Growth, phase and doping control in ZnO and In2O3 thin ... · Growth, phase and doping control in...

Growth, phase and doping control in ZnO and In2O3 thin filmsprepared by atomic layer depositionCitation for published version (APA):Wu, Y. (2016). Growth, phase and doping control in ZnO and In2O3 thin films prepared by atomic layerdeposition. Eindhoven: Technische Universiteit Eindhoven.

Document status and date:Published: 22/06/2016

Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.

Download date: 27. Jun. 2020

https://research.tue.nl/en/publications/growth-phase-and-doping-control-in-zno-and-in2o3-thin-films-prepared-by-atomic-layer-deposition(29b7f526-1d8a-4b03-a25f-15adfe3a515c).html

Growth, phase and doping control in ZnO and In2O3

thin films prepared by atomic layer deposition

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de rector magnificus prof.dr.ir. F.P.T. Baaijens,

voor een commissie aangewezen door het College voor Promoties, in het openbaar te verdedigen op donderdag 22 juni 2016 om 16:00 uur

door

Yizhi Wu

geboren te Xiamen, China

Dit proefschrift is goedgekeurd door de promotoren en de samenstelling van de promotiecommissie is als volgt: voorzitter: prof.dr. H.J.H. Clercx 1e promotor: prof.dr. ir. W.M.M. Kessels 2e promotor: prof.dr. F. Roozeboom copromotor: dr. M.A. Verheijen leden: prof.dr. B. Noheda (RUG)

Prof.Dr. J. Bachmann (Friedrich-Alexander-Universität-Erlangen-Nürnberg)

prof.dr. P.M. Koenraad adviseur: dr.ir. R.T.F. van Schaijk (Holst Centre/IMEC) Het onderzoek dat in dit proefschrift wordt beschreven is uitgevoerd in

overeenstemming met de TU/e Gedragscode Wetenschapsbeoefening.

The research project described in this dissertation has been supported by Holst Centre/imec-nl in Eindhoven, The Netherlands

Printed and bound by: Ipskamp Printing (www.ipskampprinting.nl) Cover designed by: Paul Verspaget (www.verspaget-bruinink.nl) A catalogue record is available from the Eindhoven University of Technology Library ISBN: 978-90-386-4089-1

Contents

Chapter 1 General Introduction ........................................... 1 1.1 Trends in future electronics .................................................................... 1 1.2 Semiconducting metal oxides ................................................................. 3 1.3 Atomic layer deposition .......................................................................... 5 1.4 Research project ..................................................................................... 6 1.5 References .............................................................................................. 7

Chapter 2 ALD of multi-component metal oxides .............. 11 2.1 Introduction to ALD ............................................................................... 11 2.2 Advanced ALD processes ....................................................................... 18 2.3 References ............................................................................................ 26

Chapter 3 Introduction to semiconducting metal oxides ........................................................................... 29

3.1 Semiconductors .................................................................................... 29 3.2 Semiconducting metal oxide ................................................................. 38 3.3 Indium oxide ......................................................................................... 41 3.4 Zinc oxide .............................................................................................. 48 3.5 References ............................................................................................ 54

Chapter 4 Electrical transport and Al doping efficiency in nanoscale ZnO films prepared by atomic layer deposition .................................................................... 61 4.1 Introduction .......................................................................................... 62 4.2 Experimental details ............................................................................. 63 4.3 Results and discussion .......................................................................... 64 4.4 Conclusion ............................................................................................ 78 4.5 Acknowledgements ............................................................................... 79 4.6 References ............................................................................................ 79

Chapter 5 Enhanced doping efficiency of Al-doped ZnO by atomic layer deposition using

dimethylaluminum isopropoxide as an alternative aluminum precursor ..................................................... 85 5.1 Introduction .......................................................................................... 86 5.2 Experimental Section ............................................................................ 87 5.3 Results and discussion .......................................................................... 88 5.4 Conclusions ........................................................................................... 93 5.5 Acknowledgement ................................................................................ 93 5.6 References ............................................................................................ 93

Chapter 6 On the factors limiting the doping efficiency in atomic layer deposited ZnO:Al thin films: a dopant distribution study by atom probe tomography .................................................................. 97 6.1 Introduction .......................................................................................... 98 6.2 Experimental details ........................................................................... 101 6.3 Results and discussion ........................................................................ 104 6.4 Conclusions ......................................................................................... 115 6.5 Acknowledgments ............................................................................... 116 6.6 References .......................................................................................... 116

Chapter 7 Compositional, structural and electrical properties of In2O3 thin films prepared by atomic layer deposition from cyclopentadienyl indium and O2/H2O ................................................................. 119 7.1 Introduction ........................................................................................ 120 7.2 Experimental details ........................................................................... 122 7.3 Results and discussions ....................................................................... 124 7.4 Conclusions ......................................................................................... 141 7.5 Acknowledgments ............................................................................... 142 7.6 References .......................................................................................... 142

Chapter 8 Amorphous-to-crystalline transition of In2O3 thin films during atomic layer deposition ........... 147 8.1 Introduction ........................................................................................ 148 8.2 Experimental details ........................................................................... 150 8.3 Results and discussion ........................................................................ 150 8.4 Summary and discussion ..................................................................... 157 8.5 References .......................................................................................... 160

Chapter 9 Concluding remarks and outlook ..................... 163

Summary ............................................................................ 169

List of publications related to this work ............................. 173

Dankwoord/Acknowledgement ......................................... 175

Curriculum Vitae ................................................................ 179

Chapter 1 Chapter 1

General Introduction

1.1 Trends in future electronics

Over the past six decades, the global society has significantly benefited from the advancement in microelectronics. More recently, modern electronics has further developed from individual electronic architectures referred to as system-on-a-chip (SoC), into architectures built from 3D heterogeneous integration of multifunctional electronic building blocks referred to as system-

in-package (SiP) approaches. The latter type of architectures, often referred to as More than Moore, aim at achieving the highest value for a single-packaged modular platform to combine digital with non-digital platform components, as shown in Fig. 1.1.

Figure 1.1 The combined need for digital and non-digital functionalities in

an integrated system: miniaturization of the digital functions (“More Moore”) vs. functional diversification (“More-than-Moore”).1

As an illustration, the current generations of smartphones contain many extra functional devices, such as (video) cameras, GPS navigation, touch-screen interactivity, larger displays with higher resolution, etc. As shown in Fig. 1.2, it is predicted that in the near future, the next generation of mobile phones may

2 General introduction

even be fully transparent,2 and provide additional features,3 such as dual-sided display and input.4

Figure 1.2 Timeline illustrating the development of mobile phones (Images from ref. 2).

Thus one can expect that future generations of mobile phones will continue to be “smarter” by incorporating more features. Fig. 1.3 illustrates a transparent display as an essential component of a transparent phone. In addition, a transparent photovoltaic cell can be incorporated to supply the power, and embedded sensors can monitor the environmentally hazardous gases as well as the health condition of the user.

Figure 1.3 Impression of future generation of mobile phones: “smarter”

phones, which are integrated with other advanced transparent electronic

functions, such as transparent displays, transparent solar cells, gas sensors, etc. (Images from refs. 2,5–7).

1.2 Semiconducting metal oxides 3

Similar to the case of mobile phones, additional developments can be expected in many other related markets. A prominent example can be taken from the solar cell market, which requires the continuous improvement in the energy conversion efficiency of solar cells in order to keep pace with the increasing energy consumption worldwide. One particular objective in today’s solar energy research is to develop transparent solar cells,8 which can selectively transmit visible light, while harvesting the electromagnetic energy from the other parts of the solar spectrum.5 This will enable the integration of solar cells into novel building and vehicle architectures, the so-called building integrated photovoltaic (BIPV)9 and vehicle integrated photovoltaic (VIPV)10 systems. In the BIPV and VIPV markets, transparent displays and sensors are also becoming increasingly popular. Displays can be incorporated into future windows of cars, so that people can, for example, read navigation data or obtain traffic information from the window parts during their travel. The smart and transparent devices described above can be integrated further into other portable and wearable devices and wireless network systems, that are eventually connected into the Internet-of-Things, so that people’s lifestyle and living ambient can be monitored and protected.

1.2 Semiconducting metal oxides

The roadmaps of future multifunctional electronic devices shortly described above, dictate the development of novel functional materials with ultimate control and tuning of the material properties. As an example, future transparent electronics will require advanced materials with superior properties beyond those of the conventional materials i.e. currently used metals, glasses and plastics.

One such advanced class of materials is that of semiconducting metal oxides. Here, two main subclasses are 1) transparent conducting oxides (TCOs) and 2) amorphous oxide semiconductor (AOSs). The first subclass of oxides can find its applications in those that require a compromise between sufficiently high optical transparency in the visible part of the spectrum and a sufficiently high electrical conductivity. The typical TCO materials used mostly are Sn-doped In2O3 (ITO) and Al-doped ZnO (ZnO:Al, AZO). In the second subclass of oxides, AOSs, the conductivity can be modified by several orders of magnitude by, for example, varying their composition and the applied operation voltage.


Moreover, the amorphous phase of the oxides can yield high surface smoothness, uniform film properties in a short-range,11 while maintaining high electron mobility.12 Therefore, semiconducting metal oxides in an amorphous phase can be used as amorphous oxide semiconductor (AOSs). The most popular AOS material is amorphous In-Ga-Zn-O (a-IGZO).11,13–15 Besides these two main subclasses, based on the sensitivity of their conductivity to the environment, some oxides, such as ZnO, SnO2 and In2O3, also find applications as the active layers in gas sensors.16,17

To better explain the functions of TCO and AOS materials, the pixel structure of organic light-emitting diode (OLED) displays is taken as an example, as shown in Fig. 1.4. The role of a transparent electrode in such a pixelated structure is to conduct electric current and to transmit the radiation generated by the emitting layers. Therefore, a high conductivity and a high transparency are desired, and thus the electrode is generally made of TCOs like ITO. AOS materials are often used as the channel layers switching the thin-film transistor (TFT) circuit in the subpixel cells. By applying different gate voltages, the channel layer is switched between its “conducting” and “insulating” state, resulting in the "on" and “off” states of the corresponding subpixel.

Figure 1.4 Schematic representation of the typical pixel array structure in

an organic light-emitting diode (OLED) display. The figure is adapted from ref. 18. The transparent cathode is usually made of transparent

conducting oxides. Amorphous oxide semiconductors are the typical

materials for the channel layers in modern thin-film transistor circuits in the back panel, replacing the traditional amorphous silicon channel

material.

Compared to binary oxides (e.g. In2O3 and ZnO), multicomponent oxides have better potential to fulfil the requirements of applications, because their

1.3 Atomic layer deposition 5

properties can be modified more widely and freely by adding more constituents and by altering the concentration of these constituents.19 For example, when used as the channel layers in TFTs, metal oxides should have a high electron mobility20 with a moderate electron density,11 and preferably the material should be amorphous.11,21,22 ZnO seems to be the best binary compound for the channel layers.11 However, ZnO is typically polycrystalline and has a high electron density. The polycrystalline phase of ZnO can be altered into amorphous by admixing gallium and/or indium atoms, as shown in Fig. 1.5 (a). Moreover, by adding Ga one will also decrease the electron density in indium zinc oxide, as shown in Fig. 1.5 (b).11 Therefore, the multicomponent oxide a-IGZO is able to meet the aforementioned requirements, giving rise to a better performance in TFTs than ZnO.

Figure 1.5 (a) Crystallinity and (b) electron transport properties of In-Ga-

Zn-O thin films. The values and colors in (b) denote the electron Hall mobility (cm2/V.s) with the values of the electron density (in 1018 cm-3) in

parentheses.13,23

1.3 Atomic layer deposition

There are several thin film deposition techniques for the preparation of metal oxide thin films, such as physical vapor deposition (PVD), chemical vapor deposition (CVD), pulsed laser deposition (PLD) and atomic layer deposition (ALD). Among these techniques, the PVD technique of sputtering is the most used one in industrial manufacturing. Compared to sputtering, ALD is a young emerging technique in commercial thin-film technology,24,25 and can potentially become a powerful tool to prepare thin films for future electronics. The self-limiting nature of the half-reactions in ALD growth allows for the


preparation of thin films with a high uniformity and conformality. The feature of high uniformity is favorable for the preparation of thin films on large substrate areas whereas the feature of high conformality is an asset in thin-film deposition on substrates with a three-dimensional surface topology. In addition, the cyclic nature of ALD growth enables atomic-precision (i.e. nanometer-scale) control of the layer thicknesses. For the preparation of multicomponent metal oxides, the so-called supercycle growth mode provides possibilities to control the compositions, hence, to precisely manipulate the film properties.

1.4 Research project

In this work, we aim at controlling the properties of semiconducting metal oxide thin films during ALD growth based on the in situ and ex situ monitoring and understanding of the film properties at an atomic scale. To achieve this goal, ZnO and In2O3 have been chosen as representative materials. The control of the film properties of these oxides by (intentional) Al-doping and (unintentional) H-doping, respectively, have been investigated with an emphasis on the control of the growth, microstructure and composition of the films.

ZnO and In2O3 were selected as study cases for the following reasons: 1) both oxides are often used as constituents of AOS and TCO materials, and as active layers in gas sensors; 2) to investigate the impact of doping on the physical properties of metal oxide thin films, the hydrogen doping in In2O3 and the Al doping in ZnO can be representative cases for unintentional and intentional doping, respectively; 3) both amorphous and crystalline In2O3 films can be obtained by simply varying the growth temperature,26 providing the opportunity to study the crystallization process and the control of the microstructure of metal oxides; 4) the control of the composition and distribution of constituents in multicomponent oxides can be investigated using ZnO:Al as a study case.

The control of film growth, microstructure and composition is studied based on the two study cases, as will be presented in Chapters 4 to 8. In Chapter 4, the control of the Al-doping level in ZnO was realized using the so-called ALD supercycle growth mode. In Chapter 5, it is demonstrated that the maximum

1.5 References 7

doping efficiency of Al in ZnO can be significantly enhanced by exploiting an alternative Al-precursor with bulkier ligands to better control the dispersion of the incorporated Al-dopant. In Chapter 6, some factors limiting the doping efficiency have been identified by measuring the three-dimensional dopant distribution in ZnO:Al at the atomic scale with the advanced technique of Atomic Probe Tomography. In Chapter 7, it was demonstrated that In2O3 thin films can be prepared in a wide temperature range with a fairly high growth rate per cycle while also yielding a high film conductivity. The resulting film properties, including the electrical properties, crystallinity and the origin and distribution of hydrogen in the films, were extensively studied. In Chapter 8, several factors affecting the microstructure of In2O3 films were discussed by investigating the amorphous-to-crystalline transition during ALD for a wide variety of substrate materials, substrate temperatures and preparation times.

