy of the

ZnO of:

Eion =hν −Esec +EVBM (5.7)

= 21.22 eV −17.23 eV + 3.37 eV

= 7.36 eV

The binding energy change of the overlayer is plotted in Fig. 5.16. Determination of a

band bending from the zinc photoelectrons is not very reliable, since they are expected

to undergo a chemical shift due to the ZnIn2Se4and ZnO formation. Hence, the O 1s

photoemission peak was used to investigate a possible band bending. Indeed, there is

a binding energy shift observed in Fig. 5.16, but its magnitude corresponds to the one

observed in the substrate, resulting in no additional band bending in the ZnO film. Now,

the valence band discontinuity of the ZnIn2Se4|ZnO interface can be calculated:

∆EZISe/ZnO

VB =EZnO

VBM −EZISe

VBM −eV ZISe

b−eV ZnO

b(5.8)

= 3.37 eV −1.8 eV −0.0 eV −0.0 eV

= 1.57 eV

The surface dipole at the ZnIn2Se4|ZnO interface is expected to be significantly larger

than at the CuInSe2|ZnIn2Se4interface. One reason is the replacement of the selenium

by more electronegative oxygen anions[9]. In addition, the interface of the chalcopyrite

and the ZnSe is expected to have a high structural quality. The same is not necessarily

true for the ZnO on ZnSe, since their crystal structures differ significantly. This results

in a surface dipole between ZnSe and ZnO of:

eD ZISe/ZnO =EZISe

ion −EZnO

ion + ∆EZISe/ZnO

VB (5.9)

= 6.48 eV −7.36 eV + 1.57 eV

= 0.69 eV

118

5.3 Band Alignment of CuInSe2(112) and ALD-ZnO

Even though the determination of the conduction band position in the ZnSe interface layer

is not possible, a comparison of the conduction band position of the ZnO with respect to

the CuInSe2might give some information, if direct ALD-ZnO deposition on chalcopyrite

absorbers can be beneficial for the solar cell performance, or not.

Combination of the valence band maxima of both materials and the observed band bending

in the absorber gives a valence band offset of ∆ECISe/ZnO

VB = 2.38 eV. The conduction band

offset can only be determined indirectly according to:

∆ECISe/ZnO

CB =ECISe

g−EZnO

g+ ∆ECISe/ZnO

VB (5.10)

= 1.05 eV −3.44 eV + 2.38 eV

=−0.01 eV

Finally, this ends up in a complete band alignment of the CuInSe2/ZnIn2Se4/ZnO system

as illustrated in Fig. 5.17. A negative spike in the conduction band is believed not to have

very beneficial effects on the solar cell efficiency. A negative conduction band offset can

lead to enhanced recombination of the electrons in the conduction band, which in turn

reduces the open circuit voltage of the device. Hence, a small positive spike in the con-

duction band up to 0.4 eV is believed to be beneficial for the solar cell performance[15;17].

Nevertheless, the calculation of the conduction band offset of ∆ECISe/ZnO

CB =−0.01 eV

can be subject to significant errors and is not necessarily true. On the one hand, we

assume the band gap of the films in contact being equal to their bulk values. This is not

necessarily true, since the band gap can vary at the surface of the material[6].

On the other hand, the interfacial ZnIn2Se4layer will cause additional conduction band

offsets at the CISe/ZISe and the ZISe/ZnO interfaces. Due to the wide scattering of re-

ported band gap values for ZnIn2Se4and its thickness in the monolayer range, a prediction

of the band gap at such small layer thicknesses would be merely speculative.

It should be mentioned that the presence of hydroxides in the ZnIn2Se4buffer layer was

not considered in the band alignment. There might be a significant influence of the hy-

droxide oxygen on the band offsets at the CuInSe2|ZISe interface. There are basically

two issues that complicate the band alignment determination considering hydroxides in

the interface layer: on the one hand it is not totally clear whether the hydroxide is in-

corporated in a ZnIn2Se4matrix or just located at the ZISe surface. On the other hand,

a closer investigation would require the deposition of a reference film having the same

composition as the intrinsically formed interface layer. Hence, the presence of the oxygen

in the ZISe buffer was neglected in this study.

119

5 ALD-ZnO as Cd-free Buffer Layer for Chalcopyrite Solar Cells

5.4 Summary

In this chapter, a detailed investigation of direct ZnO atomic layer deposition on epi-

taxial CuInSe2(112) chalcopyrite absorber materials was presented. In contrast to ZnO

deposited on epitaxial CuInSe2by a MOMBE process, where an intrinsic ZnSe layer

formed, this is not observed for the ALD process. In the following, the results of this

chapter are summarized shortly:

interface formation: While previous photoemission studies of MOMBE ZnO on epi-

taxial chalcopyrite layers revealed a ZnSe phase formed at the interface, using the ALD

process a very thin ZnIn2Se4intrinsic buffer layer was observed. The formation mecha-

nism is believed to be induced by self-compensation effects on doping the chalcopyrite.

Self-compensation results in the formation of copper vacancies and a copper diffusion away

from the surface. X-ray and synchtrotron-radiation photoelectron spectroscopy indicate

the formation of ZnIn2Se4, while the copper photoemission rapidly decreases. Besides

the significantly reduced deposition temperature of the ALD process, the formation of

ZnIn2Se4can result from the different growth kinetics of the sequential growth mode in

ALD.

boundary layer thickness: While the MOMBE deposition of ZnO on CuInSe2resulted

in a buffer layer thickness around 2 nm, the ZnIn2Se4film formed during ALD-ZnO de-

position is at maximum one monolayer (0.5 nm) thick. This reduced thickness can be

explained by the significantly lower deposition temperature of the substrate during atomic

layer deposition, resulting in lower diffusivity of the atoms during interface formation.

oxygen-contribution: Another difference observed in the interface formation is the

presence of oxygen from the initial atomic layer deposition step on. In contrast, MOMBE

results showed a completely oxygen-free ZnSe boundary layer. The additional oxygen

content in the buffer layer might be explained by the different growth kinetics of the

ALD process. While in MOMBE both precursor materials, diethylzinc and water, are

introduced simultaneously into the growth chamber, only one precursor is provided by the

sequential deposition mode of ALD. This leads to incorporation of small amounts of oxygen

into the buffer layer. Photoemission studies identified the oxygen component originating

mainly from hydroxide oxygen. This is a quite interesting result, since several studies

indicated, that buffer layers containing hydroxide components reach promising efficiencies

of the solar cell devices, such as Zn(OH,Se), Zn(OH,S) or In(OH,S)[5;10;11;16;91].

120

5.4 Summary

electronic band line-up: Furthermore, we have shown that the ability of stepwise

deposition and in situ analysis is not only a very powerful method to investigate the

boundary layer formation of the CuInSe2/ZnO system, but can also give a detailed picture

of the electronic band line-up at the interface. The intrinsic formation of a ZnIn2Se4

(ZISe) buffer layer results in a more complex band alignment, since there is not only one

CuInSe2/ZnO interface, but two interfaces CuInSe2/ZISe and ZISe/ZnO have to be taken

into account. The experimentally determined band line-up at the CISe/ZISe interface

shows a valence band offset of 0.8 eV and an induced band bending of about 0.2 eV.

As expected from the similarity of the two materials, only a small surface dipole was

observed. At the ZISe/ZnO interface, the dipole was significantly enhanced due to the

more electronegative oxygen in the ZnO and the more imperfect interface quality. While

a valence band offset of 1.57 eV was determined, the thickness of the ZISe buffer of only

one monolayer does not allow to draw conclusions about the conduction band behavior of

the CuInSe2/ZnIn2Se4/ZnO system.

indium diffusion: annealing experiments indicate a significant diffusivity of the indium

into the ZnO overlayer. The same behavior has already been observed in the MOMBE

ZnO deposition experiments by Hofmann[9]. Reported experiments with different types of

alternative buffer layers indicated, that indium-containing films often yield highest device

efficiencies. According to Rockett, this can result from interdiffusion of the indium. This

interdiffusion can reduce interfacial defects, which would lower interface recombination of

the electrons and can also lead to a grading of the junction, reducing the effects of possible

band offsets[5]. Hence, the presence of mobile indium found can have beneficial effects on

the properties of the electronic junction.

121

6 Water-free ZnO ALD

Probably the most challenging task in the field of ZnO research is to achieve reproducible

p-type doping. As already mentioned earlier, hydrogen impurities in the crystal play a

crucial role since they are able to compensate the effects of the p-dopants (such as nitro-

gen) incorporated into the crystal at an oxygen site[92]. More on the role of hydrogen in

ZnO is reported in detail in the work of van der Walle[93] or Nickel et al[94–97].

In conventional atomic layer deposition of ZnO, water is used as the oxygen precursor

that can act as a possible hydrogen source, since during its reaction with diethylzinc, a

hydroxide is formed in an intermediate reaction state. Hence, a new ALD process having

an alternative oxygen precursor than water, producing films of same structural, chemical

and electrical quality would be advantageous for ZnO ALD.

There are only few reports on alternative oxygen precursors used in ZnO atomic layer

deposition. Most of those techniques use either an ozone (O3)[68] or an O2plasma[69].

This might be advantageous in terms of lowering the substrate’s temperature during de-

position, but it usually results in lower quality ZnO compared to thermal-ALD. There

are several reports about successful ZnO atomic layer deposition using ZnCl2as metal

precursor[55;98]. These groups used molecular oxygen (O2) as oxidizing agent. There is

one only study on ZnO atomic layer deposition using organometallic diethylzinc and O2

by Park et al[70]. They report a very low growth per cycle of 0.36 ˚

A, not indicating real

atomic layer deposition characteristics.

In the following chapter, the first successful study of a new water-free atomic layer de-

position process will be presented. Zinc oxide films are deposited using a precursor com-

bination of metal-organic diethylzinc and O2as the oxygen source. The initial growth

on Si(111)-H substrates is investigated and compared to the standard ALD process using

water as oxidizing agent. A comparison of the two different ZnO films will yield informa-

tion about the reaction mechanism of the new water-free ALD process and the chemical

and electronic properties of the ZnO films achieved.

123

6 Water-free ZnO ALD

6.1 Growth Parameters for ZnO-ALD using O2precursor

As previously mentioned in Sec. 4.2, molecular O2is provided by a 1 l lecture bottle hav-

ing an initial pressure of 10 bar. It is connected to an additional ALD diaphragm valve

by UHV-compatible tubing. Hence, the ALD reactor is capable to deposit ALD-ZnO

by use of two different oxygen precursors: water and molecular oxygen, respectively. To

differentiate between both precursors, the terms H2O-ALD and O2-ALD are used from

now on, depending on what oxidizing agent has been used.

Compared to using water as oxygen precursor, the pressure in the deposition chamber

during O2exposure is significantly higher. To prevent an automatic shut down or even

damage of the turbo molecular pump, the shortest operable valve opening time is chosen

for O2exposure. As already discussed in Sec. 4.2.3, the error in response time of the valve

used for O2control is 12 ms. Hence, 13 ms is the shortest possible exposure time for this

valve, resulting in a maximum pressure of 1 mbar in the deposition chamber during oxy-

gen exposure. Because of those high pressures, a complete surface saturation is expected

even after this short exposure times. Since no modifications are made with respect to the

DEZn exposure, its saturation behavior as presented in Fig. 4.18 will remain the same.

To determine the ideal substrate temperature for O2-ALD, the same procedure as in

Sec. 4.3.2 for H2O-ALD was carried out. Hydrogen-terminated Si(111) is selected as sub-

strate material to achieve comparableness to H2O-ALD. As described in Chp. 4, a thin

ZnO film was deposited onto the Si(111)-H to rule out effects of initial monolayer forma-

tion on the growth-per-cycle. Afterwards, five ALD cycles were deposited and the growth

rate was measured by X-ray photoelectron spectroscopy. The temperature range in which

the growth-per-cycle has been determined started at 150 ◦C and was increased in steps of

20 ◦C. The resulting growth rates per ALD cycle are plotted as red triangles against the

substrate’s temperature in Fig. 6.1. For ease of comparison, the temperature dependence

using water as oxygen precursor is also plotted.

First of all, the presence of a temperature region with the typical temperature-independent

GPC in Fig. 6.1 between 185 −210 ◦C validates the existence of self-limiting atomic layer

deposition. It is the first successful study of ZnO atomic layer deposition using the pre-

cursor combination diethylzinc and O2. The O2-ALD window is, compared to the one

observed in H2O-ALD, just slightly shifted by 15 ◦C towards lower substrate tempera-

tures. Hence, a substrate temperature of 200 ◦C is ideal for successful O2-ALD.

The most prominent difference between both precursor combinations is found in the de-

position rate per ALD cycle. For H2O-ALD the GPC is 3.0˚

A corresponding to roughly

124

6.1 Growth Parameters for ZnO-ALD using O2precursor

Figure 6.1: Growth-per-cycle (GPC) dependence on deposition temperature of the ZnO-

ALD using DEZn and molecular oxygen (O2). The characteristic ALD window

indicates self-limited ideal growth behavior. The increase in deposition rate is

the result of a different reaction mechanism, not limited by steric hindrance

effects of the ethyl-ligands.

60 % of one complete ZnO monolayer in preferred (0001) orientation, whereas using O2as

oxidizing agent the deposition rate per ALD cycle is increased significantly. In comparison

to H2O-ALD, the GPC in O2-ALD increases by a factor of more than 1.5 reaching a value

of 5.0˚

A. Accordingly, each ALD cycle almost one entire ZnO monolayer is formed using

O2as oxidizing agent.

While the GPC was limited due to steric hindrance effects of the ethyl ligands of the

diethylzinc in case of H2O-ALD, this effect is obviously not observed in O2-ALD. It is

rather a complete removal of both ethyl ligands during DEZn exposure that leads to the

strong increase in growth rate. Furthermore, the chemical reactions at the surface are

expected to be much more complicated as compared to reaction 4.2 and 4.3, due to the

absence of hydroxides in the reaction channel. As a result, the formation of ethane by a

ligand exchange mechanism as in case of reaction 4.2 is most likely not going to occur.

Instead, other reaction products than ethane have to form. Further analysis to identify

possible reaction mechanisms is made in the following sections.

125

6 Water-free ZnO ALD

6.2 Initial Growth of O2-ALD ZnO on Si(111)-H

After successful water-free atomic layer deposition has been demonstrated in the pre-

vious section, the aim of the following one is to gain insight into the initial growth of

ZnO deposited by O2-ALD. To compare the O2-ALD ZnO with the results of initial ZnO

ALD using H2O as oxidizing agent, ZnO layers are deposited monolayer-wise on Si(111)-H

substrates and characterized in situ by XPS, as described earlier in Sec. 4.5. The photoe-

mission spectra of the involved elements Si, Zn and O are presented in Fig. 6.2.