1.5 References

1 W. Arden, M. Brillouët, P. Cogez, M. Graef, B. Huizing, and R. Mahnkopf, “More-than-Moore” White Paper, Version 2 (2010). 2 http://www.futuretimeline.net/21stcentury/2020.htm#mobilestandard 3 J.D. Hincapié-ramos, S. Roscher, W. Büschel, U. Kister, R. Dachselt, and P. Irani, in Proc. 2014 Conf. Des. Interact. Syst. (ACM, New York, 2014), pp. 161–170. 4 http://www.theverge.com/2013/2/15/3966950/will-we-see-a-transparent-phone-polytron-prototype-display 5 http://www.extremetech.com/extreme/188667-a-fully-transparent-solar-cell-that-could-make-every-window-and-screen-a-power-source 6 http://news.oled-display.net/lg-display-plans-to-secure-the-no-1-position-in-oled-tv/ 7 http://www.wearable.nl/mhealth/slechte-adem-deze-sensor-vertelt-je-wat-er-aan-scheelt/ 8 J. Lim, S.J. Yun, S.H. Lee, and J.H. Kim, “Transparent Sol. Cell” U.S. Pat. Appl. No. 12/684,768 (2010). 9 B.P. Jelle and C. Breivik, Energy Procedia 20, 68 (2012). 10 D.B. Richardson, Renew. Sustain. Energy Rev. 19, 247 (2013). 11 E. Fortunato, P. Barquinha, and R. Martins, Adv. Mater. 24, 2945 (2012).


12 K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, and H. Hosono, Nature 432, 488 (2004). 13 T. Kamiya and H. Hosono, NPG Asia Mater. 2, 15 (2010). 14 T. Kamiya, K. Nomura, and H. Hosono, Sci. Technol. Adv. Mater. 11, 044305 (2010). 15 J. Yeon Kwon and J. Kyeong Jeong, Semicond. Sci. Technol. 30, 024002 (2015). 16 J. Mizsei, Sensors Actuators B Chem. 23, 173 (1995). 17 M.E. Franke, T.J. Koplin, and U. Simon, Small 2, 36 (2006). 18 http://www.provideocoalition.com/sony_the_non-technical_technical_guide_to_oled_technology/page-2 19 T. Minami, Semicond. Sci. Technol. 20, S35 (2005). 20 J.S. Park, W.-J. Maeng, H.-S. Kim, and J.-S. Park, Thin Solid Films 520, 1679 (2012). 21 J.-Y. Kwon, D.-J. Lee, and K.-B. Kim, Electron. Mater. Lett. 7, 1 (2011). 22 D.B. Buchholz, L. Zeng, M.J. Bedzyk, and R.P.H. Chang, Prog. Nat. Sci. Mater. Int. 23, 475 (2013). 23 K. Nomura, A. Takagi, T. Kamiya, H. Ohta, M. Hirano, and H. Hosono, Jpn. J. Appl. Phys. 45, 4303 (2006). 24 S.M. George, Chem. Rev. 110, 111 (2010). 25 H.B. Profijt, S.E. Potts, M.C.M. van de Sanden, and W.M.M. Kessels, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 29, 050801 (2011). 26 J.A. Libera, J.N. Hryn, and J.W. Elam, Chem. Mater. 23, 2150 (2011).

Chapter 2

ALD of multi-component metal oxides

Abstract

This chapter focuses on the ALD growth of multi-component metal oxide films. In the first section, some basic concepts of ALD are explained, including the characteristics of ALD growth, a standard ALD cycle, growth per cycle, and recipe development. In the second section, two variations from a standard ALD cycle are introduced: the multi-step ALD cycle and the supercycle. This section also explains the control of the composition and distribution of constituents in multi-component metal oxides by using multi-step ALD cycles and supercycles.

2.1 Introduction to ALD

Atomic layer deposition is a thin-film deposition technique. It has similarities with the conventional chemical vapor deposition (CVD) technique: vapor phase reactants are introduced into a reaction chamber, and then react on the substrate surface to deposit the desired thin film. However, the way of introducing the individual chemical precursors into deposition chambers is different between these two deposition techniques. In the case of CVD, the reactants are dosed simultaneously into the chamber, and contribute to film growth via chemical reactions on the substrate surface in a continuous manner. In the case of ALD, the reactants are introduced sequentially with an inert gas purge between the precursor pulses. Instead of continuous film growth, the film growth is intermitted by the inert gas purge steps. After one ALD cycle, one atomic layer of material is deposited. In this way, ALD gives rise to a so-called layer-by-layer growth.

ALD exhibits the following features for thin-film growth, as illustrated in Fig. 2.1: (1) the cyclic nature of ALD growth allows accurate thickness control; (2)

12 ALD of multi-component metal oxides

the self-limiting growth enables the deposition of thin films with a good uniformity over a large area; (3) the self-limiting nature also allows a conformal growth on substrates with three-dimensional topology; (4) deposition at low substrate temperatures (typically around 100 to 400 °C) is possible. 1–3 In some cases, the substrate temperature can be as low as room temperature (25°C).4 Besides these general features, for the case of preparation of complicated films containing multiple components, ALD also provides the possibilities to tune the composition, as well as to control the distribution of the individual constituents in the films. This aspect will be addressed extensively later in this chapter.

Figure 2.1 Schematic representation of a thin film on a substrate with

three-dimensional topology. The four major merits of thin-film growth by ALD are illustrated: accurate thickness control, high conformality,

superior uniformity and low growth temperature.3

2.1.1 Standard ALD cycle

The aforementioned excellent growth features can be attributed to the self-limiting nature of the ALD process. To understand the self-limiting nature, the well-known model system of Al2O3 growth using Al(CH3)3 and H2O as reactants can be taken to explain a standard ALD cycle. A standard ALD cycle consists of two self-limiting half-cycles, as shown in Fig. 2.2. In the first half-cycle, the first step is the dosing of metal precursor Al(CH3)3, as shown in Fig. 2.2 (a). The introduced Al(CH3)3 reacts with the -OH surface groups, to form Al-O bonds at the film surface while releasing the volatile byproduct CH4. After such reaction, the surface is terminated by methyl (-CH3) groups, which will not

2.1 Introduction to ALD 13

further react with incident Al(CH3)3 precursors, i.e. the surface reaction is self-

limiting. In the second step, as shown in Fig. 2.2 (b), the residual precursor molecules and byproducts from the surface reactions are purged or pumped out. In the second half-cycle, H2O vapor is dosed into the reaction chamber, as shown in Fig. 2.2 (c). The H2O reacts with the CH3 surface groups, releasing the byproduct CH4. The surface becomes again terminated with OH groups, which will not react with upcoming H2O molecules anymore, i.e. the second-half reaction is also self-limiting. In the last purge step, as shown in Fig. 2.2 (d), the byproduct CH4 and the residual H2O are purged or pumped out. By repeating the cycle, the desired materials are then obtained in a layer-by-layer mode: one atomic layer is deposited by one ALD cycle. The increase of the film thickness in one Al2O3 cycle is approximately 0.1 nm.

Figure 2.2 Schematic representation of an ALD cycle for the growth of

Al2O3. The first half-cycle consists of (a) self-limiting adsorption of the Al precursor Al(CH3)3 through a reaction with the -OH surface group, and (b)

a purge step to remove the volatile reaction by-products and the excess of precursor dosed. The second half-cycle consists of (c) a dosing step of

the H2O reactant, which reacts with the -CH3 surface group in a self-limiting way, and (d) another purge step to remove the volatile reaction

by-products and the excess of reactant dosed.5

To generalize from the Al2O3 example, a typical ALD cycle consists of two half-cycles, as illustrated in Fig. 2.3. In the first half-cycle, metal atoms are introduced to the film surface by the first reactant A, which is called the (metal)

precursor. In the second half-cycle, the film surface is changed into the desired material via the surface reactions initiated by the second reactant B, which is


called the co-reactant. During each half-cycle, the surface reactions are self-limiting. Therefore, by working under saturation conditions, the film growth does not depend on the duration or the pressure of the reactant fluxes.

Figure 2.3 Schematic representation of a standard ALD cycle, which

consists of 4 steps: dosing of precursor A/ purge/ dosing of co-reactant B/ purge.

Note, for different targeted materials, the surface reaction mechanisms are different. Even for the same targeted material, the reaction mechanisms can also be different when different precursors or co-reactants are used. In the aforementioned example of Al2O3, the surface reactions occur via ligand exchange between the existing surface groups and the incoming molecules of the reactants. Besides such a ligand exchange, other reaction mechanisms also exist. For instance, 1) the precursor ligands can be combusted at the film surface, and subsequently, be removed by the co-reactant. The growth of Pt using MeCpPtMe3 and O2 is an example of such reaction mechanism.6 2) Reduction-oxidation (red-ox) reactions can be involved in addition to the ligand exchange. In the case of the growth of In2O3 using InCp and O3, O3 triggers the ligand exchange reaction as well as the red-ox reaction.7 3) Surface reactions can also be triggered by energy-enhanced co-reactants, such as O2, H2 and N2 plasmas.

Apart from differences in reaction mechanisms, the structure of an ALD cycle can also deviate from the basic Al2O3 example. Compared to the standard one, a more complicated ALD cycle can be required for the growth of certain materials. This aspect will be addressed extensively later in this chapter.


2.1.2 Development of ALD process

Deposition of a new material by ALD begins with developing the recipe for the corresponding ALD process. Such a process development starts with the selection of a suitable precursor and its corresponding co-reactant(s). A suitable precursor should have a sufficiently high vapor pressure. Moreover, the precursor should also have a good chemical stability so that it does not decompose thermally at the substrate temperature used. A precursor together with its corresponding co-reactant(s) should ensure a self-limiting growth behavior. Furthermore, they should yield a reasonable growth per cycle within a sufficiently wide range of processing temperatures, the so-called ALD

window. In some applications (e.g. depositing thin films on organic substrates), a low processing temperature (below 200 °C or even at room temperature) is desired.

After the selection of precursor and co-reactant, the optimum duration of each step in an ALD cycle should be determined. The dosing steps of precursor and co-reactant should be long enough to ensure that sufficient amounts of precursor and co-reactant are supplied to the film surface, so that the surface reactions are saturated. The purging steps should also be long enough to avoid CVD-like reactions between the residual reactant from the previous dosing step and the incoming reactant from the next dosing step. On the other hand, in order to minimize the usage of reactants as well as to increase the time efficiency of ALD growth, each step should not be unnecessarily long. The duration of each step can be determined by measuring the so-called saturation

curves.

2.1.3 Growth per cycle

The growth per cycle (GPC) is an important parameter to evaluate how effectively a certain material can be prepared using ALD. It is defined as the amount of material deposited per ALD cycle. The amount of material can be quantified in different ways, and the expression of GPC can vary accordingly. For example, by measuring the increase of film thickness, the GPC can be expressed in nm; 8 by measuring the increase of film mass, the GPC can be expressed in ng/cm

2; 9 by measuring the increase of areal atomic density, the GPC can be expressed in atoms/cm

2.8,10 In literature, nm (or Å) is the most


common expression of GPC, because, in most cases, the film thickness is measured. Furthermore, it is also easy to estimate the number of cycles required for a targeted film thickness when using nm.

When ALD growth is in a steady state, the surface chemistry and the reaction mechanism are identical during every ALD cycle. Therefore, each ALD cycle yields the same increment in film thickness. i.e. the value of GPC is constant, and film thickness is linearly dependent on the number of ALD cycles. Such a linear dependence enables an accurate control of film thickness. Assuming a GPC of 0.1 nm, as illustrated in Fig. 2.4 (a), the film thickness can be “digitally” controlled with an accuracy of 0.1 nm by varying the number of ALD cycles. The targeted film thickness can be achieved by setting the right total number of cycles, e.g. 10 nm requires 100 cycles for a GPC of 0.1 nm.

Figure 2.4 (a) Film thickness as a function of the number of ALD cycles,

showing a linear dependence. The GPC is assumed to be 0.1 nm. (b) Film thickness as a function of the number of ALD cycles. During the first few

cycles, two deviating modes from linear growth can be distinguished:

accelerated and delayed growth.

Often, deviations from linear growth can be observed during the first few cycles when a material is deposited on certain substrates. Such deviations can result from the difference in the density and/or reactivity of the surface sites between the substrate and the targeted material. Depending on the exact surface chemistry, the deviation can take the form of accelerated growth or delayed growth, as illustrated in Fig. 2.4 (b). In the case of accelerated growth, the GPC in the first few cycles is higher than the value for normal linear growth.


In contrast, in the case of delayed growth, the GPC in the first few cycles is lower. In both cases, after the first few cycles, the surface is completely covered by the targeted material. From then on, the surface chemistry becomes identical during every ALD cycle. As a result, the GPC returns to the value for linear growth.

As introduced before, one ALD cycle yields one atomic layer of the deposited material. We note, that one atomic layer is mostly less than one monolayer. This can be ascribed to the limited number of reactive sites on the surface, as well as to the steric hindrance of surface ligands and adsorbed reactant molecules. It is also worthwhile to note that one monolayer is well defined for single-crystalline materials, but is ill-defined for amorphous and polycrystalline material.11 In most practical cases, the films grown using ALD are amorphous, while in some other cases, films are polycrystalline with different grain orientations. Therefore, it is difficult to define “one monolayer” in most cases. However, the thickness of one monolayer can still be estimated based on the corresponding single-crystalline lattice structure. For example, as-deposited Al2O3 films are amorphous. As shown in Fig. 2.5, the sapphire crystal structure can be taken as the single-crystalline structure base of Al2O3 films for the estimation. In this structure, the (0001) planes are single-element terminated, and the interspacing between adjacent (0001) planes containing Al (or O) atoms is c/6=0.21 nm. Therefore, the thickness of one monolayer in the Al2O3 case can be estimated to be 0.21 nm. Compared to this value, the GPC of Al2O3 (~0.1 nm) would indicate that one ALD cycle results in the growth of ~0.5 monolayer.


Figure 2.5 Schematic representation of the sapphire structure of Al2O3.12 The red and green spheres represent O and Al atoms, respectively. The

lattice constants are a=0.476 nm and c=1.299 nm. Along the c-axis, the thickness of one monolayer is defined as 1/6 of the lattice constant c.

2.2 Advanced ALD processes

An advanced ALD process is a variation of a standard ALD process, which can provide additional features to ALD film growth. For example, multistep ALD cycles can widen process windows, whereas supercycles can be applied to prepare multi-components compounds.

2.2.1 Multistep ALD cycle

Not all materials can be prepared using a standard ALD cycle as shown in Fig. 3. In some cases, more steps in one ALD cycle are required to accomplish the growth of certain materials. These additional steps can be added to a standard ALD cycle, forming a so-called multistep ALD cycle. One can compose various forms of multistep ALD cycles. Here, we introduce a typical form ABC cycle, as well as its special manifestation: the co-dosing cycle.