As expected, the substrate’s signal is attenuated after each deposition step and disap-

pears completely when the film thickness exceeds the information depth of the photoelec-

trons. At the same time, both film elements rise until their intensities reach a maximum.

The normalized intensities of the particular peaks are plotted in Fig. 6.3. As in case of

H2O-ALD, the observed attenuation and increase behavior demonstrate ideal ZnO layer-

by-layer growth also in case of O2-ALD. Comparing the peak intensity evolution with

Fig. 4.24, shows no major differences. Only the initial growth of the first ZnO monolayer

seems to complete sooner in case of O2-ALD. Furthermore, it seems to take longer until

the substrate’s photoemission signal is completely attenuated. However, especially for

film thicknesses in the range of the information depth, small intensity variations of the

substrate’s photoemission signals will have a large influence on the calculated film thick-

Figure 6.2: X-ray photoelectron spectra of initial ZnO growth using O2as oxygen precur-

sor. Al Kαradiation is used to excite the photoelectrons.

126

6.2 Initial Growth of O2-ALD ZnO on Si(111)-H

Figure 6.3: Normalized photoemission intensities of Si, Zn and O during initial ALD of

ZnO using O2as oxygen precursor. To gain insight into the growth behavior,

the intensities are plotted over (a) the number of ALD cycles and (b) the

corresponding ZnO film thickness.

ness. This results in an increased standard deviation of the calculated film thickness for

film thicknesses in the range of the sampling depth. Anyway, comparing the measured

attenuation behaviors of both O2-ALD and H2O-ALD with the expected PES characteris-

tics during layer-by-layer growth (cf. Fig. 3.4) results in a more ideal attenuation behavior

for O2-ALD.

Peak fitting results of the Zn 2p3/2photoemission line are shown in Fig. 6.4 (a). As for

H2O-ALD, no additional chemical component can be identified. The resulting peak width

are plotted over the film thickness in Fig. 6.4 (b). For ease of comparison, the H2O-ALD

results presented in Chp. 4 are also shown.

At first glance, there are no significant differences between both films. The one grown

by the standard ALD process seems to show slightly increased peak width during the

first deposition step but that can be assigned to the smaller film thickness of only 0.96 ˚

compared to 1.55 ˚

A in case of O2-ALD. At later deposition steps, the peak width of the

Zn 2p3/2photoemission line converges to 1.5 eV while in case of H2O-ALD only 1.55 eV

is achieved. Hence, the disorder in the films seems to be slightly reduced using O2as

the oxidant. There are no significant differences observed in the zinc component of the

film compared to H2O-ALD ZnO. More interesting are possible differences in the oxygen

components of the ZnO grown by O2-ALD with respect to those grown under standard

conditions.

127

6 Water-free ZnO ALD

Figure 6.4: (a) Zn 2p3/2photoemission spectrum and associated peak fitting; (b) evolution

of the Zn 2p3/2peak width during initial ZnO growth in an O2-ALD process.

Fig. 6.5 is illustrating the development of the O 1s during initial O2-ALD of ZnO on

Si(111)-H. Therefore, three photoemission spectra and their corresponding peak fits are

presented as in case of H2O-ALD in Chp. 4, representing three different stages of ZnO

ALD. Part (a) shows the resulting O 1s photoemission after the first deposition step. The

calculated film thickness of 1.55 ˚

A indicates a very thin and not yet closed ZnO layer.

Again, as in case of H2O-ALD, there are two different oxygen components observed. The

one at a binding energy of BE = 530.5 eV is again assigned to [O2−]. In contrast to H2O-

ALD, where the second component originates from hydroxides, the reaction mechanism

of O2-ALD is unknown. Until the chemical origin of the second oxygen component in

O2-ALD is identified and investigated in more detail later in this work, it is referred to

as the negatively charged [O−] component. Peak fitting results in a somewhat increased

peak width of both components. While the [O2−] species of H2O-ALD grown ZnO had

a FWHM of 1.07 eV, using an O2precursor shows a peak width of 1.12 eV. The same

arises for the [O−] species that exhibit a FWHM of 2.08 eV compared to 1.97 eV of the

hydroxide peak using water as oxidizing agent.

With increasing film thickness, those values do not change significantly. Instead, the in-

tensity of the [O2−] rises with respect to the second oxygen component. The development

of the [O−] : [O2−] with increasing film thickness is plotted in Fig. 6.6 (a). At early stages

of ZnO deposition, the [O−] component actually exceeds the [O2−] resulting in a ratio of

[O−] : [O2−]≈1.46, while in case of H2O precursor a value around 1.0 was determined.

128

6.2 Initial Growth of O2-ALD ZnO on Si(111)-H

Figure 6.5: O 1s photoemission spectra at three different deposition stages: (a) a very

thin O2-ALD ZnO layer; (b) after the first monolayer has been completed; (c)

at film thicknesses larger than the information depth of the photoelectrons.

The thicker the film, the more the ratio between both oxygen species decreases, approach-

ing [O−] : [O2−]≈0.36 which is almost the same value than in H2O-ALD zinc oxide. A

closer investigation of the oxygen in the films produced by both H2O- and O2-ALD and

their behavior on annealing is presented later in Sec. 6.3.

At first, characterization of the initial O2precursor ZnO atomic layer deposition is to be

brought to an end. Therefore, the [Zn] : [O] ratio is determined and compared to the initial

growth of H2O-ALD. The results of both ZnO layers are shown in Fig. 6.6 (b). While the

[Zn] : [O] ratio during deposition of the first 5 ˚

A (i.e. the first ZnO monolayer) is identical

for both precursors, a constant composition of [Zn] : [O] ≈0.94 is already achieved after

deposition of roughly 1 nm ZnO using O2-ALD. This is almost a factor three earlier than

in case of H2O-ALD. The resulting film ends up with an overall oxygen-rich composition.

If this is also affected by the interruption after each monolayer deposition step and the

subsequent photoelectron spectroscopy, as discussed in Chp. 4, is clarified in the next

section. There, ZnO films thicker than the information depth of the photoelectrons are

characterized to gain more information about the oxygen component.

Analysis of the combined Auger parameter completes the comparison of initial growth

behavior of O2-ALD and H2O-ALD. Therefore, the kinetic energies of the Zn L3M45M45

129

6 Water-free ZnO ALD

Figure 6.6: (a) [O−] : [O2−] ratio of the initial growth of ALD-ZnO using molecular oxygen

(O2) as oxidizing agent; (b) [Zn] : [O] ratio plotted over the ZnO thickness.

have been fitted with the same method as described earlier in Sec.4.5. The corresponding

spectra of the X-ray induced Auger electrons are shown in Fig. 6.7 (a), while the combined

Auger parameter of the Zn L3M45M45 and the Zn 2p3/2photoelectrons is plotted against

the ZnO thickness in Fig. 6.7 (b). The Auger parameter of the initially formed partial

monolayer is determined to be 2009.78 eV. At first sight this value seems to be more than

0.2 eV lower than in case of H2O-ALD. Taking into account that in case of O2-ALD the

average film thickness of the initially formed layer is slightly higher than the H2O-ALD

Figure 6.7: (a) Zn LMM Auger electron spectra after each monolayer deposition step and

(b) the resulting combined Auger parameter with respect to the film thickness.

130

6.3 H2O-ALD vs. O2-ALD ZnO

layer, comparison of the Auger parameter results in no significant changes during initial

growth. After formation of the first ZnO monolayer, the Auger parameter exhibits a

minimum value of 2009.5 eV and increases afterwards constantly until it settles at a value

of 2009.7 eV, as in case of H2O-ALD. In general, the Auger parameter vs. film thickness

dependency shows the same characteristic behavior for both oxidizing precursors. While

in case of water-ALD the increase of the Auger parameter was attributed to the reduced

amount of oxygen in the films, this does not apply for the use of molecular oxygen as

precursor. As shown earlier, the final [Zn] : [O] ratio was already reached shortly after

formation of the initial ZnO monolayer.

6.3 H2O-ALD vs. O2-ALD ZnO

As shown in the previous section, the initial growth of O2-ALD showed no significant

differences in growth behavior compared to standard ZnO-ALD. Nevertheless, there were

indications that the PES analysis after each monolayer deposition influenced the film

growth. Hence, the following section focuses on the investigation and comparison of ZnO

films that are on the one hand thicker than the escape depth of the photoelectrons and

on the other hand were deposited in one single deposition step. Therefore, two ZnO films

of roughly the same film thickness are deposited onto Si(111)-H, 50 cycles in case of H2O-

ALD and 30 ALD cycles using molecular oxygen as precursor. According to the growth-

per-cycle determined, the resulting thicknesses of both films should be dZnO ≈15 nm.

Photoelectron spectra of the H2O-ALD ZnO film have already been shown in Fig. 4.20

of Sec. 4.4. The corresponding photoemission spectra of the O2zinc oxide are shown in

Fig. 6.8 for the Zn 2p3/2(left) and O 1s (middle) photoelectrons and the Zn L3M45M45

Auger electrons (right), respectively.

As already pointed out in the previous section, there is only one zinc component present

during initial growth of ZnO. As expected, the same is true for thick ZnO layers deposited

in one single step and only one Zn component is observed at 1021.89 eV in the Zn 2p3/2

photoelectron spectrum of Fig. 6.8, being shifted by 0.2 eV towards higher binding ener-

gies.

As in case of H2O-ALD, the oxygen 1s photoelectrons show one additional component

at the high binding energy side of the [O2−] peak. As mentioned earlier, the second

component cannot be attributed to a surface hydroxide, since no water is present in the

131

6 Water-free ZnO ALD

Figure 6.8: Regional spectra of a 15 nm O2-ALD ZnO film. As in case of water-ALD, the

Zn 2p3/2photoemission peak contains no additional components. The O 1s

consists of two chemically shifted peaks. In contrast to H2O-ALD, the second

component at the high binding energy side cannot originate from surface hy-

droxides due to the absence of water. The Zn LMM Auger peak is only slightly

shifted and the combined Auger parameter is close to the value determined

for H2O-ALD.

ALD process using O2as an alternative oxygen precursor. Detailed analysis of the second

oxygen component using synchrotron radiation and their behavior on annealing is given

at the end of this section.

As shown in Fig. 6.8, the Zn L3M45M45 Auger electron peak is located at 987.90 eV.

Compared to the H2O-grown film this value is shifted by 0.1 eV to lower kinetic ener-

gies. Hence, combining the photoelectron’s binding energy and the kinetic energy of the

corresponding Auger electrons of the zinc component in the O2-ZnO layer results in a

combined Auger parameter of:

AP (Zn) = BE (Zn 2p3/2) + KE (Zn L3M4,5M4,5) (6.1)

= 1021.89 eV + 987.90 eV

= 2009.79 eV

This is almost the same value as in case of using water as oxidizing agent (2009.68 eV)

and therefore the local chemical environment of the Zn can be expected being the same

for both oxygen precursors.

132

6.3 H2O-ALD vs. O2-ALD ZnO

6.3.1 Electronic Structure and Reaction Mechanism

The capability of using a semiconductor material for certain applications is depending

on its electronic structure[20]. Ultraviolet photoelectron spectroscopy (UPS) is usually

applied to investigate the valence band states of semiconductor materials giving insight

into the electronic properties of a certain material. As mentioned earlier in Sec. 2.2.2, in

UPS photons with an energy of 21.21 eV obtained by a He discharge lamp are used to

excite the photoelectrons. Due to the low energy of the photons, only electrons located

in the valence band contribute to the photoemission spectrum.

Fig. 6.9 presents the valence band spectra of ZnO as-grown by atomic layer deposition

using both H2O (black line) and molecular oxygen (red line) as oxygen precursors, re-

Figure 6.9: Valence band spectra of ALD ZnO deposited with both, H2O (black line) and

molecular oxygen (red line) as oxygen precursor. All spectra were recorded

using He I radiation of hν = 21.22 eV. The difference spectra of both precursor

combinations is illustrated in the inset. The broad feature between 6 −10 eV

in the O2-ALD process is likely to arise due to adsorbed hydrocarbon reaction

by-products of the DEZn + O2reaction.

133

6 Water-free ZnO ALD

spectively. Both spectra show significant differences in their valence band structure. In

case of H2O-ALD, the valence band maximum (VBM) was determined to be located at

VBM = 3.37 eV with respect to the Fermi energy (at BE = 0 eV). There are a couple

of features visible that are discussed shortly. The prominent feature at a binding energy

of 11.0 eV arises from Zn 3d photoelectrons[99–102]. According to Powell et al., the Zn 3d

core-electrons are always located at 7.5 eV below the VBM[20]. Relating the Zn 3d binding

energy to the valence band maximum, a binding energy value of 7.52 eV is determined

for H2O-ALD, which is in perfect agreement with the value reported by Powell et al. The

features in the energy region between the Zn 3d peak and the valence band maximum

contribute from O 2p states around 4.5 eV and from Zn 4s - O 2p mixed states around

7.9 eV with respect to the Fermi level[99–101].

The valence band spectrum of the ZnO film deposited using an O2precursor clearly differs

in its appearance with respect to the standard precursor. While the position and intensity

of the Zn 3d peak does not show any changes, the O 2p states are significantly attenuated

and an additional broad feature appears 8 eV below the Fermi level. To illustrate the

differences between both spectra more clearly, the inset in Fig. 6.9 shows the difference

spectrum of both valence bands. The broad feature between binding energies of 6 −10 eV

seems to consist of at least three components. Since only diethylzinc and molecular oxy-

gen are present during the deposition, additional peaks in the valence band difference

spectrum can only originate either from carbon compounds or the oxygen adsorbed at the

surface itself. An UPS study of different carbon containing molecules done by Bournel et

al. indicate that different carbon compounds generate features in this energy region[103].

On the other hand, a study of point defects on non-polar ZnO(1010) by G¨

opel et al.

attributes two broad features at 4 eV and 6 eV below the VBM (i.e. 7.5 eV and 9.5 eV

below EF, respectively) to the 2 p levels of adsorbed O−

2[104]. Thus, the UPS results can

not clearly identify the origin of the additional peaks in the valence band spectrum.