As shown in Fig. 2.6 (a), compared to a standard ALD cycle, an ABC cycle consists of three “half-cycles”, instead of two. Here, the additional third half-cycle consists of a dosing step of another co-reactant C and a subsequent

2.2 Advanced ALD processes 19

purge step. Depending on the material and the growth mechanism, such multiple co-reactants can serve different functions. For example, they can be applied to widen the process window. In the case of platinum, the temperature window is limited at 200-300 °C when MeCpPtMe3 and O2 are used as a precursor and a co-reactant.13 By adding an additional half-cycle of H2 gas or H2 plasma dosing after the second half-cycle of O2 plasma dosing, room-temperature growth can be realized. 4 In some other cases, multiple co-reactants can accomplish the growth collaboratively. For example, in the case of indium oxide, when InCp is used as indium precursor, one co-reactant H2O is used to change the surface ligands from -Cp to -OH, while the other co-reactant O2 is used to oxidize the indium atoms from the +1 charge state to the +3 charge state.14

As shown in Fig. 2.6 (b), a co-dosing cycle has a similar structure as a standard ALD cycle, but two co-reactants B and C are simultaneously dosed into the reaction chamber in the second half-cycle. It is possible to convert an ABC cycle into a co-dosing cycle, provided that the two co-reactants can be co-injected without reacting with each other in the gas phase. For example, O2 and H2 plasmas cannot be simultaneously dosed in the aforementioned case of platinum, while a co-dosing of H2O and O2 is feasible in the case of indium oxide.14

Figure 2.6 Schematic representations of a typical multistep ALD cycle, the so-called ABC cycle, and its special manifestation, the so-called co-dosing

cycle. (a) In an ABC cycle, a second co-reactant C is introduced in a third half-cycle; and (b) in a co-dosing cycle, the two co-reactants B and C are

simultaneously dosed in the second half-cycle.


2.2.2 Supercycle

The aforementioned standard and multistep ALD processes are usually applied to prepare single-element metals (e.g. Pt) and binary compounds (e.g. Al2O3). As discussed in Chapter 1, compared to binary compounds, multi-component compounds provide more freedom to design “new” films and tune the film properties for certain applications. To realize the growth of multi-component compounds, another advanced ALD process mode, the so-called supercycle, can be applied. To exemplify the use of supercycle recipes, we will describe the growth of a ternary metal oxide film. However, the content introduced in this section can be generally applied to other multi-component compounds, such as metal alloys, ternary metal nitrides, and quaternary metal oxides. As shown in Fig. 2.7, the growth of a binary metal oxide MOx (or NOy) can be realized by its standard ALD process. To grow a ternary metal oxide MaNbOc, these two standard ALD processes can be combined in one supercycle: m cycles of MOx followed by n cycles of NOy. The ratio between m and n is called the cycle ratio. By tuning the cycle ratio, the composition, i.e. the percentage (or ratio) of the constituents MOx and NOy in the MaNbOc film, can be modified. The number of supercycles is chosen to reach the targeted film thickness.

Figure 2.7 Schematic representation of a typical supercycle for the

growth of a ternary metal oxide (MaNbOc). The growth of the constituents

MOx and NOy can be realized by their own standard ALD processes. One supercycle is composed by m cycles of the ALD process of MOx followed by n cycles of the ALD process of NOy. The targeted film thickness can be

achieved by repeating one supercycle a number of times.


In addition to the control of the film compositions by varying the cycle ratio, the distribution of the constituents in the direction of film growth can also be controlled by varying the values of m and n, resulting in different film structures. Here, four typical constituent distributions are presented: (i) Using large values of m and n, a nanolaminate structure consisting of two alternating materials can be obtained, as shown in Fig. 2.8 (a). (ii) In contrast, small values of m and n can lead to a relatively homogeneous mixture, as shown in Fig. 2.8 (b). (iii) A large value of m together with n=1 can be used to incorporate a small amount of dopants into a host material, as shown in Fig. 2.8 (c). The single cycle in the second process, that introduces dopants into the host films, is normally called the doping cycle. (iv) By changing the values of m and n with film thickness, a graded mixture of the two constituents can be obtained, as shown in Fig. 2.8 (d).

Figure 2.8 Schematic representations of four typical constituent

distributions of multi-component thin films that can be prepared by ALD:

(a) nanolaminate, (b) homogeneous mixture, (c) doped film and (d) graded mixture.

2.2.3 Non-linear growth effects when using supercycles

Even though the composition and distribution of the constituents in multi-component films can be controlled using supercycles, deviations from the linear growth characteristic of ALD can occur at the same time. To illustrate the


deviations, the growth of a ternary metal oxide (MaNbOc) in a nanolaminate structure, i.e. using large values of m and n in one supercycle, is taken as an example. The knowledge introduced in this section is also valid for other multi-component thin films grown using supercycles, as well as for other types of film structures represented in Fig. 2.8. As shown in Fig. 2.9 (a), the growth of a pure binary oxide MOx (or NOy) has a certain GPC, and shows a linear behavior with increasing number of cycles. If the growth of MaNbOc would be “linear”, the film thickness of MaNbOc as a function of the number of ALD cycles would be like the solid line represented in Fig. 2.9 (a). However, non-linear growth can be expected during the first few cycles of the processes of MOx and NOy, when one constituent is deposited on the matrix of the other. The zoom-in on these first few cycles of NOy after the MOx is represented in Fig. 2.9 (b). During these cycles, several deviations from linear growth can be expected, such as accelerated growth, delayed growth, and a so-called etching effect. The actual reaction mechanisms resulting in these deviations can vary from case to case. Here, some possible mechanisms are briefly introduced:

Figure 2.9 (a) Schematic representation of the linear growth of pure MOx

and NOy films, as well as the linear growth of the ternary compound MaNbOc. (b) A zoomed-in part of (a), which shows the first few cycles of

NOy after the MOx process. Several possible deviations from linear growth

are indicated: accelerated growth, delayed growth and the so-called etching effect.

Accelerated growth. As mentioned before, the areal density and chemical reactivity of the surface functional groups, such as hydroxyl groups during the growth of most metal oxides, can have an important effect on the GPC. Similar to the accelerated growth on substrates, an accelerated growth of NOy can result from a higher areal density and/or an enhanced reactivity of M-OH groups on a MOx matrix, compared to the N-OH groups on an NOy matrix.


Delayed growth. Similarly to the delayed growth on substrates, a lower areal density and/or a reduced reactivity of M-OH groups can give rise to delayed growth. Moreover, it is also possible that the surface groups M-OH and N-OH react with each other on the film surface, depleting the surface from hydroxyl groups.9,15 As a result, the reduced areal density of hydroxyl groups can lead to a decreased GPC in the subsequent NOy cycles.

Etching effect. An etching effect occurs when previously deposited atoms on a film surface are removed by the new incoming reactants, which consequently results in a lower or even negative “apparent” GPC. For example, it has been reported that in the case of Al-doped ZnO, during the Al2O3 cycles, the zinc and oxygen atoms on the surface can be removed by the Al precursor Al(CH3)3.9,15 The details of this phenomenon will be presented in Chapter 5.

When using supercycles to grow a certain material, it is challenging to predict the relevant type of deviation from linear growth, as well as the extent of the deviation, because such predictions would require a quantitative understanding of the surface reactions between the two specific constituents. As a result, it is difficult to accurately estimate the number of supercycles needed for a targeted film thickness, as well as the cycle ratio for the targeted film composition. In practice, the cycle ratio and the number of supercycles are experimentally determined, i.e. the values are adjusted based on the measured composition and film thickness.

2.2.4 Combination of a multi-step ALD cycle and a supercycle

For the growth of doped metal oxides, it is important to control the dopant distribution in the host thin films. As shown in Fig. 2.8 (c), the dopant distribution is typically represented as a δ function, i.e. the dopants are concentrated at the “film depths” where the doping cycles have been applied.16 The control of the dopant distribution can be realized by modifying the interspacing between the dopant layers, as well as the lateral spacing between dopant atoms in one dopant layer. The former can be achieved by simply tuning the cycle ratio. The latter can possibly be realized by the combination of multi-step ALD cycles and supercycles, i.e. by advanced doping cycles.


Below, we shortly present three examples of advanced doping cycles, which have been reported in literature to reduce the amount of incorporated dopant atoms per doping cycle: sequential doping, co-doping and poisoning. Meanwhile, the limitation of each example will also be addressed. Note that these examples may not be valid for all doping systems, but the following discussions can still indicate how advanced doping cycles can assist the modification of a dopant distribution. To illustrate the mechanisms of the surface reactions during the doping cycle, the ternary oxide B-doped AOx is taken as an example. Similar to the aforementioned surface reaction mechanism of Al2O3, here we assume that the precursors of host (A) and dopant (B) elements react with the hydroxyl groups on the film surface via ligand exchange, and that the co-reactant (C) turns the surface groups back into hydroxyl groups.

Sequential doping.17 Compared to the regular doping cycle shown in Fig. 2.10

(a), in a sequential doping cycle, a half-cycle of the precursor A is inserted before the regular doping cycle, as shown in Fig. 2.10 (b1). The hydroxyl groups on the surface react with the host precursor during this additional half-cycle, leaving A-R1 groups on the film surface. In the subsequent regular doping cycle, the dopant precursor reacts with the -R1 surface groups, rather than with the hydroxyl groups. The -R1 groups can be less reactive than the hydroxyl groups. Consequently, the dopant precursor reacts with fewer surface sites, resulting in fewer incorporated dopant atoms, compared to the regular doping cycle. However, to realize such a reaction, the reactivity of the dopant precursor should be high enough to trigger the ligand exchange reaction with the -R1 group. Such a condition cannot be satisfied in all combinations of host and dopant precursors.

Co-doping.18 As shown in Fig. 2.10 (b2), a simultaneous dosing of the host and

dopant precursors is applied in the co-doping mode. During the dosing step, there is a competition between the two precursors, i.e. the ensemble of hydroxyl groups is consumed by these two precursors at the same time. Less hydroxyl groups react with the dopant precursor compared to the regular doping case, resulting in less incorporated dopant atoms. However, the ratio between the incorporated host and dopant atoms will depend on the partial pressures of the two precursors. If the partial pressures are not uniform in a large deposition area, the resulting dopant concentration can vary from one


Figure 2.10 Schematic representations of the ALD steps and the

corresponding surface reactions during the doping cycles in different

modes of doping: (a) regular doping, (b1) sequential doping, (b2) co-doping and (b3) poisoning. In the schematic representations of the

surface reactions, the letters “A” and “B” represent host and dopant

atoms, respectively, “R1” and “R2” represent the ligands bonded to A and B atoms, respectively, and “D” represents unreactive surface sites. In the

regular doping mode, the dopant precursor reacts with hydroxyl groups at the film surface, and the dopant atoms B are incorporated. In the

sequential doping mode, the hydroxyl groups are consumed by the host precursor, and the dopant precursor reacts with the surface groups -R1,

which are less reactive than the hydroxyl groups. In the co-doping mode,

the hydroxyl groups are consumed simultaneously by the host and dopant precursors in the same time. In the poisoning mode, some

hydroxyl groups are deactivated by unreactive sites D. Consequently, all the three sorts of advanced doping cycles result in a lower density of

incorporated dopant atoms than in the regular mode.


deposition area to another. This is in contrast with one of the intrinsic ALD features: an ALD growth is independent on the flux and pressure of reactants.

Poisoning.19 As shown in Fig. 2.10 (b3), a half-cycle of another co-reactant D is

inserted before the regular doping cycle. During this half-cycle, a part of the hydroxyl groups is deactivated by this co-reactant (i.e. these surface sites are “poisoned”), leaving less hydroxyl groups available to react with the dopant precursor. The deactivated surface groups should be refreshed later during the dosing of the regular co-reactant C, and become reactive again for the upcoming ALD cycles. However, it can be challenging to refresh all the deactivated sites by the co-reactant C. The unrefreshed sites can remain in the film, and act as defects.

2.3 References

1 S.M. George, Chem. Rev. 110, 111 (2010).

2 M. Leskelä and M. Ritala, Thin Solid Films 409, 138 (2002).

3 H.C.M. Knoops, S.E. Potts, A.A. Bol, and W.M.M. Kessels, Handbook of Crystal

Growth, Second Edi (Elsevier B.V., North Holland, 2015).

4 A.J.M. Mackus, D. Garcia-Alonso, H.C.M. Knoops, A.A. Bol, and W.M.M.

Kessels, Chem. Mater. 25, 1769 (2013).

5 H.C.M. Knoops, Atomic Layer Deposition: From Reaction Mechanisms to 3D-

Integrated Micro-Batteries, Eindhoven University of Technology, 2011.

6 A.J.M. Mackus, N. Leick, L. Baker, and W.M.M. Kessels, Chem. Mater. 24,

1752 (2012).

7 J.W. Elam, A.B.F. Martinson, M.J. Pellin, and J.T. Hupp, Chem. Mater. 18, 3571

(2006).

8 Y. Wu, S.E. Potts, P.M. Hermkens, H.C.M. Knoops, F. Roozeboom, and W.M.M.

Kessels, Chem. Mater. 25, 4619 (2013).

9 J. Elam and S. George, Chem. Mater. 15, 1020 (2003).

10 S.E. Potts, G. Dingemans, C. Lachaud, and W.M.M. Kessels, J. Vac. Sci.

Technol. A 30, 021505 (2012).

2.3 References 27 11 T. Faraz, F. Roozeboom, H.C.M. Knoops, and W.M.M. Kessels, ECS J. Solid

State Sci. Technol. 4, N5023 (2015).

12 http://som.web.cmu.edu/structures/S058-Al2O3.html

13 T. Aaltonen, M. Ritala, Y.-L. Tung, Y. Chi, K. Arstila, K. Meinander, and M.

Leskelä, J. Mater. Res. 19, 3353 (2004).

14 J.A. Libera, J.N. Hryn, and J.W. Elam, Chem. Mater. 23, 2150 (2011).

15 J.-S. Na, G. Scarel, and G.N. Parsons, Adv. Funct. Mater. 21, 448 (2011).

16 D.-J. Lee, H.-M. Kim, J.-Y. Kwon, H. Choi, S.-H. Kim, and K.-B. Kim, Adv. Funct.

Mater. 21, 448 (2011).

17 J.-S. Na, Q. Peng, G. Scarel, and G.N. Parsons, Chem. Mater. 21, 5585 (2009).

18 A. Illiberi, R. Scherpenborg, Y. Wu, F. Roozeboom, and P. Poodt, ACS Appl.

Mater. Interfaces 5, 13124 (2013).

19 A. Yanguas-Gil, K.E. Peterson, and J.W. Elam, Chem. Mater. 23, 4295 (2011).

Chapter 3

Introduction to semiconducting metal oxides

Abstract

In this chapter, some basic physical properties of semiconductors are introduced in the first place. Later, the introduction focuses on semiconducting metal oxides. Next, we introduce the relevant physical properties of zinc oxide and indium oxide, as well as the details related to the preparation of these materials by ALD.

3.1 Semiconductors

3.1.1 Fundamentals of semiconductors

a) Bandgap

The differences in electron conductivity behavior between a metal, an insulator and a semiconductor can be related to their electronic bandgaps. Fig. 3.1 shows the band structures of these three classes. An electronic bandgap, or often just called bandgap, is an energy range where no electron states can exist. The value of a bandgap (Eg) is generally referred to as the energy difference (in electron volts, eV) between the valence band maximum (VBM, or Ev) and the conduction band minimum (CBM, or Ec), i.e. Eg=Ec-Ev. In a metal, the conduction band and the valence band are overlapped, and a bandgap does not exist. In a semiconductor and an insulator, the conduction band and the valence band are separated, i.e. Eg>0. An insulator has a larger bandgap than a semiconductor. Typically, the most conventional semiconductor silicon has a bandgap of ~1.1 eV, while silicon dioxide, the conventional insulator, has a bandgap of ~9 eV.

30 Introduction to semiconducting metal oxides

Figure 3.1 Schematic representations of the band structures near

bandgap of three material categories: (a) a metal, (b) a semiconductor and (c) an insulator. The value of a bandgap (Eg) is defined as the energy

difference between a valence band maximum (Ev) and a conduction band minimum (Ec). In a metal, the conduction band and the valence band are

overlapped, and a bandgap does not exist. A bandgap is small in a semiconductor, and is large in an insulator.

In the class of semiconductors, there is subclass, called wide-bandgap

semiconductors. The term “wide-bandgap” refers to a bandgap larger than ~3 eV,1 which is much larger than that of silicon. Compared to silicon, wide-bandgap semiconductors are transparent in a wider range of the spectrum. Moreover, such wide-bandgap semiconductors allow devices (e.g. solar and display devices) to be operated at higher voltages, frequencies and temperatures. Zinc oxide and indium oxide, the two oxides addressed in this work, are both wide-bandgap semiconductors.

b) n-type and p-type semiconductors

The main charge carriers in semiconductors are electrons (e-) in the conduction band and holes (h+) in the valence band. In non-degenerate semiconductors (the degeneracy of semiconductors will be explained later), the density of electrons (ne) can be calculated by the following integration:

CC FD( ) ( )

eE

n g f dε ε ε∞

= ∫ (Eq. 3.1)

3.1 Semiconductors 31

where gC stands for the density of states in the conduction band, and fFD stands for the Fermi-Dirac distribution. The result of the integration can be expressed as:

3/2C B C F

2

B

2( ) exp( )2

e

m k T E En

k Tπ

−= −

(Eq. 3.2)

where mC is the effective mass of electrons at the CBM, kB is the Boltzmann constant, ħ is the Dirac constant, and T is the temperature, . EF is the Fermi level, i.e. the highest energy level that electrons can occupy at 0 K. Similarly, the density of holes (nh) can be expressed as:

3/2V B F V

2

B

2( ) exp( )2

h

m k T E En

k Tπ

−= −

(Eq. 3.3)

where mV is the effective mass of holes at the VBM.

The law of mass action can be concluded from the above expressions:2

g Bexp( / )e h

n n E k T⋅ ∝ − (Eq. 3.4)

i.e. the product of ne and nh is related to the bandgap and the temperature, but is independent of the position of the Fermi level. For example, if the density of electrons is increased by a factor of 10 by doping, the density of holes will decrease by the same factor.

The more abundant charge carriers are called majority carriers, which are primarily responsible for the electrical transport. In a typical intrinsic semiconductor, the densities of electrons and holes have comparable levels (ne~nh), with the Fermi level EF located at the center of the bandgap, as shown in Fig. 3.2 (a). In an n-type semiconductor, electrons are the majority carriers (ne»nh), and the position of the Fermi level is closer to the CBM, as shown in Fig. 3.2 (b); in a p-type semiconductor, holes are the majority carriers (ne«nh)., and the Fermi level is located closer to the VBM, as shown in Fig. 3.2 (c). As will be discussed later in this chapter, semiconducting metal oxides are mostly n-type. Therefore, the following introduction will mainly focus on n-type semiconductors, and (free) electrons will also be referred to as carriers. However, the properties of semiconductors introduced in this chapter can also apply to p-type semiconductors.


A semiconductor is non-degenerate when the Fermi level lies within the bandgap (EV < EF < EC). The cases shown in Figs. 3.2 (a) to (c) are non-degenerate semiconductors. In an n-type degenerate semiconductor, the carrier density is higher than a critical value (nC). As a result, the Fermi level rises up to higher than the CBM (EF>EC), as shown in Fig. 3.2 (d). The value of nC can be estimated by Mott’s criterion:3

1/3

0* 0.25cn a⋅ ≈ (Eq. 3.5)

where a0* is the effective Bohr radius of the material.

Figure 3.2 Schematic representations of band structures showing the

position of the Fermi level EF and the densities of electrons and holes in

different types of semiconductors: (a) in a typical intrinsic semiconductor, the densities of electrons and holes have comparable levels, and the

Fermi level lies in the middle of the bandgap; (b) in an n-type semiconductor, electrons are the majority charge carriers, and the Fermi

level is closer to the CBM; (c) in a p-type semiconductor, holes are the

majority charge carriers, and the Fermi level is closer to the VBM; (d) in a degenerate n-type semiconductor, the Fermi level is above the CBM.

c) Electron donors and acceptors

Electron donors are ionized impurities which can “donate” their extra valence electrons into the conduction band, providing excess carriers. Electron acceptors are ionized impurities which “accept” electrons from the valence band, reducing the carrier density. In intrinsic semiconductors, interstitials and vacancies of the compositional atoms are the most common forms of donors


and acceptors. In addition, in extrinsic semiconductors, foreign atoms can also act as donors and acceptors, by forming, for example, substitutional or interstitial impurities in the host lattice. Table 3.1 shows some typical donors and acceptors in indium oxide as an example.

Table 3.1 Typical donors and acceptors in intrinsic and extrinsic indium oxide. Note, that adding acceptors into In2O3 normally cannot give rise to

a p-type conductivity, as will be explained later.

Intrinsic In2O3 Extrinsic In2O3

Donors Ini3+, VO

2+ SnIn+, FO

+, Hi+

Acceptors Oi2-, VIn

3- ZnIn-,NO

-

The energy levels of electron donors and acceptors in a band structure, the so-called donor levels (ED) and acceptor levels (EA), are indicated in Fig. 3.3. In order to transfer excess electrons from donor levels to the conduction band, the donor impurities need to be ionized. The energy required for such ionization is called the ionization energy (ɛi), and is defined as the energy difference between the donor levels and the CBM, i.e. ɛi= EC- ED. The ionization energy of acceptors is defined similarly, as shown in Fig. 3.3.

Figure 3.3 Schematic representation of the donor levels (ED) and the

acceptor levels (EA) in a semiconductor band structure. The ionization energy (ɛi) is defined as the energy difference between Ec and ED for

donors, and between EV and EA for acceptors.

The ionization energy is generally provided by thermal energy (kBT), which is 25.7 meV at room temperature (25 °C). An electron donor can be classified as a


shallow donor or as a deep donor depending on its ionization energy. A shallow donor generally has an ionization energy around or lower than the thermal energy of 25.7 meV, and can hence efficiently donate electrons. In contrast, a deep donor has an ionization energy much higher than the thermal energy, and is unlikely to be ionized and to contribute to the carrier density. Along the same lines an acceptor can also be classified as a shallow acceptor or as a deep

acceptor.

3.1.2 Electrical properties

a) Conductivity

The electrical conductivity σ is used to quantify how strongly a material can conduct electrical currents, and can be expressed by the following equation:

1e nσ µ ρ−= ⋅ ⋅ = (Eq. 3.6)

where n is the carrier density, µ is the carrier mobility, e is the elementary charge. The resistivity ρ is defined as the reciprocal of the conductivity σ. For thin films with a uniform thickness, the electrical resistance is also often described by the sheet resistance:

/sR dρ= (Eq. 3.7)

where Rs and d stand for the film’s sheet resistance and thickness, respectively.

b) Carrier mobility

When an electric filed (E) is applied across a semiconductor, carriers will be accelerated until they reach a constant averaged velocity called the drift velocity (νd). The carrier mobility is defined as the ratio of the drift velocity to the magnitude of the electric field:

/dv Eµ = (Eq. 3.8)


The carrier mobility can also be expressed as:

* *

mfp

d

ee

m m v

λτµ

⋅⋅= =

⋅ (Eq. 3.9)

where e is the elemental charge, m* is the effective mass of the carriers, and the relaxation time τ depends on the drift velocity νd and mean free path λmfp. The mobility is affected by several scattering mechanisms, as will be discussed below. The relationship between the effective mobility and the scattering mechanisms can be approximately expressed by Matthiessen’s rule:

1 1

i

i

µ µ− −=∑ (Eq. 3.10)

where µi stands for the mobility limited by an individual scattering mechanism.

c) Scattering mechanisms

Ionized impurity scattering. As introduced before, the donors and acceptors in semiconductors are ionized impurities. The Coulomb interaction between impurities and carriers will deflect carriers approaching the impurities. Ionized impurity scattering is the dominant scattering mechanism in doped materials, in case the carrier density is above 1019 cm-3.4

Grain boundary scattering. In polycrystalline semiconductors, grain boundaries are defects, which usually consist of a few atomic layers of disorder between grains, because of orientational mismatch of the atomic positions of the two neighboring crystals at the grain boundary. There, ionized atoms and/or carriers can be trapped, creating potential energy barriers. As a result, the carrier mobility is reduced when carriers travel from one grain to the other.

Neutral impurity scattering. Neutral complexes within grains, such as non-ionized dopant atoms, can cause neutral impurity scattering. This scattering effect is substantial for wide-bandgap semiconductors.5

Phonon scattering. Above the absolute zero temperature, the atoms in a lattice vibrate elastically. The excited states of the vibrations are often represented by phonons, a sort of quasiparticle. During electrical transport, carriers can be scattered by phonons, resulting in a reduced carrier mobility. Phonon


scattering is temperature-dependent.6 At higher temperatures, phonon scattering becomes stronger due to an increase in the number of phonons.

Compared to the aforementioned mechanisms, other scattering mechanisms, such as dislocation scattering, have minor impacts on the carrier mobility.7 All the aforementioned mechanisms exist in intrinsic semiconductors. Admixing dopants into host materials can introduce charged and neutral defects, as well as lattice deformations and, hence, enhance the scattering.

d) Improvement of conductivity by doping

The conductivity of semiconductors can be enhanced by doping, i.e. admixing electron donors into the host materials. However, in addition to contributing to the carrier density, dopants can also act as impurities in the host lattice. These impurities can enhance the ionized impurity scattering and the neutral impurity scattering, resulting in a lower carrier mobility. Therefore, in order to achieve the maximum conductivity, the concentration of dopants should be chosen at an optimized level: on the one hand, there should be a sufficient number of dopants, which can give rise to a significant increase in carrier density; on the other hand, the dopant concentration should not be too high, so that a significant decrease of carrier mobility can be avoided. In other words, dopant atoms should act more as effective donors than as ineffective impurities, i.e. a high doping efficiency should be realized.

3.1.3 Optical properties

a) Optical properties

In inorganic semiconductors, an optical bandgap (Ego) is the threshold of energy for a photon to assist the transition of an electron from a valence band to a conduction band, the so-called optical transition. In non-degenerate semiconductors, as shown in Fig. 3.4 (a), the value of the optical bandgap is approximately equal to the electronic bandgap Eg. In degenerate semiconductors, the energy states at the bottom of the conduction band are populated. A higher phonon energy (hν’>hν) is required to assist electrons to occupy higher energy levels at the conduction band, as shown in Fig. 3.4 (b). As


a result, the optical bandgap is larger than the electronic bandgap, i.e. Ego>Eg. This phenomenon is called Burstein-Moss shift.8

Figure 3.4 Schematic representations of the optical transitions in (a) non-degenerate and (b) degenerate n-type inorganic semiconductors. EC, EV

and EF indicate the positions of the conduction band minimum, the valence band maximum and the Fermi level in the band structures,

respectively. Eg and Ego stand for the electronic bandgap and the optical bandgap, respectively. In degenerate semiconductors, photons with a

higher energy are required (hν’>hν) to induce an optical transition. As a

result, the optical bandgap is larger than the electronic bandgap (Ego>Eg).

b) Transparency

Transparency is an important feature of wide-bandgap semiconductors, and results from their bandgaps. Upon irradiating semiconductors, photons can be reflected, absorbed or transmitted. Typical reflection, absorption and transmittance spectra of a wide-bandgap semiconductor are schematically presented in Fig. 3.5. Reflection takes place over the whole spectrum, and is significant at the (near) infrared region due to the Drude effect9. Absorption mainly occurs at two regions: in the ultra-violet region, photons with an energy higher than the optical bandgap can be absorbed to trigger the optical transition; in the (near) infrared region, photons with a frequency close to the plasma frequency (ωp) can be absorbed by free electron due to the Drude effect. An increase in carrier density can result in a stronger Drude effect, leading to a lower transparency in the (near) infrared region. In wide-bandgap


semiconductors, transmission mainly takes place in the visible region (around 400 to 700 nm), and the transparency is generally above 80% in this region.

Figure 3.5 Schematic reflection, absorption and transmittance spectra of

a wide-bandgap semiconductor. The spectra consist of three regions:

infrared (IR), visible (Vis) and ultraviolet (UV). Ego and ωp are the optical bandgap and the plasma frequency of a semiconductor, respectively. ħωp

is the photon energy where Drude absorption and reflection take place.

3.2 Semiconducting metal oxide

Semiconducting metal oxides compose a major group of semiconductors (the other groups are group IV semiconductors, III-V compound materials, etc.). Besides the common features of semiconductors introduced in the previous section, most semiconducting metal oxides exhibit specific properties, such as a degenerate n-type conductivity and a high carrier mobility in an amorphous phase. The degenerate conductivity makes semiconducting metal oxides very suitable as transparent conducting oxides (TCOs), while the high electron mobility favors their application as an amorphous oxide semiconductor (AOS). In this section, these two properties as well as the two applications will be addressed.

3.2.1 Degenerate n-type conductivity

Semiconducting metal oxides are generally n-type, even in nominally undoped states.10 The intrinsic defects, such as metal interstitials and oxygen vacancies, can contribute to the carrier density. In fact, it is challenging to convert most metal oxides from n-type to p-type by introducing electron acceptors into the materials, because the resulting p-type state can be unstable, i.e. the holes

3.2 Semiconducting metal oxide 39

generated by acceptors can be compensated by the electrons generated by intrinsic defects.11 Moreover, the hole mobility of p-type oxides is generally lower than the electron mobility in the case of n-type conductivity.12

The carrier density levels of metal oxide thin films reported in literature are usually so high that, according to the Mott’s criterion, these films are degenerate n-type even without intentional doping. For example, the effective Bohr radii for In2O3 and ZnO are around 1.313 and 1.4 nm14, which correspond to critical carrier density values of nC ~6.4×1018 and 5.7×1018 cm-3, respectively. The carrier densities of atomic layer deposited ZnO and In2O3 thin films are higher than 1019 cm-3.15,16 Therefore, according to the Mott’s criterion, such ZnO and In2O3 thin films are degenerate n-type semiconductors.