In case of carbon species being present, they should also be visible in the X-ray photoelec-

tron spectra by emission of C 1s photoelectrons. Fig. 6.10 plots the regional photoemission

spectra of C 1s of both ALD-ZnO layers. Spectrum (a) at the left hand side of Fig. 6.10

corresponds to the ZnO film deposited using H2O as oxygen precursor. No indication of

carbon adsorbates is found in the spectrum. This is consistent with the UPS spectrum

given in Fig. 6.9, where also no spectral features of carbon were identified. These re-

sults reveal the purity of the ALD-ZnO layers using water as precursor and indicates that

the ligand exchange reaction mechanism during water exposure completely removes the

remaining ethyl groups of the monoethylzinc.

134

6.3 H2O-ALD vs. O2-ALD ZnO

A completely different situation is present in the case of using O2precursor gas. Fig. 6.10

(b) plots the C 1s photoemission spectrum of the O2-ALD ZnO film. In contrast to

H2O-ALD, at least two separated peaks are observed. There might be a third com-

ponent present, indicated by a weak shoulder appearing at 287 eV. Unfortunately, the

signal-to-noise ratio is pretty low due to the low amount of carbon at the surface, so that

quantification of the results is difficult. To yield better results, the sample has been trans-

ferred under UHV conditions to the U49/2 PGM-2 beamline of the BESSY II synchrotron

source. The tunable excitation energy (85 −1890 eV) allows very surface sensitive photo-

electron spectroscopy since the information depth is directly dependent on the excitation

energy of the incident light (cf. Chp. 2).

The C 1s spectrum shown in Fig. 6.11 was recorded using an excitation energy of

hν= 415 eV, resulting in a kinetic energy of the excited photoelectrons of KE = 130 eV.

At this kinetic energy, the inelastic mean free path of the photoelectrons is about 9.2˚

i.e. surface sensitive to the topmost atomic layers. The spectrum shows the same two

photoemission lines observed in Fig. 6.10 (b) and no indication for a third component,

which might be located deeper in the material and not contributing to the spectra due

to the increased surface sensitivity. After background removal, both peaks were fitted by

Figure 6.10: (a) Carbon 1s photoelectron spectra recorded with Al Kαradiation. The

H2O-ZnO in spectrum (a) is not showing any carbon surface contamination,

indicating the complete reaction between DEZn and water into ZnO and

ethane. In contrast, there is a considerable amount of carbon in case of O2-

ALD shown in (b), which indicates a more complex reaction between DEZn

and O2with unwanted reaction by-products adsorbed at the surface.

135

6 Water-free ZnO ALD

Figure 6.11: O2-ALD ZnO C 1s spectrum recorded at the BESSY II synchrotron radiation

source with an excitation energy of hν = 415 eV resulting in a kinetic energy

of the C 1s electrons of KE = 130 eV. Two different carbon states are clearly

identified. The peak at lower binding energies is attributed to some kind of

hydrocarbons, while the higher binding energy peak’s origin is a carbon in a

carboxylic environment.

Voigt functions, respectively. The binding energy of the more intense peak was deter-

mined to be BE = 285.25 eV. With a peak width of FWHM = 1.8 eV the peak consist

only of one single carbon component.

The second peak observed is chemically shifted by 4.1 eV to higher binding energies, in-

dicating a very different bonding situation of the carbon atom. Its binding energy results

in BE = 289.35 eV and a peak width of FWHM ≈2.0 eV. An intensity ratio between

both carbon species of 4.6 was calculated.

To identify the two carbon components, the results of the ultraviolet photoelectron spec-

troscopy and the synchrotron radiation photoelectron spectroscopy of the C 1s peak have

to be compared with studies reported by other groups. Identification of the carbon species

would help to determine the chemical reaction during the atomic layer deposition. Even

though there are no reports in literature that deal with the electronic structure of ZnO

grown using DEZn and molecular oxygen precursors, there are various studies on the

adsorption behavior of gaseous molecules on surfaces.

136

6.3 H2O-ALD vs. O2-ALD ZnO

Reaction product Structure Formula Source

ethoxyethylzinc Et - Zn - O - Et C2H5OZnC2H5[108;109]

diethoxyzinc (DEOZn) Et - O - Zn - O - Et Zn(OC2H5)2[108;109]

ethylethylperoxyzinc Et - Zn - O - O - Et C2H5ZnOOC2H5[108;110]

Table 6.1: Summary of reaction products during oxidation of diethylzinc (Et-Zn-Et) re-

ported by other groups.

The diethylzinc (Zn(C2H5)2) precursor is the only possible carbon source during the ZnO

forming reaction. In their XPS study of the adsorbed state of DEZn on ZnO (0001), Vohs

and Barteau determined a broad C 1s peak located at a binding energy of 284.2 eV[105].

Due to its peak width of FWHM = 2.7 eV, and the different bonding situation of the

two carbon atoms in the molecule they considered this peak being a convolution of two

peaks of FWHM = 1.7 eV, each. They determined the binding energy of the carbon atom

bound to the zinc being located at 283.4 eV, while the methyl group carbon (-CH3) is

located at BE = 284.6 eV. A similar value was reported by Kliv´enyi et al. for DEZn

adsorption at low temperatures on rhodium[106]. They determined a broad C 1s peak at

284.4 eV for high DEZn exposures. These results do not match to our measured values

leading to the conclusion that the origin of the C 1s peak neither is adsorbed diethylzinc

nor its ethyl ligands. Even if the binding energy difference is not differing significantly,

the peak width and the absence of a second component rule out DEZn or an ethyl group

being the source of carbon contamination. Atomic carbon with a reported binding energy

of 284.3 eV[105] can also be excluded due to the significant difference in binding energy.

It is also unlikely that the carbon originates from thermal dissociation of the DEZn

molecule at the substrate’s surface. On the one hand, the initial dissociation of DEZn is

reported to start at temperatures above 300 ◦C[107;108]. On the other hand, the deposition

temperature of O2-ALD and H2O-ALD differs only by 10 ◦C. If the DEZn would dissoci-

ate, a C 1s photoemission peak would have been observed in case of water-ALD, too.

Hence, the carbon has to be a by-product of the reaction between Zn(C2H5)2and O2.

There are only few studies on the oxidation of diethylzinc[108–110] and most of them are

theoretical works. All of these studies report that one or even two oxygen atoms are

inserted between the zinc atom and the ethyl ligand. Depending on the number and

position the oxygen atom(s) are inserted, this results in the following reaction products,

summarized in Tab. 6.1.

137

6 Water-free ZnO ALD

In the first case, one oxygen atom is inserted between the zinc and one ethyl group. The

resulting molecule is an ethoxyethylzinc (Et-Zn-O-Et) compound. The second case can

be seen as an oxidation of this ethoxyethylzinc by inserting an additional oxygen atom

between the other ethyl ligand and the zinc atom. This results in a diethoxyzinc (DE-

OZn). According to Maejima et al., this compound is stable up to temperatures of 300 ◦C

after it dissociates into C2H5OZnOH and ethylene (C2H4)[108]. The third reported reac-

tion product is an ethylethylperoxyzinc compound, where both oxygen atoms are located

between the zinc and one of the ethyl ligands (Et-Zn-O-O-Et).

There are no photoelectron spectroscopic studies available for those reaction products.

Anyhow, Verhoeven et al. investigated the interaction of different hydrocarbons[111].

There, they investigated the chemical shift of C 1s photoelectrons depending on their

bonding partner. Hydrocarbons like C2H4, C2H6or •C2H5appear at binding energies

of about 285 eV. This is quite the same binding energy as observed at the intense C 1s

photoemission line in case of the O2-ALD grown ZnO layers and therefore we attribute

this peak to some type of hydrocarbons. Verhoeven also states, that if one of the carbon

atoms of a hydrocarbon molecule forms a single covalent bond with an oxygen atom, its

binding energy shifts about ∆BE ≈1.5 eV[111]. This situation is present in all three

possible reaction products after DEZn oxidation. Hence, a C 1s signal at approximately

286.6 eV would be present in Fig. 6.11 if residuals of oxidized DEZn are present at the

surface. This additional peak is not observed and so we exclude the presence of one of

the reaction products summarized in Tab. 6.1 at the ZnO surface.

If an oxygen atom reacts with an hydrocarbon in a way that a double bond is formed,

the resulting chemical shift is approximately ∆BE ≈3.0 eV. Such a peak is also not

present in Fig. 6.11. The third bonding type Verhoeven investigated is the formation of

a carboxyl (-COOH) compound, where three covalent bonds are formed between C and

the two O atoms. This configuration will shift the binding energy of the carbon atom

by 4 −4.5 eV, which is in very good agreement with the observed binding energy shift of

4.1 eV determined for the second carbon component. Another XPS study of Dilks about

peroxy features at polymer surfaces oxidized by O2also attributes high binding energy

C 1s components at 289.1 eV to carbon in a carboxylic environment[112]. Unfortunately,

combined XPS, UPS and SR-PES investigation cannot clearly identify the exact origin

of the carbon components due to the absence of information about hydrogen atoms in

the spectra. Nevertheless, photoelectron spectroscopy results help to narrow down the

possible carbon configurations: We suggest the more intense carbon component being

attributed to some kind of hydrocarbons. These can be fragments of the ethyl ligand like

138

6.3 H2O-ALD vs. O2-ALD ZnO

a methyl radical (-CH3) for instance, or a reaction product like ethylene (C2H4)[113]. The

high binding energy component is most likely arising due to the presence of a carbon in

a carboxylic environment, i.e. bond to two oxygen atoms for instance a carboxyl-radical

(O=C-OH)[111] or carbon in a perester functionality (O=C-OO)[112]. These results indi-

cate a much more complicated reaction mechanism at the surface during O2-ALD with

respect to H2O-ALD.

All of the carbon species suggested above as possible products of the ALD process should

be loosely bound to the surface. Hence, it is expected that they are easily removed by

an heat treatment of the deposited ZnO films. Therefore, both films were annealed un-

der UHV conditions for 60 min at 400 ◦C. A complete removal of any carbon species

should have a significant influence on the valence band structure of the annealed ZnO.

For comparison, Fig. 6.12 shows both, the as-grown and annealed ultraviolet photoemis-

sion spectra for H2O- and O2-ALD, respectively.

All spectra are normalized to the Zn 3d photoemission peak located at approximately

BE = 11 eV. The valence band spectra of the ZnO films deposited using water as oxygen

precursor are shown in the upper part of Fig. 6.12. It shows two noticeable changes of

the photoemission features between the as-grown ZnO (top left) and after annealing (top

right). First, the intensity of the Zn 3d peak increased strongly with respect to the oxygen

induced peaks close to the valence band edge. The valence band edge itself shifted from

3.37 eV to 3.18 eV after annealing. This is almost the same value determined for annealed

O2-ALD, which has been determined being located 3.23 eV below the VBM as shown in

the lower right part of Fig. 6.12.

The annealed valence band spectrum of O2-ALD shows significant changes compared with

the film without heat treatment. All additional peaks vanished and the valence band

shows the same structural features as the H2O-ZnO. This is attributed to the removal of

adsorbed carbon species by the annealing process as discussed above. Both annealed ZnO

films show similar valence band characteristics. Only the intensity of the O 2p electrons

at BE ≈4.5 eV is more pronounced. This could indicate a higher amount of oxygen at

the surface of the O2-ALD ZnO.

6.3.2 Annealing Behavior of the Oxygen Component

Due to the absence of hydroxides and the different reaction mechanism it was not expected

that the two oxygen components of both films look quite similar (cf. Fig. 6.8 and Fig. 4.20)

139

6 Water-free ZnO ALD

Figure 6.12: He I valence band spectra of H2O-ALD (top) and O2-ALD ZnO (bottom)

both, before (left) and after annealing (right). Significant changes are ob-

served in the annealed O2-ZnO valence band spectrum, indicating the re-

moval of adsorbed carbon compounds from the surface.

140

6.3 H2O-ALD vs. O2-ALD ZnO

at first sight. Especially the second component that is usually attributed to surface

hydroxides in case of H2O-ALD is unlikely to have the same origin in case of O2-ZnO.

Hence, the following part of this section concentrates on the oxygen component in the

particular zinc oxide films, their differences and similarities. The effects of annealing on

the valence band structures was shown just recently. In Fig. 6.13 the influence of heating

the sample on the O 1s photoemission signal is presented. The top left picture represents

the ALD-ZnO film as deposited using H2O as oxidizing agent. Two peaks are fitted after a

Shirley background removal of the scattered electrons. The more intense one attributed to

[O2−] forming the ZnO compound is located at BE = 530.49 eV and exhibits a peak width

of FWHM = 1.07 eV. The second component, attributed to surface hydroxides [O−OH], is

chemically shifted by ∆BE = 1.24 eV to a higher binding energy of BE = 531.73 eV. The

[O−OH] peak width is increased with respect to the [O2−], resulting in a FWHM of 1.97 eV.

After annealing the sample for 60 min at 400 ◦C, there are no big changes observed in

the peak binding energies. The [O2−] peak position has been determined being located at

BE = 530.42 eV. The difference in binding energy was kept constant at ∆BE = 1.24 eV

during the fitting routine. While the peak width of the [O2−] photoelectrons remain

constant at FWHM = 1.07 eV, the [O−OH] peak FWHM decreases a little, resulting in

a value of FWHM = 1.86 eV after annealing. The corresponding O 1s spectrum of the

annealed film is plotted at the upper right corner of Fig. 6.13. A [O−OH] : [O2−] intensity

ratio of the as-grown film of 0.36 has been calculated, i.e. the hydroxide bound oxygen

amounts to 26 % of the total oxygen in the H2O-ZnO. After annealing, the total amount

of hydroxide oxygen decreases, resulting in a reduced [O−OH] : [O2−] ratio of 0.20 (i.e.

16 % of the total oxygen).

Almost the same behavior is observed for the O2precursor deposited zinc oxide. The cor-

responding O 1s photoemission spectrum of the as-grown ZnO is shown in Fig. 6.13 (c).

Initially, an oxygen ratio of [O−] : [O2−] = 0.35 is determined, which is almost the same

amount of [O−] oxygen than in the H2O-grown ZnO samples. After heating to 400 ◦C,

the high binding energy oxygen component is reduced by the same amount than in case

of H2O-ALD, resulting in a [O−] : [O2−] ratio of 0.20.