3.2.2 High mobility in an amorphous phase

It is believed that the carrier mobility in amorphous semiconductors is considerably degraded compared to their corresponding crystalline phases. This is actually the case for the conventional semiconductor silicon.10 Crystalline silicon (c-Si) exhibits a carrier mobility of 1500 cm2/Vs, while amorphous silicon (a-Si) shows a mobility lower than 1 cm2/Vs.10 However, non-transition metal oxides, such as In2O3 and SnO2, can show a large electron mobility (>10 cm2/Vs) in their amorphous phase. Such a difference comes from their carrier transport paths, as illustrated in Fig. 4.17 In both a-Si and c-Si, the bottom of the conduction band is composed of hybridized sp

3 orbitals which are strongly directive. In c-Si, the overlap of the sp

3 orbitals forms the carrier path, as shown in Fig. 3.6 (a). In contrast, in a-Si, the disorder greatly degrades the magnitude of bond overlapping (Fig. 3.6 (b)), resulting in a low carrier mobility. In the case of non-transition metal oxides, the chemical bonds between the metal and oxygen atoms are highly ionic. As a result, the bottom of the conduction band is mainly composed of the spatially spread metal ns orbitals with isotropic shapes. The disorder in an amorphous phase does not interrupt the electron transfer paths, as shown in Fig. 3.6 (d). Therefore, non-transition metal oxides in an amorphous phase can have a similar carrier mobility as in the corresponding crystalline phases. Note, that this theory is not limited to non-transition metal oxides. For example, ZnO is a transition metal oxide, but due to the ionic nature of the Zn-O bond, the theory still applies to ZnO.18


Figure 3.6 Schematic orbital drawings for the carrier transport paths, i.e. conduction band bottoms, in (a) crystalline and (b) amorphous covalent

semiconductors, such as Si, and in (c) crystalline and (d) amorphous non-transition metal oxides, such as In2O3.17

3.2.3 Requirements for applications

The application of a semiconducting metal oxide as a TCO, requires films with a high conductivity and a high transparency. The conductivity can be enhanced by increasing the carrier density and/or mobility. However, the increased carrier density can result in a lower transparency due to the Drude effect.9 Therefore, a high carrier mobility with a moderate carrier density is preferred to yield a high conductivity, while maintaining a high transparency.

In the application of AOS, a high carrier mobility, an amorphous phase and a moderate carrier density are required. A high field effect mobility (> 3 cm2/(V∙s)) is required for displays with an ultra-definition, large size and high frame rate operation.19 A moderate electron density ensures that the channel layer can be depleted by a reasonable gate voltage.12 Also, an amorphous phase provides uniform electrical properties, low surface roughness and consistent film thickness over large areas, which are required for the mass-production of displays.20,21 In addition, it avoids the presence of

3.3 Indium oxide 41

defects at (the non-existent) grain boundaries that can deteriorate the stability of devices.12 As stated before, the carrier density in semiconducting metal oxides is generally so high that it will be difficult to switch off the device by depleting the carriers in the active layer. Therefore, in order to decrease the carrier density to a moderate level, electron acceptors are often doped into the oxides.

3.3 Indium oxide

Indium oxide is a popular semiconducting metal oxide, although the abundance of indium is low (~160 ppb22) in the earth crust. When used as a TCO, indium oxide is often doped with tin, forming indium tin oxide (ITO). In the class of TCO materials, ITO is favored because it combines a high transparency with a high electrical conductivity.11 Recent research suggests that hydrogen-doped indium oxide (In2O3:H) can potentially be an alternative for ITO, since In2O3:H can yield a better carrier mobility and a higher transparency than ITO.23,24 When used as an AOS, In2O3 is an essential component of indium gallium zinc oxide (IGZO).25

Indium oxide is a wide-bandgap semiconductor. Its electronic and optical bandgaps are ~2.9 and 3.9 eV, respectively. The difference between these two values will be explained later in this section. In2O3 thin films generally show a degenerate n-type conductivity, which may result from an unintentional incorporation of hydrogen impurities during growth, as will be addressed extensively in Chapter 7.

3.3.1 Basic physical properties of indium oxide

a) Crystal structure

Amorphous In2O3 thin films can be obtained at low substrate temperatures (<130 °C),16,26 while crystalline films can be formed at high substrate temperatures (>130 °C)16,26 or by post-annealing of as-deposited amorphous films23. Crystalline In2O3 can exist in different phases, such as cubic, rhombohedral and orthorhombic phases.27 Under ambient conditions, In2O3 commonly crystallizes in a cubic bixbyite structure (space group symmetry: Ia3) as shown in Fig. 3.7.28 The unit cell of bixbyite contains 80 atoms (48 oxygen


atoms and 32 indium atoms), and has a lattice constant of 1.0117 nm. Along the [001] direction, one unit cell consists of four (004) bi-layers: four are composed of only indium atoms, and the other four of only oxygen atoms. As indicated in Fig. 3.7, along the [001] direction, the thickness of one monolayer can be defined as a quarter of the lattice constant a, i.e. a/4 = 0.25 nm. The areal densities of indium and oxygen atoms in one monolayer are 0.8×1015 and 1.2×1015 atoms/cm2, respectively.

Figure 3.7 Unit cell of the In2O3 bixbyite structure. The red spheres

represent oxygen atoms, while the dark and light green ones represent

indium atoms in two different bonding states.28 Along the [001] direction, the thickness of one monolayer is defined as a quarter of the lattice

constant a.

b) Electron donors

As listed in Table 3.1, indium interstitials (Ini3+) and oxygen vacancies (VO

2+) can act as electron donors in intrinsic In2O3. Today, a debate exists about the main electron donors in In2O3 films. Oxygen vacancies were believed to be the main donor.16,29 However, both theoretical30 and experimental24 studies indicate that hydrogen impurities may also act as electron donors in In2O3 films. Even though a hydrogen impurity is not an intrinsic defect, it can be easily incorporated into In2O3 films, for example, from ambient water or reactants. A theoretical study points out that both interstitial (Hi

+) and substitutional (HO+)

hydrogen have lower formation energies than oxygen vacancies (VO2+).30 The

configurations of interstitial and substitutional hydrogen in an In2O3 lattice are

3.3 Indium oxide 43

shown in Fig. 3.8. In Koida’s experimental work,24 the In2O3 films were prepared by sputtering with intentional hydrogen doping. They claimed that in their annealed and consequently crystallized In2O3 films, hydrogen impurities form the main electron donor.

Figure 3.8 Schematic configurations of two hydrogen defects in an In2O3

lattice: (a) a hydrogen interstitial at an anti-bonding site next to an oxygen atom (Hi

+); (b) a substitutional hydrogen located at an oxygen

vacancy (HO+).30 Similar to Fig. 12, In1 and In2 indicate indium atoms in

two different bonding states. The indexes indicate the lattice orientations.

The percentages represent the deviations of the corresponding In-O bond lengths from the average value of 2.169 Å.

c) Optical bandgap

The electronic bandgap of In2O3 is ~2.9 eV31, but the optical bandgap is ~3.9 eV.32 The difference of ~ 1 eV cannot simply be explained by the aforementioned Burstein-Moss shift. Different from the optical transition as illustrated in Fig. 3.4 (b), a direct optical transition from the VBM to the CBM is forbidden due to the parity selection rules, as shown in Fig. 3.9.33 In2O3 has a centrosymmetric crystal structure. According to Laporte’s selection rule for centrosymmetric molecules, only electronic transitions between states with opposite parities are allowed. Both the VBM and the CBM have even parity, so a direct transition between these two states is forbidden. The first allowed strong optical transition is from the states 0.81 eV below the VBM to the CBM. This phenomenon is the main reason for the ~1 eV difference between the values of the electronic bandgap and the optical bandgap.


Figure 3.9 Band structure of In2O3 and related optical transitions. A direct

optical transition from the VBM to the CBM is forbidden due to the parity selection rule. The first allowed optical transition occurs from the band

0.81 eV below the VBM.34

d) Electron properties

As mentioned before, In2O3 is often doped with tin, forming ITO. The electrical properties of intrinsic and tin-doped In2O3 depend on the growth conditions, such as growth methods, dopants, reactants, substrate temperatures, etc. In Table 3.2, we present several examples to indicate some typical values of the electrical properties of ITO layers.

3.3 Indium oxide 45

Table 3.2 Examples of the properties of intrinsic and Sn-doped In2O3 films

prepared by different deposition methods. The deposition methods listed are dc magnetron sputtering (dc MS), Pulsed Laser Deposition (PLD),

MOCVD, spray CVD and ALD. Tsub means substrate temperature, and d means film thickness. The electrical properties are resistivity (ρ), carrier

density (n) and carrier mobility (µ).

Deposition method

Precursors & co-reactant

Sn content

Tsub (°C)

d

(nm) ρ

(mΩ∙cm) n

(1020 cm-3) µ

(cm2/Vs) Ref.

dc MS In2O3 target 0 <37 150 3.1 2.0 10 35 PLD In2O3 target 0 490 350 0.5 2.4 50 36

MOCVD TMIn, H2O 0 350 74 3.6 0.7 25 37 ALD InCp, H2O&O2 0 100 40 0.4 4.5 40 16

dc MS ITO target 10 wt.% 400 120 0.2 6.7 46 38 PLD ITO target 5.7 wt.% 600 400 0.08 19 42 39

Spray CVD InCl3∙3.5H2O; SnCl2∙2H2O

5.8 at.% 350 215 0.19 9 36 40

ALD InCp, O3; TDMASn, H2O2

5 % SnO2 cycles

275 51 0.39 4.0 49 41

3.3.2 Indium oxide thin films prepared by ALD

The current research on atomic layer deposited In2O3 focuses on the selection of an appropriate indium precursor and its corresponding co-reactant(s), in order to establish growth recipes, which can yield high values of growth per cycle, wide ALD windows and a high film conductivity. In this section, we will give a literature overview of some reported indium precursors, and then choose a representative example to explain the growth mechanism.

a) Overview of indium precursors

The criteria about how to select an appropriate precursor have been described in Chapter 2. The growth properties of several ALD processes reported in literature are listed in Table 3.3. Most of these growth properties cannot fulfil all criteria at the same time. For example, the growth using indium chloride (InCl3) and water requires high substrate temperatures (400-500 °C). Moreover, the growth per cycle is relatively low (0.027 nm), and the by-product HCl can etch the deposited In2O3 films. Some of the other recipes listed in Table 3.3 result in no growth, a growth that is not self-limiting or a high film resistivity, etc. Compared to these recipes, indium cyclopentadienyl (InCp) with H2O and O2 as co-reactants has been reported to fulfil the desired criteria:16,42 high


growth per cycle (0.12 nm), low substrate temperature (100 °C) and low film conductivity (~ 1 mΩ∙cm). Therefore, we choose this recipe as a representative example to further introduce the ALD growth of In2O3.

Table 3.3 Growth properties of selected ALD processes for In2O3 reported in literature.

b) Growth mechanism when using InCp

The growth mechanism introduced here is based on the growth recipe mentioned in the previous section, i.e. using InCp with a co-dosing H2O and O2. Similar to the textbook case of Al2O3 introduced in Chapter 2, the surface reactions are based on ligand exchange. However, the oxidation state of the indium atoms in InCp is +1, instead of +3 as in the other precursors listed in Table 3.3. Therefore, in addition to ligand exchange, a red-ox reaction is required to oxidize indium atoms from the +1 to the +3 state. The growth mechanism is proposed by Libera et al.,16 and illustrated in Fig 3.10. In the first half-cycle, InCp molecules react with the -OH surface groups. As a result, part of the -Cp ligands are removed, forming volatile HCp molecules, while the

Precursor

Co-reactant

Growth per cycle (nm)

Substrate temp.

(°C)

Ref.

InCl3 H2O 0.023-0.027 400-500 43 InCl3 H2O2 0.04 300-500 44

In(hfac)3 H2O or H2O2

No growth / 45

In(thd)3 H2O or H2O2

No growth / 45

In(acac)3 H2O 0.015-0.028 165-250 46 In(acac)3 O3 0.012-0.060 165-300 46 In(CH3)3 H2O 0.01-0.08 150-320 29 In(CH3)3 H2O2 0.05-0.20 350-550 47

In[(iPr)2CN(CH3)2] H2O 0.01-0.04 160-320 48 InCA O3 0.08-0.10 50-250 49

In(C2H5)3 O3 0.05-0.11 50-250 49 DADI O3 0.04-0.08 50-250 49 InCp H2O/O2 0.13-0.16 115-250 16

3.3 Indium oxide 47

unreacted -Cp ligands remain on the surface. In the purge step, the remaining InCp and the by-product HCp molecules are purged away from the reaction chamber. In the second half-cycle, the co-reactant H2O reacts with the resulting surface to release the remaining -Cp ligands and to generate the initial -OH surface groups. Meanwhile, indium is oxidized from the +1 oxidation state to the +3 state by the other co-reactant O2. In the purge step, the remaining H2O and O2 co-reactants and the by-product HCp are purged away. In this mechanism, H2O and O2 each perform separate but necessary functions: H2O releases -Cp ligands and O2 oxides the indium.

Figure 3.10 Schematic representation of the growth mechanism of In2O3 with InCp and a co-dosing of water and oxygen. The indium atoms in the

different oxidation states +1 and +3 are denoted by the red and purple colors, respectively. (a) In the first half-cycle, InCp reacts with -OH surface

groups, whilea fraction of the -Cp ligands is removed at the same time. (b) In the second half-cycle, H2O removes the remaining -Cp ligands, while O2

oxidizes the indium atoms from +1 to +3.16,32

In our work, the growth recipe consists of 5 s InCp dose/2 s purge/5 s stabilization of O2/0.25 s co-dose of H2O and O2/10 s purge. The duration of each step is determined by the corresponding saturation curve, as will be presented later in Chapter 7. A relatively constant growth per cycle of 0.12 nm is obtained within the ALD window of 100-350 °C. Based on the thickness of one monolayer as defined before, one ALD cycle yields around 0.5 monolayer. The resulting film properties, such as the microstructure, composition and electrical properties, will be addressed extensively in Chapter 7 and 8.


3.4 Zinc oxide

ZnO has a direct bandgap of ~ 3.4 eV, and shows a degenerate n-type conductivity even in nominally undoped states. When used as a TCO, ZnO is often doped, typically by Al, Ga, In, B and F,50–53 to improve the conductivity. Al-doped ZnO (ZnO:Al) is considered as an alternative to ITO, because zinc is more abundant and thus cheaper than indium (the abundance of zinc in the earth crust is 132 ppm).11 In the field of AOS, intrinsic ZnO (i-ZnO) and IGZO are the most common materials.12

3.4.1 Basic physical properties of zinc oxide

a) Crystal structure

ZnO has two main crystal structures: hexagonal wurtzite and cubic zinc blende. The wurtzite structure is more stable under ambient conditions, like room temperature and atmospheric pressure.7,52,54 The wurtzite structure (space group symmetry: P63mc) is shown in Fig. 3.11. One hexagonal unit cell contains two formula units of ZnO. Along the c-axis, one unit cell consists of two (0002) layers, each consisting of two sub-layers composed of the individual elements, zinc and oxygen. As indicated in Fig. 3.11, along the c-axis, one monolayer of ZnO can be defined as half of the lattice constant c, i.e. c/2 = 0.23 nm, with an areal zinc (or oxygen) atomic density of 1.1×1015 atoms/cm2.