The peak positions and their particular widths of both oxygen components are also de-

termined for the as-grown and annealed O2-ZnO film. The intense [O2−] 1s photoelec-

trons have a binding energy of BE = 530.61 eV, which is almost equal to the H2O-ALD

[O2−] photoelectrons. In contrast, the peak is noticeable broader, having a peak width

of FWHM = 1.12 eV. In addition, the chemical shift of the [O−] component is more

pronounced in case of O2-ALD. Its binding energy has been determined being 532.07 eV,

141

6 Water-free ZnO ALD

Figure 6.13: O 1s regional photoemission spectra of ALD-ZnO using both oxygen pre-

cursors and their dependence on annealing. (a) H2O-ALD, as-grown; (b)

H2O-ALD, annealed to 400◦; (c) O2-ALD, as-grown; (d) O2-ALD, annealed

to 400◦. All spectra are recorded using monochromized Al Kαradiation.

resulting in a chemical shift of ∆BE = 1.46 eV. After annealing, the [O2−] decreases to

1.05 eV, a value similar to the one of [O2−] in H2O-ZnO. Both fitted peaks of the an-

nealed O2-ZnO O 1s photoemission line are shown in the lower right part of Fig. 6.13.

After the annealing step, the chemical shift of both oxygen species remains constant at

∆BE = 1.46 eV. The different chemical shift of the high binding energy oxygen compo-

nent in O2-ALD and H2O-ALD indicates a different chemical state of the [O−] oxygen in

comparison with [O−OH]. Hence, the [O−] obviously is not originating from hydroxides as

reaction products of the DEZn + O2reaction.

142

6.3 H2O-ALD vs. O2-ALD ZnO

In case of not being a reaction product of the DEZn + O2reaction, this oxygen component

is likely to originate from a defect in the ZnO crystal structure. There are several studies

about native defects in zinc oxide and their formation energies in literature[114–117]. Erhart

et al. reported about oxygen interstitial defects where two oxygen atoms occupy one

regular lattice site in a dumbbell configuration[114;117] as shown schematically in Fig. 6.14.

For oxygen-rich growth conditions the [O2−

2] dumbbell defects have the lowest formation

enthalpies of all intrinsic point defects[117]. Due to the sequential growth mode of atomic

layer deposition, the growth conditions during the oxidizing half-reaction can be regarded

as very oxygen-rich growth conditions. Therefore, the formation of [O2−

2] dumbbell defects

is very likely and they should be present in significant amounts in O2-ALD ZnO films.

Its singly negative charge per atom is comparable to the bonding conditions of hydroxide

oxygen in case of H2O-ALD. Therefore it is expected that the [O2−

2] dumbbell oxygen also

exhibits a comparable chemical shift towards higher binding energies and we can attribute

the high binding energy peak in case of O2-ALD to [O2−

2] dumbbell oxygen.

6.3.3 Non-destructive Photoemission Depth-Profiling

The oxygen components of both films show a similar behavior on annealing, as described

earlier in this section. For H2O-ALD, the decrease of the [O−OH] photoemission signal

can be attributed to a removal of hydroxides from the surface of the ZnO film. If all

hydroxides are located at the surface or if there is also a bulk hydroxide component

will be investigated in the following. The same question arises for the [O2−

2] dumbbell

oxygen in case of O2-ALD. To answer these questions, non-destructive depth profiling

Figure 6.14: Illustration of the ideal wurtzite structure of ZnO (left) and the O2−

2inter-

stitial dumbbell defect (right). The picture has been taken from Erhart et

al.[114]

143

6 Water-free ZnO ALD

using photoelectron spectroscopy is applied. Even though all photoelectron spectroscopy

techniques are very surface sensitive, it is possible to decrease the information depth of the

photoelectrons in two ways: Either, one tilts the surface under investigation to decrease

the effective escape depth of the photoelectrons (referred to as angle-resolved photoelectron

spectroscopy (AR-PES)). One other possibility is to change the excitation energy of the

X-rays and therefore the kinetic energy of the emitted photoelectrons. This requires a

tunable X-ray source and therefore a synchrotron radiation facility. Further information

about depth profiling was discussed earlier in Chp. 2. In the following, both techniques are

primary used to do a depth profiling of the oxygen in both ZnO films, respectively. This

will help to identify if surplus oxygen is located at the surface or constantly distributed

into the crystal for both oxygen precursors.

Figure 6.15: Angle-dependent X-ray photoemission spectra of ALD-ZnO using O2precur-

sor. Shown are the Zn 2p3/2, O 1s and Zn LMM photoemission spectra for

three exemplary exit angles. The exit angle is denoted with respect to the

film’s surface.

144

6.3 H2O-ALD vs. O2-ALD ZnO

After deposition, an angular-resolved depth profile of the two as-grown ZnO films was

recorded using monochromized Al Kαradiation and tilting the sample in the laboratory.

The left part of Fig. 6.15 plots the Zn 2p3/2at different exit angles, with respect to the

surface. Here, 90◦characterizes the emission of photoelectrons normal to the substrate’s

surface, while 10◦describes a grazing exit angle resulting in a very surface sensitive in-

formation depth. Fig. 6.15 only plots the spectra of three different emission angles to

illustrate the general behavior of the photoemission peaks. For detailed characterization

of the ZnO films, the exit angle of the photoelectrons was varied in steps of 10◦in the

range between 90◦- 50◦. Afterwards, in a range between 45◦- 10◦spectra were recorded

every 5◦because of the increased influence of the emission angle on the information depth.

The effective information depth of the photoelectrons entering the spectrometer depends

on the emission angle, the excitation energy and the binding energy of the photoelectrons.

For Al Kαexcited Zn 2p3/2photoelectrons, the effective information depth is in a range

between 16.9˚

A for an exit angle of 90◦with respect to the surface and 2.9˚

A for very flat

angles of 10◦.

Due to the drop in the photoemission intensities at small exit angles all spectra in Fig. 6.15

were normalized to 1 for ease of comparability. The Zn 2p3/2photoemission spectra do not

show any binding energy dependence on the exit angle. Besides the Zn 2p3/2, also the O 1s

photoelectrons were recorded as shown in the middle of Fig. 6.15 for three exemplary exit

angles. Its lower binding energy of about 530 eV leads to a slightly increased information

depth with respect to the Zn 2p photoelectrons. This results in an effective information

depth between 24.3−4.2˚

A for O 1s photoelectrons. In addition to the Zn 2p and O 1s

photoelectrons, the Zn L3M45M45 Auger electron peak was measured to determine the

depth dependence of the combined Auger parameter of the zinc component.

The spectra shown in Fig. 6.15 originate from a ZnO film using molecular oxygen as pre-

cursor gas. The same experiments described above were also done for H2O-ALD deposited

ZnO. Quantification of the measured spectra results in depth profiles of the composition,

contains information about the distribution of the two oxygen components present in the

ZnO crystal, and shows the depth dependence of the Auger parameter of the zinc compo-

nent. Fig. 6.16 plots all results for both oxygen precursors and is now discussed in detail.

Fig. 6.16 (a) shows the [Zn]:[O] composition over the effective information depth of the

photoelectrons before they undergo an inelastic scattering event. The results for H2O-

ALD are illustrated by the red markers, whereas the blue ones stand for the O2-ALD ZnO

film. In case of H2O-ALD, the composition remains constant at a [Zn]:[O] ratio around a

145

6 Water-free ZnO ALD

Figure 6.16: Results of angle-resolved XPS investigation of H2O-ALD ZnO (blue) and O2-

ALD ZnO (red). Reducing the exit angle of the photoelectrons with respect

to the detector increases the surface sensitivity of the measurements.

146

6.3 H2O-ALD vs. O2-ALD ZnO

stoichiometric value of 1.0 over a wide range. Only directly at the surface the composition

seems to increase slightly, resulting in zinc-rich conditions. The opposite is the case for

photoemission normal to the surface, where the [Zn]:[O] ratio suddenly drops to a value

around 0.90. If this decrease originates from oxygen-rich conditions deeper in the material

or is due to inaccuracy of the measurement and quantification process is not clear at this

moment.

The composition depth profile of the O2-ALD grown sample yields even more unclear re-

sults. The composition at the surface and deep in the film seems to be oxygen-rich, while

in between the composition drops to a value of [Zn]:[O] = 0.70. The scattering of the data

indicates an increased error of the AR-XPS results in case of O2-ALD and therefore, the

AR-XPS analysis of the ZnO layers seem not to provide reliable depth information about

the film composition.

The depth profile of the combined Auger parameter is shown in Fig. 6.16 (b). These

results look much more reliable than the composition depth profile. There is only slight

scattering of the data and the determined Auger parameters for both ZnO films lie on top

of each other at a value of 2009.8 eV. This value is exactly the one determined earlier for

both precursor combinations. It does not show any depth dependence and that is why it

is expected that the local chemical environment of the zinc component of the ZnO film

does not change strongly, but remains constant in the depth region under investigation.

Fig. 6.16 (c) shows the results of the O 1s analysis. There, the ratio of the high binding

energy component (i.e. [O2−

2] in case of O2-ALD and [O−OH] for H2O-ALD) and the [O2−]

oxygen is plotted against the information depth. In case of water as oxygen precursor, the

amount of hydroxides deeper in the ZnO crystal is about 35 % of the [O2−] component.

This corresponds to 26 % of the total amount of oxygen in the zinc oxide layer, as shown

earlier. Closer to the surface, a constant increase of the amount of hydroxides is observed,

reaching a level of 45 % with respect to the [O2−] oxygen and accordingly 30 % of the total

amount of oxygen in the ZnO. This is just a statistical value, but indicates the tendency

of hydroxides being located more at the surface than in the bulk of the ZnO. In case

of O2-ALD the scattering of the data is increased with respect to the H2O-ALD data.

Photoelectrons emitted from deeper regions of the ZnO (exit angles between 90 −70◦)

show a similar ratio of the two oxygen components as in case of water-ALD. In contrast,

the depth profile of the oxygen ratio is not as constant as the one for H2O-ALD. Similar

as in composition analysis shown above, the information content of the O2-ALD depth

profile is not very reliable.

147

6 Water-free ZnO ALD

Figure 6.17: Depth-profiles of the O 1s photoelectrons of H2O-ALD (left) and O2-ALD

(right). Changing the excitation energy results in different IMFPs of the

photoelectrons and therefore containing depth information of the particular

element under investigation.

148

6.3 H2O-ALD vs. O2-ALD ZnO

From the results just presented follows that we have to proof whether this inaccuracy is

caused by the sample itself or due to erroneous acquisition of the data during the measure-

ments. Therefore, the same samples investigated in the laboratory were transferred under

ultra-high vacuum conditions to the U49/2-PGM2 beamline of the BESSY II synchrotron

radiation facility. Using synchrotron radiation it is possible to acquire depth profiling in

two different ways. One the one hand, the analytic system allows to do angle-dependent

measurements using constant excitation energy as in case of AR-XPS measurements at

the Integrated System. Its advantage over Al Kαradiation used in the laboratory is

the possibility of choosing a constant kinetic energy of each element’s photoelectrons by

choosing different excitation energies. Having same kinetic energies, the inelastic mean

free path is the same for all photoelectrons, independent on their binding energies. Thus,

all photoelectrons detected are created at same depths and hence are better to compare

with each other.

The second possibility of recording a depth profile using synchrotron radiation is to vary

the excitation energy at normal emission from the surface. Thus, the information depth

where photoelectrons are emitted from is changed resulting in depth profiles of the par-

ticular elements. To distinguish between those two synchrotron-based depth profiling

methods, the term angle-resolved photoelectron spectroscopy (AR-PES) depth profiling

is used in case of changing the exit angle of the detected photoelectrons. In contrary,

if the kinetic energy of the photoelectrons is varied under normal emission with respect

to the surface, we use the general term synchrotron-radiation photoelectron spectroscopy

(SR-PES) depth profiling.

Fig. 6.17 shows such a series of O 1s spectra recorded by SR-PES, varying the kinetic

energy of the emitted photoelectrons between 80 and 750 eV. On the left, the spectra of

H2O-ZnO are presented, while the spectra on the right show an O2-ALD ZnO film. The

energy range is limited by two factors. For high kinetic energies, the signal intensity drops

significantly due to the reduced photoionization cross-section at high excitation energies.

In addition, the maximum excitation energy of the U49/2-PGM2 beamline is limited to

1890 eV. Hence, 750 eV is the highest kinetic energy that can be achieved for Zn 2p3/2with

its binding energy of about 1021 eV. The excitation of lower kinetic energy electrons is not

limited by the beamline, but by the increasing amount of secondary electron background

for such low kinetic energies. At kinetic energies below 80 eV, a background removal of

the secondary electrons just using a Shirley routine is not sufficient and therefore 80 eV

is chosen as the lower limit for the depth profiling experiments. In terms of inelastic

149

6 Water-free ZnO ALD

mean free paths, the depth information can be varied between 21.5˚

A and 7.8˚

A using this

method. The maximum intensity of the [O2−] peak of the spectra plotted in Fig. 6.17 is

normalized to 1. Two peaks are fitted, related to [O2−] and either [O−OH] or [O2−

2], as

demonstrated for the 80 eV photoemission spectrum in Fig. 6.17.

To reduce the surface sensitivity even more, AR-PES depth profiling at the same samples

was carried out. A kinetic photoelectron energy of 200 eV was chosen. At the one hand, at

this energy there are no Auger features overlapping the photoemission peak. On the other

hand, the secondary electron background is still removable by using a Shirley routine. The

smallest achievable exit angle of the photoelectrons was 15◦with respect to the electron

detector. This results in information depths of 11.4˚

A at normal emission and decreases

down to 2.9˚

A for flat angles.

The evaluation of the O 1s depth-profiling in terms of the [O−] : [O2−] ratio is presented

in Fig. 6.18. Here, the results for both synchrotron-based depth profiling methods are

combined. The left column of Fig. 6.18 presents the more surface sensitive AR-PES

results for both oxidizing agents. In addition, the right column illustrates the SR-PES

depth profiles containing information from deeper regions of the material.

The upper part of Fig. 6.18 presents the standard ZnO-ALD layers using water as oxygen

precursor, while the O2-ALD ZnO film is shown on the lower part. The filled circles

indicate the data of the as-grown ZnO samples and are discussed first for the case of

H2O-ALD. The corresponding depth-profiles of the [O−OH] : [O2−] ratio are plotted in

Fig. 6.18 (a) and (b), respectively. The AR-PES results show a very constant distribution

of the [O−OH] in a region directly at the surface. Here, the number of hydroxides amount

to 60 % of the [O2−] oxygen corresponding to 38 % of the total amount of oxygen in the

ZnO. At a depth of about 11 ˚

A the amount of [O−OH] starts decreasing constantly with

respect to the [O2−] oxygen component as the data of Fig. 6.18 (b) show. The kink visible

in the [O−OH] : [O2−] AR-PES depth profile at approximately 11 ˚

A is also observed for

the depth profile using the SR-PES method. This is giving an indication that the two

methods are providing reliable and consistent results of the material under investigation.