Figure 3.11 Schematic representation of the wurtzite structure of ZnO. 7 A

hexagonal unit cell is indicated by the thick lines. The black and grey

spheres represent Zn and O atoms, respectively. The lattice constants are a = 0.325 nm and c = 0.521 nm. Along the c-axis, the thickness of one

monolayer is defined as half of the lattice constant c.

3.4 Zinc oxide 49

b) Electron donors

In i-ZnO thin films, interstitial zinc (Zni2+) and oxygen vacancies (VO

2+), as shown in Figs. 3.12 (a) and (b), can act as electron donors.51 Similar to the case of In2O3, historically, oxygen vacancies were believed to be the main electron donor.55 This is based on the observation that a change of the O2 partial pressure during deposition or post-annealing can result in a change in the film conductivity. However, some recent theoretical studies suggest that oxygen vacancies can only act as a deep donor in ZnO films, so they may not be the primary source contributing to the n-type conductivity.51

Hydrogen impurities can be easily incorporated into ZnO films during preparation. Hydrogen interstitials (Hi

+) and hydrogen at a substitutional site of oxygen (HO

+), as shown in Figs. 3.12 (c) to (e), have been proposed to be shallow donors, based on theoretical studies.56,57 In an experimental study, we observed the vibration mode of O-H bonds by infrared spectroscopy.58 Such an O-H bond may be formed between an oxygen atom in the lattice and a hydrogen interstitial (Hi

+). Moreover, a change in the HO+ density can also

explain the aforementioned change of the conductivity under different oxygen partial pressures.57

Figure 3.12 Calculated local atomic geometries of some electron donors

in ZnO: (a) zinc interstitial; (b) oxygen vacancy; (c) hydrogen interstitial at a bond-center site; (d) hydrogen interstitial at an anti-bonding site; (e)

hydrogen at a substitutional site of oxygen.51


c) Electrical properties

The aforementioned oxygen vacancies and hydrogen impurities can contribute to the n-type conductivity of unintentionally doped ZnO films. As mentioned before, to further improve the conductivity, ZnO films are often intentionally doped. Similar to In2O3, the resulting electrical properties depend on the growth conditions. In Table 3.4, we present the electrical properties of several typical ZnO thin films.

Table 3.4 Electrical properties of several ZnO films doped by different elements and prepared by different deposition methods.59 The

deposition methods are radio frequency magnetron sputtering (RFMS), pulsed laser deposition (PLD), metal-organic chemical vapor deposition

(MOCVD), plasma-enhanced chemical vapor deposition (PECVD) and ALD. Tsub is substrate temperature, and d is film thickness. The resulting

electrical properties are resistivity (ρ), carrier density (n) and carrier

mobility (µ). n/a means not applicable.

3.4.2 Zinc oxide thin films prepared by ALD

The ALD growth of intrinsic ZnO thin films is already well established. As listed in Table 3.4, diethyl zinc ((C2H5)2Zn, DEZ) and water are commonly used as the zinc precursor and the corresponding co-reactant, respectively. This common recipe can give rise to a wide ALD window, a high growth per cycle and reproducible film properties. In this section, the common growth recipe will be introduced. After this, ZnO:Al is chosen as an example to illustrate how doping can modify the electrical properties of ZnO thin films.

Methods Precursors & co-reactant

Doping element

Tsub (°C)

d

(nm) ρ

(mΩ∙cm) n

(1020 cm-3) µ

(cm2/Vs) Ref.

RFMS ZnO:Al2O3 target n/a 150 300 0.14 13 34 60 PLD ZnO:Ga2O3 target n/a 300 300 0.08 25 31 61

MOCVD (C2H5)2Zn, C2H5O F 400 780 0.6 4 25 62 PECVD (C2H5)2Zn, H2O Ga 350 410 0.75 5.5 15 63 PECVD (C2H5)2Zn, H2O Al 200 300 0.4 9 18 64

ALD (C2H5)2Zn, H2O Al 200 75 0.7 7 13 65 ALD (C2H5)2Zn, H2O B 150 45 3.5 2.2 8 66

3.4 Zinc oxide 51

a) Growth mechanism

When DEZ and H2O are used as reactants, the growth mechanism is based on ligand exchange, and is very similar to the textbook example of Al2O3 presented in Chapter 2. As shown in Fig. 3.13, in the first half-cycle, DEZ molecules react with the -OH surface groups, forming Zn-O bonds on the surface and releasing the volatile byproduct C2H6. At the end of this half-cycle, the surface is terminated by -C2H5. In the second half-cycle, H2O react with the -C2H5 surface groups. At the end of the second half-cycle, the surface groups are changed back to -OH.

Figure 3.13 Schematic representation of the growth mechanism of ZnO

thin films: (a) in the first half-cycle, diethyl zinc ((C2H5)2Zn, DEZ) is dosed as a zinc precursor, and reacts with -OH groups on surface; (b) in the

second half-cycle, H2O is dosed as a co-reactant, and reacts with -C2H5

groups on surface.

b) ALD window

In our work, the growth recipe consists of 30 ms DEZ dose/ 5 s purge/ 20 ms H2O dose/ 6 s purge.67 The growth per cycle as a function of substrate temperature is shown in Fig. 3.14. A relatively constant growth per cycle of 0.17 to 0.22 nm is obtained at substrate temperatures ranging from 100 °C to 250 °C, i.e. the ALD window is 100-250 °C. The deposited films have a polycrystalline wurtzite structure with a preferred <0002> grain orientation67 By comparing the growth per cycle with the thickness of one monolayer


defined as before, one ALD cycle in the ALD window approximately yields a growth of 0.65-0.85 monolayer.

Figure 3.14 Growth per cycle of i-ZnO thin films as a function of substrate

temperature.67

c) Literature overview of ZnO:Al by ALD

Among the doping elements (aluminum14,68–80, boron66,81,82, gallium83 etc.) reported in literature, aluminum is the most reported one. For the growth of ZnO:Al, trimethyl-aluminum (Al(CH3)3, TMA) is commonly used as the Al precursor. Fig. 3.15 shows the electrical properties of ZnO:Al films with different Al doping levels as selected from literature.69–71,73 The doping level is described by the atomic percentage of metal atoms being Al, the so-called Al

fraction. As shown in Fig. 15 (a), in all series, the optimized Al fractions with respect to the lowest resistivities are consistently around 2-3 at.%. At low Al fractions, a significant increase of carrier density with increasing Al doping level is observed, while at high Al fractions, the carrier density decreases. The carrier mobility decreases with increasing Al fraction as observed in the entire doping range. The reasons for these trends in the electrical properties will be discussed extensively in the Chapters 4, 5 and 6.

3.4 Zinc oxide 53

Figure 3.15 (a) Resistivity, (b) carrier concentration and (c) mobility of Al-

doped ZnO films grown by ALD as a function of the Al fraction from

selected literature references67,69–71,73 and our own work15.


3.5 References

1 http://users.ece.gatech.edu/~shensc/research.html 2 W. Dan, Solid State Physics (Chinese Edition), 2nd ed. (Tsinghua University Press, Beijing, 2007). 3 N.F. Mott, Can. J. Phys. 34, 1356 (1956). 4 K. Ellmer and R. Mientus, Thin Solid Films 516, 4620 (2008). 5 T. Ouisse, Phys. B 270, 262 (1999). 6 B. Macco, H.C.M. Knoops, and W.M.M. Kessels, ACS Appl. Mater. Interfaces 7, 16723 (2015). 7 K. Ellmer, A. Klein, and B. Rech, Transparent Conductive Zinc Oxide: Basics and

Applications in Thin Film Solar Cells (Springer, New York, 2008). 8 B. Elias, Phys. Rev. 93, 632 (1954). 9 H.C.M. Knoops, B.W.H. van de Loo, S. Smit, M. V. Ponomarev, J.-W. Weber, K. Sharma, W.M.M. Kessels, and M. Creatore, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 33, 021509 (2015). 10 T. Kamiya and H. Hosono, NPG Asia Mater. 2, 15 (2010). 11 D.S. Ginley, H. Hosono, and D.C. Paine, Handbook of Transparent Conductors (Springer, New York, 2010). 12 E. Fortunato, P. Barquinha, and R. Martins, Adv. Mater. 24, 2945 (2012). 13 J.M. Dekkers, Transparent Conducting Oxides on Polymeric Substrates, Twente University, The Netherlands, 2007. 14 J.-S. Na, G. Scarel, and G.N. Parsons, Adv. Funct. Mater. 21, 448 (2011). 15 Y. Wu, P.M. Hermkens, B.W.H. van de Loo, H.C.M. Knoops, S.E. Potts, M.A. Verheijen, F. Roozeboom, and W.M.M. Kessels, J. Appl. Phys. 114, 024308 (2013). 16 J.A. Libera, J.N. Hryn, and J.W. Elam, Chem. Mater. 23, 2150 (2011). 17 K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, and H. Hosono, Nature 432, 488 (2004). 18 T. Kamiya, K. Nomura, and H. Hosono, J. Disp. Technol. 5, 273 (2009). 19 J.S. Park, W.-J. Maeng, H.-S. Kim, and J.-S. Park, Thin Solid Films 520, 1679 (2012). 20 J.-Y. Kwon, D.-J. Lee, and K.-B. Kim, Electron. Mater. Lett. 7, 1 (2011).

3.5 References 55 21 D.B. Buchholz, L. Zeng, M.J. Bedzyk, and R.P.H. Chang, Prog. Nat. Sci. Mater. Int. 23, 475 (2013). 22 Http://periodictable.com/Properties/A/CrustAbundance.an.html, (n.d.). 23 B. Macco, Y. Wu, D. Vanhemel, and W.M.M. Kessels, Phys. Status Solidi - Rapid Res. Lett. 8, 987 (2014). 24 T. Koida, H. Fujiwara, and M. Kondo, Jpn. J. Appl. Phys. 46, L685 (2007). 25 K. Nomura, A. Takagi, T. Kamiya, H. Ohta, M. Hirano, and H. Hosono, Jpn. J. Appl. Phys. 45, 4303 (2006). 26 F.O. Adurodija, L. Semple, and R. Brüning, J. Mater. Sci. 41, 7096 (2006). 27 M.F. Bekheet, M.R. Schwarz, S. Lauterbach, H.J. Kleebe, P. Kroll, R. Riedel, and A. Gurlo, Angew. Chemie - Int. Ed. 52, 6531 (2013). 28 E.H. Morales, Y. He, M. Vinnichenko, B. Delley, and U. Diebold, New J. Phys. 10, 125030 (2008). 29 D.-J. Lee, J.-Y. Kwon, J. Il Lee, and K.-B. Kim, J. Phys. Chem. C 115, 15384 (2011). 30 S. Limpijumnong, P. Reunchan, A. Janotti, and C. Van de Walle, Phys. Rev. B 80, 193202 (2009). 31 P.D.C. King, T.D. Veal, F. Fuchs, C.Y. Wang, D.J. Payne, A. Bourlange, H. Zhang, G.R. Bell, V. Cimalla, O. Ambacher, R.G. Egdell, F. Bechstedt, and C.F. McConville, Phys. Rev. B - Condens. Matter Mater. Phys. 79, 205211 (2009). 32 D. Vanhemel, Atomic Layer Deposition of In2O3: Growth, Morphology and Hydrogen Doping, Master thesis, Eindhoven University of Technology, The Netherlands, 2015. 33 A. Walsh, J.L.F. Da Silva, S.H. Wei, C. Körber, a. Klein, L.F.J. Piper, A. Demasi, K.E. Smith, G. Panaccione, P. Torelli, D.J. Payne, a. Bourlange, and R.G. Egdell, Phys. Rev. Lett. 100, 2 (2008). 34 O. Bierwagen, Semicond. Sci. Technol. 30, 024001 (2015). 35 P.K. Song, H. Akao, M. Kamei, Y. Shigesato, and I. Yasui, Japanese J. Appl. Physics, Part 1 Regul. Pap. Short Notes Rev. Pap. 38, 5224 (1999). 36 E.J. Tarsa, J.H. English, and J.S. Speck, Appl. Phys. Lett. 62, 2332 (1993). 37 C. Wang, V. Cimalla, G. Cherkashinin, H. Romanus, M. Ali, and O. Ambacher, Thin Solid Films 515, 2921 (2007). 38 Y. Shigesato and D.C. Paine, Thin Solid Films 238, 44 (1994). 39 H. Ohta, M. Orita, M. Hirano, H. Tanji, H. Kawazoe, and H. Hosono, Appl.