The open circles in Fig. 6.18 represent the same samples after an annealing for one hour

at approximately 500 ◦C. As expected, the amount of hydroxides at the surface is reduced

significantly by 20 % with respect to the [O2−] component. Interestingly, deeper in the

material (starting around 14 ˚

A) the amount of hydroxides is not reduced but remains

constant at the level before annealing. This indicates, that only the hydroxides located

directly at the surface are removed, but those buried deeper in the material are not subject

to any changes.

150

6.3 H2O-ALD vs. O2-ALD ZnO

Figure 6.18: Depth-profiling results of the O 1s of H2O-ZnO and O2-ZnO investigated by

synchrotron radiation before and after annealing the samples. Using differ-

ent depth-profiling methods, information from different depth-regions can be

investigated.

The depth profiling results of the O2-ALD ZnO show a different picture of the [O2−

2]:[O2−]

ratio, presented in Fig. 6.18 (c) and (d). In contrast to H2O-ALD, where a constant

amount of hydroxides is distributed over the first atomic layers, the number of [O2−

dumbbell oxygen is strongly increased. While the number of hydroxides in the H2O-ZnO

account for approx. 38 % of the total amount of oxygen in the ZnO, almost every second

oxygen species at the surface of O2-ALD ZnO belongs to dumbbell oxygen. Instead

151

6 Water-free ZnO ALD

Figure 6.19: Depth-profile of the two ALD-ZnO films using different oxygen precursors.

(a) presents the depth-dependent [Zn]:[O] ratio of the as-grown and annealed

H2O-ALD film, respectively, while (b) illustrates the O2-ALD ZnO layer.

of a constant distribution as in case of the hydroxides, the concentration of dumbbell

oxygen constantly decreases going deeper into the material. As in case of H2O-ZnO, an

equilibrium of the [O2−

2] concentration with respect to the [O2−] is not observed.

Annealing of the sample has a clearly increased effect on the oxygen dumbbell defect as

on the hydroxides. While the number of hydroxides were were decreased by 20 % with

respect to the [O2−] component, the oxygen dumbbell defects are strongly reduced after

annealing the sample for one hour at 500 ◦C. Furthermore, the remaining defects are

distributed more uniformly into the crystal. In contrast to the hydroxide oxygen, the

dumbbell defects are not only reduced at the surface but also deeper in the material,

where they almost completely disappear after annealing. Hence, it is believed that during

the annealing process the interstitial dumbbell oxygen is incorporated into the ZnO crystal

lattice. This results in a reduced amount of dumbbell defects and indicates the formation

of ideal wurtzite ZnO shown in Fig. 6.14.

In addition to the O 1s photoemission, the zinc component of the two ZnO films has

been depth-profiled by the AR-PES method. This allows to determine depth-dependent

composition profiles of the ZnO films. For H2O-ZnO, the [Zn]:[O] depth-profiles for the

as-deposited and the annealed film are shown in Fig. 6.19 (a), whereas the O2-ZnO case

152

6.3 H2O-ALD vs. O2-ALD ZnO

is illustrated in Fig. 6.19 (b). Ideal stoichiometric composition is indicated by the dashed

line in both plots. Again, the data of the annealed layers are illustrated by open circles.

For the as-grown H2O-ZnO in Fig. 6.19 (a), a surface composition of 0.9 is determined,

i.e. the film is slightly oxygen-rich. Deeper in the material, a considerable change of

the film’s composition is observed. In this region around 10 ˚

A, the films is closer to

stoichiometry, but still slightly oxygen-rich. At even deeper regions, the ZnO composition

again changes to more oxygen-rich conditions. Annealing of the sample did not change

its composition characteristics. As expected, the film turns more stoichiometric at the

surface due to the removal of surface hydroxides. In agreement with the O 1s depth-

profiling results presented in Fig. 6.18, the composition remains constant deeper in the

material, also showing the characteristic compositional change deeper in the material.

It is assumed that these variations in composition originate from undefined deviations

during the deposition process and does not represent all ZnO films grown by atomic layer

deposition using water as oxidizing agent.

The zinc oxide film deposited by O2-ALD does not show such variations in the composition

depth-profile. The as-deposited ZnO shows a comparable surface composition to the H2O-

ALD ZnO of [Zn]:[O] ≈0.9, slightly increasing towards a more stoichiometric composition

deeper in the material. As previously shown, the interstitial oxygen is drastically reduced

during annealing of the sample. Above it was assumed that the interstitial dumbbell

oxygen is incorporated into the ZnO crystal. If this is the case, it should result in an

observed composition close to the stoichiometric value of [Zn]:[O] = 1.00. Surprisingly,

this is not the case. Fig. 6.19 (b) clearly reveals that the composition of the ZnO film even

turns zinc-rich after annealing. A constant composition of [Zn]:[O] = 1.06 is determined

throughout the whole information depth of the AR-PES depth profiling experiments.

This might be explained by not only incorporating the oxygen in the ZnO lattice, but

the complete removal of those dumbbell structures from the crystal. This would result in

oxygen vacancies [VO]. As mentioned in the beginning of this chapter, oxygen vacancies

are one of those point defects that are believed to prevent p-type doping of the ZnO. This

would also explain, why there was no change observed in the valence band structure of

the annealed O2-ZnO in Fig. 6.12.

153

6 Water-free ZnO ALD

6.4 Summary

In this chapter, the first successful study of a water-free ZnO ALD process using

pure O2as oxidizing precursor and organo-metallic DEZn as metal-precursor has been

presented. Even though Nalwa et al. stated that molecular O2is too inert to react with

an organometallic precursor[54], it was shown that this is obviously not the case for the

reaction O2+ Zn(C2H5)2. Instead, the typical self-limited ALD growth behavior was

demonstrated. A strong increase in growth rate in the ALD window temperature

regime that is close to one completed ZnO monolayer per ALD cycle was observed. This

indicates a completely different reaction mechanism that is not limited by steric hin-

drance effects of the ethyl-ligands. Further evidence of the differences in growth mech-

anism is given by the appearance of two additional carbon features in the photoelectron

spectra. These are attributed to hydrocarbons and carbon in a carboxylic environment,

respectively.

Investigation of the initial growth on Si(111)-H show no significant differences between

the ZnO films using H2O and O2precursors. As in case of H2O-ALD, a second feature

at high binding energies of the [O2−] lattice oxygen is present in the O 1s photoemission

spectra. While using water as oxygen precursor the second component clearly can be

attributed to hydroxide oxygen, its origin is not clear in case of O2-ALD.

To identify the additional oxygen component, the two particular ZnO films were investi-

gated at the synchrotron radiation facility BESSY II. Photoelectron spectroscopy in the

laboratory and non-destructive depth-profiling of the as-deposited and annealing samples

at the BESSY II suggest the high binding energy oxygen component originating from

an interstitial dumbbell defect as stated by Erhart et al.[114;117]. Heat treatment of

the samples not only removes the surface carbon contamination but almost completely

removes the interstitial oxygen defects by incorporation into the ZnO lattice. While the

H2O-ZnO films usually are oxygen-rich, the annealed O2-ZnO ALD films show a slightly

zinc-rich composition. This, and the similarities of the valence bands indicate the pres-

ence of oxygen vacancies in case of annealed O2-ZnO. Depth-profiling reveals that the

composition is spread across the information depth observable by synchrotron radiation

depth-profiling very homogeneously. Hence, the water-free ALD of ZnO can be regarded

being successful producing films of the same qualities as the standard ALD films.

154

7 Conclusion and Outlook

Atomic layer deposition is possibly the most promising technique for alternative buffer

layer deposition in chalcopyrite solar cell production. Not only its potential of depositing

high-quality material films with utmost film conformity even in the monolayer thickness

regime, but also the promising results in terms of solar cell efficiency shown by other

research groups in the recent years.

The primary goal of this thesis was the investigation of the interface formation and char-

acterization of the interface’s electronic properties between ZnO and CuInSe2, grown epi-

taxially in the (112) orientation. Since all experiments performed in this thesis are very

surface sensitive, any surface modifications have to be prevented until the sample was

characterized by photoelectron spectroscopy. Hence, all deposition and analysis experi-

ments were performed under ultra-high vacuum conditions without exposing the samples

to air. For this purpose, a UHV-compatible ALD reactor was designed, assembled and

commissioned for the deposition of zinc oxide using the standard precursor combination

diethylzinc and water as reactive gases. The attachment of the atomic layer deposition

chamber to the integrated deposition and analysis system in the laboratory allows in situ

measurements of the specimen. It has been demonstrated that the deposition system

shows the characteristic self-limited film growth in a temperature regime between approx-

imately 200 −225 ◦C, the so-called ALD window. In this temperature regime, a constant

deposition rate of 3.0˚

A per ALD cycle is observed. Characterization of stepwise deposited

ZnO films clearly reveal a layer-by-layer growth mode as it is expected for atomic layer

deposition. The initial growth of ZnO on hydrogen-terminated Si(111)-H substrates indi-

cate an oxygen-rich interface formation of the ZnO films, but does not show any oxidation

of the subjacent silicon. After having the thickness of a few monolayers, the [Zn]:[O] ratio

shows stoichiometric composition of the zinc oxide films.

155

7 Conclusion and Outlook

After demonstrating successful ZnO atomic layer deposition, the interface formation of

ZnO and epitaxial CuInSe2(112) was investigated. Earlier studies already investigated the

ZnO deposition by metal-organic molecular beam epitaxy (MOMBE) on epitaxial chal-

copyrite absorbers. In contrast to ALD, the MOMBE process exposes the sample to the

two precursor gases simultaneously and the deposition takes place at significantly higher

substrate temperatures. Photoemission spectroscopy results indicated the formation of an

intrinsically formed ZnSe interface layer between CuInSe2and ZnO. The formation mech-

anism is believed to be induced by self-compensation effects on doping the chalcopyrite.

Self-compensation results in the formation of Cu vacancies, while the copper diffuses away

from the surface at the same time. An intrinsic interface formation is also found in this

study, even though some major differences are observed either due to the lower substrate’s

temperature or the differing growth kinetics of the ALD process. On the one hand, in-

stead of ZnSe a ZnIn2Se4(ZISe) layer forms at the interface of CuInSe2and ALD-ZnO.

Its layer thickness is considerably reduced in case of atomic layer deposition. While in

the MOMBE case the buffer layer thickness was determined being 2 nm, Auger parameter

analysis of the zinc component clearly reveals a ZnIn2Se4thickness of only 0.5 nm, i.e. the

range of one monolayer. Photoelectron spectroscopy indicates a promising effect of the

intrinsic ZISe buffer layer on the electronic band alignment of the CuInSe2|ZnIn2Se4|ZnO

system for chalcopyrite thin-film solar cell devices. In addition, MOMBE results showed

a completely oxygen-free ZnSe buffer layer, while during the ALD process small amounts

of hydroxide are found in the boundary layer. Their incorporation into the ZISe are most

likely to be explained by the sequential deposition mode of the precursor gases in atomic

layer deposition. According to several other studies, buffer layers containing consider-

able amounts of hydroxides, such as Zn(OH,Se), Zn(OH,S) or In(OH,S), show increased

solar cell efficiencies. Hence, the additional hydroxide oxygen found in the intrinsically

formed buffer layer is a very interesting result of this study. Annealing experiments of

ZnO deposited on CuInSe2indicated the same indium diffusivity as in case of MOMBE

ZnO deposition. This indium diffusion might also have beneficial effects on the device

efficiencies, since it can reduce interfacial defects and hence lower possible interface re-

combination of the electrons.

In the final chapter of this thesis, successful water-free atomic layer deposition using metal-

organic diethylzinc and molecular oxygen (O2) has been demonstrated for the first time.

A water-free ZnO-ALD process might be able to reduce the amount of interstitial hydro-

gen atoms acting as donors in ZnO. Just like the standard ALD precursor combination of

diethylzinc and water, water-free atomic layer deposition shows the typical self-limited de-

156

position rate behavior in the temperature regime known as the ALD window. In contrast,

the deposited amount of material per ALD cycle is significantly increased to 5.0˚

A/cycle,

i.e. almost one complete ZnO monolayer per ALD cycle. While in standard ALD, steric

hindrance effects of the ethyl ligands are responsible for a limitation of the growth rate,

the high deposition rate in water-free ALD indicate a completely different growth mech-

anism. This is also emphasized by the appearance of two additional carbon features

observed in photoelectron spectroscopy. Those are attributed to hydrocarbons and car-

bon in an carboxylic environment and can be removed almost completely by annealing

the samples. While the initial growth on Si(111)-H does not show significant differences

to H2O-ALD, the origin of a second oxygen component visible in the photoemission spec-

tra is not likely originating from hydroxides. Instead, non-destructive depth profiling

and annealing experiments at the BESSY II synchrotron radiation source led to the con-

clusion that this additional oxygen component originates from oxygen-dumbbell defects,

that are almost complete removed after annealing the samples for 1 hour at 500 ◦C. In

contrast to H2O-ALD, the annealed ZnO films grown by water-free ALD show a slightly

zinc-rich composition and synchrotron-radiation depth profiling indicates very conformal

ZnO layers. Hence, water-free ALD of ZnO using diethylzinc and molecular oxygen do

not show any structural, chemical or electronic disadvantages to the standard precursor

combination.

Outlook

After successful demonstration of water-free atomic layer deposition of zinc oxide in this

thesis, further work on the fundamental properties of those layers are of interest. Like

several other wide band gap materials, ZnO is naturally n-doped and it is very difficult

to obtain an effective and reproducible p-type doping. This is mainly resulting from

compensation effects of point defects such as interstitial zinc atoms or oxygen vacancies.

But also intrinsic hydrogen might play a crucial role, as it acts as an donor in the zinc

oxide crystal. In atomic layer deposition of ZnO, water is often used as oxygen source in

the oxidizing half-reaction. Preventing the use of water, which acts as a hydrogen source,

might result in reduced amounts of hydrogen in the ZnO crystal. It is recommended to

investigate if this is the case in water-free ALD zinc oxide layers in future work.

157

Bibliography

[1] REN21. Renewables 2011 Global Status Report, 2011.

[2] Razykov, T., Ferekides, C., Morel, D., Stefanakos, E., Ullal, H. and

Upadhyaya, H. Solar photovoltaic electricity: Current status and future prospects.

Solar Energy, 85 (8), (2011), pp. 1580 – 1608. ISSN 0038-092X.

[3] ZSW Press Release 10/2011; http://www.zsw-bw.de/fileadmin/

ZSW_files/Infoportal/Presseinformationen/docs/pi11-2010-e_

ZSW-Weltrekord-DS-CIGS.pdf.

[4] Kasap, S., Koughia, C., Ruda, H. and Johanson, R. Electrical Conduction

in Metals and Semiconductors. In S. Kasap and P. Capper (Editors), Springer

Handbook of Electronic and Photonic Materials, pp. 19–45 (Springer, 2007). ISBN

978-0-387-29185-7.

[5] Rockett, A. Current status and opportunities in chalcopyrite solar cells. Current

Opinion in Solid State and Materials Science, 14 (6), (2010), pp. 143–148. ISSN

1359-0286.

[6] Rau, U. and Schock, H.W. Cu(In,Ga)Se2thin-film solar cells. In T. Markvart

and L. Casta˜ner (Editors), Solar Cells, pp. 303–349 (Elsevier Science, Oxford, 2005).

[7] Shay, J. and Wernick, J. Ternary chalcopyrite semiconductors: growth, elec-

tronic properties, and applications. International series of monographs in the science

of the solid state (Pergamon Press, 1975). ISBN 978-0-080-17883-7.

[8] Zhang, S.B., Wei, S.H., Zunger, A. and Katayama-Yoshida, H. Defect

159

Bibliography

physics of the CuInSe2chalcopyrite semiconductor. Phys. Rev. B, 57 (16), (1998),

pp. 9642–9656.

[9] Hofmann, A. Elektronische Struktur epitaktischer CuInSe2-Schichten und deren

Heterokontakte f¨

ur die Photovoltaik. Ph.D. thesis, Brandenburgische Technische

Universit¨

at Cottbus, 2011.

[10] Hariskos, D., Spiering, S. and Powalla, M. Buffer layers in Cu(In,Ga)Se2

solar cells and modules. Thin Solid Films, 480-481, (2005), pp. 99–109. ISSN 0040-

6090.

[11] Bhattacharya, R. and Ramanathan, K. Cu(In,Ga)Se2thin film solar cells

with buffer layer alternative to CdS. Solar Energy, 77 (6), (2004), pp. 679 – 683.

[12] King, P.D.C., Veal, T.D., Fuchs, F., Wang, C.Y., Payne, D.J.,

Bourlange, A., Zhang, H., Bell, G.R., Cimalla, V., Ambacher, O.,

Egdell, R.G., Bechstedt, F. and McConville, C.F. Band gap, electronic

structure, and surface electron accumulation of cubic and rhombohedral In2O3.

Physical Review B, 79 (20), (2009), p. 205211.

[13] Spiering, S., Eicke, A., Hariskos, D., Powalla, M., Naghavi, N. and

Lincot, D. Large-area Cd-free CIGS solar modules with In2S3buffer layer de-

posited by ALCVD. Thin Solid Films, 451-452 (0), (2004), pp. 562 – 566. ISSN

0040-6090. Proceedings of Symposium D on Thin Film and Nano-Structured Ma-

terials for Photovoltaics, of the E-MRS 2003 Spring Conference.

[14] Yousfi, E.B., Weinberger, B., Donsanti, F., Cowache, P. and Lincot,

D. Atomic layer deposition of zinc oxide and indium sulfide layers for Cu(In,Ga)Se2

thin-film solar cells. Thin Solid Films, 387 (1-2), (2001), pp. 29 – 32. ISSN 0040-

6090.

[15] Platzer-Bj¨

orkman, C. Zn(O,S) buffer layers by atomic layer deposition in

Cu(In,Ga)Se2based thin film solar cells: Band alignment and sulfur gradient. Jour-

nal of Applied Physics, 100 (4), (2006), p. 044506.

[16] Yousfi, E.B., Asikainen, T., Pietu, V., Cowache, P., Powalla, M. and

Lincot, D. Cadmium-free buffer layers deposited by atomic later epitaxy for

160

Bibliography

copper indium diselenide solar cells. Thin Solid Films, 361-362, (2000), pp. 183 –

186. ISSN 0040-6090.

[17] Platzer-Bj¨

orkman, C., Lu, J., Kessler, J. and Stolt, L. Interface study of

CuInSe2/ZnO and Cu(In,Ga)Se2/ZnO devices using ALD ZnO buffer layers. Thin

Solid Films, 431-432 (0), (2003), pp. 321 – 325. ISSN 0040-6090. Proceedings

of Symposium B, Thin Film Chalcogenide Photovoltaic Materials, E-MRS Spring

Meeting.

[18] Contreras, M.A., Egaas, B., Ramanathan, K., Hiltner, J., Swartz-

lander, A., Hasoon, F. and Noufi, R. Progress toward 20% efficiency in

Cu(In,Ga)Se2polycrystalline thin-film solar cells. Progress in Photovoltaics: Re-

search and Applications, 7 (4), (1999), pp. 311–316.

[19] Janotti, A. and van de Walle, C.G. Fundamentals of zinc oxide as a semi-

conductor. Reports on Progress in Physics, 72 (12), (2009), p. 126501.

[20] ¨

Ozg¨

ur, U., Alivov, Y.I., Liu, C., Teke, A., Reshchikov, M.A., Dogan,

S., Avrutin, V., Cho, S.J. and Morkoc, H. A comprehensive review of ZnO

materials and devices. Journal of Applied Physics, 98 (4), 041301.

[21] Andres, S., Lehmann, C. and Pettenkofer, C. Epitaxial growth of ZnO on

CuInS2(112). Thin Solid Films, 518 (4), (2009), pp. 1032 – 1035.

[22] Hofmann, A. Surface orientation dependent band alignment for CuInSe2- ZnSe -

ZnO. Applied Physics Letters, 98 (11), (2011), p. 113503.

[23] Rockett, A., Abou-Elfotouh, F., Albin, D., Bode, M., Ermer, J.,

Klenk, R., Lommasson, T., Russell, T., Tomlinson, R., Tuttle, J.,

Stolt, L., Walter, T. and Peterson, T. Structure and chemistry of CuInSe2

for solar cell technology: current understanding and recommendations. Thin Solid

Films, 237 (1-2), (1994), pp. 1 – 11. ISSN 0040-6090.

[24] Einstein, A. ¨

Uber einen die Erzeugung und Verwandlung des Lichtes betreffenden

heuristischen Gesichtspunkt. Annalen der Physik, 322 (6), (1905), pp. 132–148.

[25] H¨

ufner, S. Photoelectron Spectroscopy - Principles and Applications (Springer,

161

Bibliography

2003), 3rd edn.

[26] Bubert, H. and Rivi`

ere, J.C. Surface and Thin Film Analysis (Wiley-VCH

Verlag GmbH, 2002). ISBN 978-3-527-60016-8.

[27] Ratner, B.D. and Castner, D.G. Surface Analysis - The Principal Techniques,

chap. Electron Spectroscopy for Chemical Analysis, pp. 47–112 (John Wiley & Sons,

Ltd, 2009). ISBN 978-0-470-72158-2.

[28] Alford, T., Feldman, L. and Mayer, J. X-ray Photoelectron Spectroscopy. In

Fundamentals of Nanoscale Film Analysis, pp. 199–213 (Springer US, 2007). ISBN

978-0-387-29261-8.

[29] NIST X-ray Photoelectron Spectroscopy Database, Version 3.5 (National Institute

of Standards and Technology, Gaithersburg, 2003; http://srdata.nist.gov/xps/.

[30] Wagner, C. and Joshi, A. The Auger Parameter, its Utility and Advantages: A

Review. Journal of Electron Spectroscopy and Related Phenomena, 47, (1988), pp.

283 – 313. ISSN 0368-2048.

[31] Sch¨

on, G. Auger and direct electron spectra in X-ray photoelectron studies of

zinc, zinc oxide, gallium and gallium oxide. Journal of Electron Spectroscopy and

Related Phenomena, 2 (1), (1973), pp. 75 – 86. ISSN 0368-2048.

[32] UK Surface Analysis Forum, http://www.uksaf.org/data/table.html.

[33] Moretti, G. Auger parameter and Wagner plot in the characterization of chem-

ical states by X-ray photoelectron spectroscopy: a review. Journal of Electron

Spectroscopy and Related Phenomena, 95 (2-3), (1998), pp. 95 – 144. ISSN 0368-

2048.

[34] Seah, M.P. and Dench, W.A. Quantitative electron spectroscopy of surfaces:

A standard data base for electron inelastic mean free paths in solids. Surface and

Interface Analysis, 1 (1), (1979), pp. 2–11. ISSN 1096-9918.

[35] Shirley, D.A. High-Resolution X-Ray Photoemission Spectrum of the Valence

Bands of Gold. Physical Review B, 5 (12), (1972), p. 4709.

162

Bibliography

[36] Wagner, C.D., Davis, L.E., Zeller, M.V., Taylor, J.A., Raymond, R.H.

and Gale, L.H. Empirical atomic sensitivity factors for quantitative analysis by

electron spectroscopy for chemical analysis. Surface and Interface Analysis, 3 (5),

(1981), pp. 211–225. ISSN 1096-9918.

[37] Jablonski, A. and Zemek, J. Overlayer thickness determination by XPS using

the multiline approach. Surface and Interface Analysis, 41 (3), (2009), pp. 193–204.

[38] Yeh, J. and Lindau, I. Atomic subshell photoionization cross sections and asym-

metry parameters: 1 ≤Z≤103. Atomic Data and Nuclear Data Tables, 32 (1),

(1985), pp. 1 – 155. ISSN 0092-640X.

[39] Hove, M., Weinberg, W. and Chan, C. Low-energy electron diffraction: ex-

periment, theory and surface structure determination. Springer series in surface

sciences (Springer-Verlag, 1986). ISBN 978-3-540-16262-9.

[40] Pendry, J. Low energy electron diffraction: the theory and its application to

determination of surface structure. Techniques of physics (Academic Press, 1974).

ISBN 978-0-125-50550-5.

[41] Ertl, G. and K¨

uppers, J. Low energy electrons and surface chemistry (VCH,

1985). ISBN 978-3-527-26056-0.

[42] Rockett, A. The Materials Science of Semiconductors (Springer Publishing Com-

pany, Incorporated, 2007), 1st edn. ISBN 978-0-387-25653-5.

[43] Ohring, M. The Materials Science of Thin Films (Academic Press, 2002).

[44] Gould, R. Thin Films. In S. Kasap and P. Capper (Editors), Springer Handbook

of Electronic and Photonic Materials, pp. 659–716 (Springer, 2007). ISBN 978-0-

387-29185-7.

[45] Klimova, A., Ananichev, V., Arif, M. and Blinov, L. Investigation of the

Saturated Vapor Pressure of Zinc, Selenium, and Zinc Selenide. Glass Physics and

Chemistry, 31, (2005), pp. 760–762. ISSN 1087-6596.

[46] Capper, P., Irvine, S. and Joyce, T. Epitaxial Crystal Growth: Methods and

163

Bibliography

Materials. In S. Kasap and P. Capper (Editors), Springer Handbook of Electronic

and Photonic Materials, pp. 271–301 (Springer, 2007). ISBN 978-0-387-29185-7.

[47] Knudsen, M. Die Gesetze der Molekularstr¨

omung und der inneren Reibungsstr¨

mung der Gase durch R¨

ohren. Annalen der Physik, 333 (1), (1909), pp. 75–130.

ISSN 1521-3889.

[48] Hampden-Smith, M.J. and Kodas, T.T. Chemical vapor deposition of metals:

Part 1. An overview of CVD processes. Chemical Vapor Deposition, 1 (1), (1995),

pp. 8–23. ISSN 1521-3862.

[49] Carlsson, J.O. and Martin, P.M. Chemical Vapor Deposition. In P.M. Mar-

tin (Editor), Handbook of Deposition Technologies for Films and Coatings (Third

Edition), pp. 314 – 363 (William Andrew Publishing, Boston, 2010). ISBN 978-0-

815-52031-3.

[50] Putz, N., Veuhoff, E., Heinecke, H., Heyen, M., Luth, H. and Balk, P.

GaAs Growth In Metal-Organic MBE. Journal of Vacuum Science & Technology

B, 3 (2), (1985), pp. 671–673.

[51] Andres, S., Pettenkofer, C., Speck, F. and Seyller, T. Morphology

and electronic properties of metal organic molecular beam epitaxy grown ZnO on

hydrogen passivated 6H-SiC(0001). Journal of Applied Physics, 103 (10), (2008), p.

103720.

[52] US Patent No. 4,058,430, 1977.

[53] George, S.M. Atomic Layer Deposition: An Overview. Chemical Reviews, 110 (1),

(2010), pp. 111–131. ISSN 0009-2665.

[54] Ritala, M. and Leskel¨

a, M. Atomic Layer Deposition (Academic Press,

Burlington, 2002). ISBN 978-0-12-512908-4.

[55] Puurunen, R.L. Surface chemistry of atomic layer deposition: A case study for

the trimethylaluminum/water process. Journal of Applied Physics, 97 (12), 121301.

[56] Suntola, T. Atomic Layer Epitaxy. Materials Science Reports, 4 (5), (1989), pp.

164

Bibliography

265–312. ISSN 0920-2307.

[57] Puurunen, R. Growth Per Cycle in Atomic Layer Deposition: Real Application

Examples of a Theoretical Model. Chemical Vapor Deposition, 9 (6), (2003), pp.

327–332. ISSN 1521-3862.

[58] Kuniya, Y., Deguchi, Y. and Ichida, M. Physicochemical properties of

dimethylzinc, dimethylcadmium and diethylzinc. Applied Organometallic Chem-

istry, 5 (4), (1991), pp. 337–348. ISSN 1099-0739.

[59] Fulem, M., Ruzicka, K., Ruzicka, V., Hulicius, E., Simecek, T.,

Melichar, K., Pangr´

ac, J., Rushworth, S.A. and Smith, L.M. Vapor

pressure of metal organic precursors. Journal of Crystal Growth, 248, (2003), pp.

99 – 107.

[60] NIST Chemistry WebBook, 2011; http://webbook.nist.gov/chemistry/.

[61] Stull, D. Vapor Pressure of Pure Substances Organic Compounds. Industrial &

Engineering Chemistry, 39, (1947), pp. 517–540.

[62] Bridgeman, O. and Aldrich, E. Vapor Pressure Tables for Water. Journal of

Heat Transfer, 86, (1964), pp. 279–286.

[63] Kopalko, K., Godlewski, M., Domagala, J.Z., Lusakowska, E.,

Minikayev, R., Paszkowicz, W. and Szczerbakow, A. Monocrystalline

ZnO Films on GaN/Al2O3by Atomic Layer Epitaxy in Gas Flow. Chemistry of

Materials, 16 (8), (2004), pp. 1447–1450.