Phys. Lett. 76, 2740 (2000). 40 Y. Sawada, C. Kobayashi, S. Seki, and H. Funakubo, Thin Solid Films 409, 46 (2002). 41 J.W. Elam, D.A. Baker, A.B.. Martinson, M.J. Pellin, and J.T. Hupp, J. Phys. Chem. C 112, 1938 (2008). 42 J.W. Elam, A.B.F. Martinson, M.J. Pellin, and J.T. Hupp, Chem. Mater. 18, 3571 (2006). 43 T. Asikainen, M. Ritala, and M. Leskeei, J. Ectrochemical Soc. 141, 3210 (1994). 44 M. Ritala, T. Asikainen, and M. Leskelä, Electrochem. Solid-State Lett. 1, 156 (1998). 45 M. Ritala, T. Asikainen, and M. Leskelä, Mater. Res. Soc. Symp. Proc. 426, 513 (1996). 46 O. Nilsen, R. Balasundaraprabhu, E.V. Monakhov, N. Muthukumarasamy, H. Fjellvåg, and B.G. Svensson, Thin Solid Films 517, 6320 (2009). 47 K. Ozasa, T. Ye, and Y. Aoyagi, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 12, 120 (1994). 48 M. Gebhard, M. Hellwig, H. Parala, K. Xu, M. Winter, and A. Devi, Dalton Trans. 43, 937 (2014). 49 W.J. Maeng, D. Choi, J. Park, and J. Park, J. Alloys Compd. 649, 216 (2015). 50 L. Schmidt-Mende and J.L. MacManus-Driscoll, Mater. Today 10, 40 (2007). 51 A. Janotti and C.G. Van de Walle, Reports Prog. Phys. 72, 126501 (2009). 52 U. Özgür, D. Hofstetter, and H. Morkoç, Proc. IEEE 98, (2010). 53 W. Beyer, J. Hüpkes, and H. Stiebig, Thin Solid Films 516, 147 (2007). 54 N. Izyumskaya, V. Avrutin, Ü. Özgür, Y.I. Alivov, and H. Morkoç, Phys. Status Solidi 244, 1439 (2007). 55 P.P. Edwards, A. Porh, M.O. Jones, D. V. Morgan, and R.M. Perks, Dalt. Trans. 19, 2995 (2004). 56 C. G. van de Walle, Phys. Rev. Lett. 85, 1012 (2000). 57 A. Janotti and C.G. van de Walle, Nat. Mater. 6, 44 (2006). 58 S. Jokela and M. McCluskey, Phys. Rev. B 72, 113201 (2005). 59 B.L. Williams, M. V Ponomarev, M.A. Verheijen, H.C.M. Knoops, A. Chandramohan, L. Duval, and M. Creatore, Plasma Process. Polym. 13, 54

3.5 References 57

(2015). 60 Y. Igasaki and H. Saito, J. Appl. Phys. 70, 3613 (1991). 61 S.-M. Park, T. Ikegami, and K. Ebihara, Thin Solid Films 513, 90 (2006). 62 J. Hu and R.G. Gordon, Sol. Cells 30, 437 (1991). 63 J.J. Robbins, J. Harvey, J. Leaf, C. Fry, and C. a. Wolden, Thin Solid Films 473, 35 (2005). 64 M. V Ponomarev, M.A. Verheijen, W. Keuning, M.C.M. Van De Sanden, and M. Creatore, J. Appl. Phys. 112, 043708 (2012). 65 B. Macco, D. Deligiannis, S. Smit, R.A.C.M.M. van Swaaij, M. Zeman, and W.M.M. Kessels, Semicond. Sci. Technol. 29, 122001 (2014). 66 D. Garcia-Alonso, S.E. Potts, C.A.A. van Helvoirt, M.A. Verheijen, and W.M.M. Kessels, J. Mater. Chem. C 3, 3095 (2015). 67 P.M. Hermkens, Atomic Layer Deposition of ZnO and Al-Doped ZnO, Master Thesis, Eindhoven University of Technology, The Netherlands, 2012. 68 V. Lujala, J. Skarp, M. Tammenmaa, and T. Suntola, Appl. Surf. Sci. 82-83, 34 (1994). 69 C.H. Ahn, H. Kim, and H.K. Cho, Thin Solid Films 519, 747 (2010). 70 P. Banerjee, W.-J. Lee, K.-R. Bae, S.B. Lee, and G.W. Rubloff, J. Appl. Phys. 108, 043504 (2010). 71 D.-J. Lee, H.-M. Kim, J.-Y. Kwon, H. Choi, S.-H. Kim, and K.-B. Kim, Adv. Funct. Mater. 21, 448 (2011). 72 J. Elam and S. George, Chem. Mater. 15, 1020 (2003). 73 J.W. Elam, D. Routkevitch, and S.M. George, J. Electrochem. Soc. 150, 339 (2003). 74 H. Saarenpää, T. Niemi, A. Tukiainen, H. Lemmetyinen, and N. Tkachenko, Sol. Energy Mater. Sol. Cells 94, 1379 (2010). 75 J.Y. Kim, Y.-J. Choi, H.-H. Park, S. Golledge, and D.C. Johnson, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 28, 1111 (2010). 76 N.P. Dasgupta, S. Neubert, W. Lee, O. Trejo, J.-R. Lee, and F.B. Prinz, Chem. Mater. 22, 4769 (2010). 77 J.W. Elam, Z.A. Sechrist, and S.M. George, Thin Solid Films 414, 43 (2002). 78 H. Yuan, B. Luo, D. Yu, A. Cheng, S.A. Campbell, and W.L. Gladfelter, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 30, 01A138 (2012).


79 Z. Baji, Z. Lábadi, Z.E. Horváth, M. Fried, B. Szentpáli, and I. Bársony, J. Therm. Anal. Calorim. 105, 93 (2011). 80 T. Dhakal, D. Vanhart, R. Christian, A. Nandur, A. Sharma, and C.R. Westgate, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 30, 021202 (2012). 81 Y. Yamamoto, K. Saito, K. Takahashi, and M. Konagai, Sol. Energy Mater. Sol. Cells 65, 125 (2001). 82 B. Sang, A. Yamada, and M. Konagai, Sol. Energy Mater. Sol. Cells 49, 19 (1997). 83 K. Saito, Y. Hiratsuka, A. Omata, H. Makino, S. Kishimoto, T. Yamamoto, N. Horiuchi, and H. Hirayama, Superlattices Microstruct. 42, 172 (2007).

Chapter 4

Electrical transport and Al doping efficiency in nanoscale ZnO films prepared by atomic

layer deposition*

Abstract

In this work, the structural, electrical and optical properties as well as chemical bonding state of Al-doped ZnO films deposited by atomic layer deposition have been investigated to obtain insight into the doping and electrical transport mechanisms in the films. The range in doping levels from 0 to 16.4% Al, was accomplished by tuning the ratio of ZnO and Al2O3 ALD cycles. With X-ray photoelectron spectroscopy depth profiling and transmission electron microscopy, we could distinguish the individual ZnO and AlOx layers in the films. For films with a thickness of 40 nm, the resistivity improved from 9.8 mΩ∙cm for intrinsic ZnO to an optimum of 2.4 mΩ∙cm at 6.9 at.% Al. The binding energy of Zn 2p3/2 increased by 0.44 eV from the intrinsic ZnO to the highest Al-doped ZnO. This shift can be ascribed to an increase of the Fermi level. Ex-situ spectroscopic ellipsometry and Fourier transform infrared spectroscopy were used to measure the optical properties, from which the carrier concentration and intra-grain mobility were extracted. The results showed that with increasing Al content, the grain boundary mobility increased at first due to an increased Fermi level, and then decreased mainly due to the scattering at AlOx/ZnO interfaces. For the same reasons, the doping efficiency of Al for highly Al-doped ZnO dropped monotonically with increasing Al. Furthermore, a blue shift of the optical band-gap ΔEg up to 0.48 eV was observed, consistent with the shifts of the Fermi level and the binding energy of the Zn 2p3/2 state.

*Published as: Y. Wu, P.M. Hermkens, B.W.H. van de Loo, H.C.M. Knoops, S.E. Potts, M.A.

Verheijen, F. Roozeboom and W.M.M. Kessels, J. Appl. Phys., 114 024308 (2013).

62 Electrical transport and Al doping efficiency in ZnO

4.1 Introduction

ZnO is a transparent semiconductor with a wide and direct band gap of 3.4 eV. For this reason, ZnO thin films are widely studied a) as semiconducting layers in thin-film transistors,1 b) as active layers in gas sensors,2–6 and c) as alternatives for indium tin oxide which is currently used as a transparent conducting oxide (TCO)7–10 in Si-based solar cells.11 For the latter application, ZnO thin films doped with B,12 Al,13–15 Ga,16 etc., have been actively investigated, due to their high conductivity, optical transparency, high thermal stability and last but not least the high material abundance. Thin ZnO films (<100 nm) with relatively low resistivity (~1 mΩ∙cm) are generally desired for the aforementioned applications. For example, when used as a semiconducting layer in gas sensors, the sensitivity of ZnO maximizes17 when the ZnO thickness is of the same scale of the Debye-length (in the order of 10nm).18

Several deposition techniques have been reported for ZnO films, such as magnetron sputtering,19 pulsed laser deposition,20 chemical vapor deposition21 and atomic layer deposition (ALD).22–25 Among these, ALD is considered to be a promising technique to deposit nanoscale ZnO films, because it is a self-limiting thin-film growth technique that guarantees excellent film conformality, uniformity, precise thickness control, sharp interfaces, as well as possibilities for creating reproducible and well-defined nanolaminate structures.26 The doping concentration of Al-doped ZnO (AZO) can be precisely tuned by careful control of the ALD cycle ratio between the Zn and the dopant precursors. Therefore, the thickness, conductivity and carrier density of the films can be controlled to meet stringent specifications.27

In the literature, AZO films prepared by ALD have been investigated for their morphological, electrical and structural properties as a function of Al concentration.1,13,27–29 The chemical environment as well as the atomic charge of elements as a function of doping concentration still need to be studied in more detail, because the chemical bonding states of the elements (e.g. Al as an effective AlZn

+ dopant in the ZnO lattice or Al within ineffective AlOx clusters) determine the amount of free charge carriers in AZO thin films. The electrical properties regarding the charge transport, such as the Fermi level, charge carrier mobility within grains and at grain boundaries, are not fully understood either, while such parameters determine the performance of the AZO films as a

4.2 Experimental details 63

semiconductor or transparent conductive oxide. Moreover, in order to understand how Al doping affects the crystallinity of ZnO films, the role of Al with respect to the structural and electrical properties of AZO films needs to be elucidated. The doping efficiency of Al should be calculated quantitatively for further characterization and comparison. All these parameters are important for a fundamental understanding of the doping mechanism, and consequently, the optoelectronic film properties of AZO as well.

Therefore, in this work, AZO films with various Al doping levels were prepared by ALD and characterized extensively. Firstly, the thickness and deposition temperature were varied to find optimized values for the resistivity. Secondly, in order to figure out how the atomic distribution and crystallinity of AZO films affect their electrical properties, depth profiling X-ray photoelectron spectroscopy (XPS) and transmission electron microscopy (TEM) were used. Thirdly, in order to obtain insight into the optical and electrical properties, spectroscopic ellipsometry (SE) and Fourier transform infrared spectroscopy (FTIR) were applied to derive the carrier density, intra-grain mobility, grain boundary mobility and optical band gap. Next, the concept of Al doping efficiency was used to characterize the doping effect quantitatively. Finally, the shifts of the Fermi level were calculated and related to both the shifts of the binding energy and the optical band gap.

4.2 Experimental details

4.2.1 Film preparation

Intrinsic ZnO (i-ZnO) and AZO films were deposited using an open-load Oxford Instruments OpALTM reactor. Si wafers with 450 nm thermally grown SiO2 on top were used as substrates. Diethyl zinc [DEZ, Zn(C2H5)2] and deionized water (DI H2O) vapor were used as precursors for the deposition of i-ZnO films and the ZnO cycles in AZO films. The dosing and purging times in one ZnO cycle were DEZ (50 ms) / purge (5 s) / DI vapor (20 ms) / purge (6 s). Similarly, trimethyl-aluminium [TMA, Al2(CH3)6] and DI H2O vapor were used as precursors for the deposition of AlOx layers in AZO films, with dosing and cycling times TMA (20 ms) / purge (3.5s) / DI H2O vapor (20 ms) / purge (3.5 s).


4.2.2 Electrical and structural analysis

The resistivity of the films was measured ex-situ at room temperature using a Signatone four-point probe (FPP), in combination with a Keithley 2400 Source Measurement Unit. Hall measurements were carried on a BioRad instrument. The XPS set-up used in this work was a Thermo Scientific K-Alpha KA1066 spectrometer using monochromatic Al Kα X-ray radiation (hν = 1486.6 eV). Photoelectrons were collected at a take-off angle of 60°, as measured from the surface normal. A 400-μm diameter X-ray spot was used in the analyses. A flood gun was used to correct for possible sample charging. Furthermore, all samples were corrected for sample charging using the Si 2p orbital from the Si substrate as an internal reference with a binding energy of 99.3 eV.30 For XPS depth-profiling, an Ar-gun with a voltage of 1000 eV and high current (17.9 μA) was applied to sputter the i-ZnO and AZO films. The sputtering rate in this setting was ~0.13 nm/s. Cross-sectional TEM studies on FIB lift-out samples were performed in bright field and in high-angle annular dark-field (HAADF) modes using a FEI Tecnai F30ST transmission electron microscope (TEM).

4.2.3 Optical analysis

The SE measurements were performed using a J.A.Woollam Co. Inc. M-2000D spectrometer with an XLS-100 light source (1.2 - 6.5 eV of photon energy).31 A Psemi-M0 model32 was applied in the data analysis to extract information on the material properties, such as the thickness, and on the optical parameters, in particular on the dielectric constants ε1 and ε2. Furthermore, the reflectance was measured by using a Bruker Tensor 27 reflectance-FTIR instrument in the photon energy range of 0.12 - 0.86 eV. A Drude oscillator model33,34 was used afterwards to extract the optical mobility and carrier density from the combined data obtained from reflectance-FTIR and the SE. Details of the modeling will be described in a separate publication.35

4.3 Results and discussion

4.3.1 Intrinsic ZnO

For i-ZnO films, the effects of film thickness and growth temperature on their resistivity were studied, as shown in Fig. 4.1. Figure 4.1 (a) shows the resistivity

4.3 Results and discussion 65

of i-ZnO films as a function of thickness. In order to avoid possible issues with the aging effect of ZnO films, the resistivity and the thickness were measured directly after the deposition (within 30 minutes) by FPP and SE, respectively. Based on the curve in Fig. 4.1 (a), a critical thickness (D0) and the corresponding resistivity (ρ0) can be defined as D0=40 nm and ρ0=11.0 mΩ∙cm, respectively. For films with thicknesses above D0, the resistivity does not improve significantly, while below D0, the resistivity increases significantly as the thickness decreased. According to Kasap’s model on polycrystalline thin films,36 the scattering of electrons at the surface, interface and grain boundaries during the electrical transport may lead to such a phenomenon. The mean free path of electrons is limited when the thickness of the film is below its critical value D0. Meanwhile, a lower degree of crystallinity and smaller grain size in the initial layer might also limit the electrical properties. Therefore, a thickness of ~40 nm was chosen as a standard for further investigation of AZO film growth. As mentioned in the Introduction, ZnO films with thickness well below 100 nm and with a relatively low resistivity are desired to meet the requirements of various applications.

Figure 4.1 (a) Resistivity of i-ZnO films grown on 450 nm SiO2 / p-Si

substrates at 300 °C as a function of film thickness. The critical thickness (D0) and the corresponding resistivity (ρ0) are indicated in the figure.

When the thickness of i-ZnO films is below the critical value, the resistivity increases abruptly. (b) Resistivity of 40 nm i-ZnO films

deposited on 450nm SiO2/p-Si substrate as a function of the substrate temperature.

The effect of the growth temperature on the resistivity of i-ZnO is shown in Fig. 4.1 (b). The optimum growth temperature with respect to minimum resistivity was between 200 °C and 250 °C. Similar results have been described in other


reports where the lower resistivity at the temperature between 200 °C and 250 °C was ascribed to a higher carrier density37–40 of i-ZnO films, compared to other growth temperatures. Furthermore, the same i-ZnO films were deposited on glass substrate for comparison. The result showed that the resistivity of the films on glass substrates was slightly higher than that of films grown on thermal oxide 450 nm SiO2/ p-Si substrate. Based on the aforementioned results, a thickness of 40 nm, growth temperature of 250 °C and substrates of thermal oxide 450 nm SiO2/ p-Si were chosen for our study on AZO films.