[64] Mar, L.G., Timbrell, P.Y. and Lamb, R.N. An XPS study of zinc oxide thin

film growth on copper using zinc acetate as a precursor. Thin Solid Films, 223 (2),

(1993), pp. 341 – 347. ISSN 0040-6090.

[65] Ye, J., Gu, S., Zhu, S., Liu, S., Liu, W., Zhou, X., Hu, L., Zhang, R.,

Shi, Y. and Zheng, Y. Comparative study of diethylzinc and dimethylzinc for

the growth of ZnO. Journal of Crystal Growth, 274 (3-4), (2005), pp. 489 – 494.

ISSN 0022-0248.

165

Bibliography

[66] Angermann, H., Henrion, W., Rebien, M. and R¨

oseler, A. Wet-chemical

passivation and characterization of silicon interfaces for solar cell applications. Solar

Energy Materials and Solar Cells, 83 (4), (2004), pp. 331 – 346.

[67] Guziewicz, E., Kowalik, I.A., Godlewski, M., Kopalko, K., Osinniy,

V., Wojcik, A., Yatsunenko, S., Lusakowska, E., Paszkowicz, W. and

Guziewicz, M. Extremely low temperature growth of ZnO by atomic layer depo-

sition. Journal of Applied Physics, 103 (3), 033515.

[68] Kim, S.K., Hwang, C.S., Park, S.H.K. and Yun, S.J. Comparison between

ZnO films grown by atomic layer deposition using H2O or O3as oxidant. Thin Solid

Films, 478 (1-2), (2005), pp. 103 – 108. ISSN 0040-6090.

[69] Kim, D., Kang, H., Kim, J.M. and Kim, H. The properties of plasma-enhanced

atomic layer deposition (ALD) ZnO thin films and comparison with thermal ALD.

Applied Surface Science, 257 (8), (2011), pp. 3776 – 3779. ISSN 0169-4332.

[70] Park, S.H.K. and Lee, Y.E. Controlling preferred orientation of ZnO thin films

by atomic layer deposition. Journal of Materials Science, 39, (2004), pp. 2195–2197.

ISSN 0022-2461.

[71] Yamada, A., Sang, B. and Konagai, M. Atomic layer deposition of ZnO

transparent conducting oxides. Applied Surface Science, 112, (1997), pp. 216 – 222.

ISSN 0169-4332.

[72] PCPDFWIN Database.

[73] Matsubara, K., Fons, P., Yamada, A., Watanabe, M. and Niki, S. Epi-

taxial growth of ZnO thin films on LiNbO3substrates. Thin Solid Films, 347 (1-2),

(1999), pp. 238 – 240. ISSN 0040-6090.

[74] Libralesso, L., Schroeder, T., Lee, T.L. and Zegenhagen, J. Initial stages

of the epitaxial growth of Pr2O3on Si(111) studied by LEED and STM. Surface

Science, 598 (1-3), (2005), pp. L347–L354. ISSN 0039-6028.

[75] Brennan, B., McDonnell, S. and Hughes, G. Photoemission studies of the

166

Bibliography

initial interface formation of ultrathin MgO dielectric layers on the Si(111) surface.

Thin Solid Films, 518 (8), (2010), pp. 1980 – 1984.

[76] Dupin, J.C., Gonbeau, D., Vinatier, P. and Levasseur, A. Systematic XPS

studies of metal oxides, hydroxides and peroxides. Physical Chemistry Chemical

Physics, 2 (6), (2000), pp. 1319–1324.

[77] Au, C., Roberts, M. and Zhu, A. Surface hydroxylation at a Zn(0001)-O

surface. Surface Science Letters, 115 (2), (1982), pp. L117 – L123.

[78] Meier, U. and Pettenkofer, C. Morphology of the Si-ZnO interface. Applied

Surface Science, 252 (4), (2005), pp. 1139 – 1146. ISSN 0169-4332.

[79] Moulder, J.F., Stickle, W.F. and Sobol, P.E. Handbook of X Ray Pho-

toelectron Spectroscopy (Perkin-Elmer, Physical Electronics Division, 1992). ISBN

0962702625.

[80] Ueno, K. Heteroepitaxial growth of layered transition metal dichalcogenides on

sulfur-terminated GaAs{111}surfaces. Applied Physics Letters, 56 (4), (1990), p.

327.

[81] Babu, P., Reddy, M.V., Revathi, N. and Reddy, K.T.R. Effect of pH on the

Physical Properties of ZnIn2Se4Thin Films Grown by Chemical Bath Deposition.

Journal of Nano- and Electronic Physics, 3 (1), (2011), pp. 85 – 91.

[82] Ohtake, Y., Okamoto, T., Yamada, A., Konagai, M. and Saito, K.

Improved performance of Cu(InGa)Se2thin-film solar cells using evaporated Cd-

free buffer layers. Solar Energy Materials and Solar Cells, 49 (1-4), (1997), pp. 269

– 275. ISSN 0927-0248.

[83] Sun, X., He, Y. and Feng, J. Growth and characterization of ZnIn2Se4buffer

layer on CuInSe2thin films. Journal of Crystal Growth, 312 (1), (2009), pp. 48 –

51. ISSN 0022-0248.

[84] Hendia, T. and Soliman, L. Optical absorption behaviour of evaporated

ZnIn2Se4thin films. Thin Solid Films, 261 (1-2), (1995), pp. 322 – 327. ISSN

0040-6090.

167

Bibliography

[85] Choe, S.H. Optical energy gaps of undoped and Co-doped ZnIn2Se4single crystals.

Current Applied Physics, 9 (1), (2009), pp. 1 – 3.

[86] Margaritondo, G., Katnani, A.D. and L˜

Al’vy, F. Electronic States of

CdIn2Se4and ZnIn2Se4: Role of the Cation Pseudovacancies. Physica Status Solidi

B, 103 (2), (1981), pp. 725–731. ISSN 1521-3951.

[87] Locmelis, S., Milke, E., Binnewies, M., Gruhl, S. and Vogt, C. Chemis-

cher Transport fester L¨

osungen. 18 [1] Der Chemische Transport von ZnIn2S4,

ZnIn2Se4und ZnIn2S4-xSex-Mischphasen. Zeitschrift f¨

ur Anorganische und Allge-

meine Chemie, 631 (9), (2005), pp. 1667–1674. ISSN 1521-3749.

[88] Persson, C., Zhao, Y.J., Lany, S. and Zunger, A. n-type doping of CuInSe2

and CuGaSe2.Physical Review B, 72 (3), (2005), p. 035211.

[89] Desnica, U.V. Doping limits in II-VI compounds - Challenges, problems and

solutions. Progress in Crystal Growth and Characterization of Materials, 36 (4),

(1998), pp. 291 – 357.

[90] Klein, A. Fermi-level-dependent defect formation in Cu-chalcopyrite semiconduc-

tors. Applied Physics Letters, 74 (16), (1999), p. 2283.

[91] Klenk, R., Klaer, J., K¨

oble, C., Mainz, R., Merdes, S., Rodriguez-

Alvarez, H., Scheer, R. and Schock, H. Development of CuInS2-based solar

cells and modules. Solar Energy Materials and Solar Cells, 95 (6), (2011), pp. 1441

– 1445. ISSN 0927-0248. Special Issue : Thin film and nanostructured solar cells.

[92] Walukiewicz, W. Defect formation and diffusion in heavily doped semiconduc-

tors. Physical Review B, 50 (8), (1994), pp. 5221–5225.

[93] Van de Walle, C. Hydrogen as a cause of doping in zinc oxide. Physical Review

Letters, 85 (5), (2000), pp. 1012–1015. ISSN 0031-9007.

[94] Friedrich, F. and Nickel, N.H. Resonant Raman scattering in hydrogen and

nitrogen doped ZnO. Applied Physics Letters, 91 (11), (2007), p. 111903.

[95] Nickel, N.H. Hydrogen migration in single crystal and polycrystalline zinc oxide.

168

Bibliography

Physical Review B, 73 (19), (2006), p. 195204.

[96] Nickel, N.H. and Brendel, K. Hydrogen density-of-states distribution in zinc

oxide. Physical Review B, 68 (19), (2003), p. 193303.

[97] Nickel, N.H. and Fleischer, K. Hydrogen Local Vibrational Modes in Zinc

Oxide. Physical Review Letters, 90 (19), (2003), p. 197402.

[98] Kaiya, K., Yoshii, N., Takahashi, N. and Nakamura, T. Atmospheric

pressure atomic layer epitaxy of ZnO on a sapphire (0001) substrate by alternate

reaction of ZnCl2and O2.Journal of Materials Science Letters, 19, (2000), pp.

2089–2090. ISSN 0261-8028.

[99] G¨

opel, W., Pollmann, J., Ivanov, I. and Reihl, B. Angle-resolved photoe-

mission from polar and nonpolar zinc oxide surfaces. Physical Review B, 26 (6),

(1982), p. 3144.

[100] Girard, R.T., Tjernberg, O., Chiaia, G., S¨

oderholm, S., Karlsson,

U.O., Wigren, C., Nyl´

en, H. and Lindau, I. Electronic structure of

ZnO(0001) studied by angle-resolved photoelectron spectroscopy. Surface Science,

373 (2-3), (1997), pp. 409 – 417. ISSN 0039-6028.

[101] Ranke, W. Separation of the partial s- and p-densities of valence states of ZnO

from UPS-measurements. Solid State Communications, 19 (7), (1976), pp. 685 –

688. ISSN 0038-1098.

[102] Sawada, K., Shirotori, Y., Ozawa, K., Edamoto, K. and Nakatake, M.

Valence band structure of the ZnO surface studied by angle-resolved photoemission

spectroscopy. Applied Surface Science, 237 (1-4), (2004), pp. 343 – 347. ISSN

0169-4332. Proceedings of the Seventh International Symposium on Atomically

Controlled Surfaces, Interfaces and Nanostructures.

[103] Bournel, F., Laffon, C., Parent, P. and Tourillon, G. Adsorption of

some substituted ethylene molecules on Pt(111) at 95 K Part 1: NEXAFS, XPS

and UPS studies. Surface Science, 350 (1-3), (1996), pp. 60 – 78.

[104] G¨

opel, W., Bauer, R. and Hansson, G. Ultraviolet photoemission studies

169

Bibliography

of chemisorption and point defect formation on ZnO nonpolar surfaces. Surface

Science, 99 (1), (1980), pp. 138 – 156. ISSN 0039-6028.

[105] Vohs, J.M. and Barteau, M.A. Photoelectron spectroscopy of diethylzinc on

the polar surfaces of zinc oxide. Journal of Electron Spectroscopy and Related Phe-

nomena, 49 (1), (1989), pp. 87 – 96. ISSN 0368-2048.

[106] Kliv´

enyi, G., Kov´

acs, I. and Solymosi, F. Thermal and photo-induced dis-

sociation of (C2H5)2Zn on Rh(111) surface. Surface Science, 442 (1), (1999), pp.

115 – 130. ISSN 0039-6028.

[107] Thiandoume, C., Sallet, V., Triboulet, R. and Gorochov, O. Decompo-

sition kinetics of tertiarybutanol and diethylzinc used as precursor sources for the

growth of ZnO. Journal of Crystal Growth, 311 (5), (2009), pp. 1411 – 1415. ISSN

0022-0248.

[108] Maejima, K. and Fujita, S. Chemical vapor reactions of ZnO growth by metal-

organic vapor phase epitaxy. Journal of Crystal Growth, 293 (2), (2006), pp. 305 –

310. ISSN 0022-0248.

[109] Bamford, C.H. and Newitt, D.M. Physical and chemical properties of organo-

metallic compounds. Part II. The oxidation of zinc alkyls. Journal of the Chemical

Society., 143, (1946), pp. 688–695.

[110] Thompson, H.W. and Kelland, N.S. The oxidation of diethylzinc. Journal of

the Chemical Society, 187, (1933), pp. 756–757. ISSN 0368-1769.

[111] Verhoeven, J. and Doveren, H.V. An XPS investigation of the interaction

of CH2, C2H2, C2H4and C2H6with a barium surface. Surface Science, 123 (2-3),

(1982), pp. 369 – 383. ISSN 0039-6028.

[112] Dilks, A. The identification of peroxy-features at polymer surfaces by ESCA.

Journal of Polymer Science: Polymer Chemistry Edition, 19 (6), (1981), pp. 1319–

1327.

[113] Ignatyev, I.S., Xie, Y., Allen, W.D. and Schaefer, H.F. Mechanism of the

C2H5+O2reaction. The Journal of Chemical Physics, 107 (1), (1997), pp. 141–155.

170

Bibliography

[114] Erhart, P., Klein, A. and Albe, K. First-principles study of the structure

and stability of oxygen defects in zinc oxide. Physical Review B, 72 (8), (2005), p.

085213.

[115] McCluskey, M.D. and Jokela, S.J. Defects in ZnO. Journal of Applied

Physics, 106 (7), (2009), p. 071101.

[116] Oba, F., Nishitani, S.R., Isotani, S., Adachi, H. and Tanaka, I. Energetics

of native defects in ZnO. Journal of Applied Physics, 90 (2), (2001), pp. 824–828.

[117] Erhart, P., Albe, K. and Klein, A. First-principles study of intrinsic point

defects in ZnO: Role of band structure, volume relaxation, and finite-size effects.

Physical Review B, 73 (20), (2006), p. 205203.

[118] Chaisitsak, S., Sugiyama, T., Yamada, A. and Konagai, M. Cu(InGa)Se2

Thin-film Solar Cells with High Resistivity ZnO Buffer Layers Deposited by Atomic

Layer Deposition. Japanese Journal of Applied Physics, 38, (1999), pp. 4989–4992.

[119] F¨

oll, H. Hyperscript Semiconductor Technology; www.tf.uni-kiel.de/matwis/

amat/semitech_en/index.html, 2011/05/29.

[120] Garcia, F. and Tomar, M. n-CdS/p-ZnIn2Se4thin film solar cell. Thin Solid

Films, 69 (2), (1980), pp. 137 – 139. ISSN 0040-6090.

[121] Grimm, A., Just, J., Kieven, D., Lauermann, I., Palm, J., Neisser,

A., Rissom, T. and Klenk, R. Sputtered Zn(O,S) for junction formation in

chalcopyrite-based thin film solar cells. Physica Status Solidi (RRL), 4 (5-6), (2010),

pp. 109–111. ISSN 1862-6270.

[122] Grimm, A., Klenk, R., Klaer, J., Lauermann, I., Meeder, A., Voigt,

S. and Neisser, A. CuInS2-based thin film solar cells with sputtered (Zn,Mg)O

buffer. Thin Solid Films, 518 (4), (2009), pp. 1157 – 1159. ISSN 0040-6090.

[123] Hansson, G.V. and Uhrberg, R.I.G. Photoelectron Spectroscopy of Surface

States on Semiconductor Surfaces. Surface Science Reports, 9 (5-6), (1988), pp. 197

– 292. ISSN 0167-5729.

171

Bibliography

[124] Heinecke, H. Die metallorganische Molekularstrahlepitaxie f¨

ur die Erzeugung

neuartiger III-V-Halbleiterstrukturen. Vakuum in Forschung und Praxis, 6 (1),

(1994), pp. 35–41. ISSN 1522-2454.

[125] H¨

ufner, S., Schmidt, S. and Reinert, F. Photoelectron Spectroscopy - An

Overview. Nuclear Instruments and Methods in Physics Research Section A: Accel-

erators, Spectrometers, Detectors and Associated Equipment, 547 (1), (2005), pp. 8

– 23. ISSN 0168-9002. Proceedings of the Workshop on Hard X-ray Photoelectron

Spectroscopy - HAXPES.

[126] Irvine, S. Solar Cells and Photovoltaics. In S. Kasap and P. Capper (Editors),

Springer Handbook of Electronic and Photonic Materials, pp. 1095–1106 (Springer,

2007). ISBN 978-0-387-29185-7.

[127] Khallaf, H., Chai, G., Lupan, O., Heinrich, H., Park, S., Schulte, A.

and Chow, L. Investigation of chemical bath deposition of ZnO thin films using

six different complexing agents. Journal of Physics D: Applied Physics, 42 (13),

(2009), p. 135304.

[128] Konovalov, I. Material requirements for CIS solar cells. Thin Solid Films, 451-

452 (0), (2004), pp. 413 – 419. ISSN 0040-6090. Proceedings of Symposium D on

Thin Film and Nano-Structured Materials for Photovoltaics, of the E-MRS 2003

Spring Conference.

[129] Kovacs, I. and Solymosi, F. Thermal and photoinduced dissociation of ethyl

iodide to yield ethyl on a palladium(100) surface. The Journal of Physical Chemistry,

97 (42), (1993), pp. 11056–11063.

[130] Margaritondo, G. and Franciosi, A. Synchrotron Radiation Photoemission

Spectroscopy of Semiconductor Surfaces and Interfaces. Annual Review of Materials

Science, 14 (1), (1984), pp. 67–93.

[131] M¨

onch, W. Electronic Properties of Semiconductor Interfaces. In S. Kasap and

P. Capper (Editors), Springer Handbook of Electronic and Photonic Materials, pp.

147–160 (Springer, 2007). ISBN 978-0-387-29185-7.

[132] Naghavi, N., Renou, G., Bockelee, V., Donsanti, F., Genevee, P.,

172

Bibliography

Jubault, M., Guillemoles, J. and Lincot, D. Chemical deposition meth-

ods for Cd-free buffer layers in CI(G)S solar cells: Role of window layers. Thin

Solid Films, 519 (21), (2011), pp. 7600 – 7605. ISSN 0040-6090. Proceedings of the

EMRS 2010 Spring Meeting Symposium M: Thin Film Chalcogenide Photovoltaic

Materials.

[133] Neumark, G., Gong, Y. and Kuskovsky, I. Doping Aspects of Zn-Based

Wide-Band-Gap Semiconductors. In S. Kasap and P. Capper (Editors), Springer

Handbook of Electronic and Photonic Materials, pp. 843–854 (Springer, 2007). ISBN

978-0-387-29185-7.

[134] Ramaiah, S. and Kodigala. Chapter 4 - Structural Properties of I-III-VI2Ab-

sorbers. In S.R. Kodigala (Editor), Cu(In1-xGax)Se2Based Thin Film Solar, Thin

Films and Nanostructures, pp. 115 – 194 (Academic Press, 2010).

[135] Ruckh, M., Schmid, D. and Schock, H.W. Photoemission studies of the

ZnO/CdS interface. Journal of Applied Physics, 76 (10), (1994), pp. 5945–5948.

[136] Shimizu, A., Chaisitsak, S., Sugiyama, T., Yamada, A. and Konagai, M.

Zinc-based buffer layer in the Cu(InGa)Se2thin film solar cells. Thin Solid Films,

361-362 (0), (2000), pp. 193 – 197. ISSN 0040-6090.

[137] Solymosi, F. Thermal stability and reactions of CH2, CH3 and C2H5 species on

the metal surfaces. Catalysis Today, 28 (3), (1996), pp. 193 – 203. ISSN 0920-5861.

Scientific Aspects of Practical Metallic Catalysts.

[138] Turner, N.H. and Schreifels, J.A. Surface Analysis: X-ray Photoelectron

Spectroscopy and Auger Electron Spectroscopy. Analytical Chemistry, 68 (12),

(1996), pp. 309–332. ISSN 0003-2700.

[139] Veuhoff, E., Pletschen, W., Balk, P. and L¨

uth, H. Metalorganic CVD of

GaAs in a molecular beam system. Journal of Crystal Growth, 55 (1), (1981), pp.

30–34. ISSN 0022-0248.

[140] Wang, J. and Isshiki, M. Wide-Bandgap II-VI Semiconductors: Growth and

Properties. In S. Kasap and P. Capper (Editors), Springer Handbook of Electronic

and Photonic Materials, pp. 325–342 (Springer, 2007). ISBN 978-0-387-29185-7.

173

Bibliography

[141] Yu, K., Spicer, W., Lindau, I., Pianetta, P. and Lin, S. UPS studies of

the bonding of H2, O2, CO, C2H4and C2H2on Fe and Cu. Surface Science, 57 (1),

(1976), pp. 157 – 183. ISSN 0039-6028.

[142] Zhou, X.L., Blass, P.M., Koel, B.E. and White, J.M. Low energy electron

induced chemistry: C2H5Cl on Ag(111). Surface Science, 271 (3), (1992), pp. 452 –

467. ISSN 0039-6028.

174

Danksagung

Zum Abschluss m¨

ochte ich mich bei allen bedanken, die in den letzten drei Jahren

(und/oder davor) zu der erfolgreichen Fertigstellung dieser Dissertation beigetragen ha-

ben.

Angefangen wird bei Prof. Dr. Christian Pettenkofer, nicht nur daf¨

ur, dass er mir

die M¨

oglichkeit gegeben hat meine Doktorarbeit am HZB anzufertigen und damit f¨

ur ein

geregeltes monatliches Auskommen gesorgt hat, sondern auch f¨

ur die entspannte Arbeit-

satmosph¨

are und seine einzigartige Doktorandenbetreuung. Danke, dass du mir jederzeit

mit Rat und Tat zur Seite gestanden hast, f¨

ur die freien Entfaltungsm¨

oglichkeiten und

die großartigen Dienstreisen (mit Ausnahme des fr¨

ankischen Tal des Todes...). Neben der

Physik habe sehr viel von dir f¨

ur das Leben gelernt. So weiß ich beispielsweise jetzt,

wie man Probleme mit kroatischen Zuh¨

altern l¨

ost oder wo man als Vegetarier die besten

Hax’n w¨

ahrend einer DPG Tagung bekommen kann. Danke!

Bei Prof. Dr. Bernd Rech m¨

ochte ich mich daf¨

ur bedanken, dass er sich - obwohl von

Arbeitsmangel bei dir wohl keine Rede sein kann - sofort bereit erkl¨

art hat die univer-

sit¨

are Betreuung an der Technischen Universit¨

at Berlin zu ¨

ubernehmen und mir damit

mein lang erarbeitetes Ziel vom Dr.-Ing. erm¨

oglicht.

Nicht nur daf¨

ur, dass er sich bereit erkl¨

art hat als externer Gutachter dieser Arbeit zu

wirken, m¨

ochte ich mich bei Prof. Dr. Wolfgang J¨

ager bedanken. Sie haben mir viel

erm¨

oglicht in den letzten Jahren und wahrscheinlich h¨

atte ich diese Dissertation nie ge-

schrieben, wenn Sie nicht stets soviel Interesse an Ihren (ehemaligen) Studenten zeigen

w¨

urden.

Meinem Leidensgenossen in Sachen Doktorarbeit, Dr. Andreas Hofmann, gilt ganz be-

sonderer Dank. Mit deinem großartigen Sinn f¨

ur Humor haben die Arbeiten im Labor und

im B¨

uro sehr viel Spaß gemacht. Das wird mir fehlen, wenn es irgendwann nicht mehr so

ist. Ohne dich w¨

are ich in den letzten drei Jahren sicherlich verr¨

uckt geworden. Oder bin

ich es gerade deswegen geworden? Wie dem auch sei: Danke, Alter!

175

Bei Dr. Carsten Lehmann und Dr. Wolfram Calvet m¨

ochte ich mich f¨

ur die vie-

len (nicht) wissenschaftlichen Diskussionen bedanken, die meist zum tiefergehenden Ver-

st¨

andnis der Materie gef¨

uhrt haben. Eure Unterst¨

utzung am Integrierten System bzw. der

SoLiAS haben am Erfolg der Experimente maßgeblich beigetragen.

Gleiches gilt f¨

ur Wolfgang Bremsteller und Herbert Sehnert, deren Bastelk¨

unste das

wartungsintensive technische Equipment am Leben gehalten haben. MacGyver w¨

urde den

Hut ziehen! Außerdem danke ich euch beiden f¨

ur das angenehme Arbeitsklima im B¨

uro

und im Labor. Den beiden ehemaligen Arbeitsgruppenmitgliedern Dr. Patrick Hoff-

mann und Julius K¨

uhn m¨

ochte ich f¨

ur die anf¨

angliche Unterst¨

utzung beim Aufbau des

ALD Systems und dessen Inbetriebnahme danken.

F¨

ur die schnelle und unkomplizierte Herstellung der ZnIn2Se4Referenzkristalle m¨

ochte

ich mich bei Prof. Dr. Michael Binnewies und seiner Arbeitsgruppe am Institut f¨

Anorganische Chemie der Leibniz Universit¨

at Hannover bedanken. Carola Klimm danke

ich f¨

ur die Unterst¨

utzung bei den SEM Untersuchungen meiner ZnO Schichten. Besonde-

rer Dank gilt Dr. Christiane Stephan f¨

ur die GI-XRD Messungen, ihr unglaubliches

Wissen was alles im und rund um das HZB angeht und f¨

ur die lustige Zeit abseits der

Arbeit.

F¨

ur ihren Beitrag an der angenehmen Arbeitsatmosph¨

are danke ich des Weiteren allen

restlichen und ehemaligen Mitarbeitern der Arbeitsgruppe E-I1 (formerly known as E-I4,

bzw. SE6) und den Kollegen der CISSY-Gruppe, Dr. Iver Lauermann,Alexander

Grimm und Britta H¨

opfner. Besonders Dr. Benjamin Johnson danke ich f¨

ur unsere

Kaffeepausen w¨

ahrend, und f¨

ur das Bier zu vern¨

unftiger Musik nach der Arbeit. The fiese

Ami rocks!

Danke allen HZB Doktoranden, mit denen ich in den letzten drei Jahren viel Spaß ge-

habt habe und von denen einige gute Freunde geworden sind.

Dr. Claudia Klingler danke ich f¨

ur die ausf¨

uhrlichen Korrekturen dieser Dissertation

und bei Thomas Gaertner bedanke ich mich f¨

ur die vielen Stunden an den Kickerti-

schen im Berliner Nachtleben, den vielen Konzerten w¨

ahrend der letzten Jahre und - vor

allem - f¨

ur seine Steuergelder.

Meiner Familie danke ich f¨

ur das gute Genmaterial und f¨

ur alles, was Sie mir in den letz-

ten 29 Jahren mit auf den Weg gegeben haben. Danke f¨

ur eure st¨

andige Unterst¨

utzung

auf meinem Werdegang.

Und da das Beste immer zum Schluß kommt: Der gr¨

oßte Dank gilt vor allem Teresa f¨

ihre st¨

andige Unterst¨

utzung, daf¨

ur dass sie mir w¨

ahrend der letzten Monate so viel vom

Hals gehalten hat und f¨

ur noch viel, viel mehr. Danke!

List of Publications

Publications in Journals

E. Janocha and C. Pettenkofer

ALD of ZnO using diethylzinc as metal-precursor and oxygen as oxidizing agent

Applied Surface Science 257 (2011), 10031 - 10035

Poster Presentations

E. Janocha and C. Pettenkofer, Atomic Layer Deposition of TCO window layers for

chalcopyrite solar cells, 481. WE-Heraeus-Seminar: Energy Materials Research by

Neutrons and Synchrotron Radiation, Bad Honnef, 09.05.2011 - 11.05.2011

E. Janocha, A. Hofmann, C. Pettenkofer, Non-destructive depth profiling of the

CuInSe2- ZnO interface by SR-PES, Joint Users’ Meeting 2010, BESSY, Berlin,

09.12.2010 - 10.12.2010

177

Oral Presentations

E. Janocha and C. Pettenkofer, Investigation of the initial interface formation be-

tween CuInSe2(112) and ZnO grown by ALD, DPG Fr¨

uhjahrstagung, Dresden,

13.03.2011 - 18.03.2011

E. Janocha and C. Pettenkofer, Initial Growth and Interface Reactions of Epitaxial

ZnO Layers Grown by Atomic Layer Deposition, MRS Fall Meeting, Boston, MA,

USA, 28.11.2010 - 04.12.2010

E. Janocha and C. Pettenkofer, In situ PES analysis of ultra-thin ZnO layers grown

by atomic layer deposition (ALD), DPG Fr¨

uhjahrstagung, Regensburg, 21.03.2010

- 26.03.2010

E. Janocha and C. Pettenkofer, Growth of ALD-ZnO on Si(111):H and initial growth

on CuInS2(112), CRG-Klausurtagung, Bischofsm¨

uhle, 12.02.2010 - 14.02.2010

C. Lehmann, E. Janocha, A- Hofmann, A. Dombrowa, C. Pettenkofer, The ZnO -

CIS interface, CRG-Klausurtagung, Bischofsm¨

uhle, 12.02.2010 - 14.02.2010

E. Janocha and C. Pettenkofer, Current status of ALD assembly, CRG-Klausurtagung,

Bischofsm¨

uhle, 30.01.209 - 01.02.2009