4.3.2 Al-doped ZnO

A series of AZO films was prepared with different aluminum concentrations. The aluminum concentration was denoted as “aluminum fraction” (AF), and defined as the atomic ratio Al/(Al+Zn), that is the fraction of Zn atoms replaced with Al. In order to prepare a particular AZO film with a certain AF, one TMA cycle was inserted after a certain number m of DEZ cycles. Thus, one “supercycle” of AZO film was defined as m cycles of DEZ plus one subsequent TMA cycle, and m is called “cycle ratio”. The total number of supercycles M and the cycle ratio m were chosen to target a nominal thickness of around 40 nm for each sample on the basis of the growth per cycle (GPC) for pure ZnO and Al2O3:

2 3

Thickness = ( )ZnO Al OM GPC m GPC× × + (Eq. 4.1)

Typical GPC values of pure ZnO films at 250 °C were 0.16 nm/cycle. The GPC of a single Al2O3 cycle on a ZnO matrix is 0.15 nm/cycle, as determined using in-

situ SE. The nominal Al fraction AFNom was calculated based on the GPCs, by

2 3

2 3

100%Al O

Nom

Al O ZnO

GPCAF

GPC GPC n= ×

+ × (Eq. 4.2)

Depth-profiling XPS was used to measure the atomic percentage of Al, Zn, C and O throughout the films (indicated by Alat.%, Znat.%, etc.), and, consequently, the actual Al fraction AFXPS was calculated by

.

. .

%100%

% %

atXPS

at at

AlAF

Al Zn= ×

+ (Eq. 4.3)


The interspacing l between adjacent AlOx layers was determined by l=D/M, where D is the total thickness of the AZO films.

In total, 11 AZO samples were prepared with varying AF values. The parameters are listed in Table 4.1. The nominal AF (AFNom) was found to deviate from the AFXPS, especially at high AFNom values, as presented in Fig. 4.2 (a). The nucleation delay of ZnO on an AlOx matrix41 has been reported to be the reason for such a deviation: the actual GPCZnO after one Al2O3 cycle was smaller than that on bulk ZnO matrix. Therefore, in our further analysis and discussion, the doping level of aluminum is described in terms of AFXPS, instead of AFNom.

Table 4.1 Growth parameters of AZO films deposited by thermal ALD at 250 °C on p-Si substrate with 450nm thermal oxide SiO2. AFNom is the

nominal aluminum fraction defined by Eq. 4.2. The actual aluminum fraction was calculated from Eq. 4.3 using the atomic percentage

obtained from depth-profiling XPS, and is denoted as AFXPS. Thicknesses

were determined from SE data which were analyzed using the Psemi-M0 model.32 The error in the thickness values is typically < 1 nm.

Sample ID Cycle ratio n

Supercycle N

AFNom

(%) AFXPS

(%) Thickness

D (nm) Interspacing of AlOx

l (nm) 01 N/A N/A 0 0 41.2 N/A

02 126 2 0.7 0.9 40.7 20.4

03 85 3 1.1 1.9 41.1 13.7

04 51 5 1.8 3.0 40.1 8.0

05 36 7 2.5 4.1 40.4 5.8

06 28 9 3.2 5.9 38.9 4.3

07 23 11 3.9 6.9 41.4 3.8

08 18 13 5.0 9.4 39.6 3.0

09 16 15 5.5 11.0 40.1 2.7

10 14 17 6.3 13.1 39.7 2.3

11 12 19 7.2 16.4 38.2 2.0

The atomic percentages of Zn, Al and O with different AFXPS are presented in Fig. 4.2 (b). Firstly, no carbon was detected throughout the films. Carbon was only present as surface contamination. Secondly, the atomic percentage of O increased at higher AFXPS. In the ideal case, if all of the aluminum atoms would be incorporated as dopants substituting the Zn atoms in the form of AlZn

+, the aluminum doping would not lead to additional oxygen atoms in the AZO films. Thus, the increase of oxygen content suggests the existence of an Al2O3-like


phase in the AZO films, since the O atomic percentage is larger in Al2O3 than in a pure ZnO.

Figure 4.2 (a) Aluminum fraction measured by XPS as a function of

nominal aluminum fraction. The relation for the nominal aluminum fraction is also indicated. (b) Atomic percentage of Zn, O and Al measured

by XPS for various aluminum fractions AFXPS.

Figure 4.3 (a) shows the distribution of Al, Zn, O and Si elements of an AZO film along the growth direction as obtained by depth-profiling XPS. The sample corresponding to Fig. 4.3 (a) was deposited with the same recipe as sample 03 in Table 1, but on a p-Si substrate with ~1.5 nm native SiO2 instead of 450 nm thermally grown SiO2. This ~40 nm AZO film deposited from 3 supercycles contains basically three separate AlOx/ZnO stacks. At around 40 nm depth, the Zn and O levels dropped to zero, and Si appeared, reaching 100 at.% at the interface between the AZO film and the Si substrate. In the XPS depth profile, Al signals appear at certain depths, consistent with the schematic representation of the sample in the graph. Such a periodic variation in atomic percentage suggests the presence of a nanolaminate structure resulting from the ALD deposition scheme.42,43 The nanolaminate structure was confirmed by TEM imaging as well, as shown in Fig. 4.4 (a). Figure 4.4 (a) shows a HAADF STEM image of sample 03. The contrast in the imaging mode is caused by the mass difference between the elements. Within the bulk of the film, the image reveals two AlOx layers with a lower atomic number, i.e. layers with a higher Al content. The high resolution TEM image of Fig. 4.4 (b) gives more information about the morphology of the AZO films. AZO grains appear separated into three regions and two AlOx layers are located at the interfaces between these regions. The image shows clearly that Al2O3 ALD cycles have interrupted the


growth of the ZnO grains and that new ZnO grains nucleated after the individual Al2O3 ALD cycles. The latter is also in agreement with X-ray diffraction patterns of the AZO films, which revealed that the crystallinity of the films decreases at higher AF as also reported in the literature.13,41

Figure 4.3 (a) Atomic percentage of Zn, O, Al and Si elements as a

function of depth from the surface as determined by depth-profiling XPS.

The schematic representation above the graph shows the structure of the AZO film. The ALD recipe used for this sample was similar to the one for sample 03 listed in Table 4.1, but here a ~1.5 nm native SiO2 / p-Si was

used as a substrate instead of 450 nm thermal oxide SiO2 / p-Si. (b) XPS spectrum of O 1s peak for sample 03 at a depth of 14 nm. The oxygen

peak was deconvoluted into two components with different binding energies. The component of 532.1 eV can be attributed to both O-Al and O-H bonds. The component at 530.8 eV can be attributed to Zn-O bonds.

The nanolaminate structure was observed in the AZO films with low AFXPS (samples 01-06), while for the AZO films with higher AFXPS (samples 07-11), the atomic distribution of Al, Zn and O could no longer be resolved with XPS. The resolution of XPS in the growth direction is limited to lr≈5-10 nm, so when the interspacing between adjacent AlOx layers l is less than lr (l<lr), the discrete nanolaminate structure can no longer be resolved by XPS.


Figure 4.4 Cross-sectional HAADF (a) and high-resolution TEM (b) images

of sample 03. The ALD recipe existed of 3 supercycles of AlOx/ZnO. The AZO film was ~40 nm thick (see Table 4.1).

The O peak in Fig. 4.3 (b) contains contributions of two components with different binding energies. The lower binding energy component of the O 1s peak at 530.8 eV can be attributed to O2- ions surrounded by Zn2+ ions indicating the Zn-O bonds25,44–48 and is denoted by OI in Fig. 4.3 (a). This O signal intensity shows the same variation as the Zn signal intensity, which confirms the chemical bonding state of O2- coordinated with Zn2+. The second O component at the binding energy of 532.1 eV denoted by OII in Fig. 4.3 (a), can be assigned mainly to hydroxyl groups (–OH)25,44–48 which spread throughout the AZO film. However, in Fig. 4.3 (a), the OII component shows the same variation with the Al content, meaning that part of this OII component at binding energy of 532.1 eV can be attributed to O2- ions coupled with Al3+. Such O-coupled Al3+ might be present as AlZn

+ in the ZnO phase, but also as AlOx clusters. In theory, it should be possible to distinguish these two Al3+


environments by the Al 2p peak fitting. Yet, in practice, the low Al doping level resulted in too low intensities for the Al 2p peak to investigate the chemical environment. The atomic percentage of the OII component reached a maximum at the interface. This part of the OII component was assigned to O in the ~1.5 nm thick native SiO2.

The binding energy of Zn 2p3/2 in AZO films with various AFXPS throughout the films was measured by depth-profiling XPS, and is presented in Fig 4.5. The binding energy at the surface (depth = 0 nm) deviated due to the surface contamination. The higher binding energy at the interface (depth = 35-40 nm) can be ascribed the different chemical bonding state of Zn on the SiO2 matrix. Excluding these two effects, the binding energies generally shifted to higher value with higher AFXPS. This shift was observed consistently for both the bulk of the AZO films (0 nm < depth < 40 nm) and for the surfaces (depth = 0 nm). Therefore, a possible sputtering effect during sample examination cannot be the reason for such a shift. Since the binding energies are referenced to the Fermi level (EF),

30,49,50 the increase can be attributed to an increase in EF.51,52

That is, the Al doping contributes free electrons to the AZO films, leading to a higher EF. Furthermore, for the samples with low AFXPS, the binding energy of Zn 2p3/2 oscillates throughout the films, with an amplitude of around 0.1 eV. Again, taking sample 03 as an example (AFXPS = 1.9%), it is clear that the binding energy reaches its local maximum where the AlOx layers are located. A possible mechanism that can be proposed to explain the oscillation is the following. The AlZn

+ in the ZnO lattice can serve as an effective positive charge and create a local static electric field.42 The electric field can affect the surrounding atoms. Core electrons from adjacent Zn atoms may shift towards the AlZn

+ centre by the electrostatic force, leading a higher atomic charge of the Zn. It was found that a higher atomic charge can cause a higher binding energy in the measurement of XPS.53–55 Therefore, the local maxima can be ascribed to the delocalization of core electrons of Zn towards AlZn

+. For samples with high AFXPS, such oscillations can no longer be observed since the interspacing between the adjacent AlOx layers is within the depth resolution of XPS. In summary, the increase of the free electron density can cause a shift in the binding energy globally which manifests as an increase in EF. Meanwhile, according to the mechanism proposed by us, the delocalization of Zn core electrons by AlZn

+ centres leads to a local increase of the binding energy.


Figure 4.5 XPS depth profiles for the Zn 2p3/2 signal of several AZO

samples.

The carrier density and optical mobility are presented in Fig. 4.6 (a). The carrier densities from Hall measurements and the optical modeling (using SE and FTIR data) are consistent with each other, which confirms the accuracy of our modeling and validates the value of the electron effective mass (m*=0.4me) that we used here. The average distance an electron travels while interacting with a photon, is much shorter than the average grain size. Hence, it can be assumed that for optical measurements, the grain boundary scattering can be neglected. Therefore, the mobility derived from modeling SE and FTIR measurements μopt, can be assumed to be equal to the intra-grain mobility, as μopt≈μintra-grain.56 The intra-grain mobility is considered to be determined by ionized and neutral impurity scattering, etc.

57 The intra-grain resistivity can be defined as:

1

intra-grain opt( )e nρ µ −= × × (Eq. 4.4)

where e, n and μopt are the elementary charge, carrier density and optical mobility, respectively. Both the optical determined intra-grain resistivity and the electrically determined effective resistivity (measured by FPP) are plotted in Fig. 4.6 (b). The difference between both curves was ascribed to scattering at grain boundaries. In the FPP series of the effective resistivity, the resistivity was improved from 9.8 mΩ∙cm for intrinsic ZnO to an optimum of 2.2 mΩ∙cm at AFXPS = 6.9%. The optically determined intra-grain resistivity also shows an optimum value at the same value of AFXPS. Thus, the AZO series can be classified into two regions: region I with sample 01-06, and region II with


sample 08-11, and with sample 07 just on the borderline, as denoted in Fig. 4.6 (a) and (b). In Fig. 4.6 (a), the carrier density increases significantly with the addition of Al doping in region I, since Al species in the form of AlZn

+ in the ZnO lattice release electrons and contribute a number of free electron carriers to the AZO films. However, in region II, the increase of the carrier density shows a soft saturation. This behavior can be explained by Lee’s model,42 as shown in Fig. 4.7. Since Al-doping mainly occurs at the interface of the ZnO and AlOx layers, AlZn

+ creates an effective electric field at the position of the AlOx layers. At high values of AFXPS, more AlOx layers are deposited within the 40 nm AZO films such that the interspacing of adjacent AlOx layers becomes smaller. When the interspacing is larger than a critical value (l > lc), as shown in Fig. 4.7 (a), the effective electric fields from adjacent AlOx layers do not overlap each other, and Al doping is relatively efficient. Therefore, the carrier density shows a significant increase with higher Al doping at region I. When l < lc, as shown in Fig. 4.7 (c), the effective electric fields overlap each other. According to Lee’s model, this overlap can inhibit further Al-doping by the repulsion between adjacent electrons or charged donors.42 Therefore, in region II, the addition of Al atoms no longer contributes to the carrier density effectively. On the basis of the different trends of carrier density between region I and II, the critical interspacing can be defined as the value of sample 07 (lc =3.8 nm), as shown in Fig.4.7 (b). Moreover, the optical mobility decreases gradually with additional doping in the entire Al-doping range, as shown in Fig. 4.6 (a). As discussed before, effective AlZn

+ dopants act as ionized impurities, while the rest of the Al atoms in ZnO lattice form neutral impurities.58 Both types of Al species are point defects and will lead to the scattering of free carriers during the electrical transport within ZnO grains.57


Figure 4.6 (a) Carrier density and optical mobility as a function of AFXPS, as extracted from SE and FTIR data. (b) Intra-grain resistivity (SE resistivity)

and effective resistivity (FPP resistivity) as a function of AFXPS.

Figure 4.7 Schematic representations of the effective electric field of AlZn

+:

(a) in region I, AlZn+ donates electrons effectively when l > lc; (b) transition

region between region I and II, lc ≈ 3.8 nm, corresponding to sample 07; (c) in region II, the effective electric field prohibits further doping when l < lc.

lc is the critical interspacing between AlOx layers in the film growth direction.

Next, in order to study the role of the Al species at the grain boundaries, the effective mobility and the grain boundary mobility can be calculated from Matthiessen’s rule, by

1

eff eff( )e nρ µ −= × × (Eq. 4.5)

1 1 1

eff intra-grain GBµ µ µ− − −= + (Eq. 4.6)


where ρeff is the effective resistivity measured by FPP, e is the elementary charge, n is the carrier density derived from optical analysis, and μeff, μintra-grain and μGB are the effective, intra-grain mobility and grain boundary mobility, respectively. As plotted in Fig. 4.8 (a), with increasing AFXPS, μGB increases slightly in region I while decreasing strongly in region II. Such a phenomenon can be explained according to the band diagram at grain boundaries, as presented in Fig. 4.8 (b).57 In AZO films, defects at grain boundaries are charged by electrons, leading to trapping states and barriers at grain boundaries. The electron transport through grain boundaries can then be described by the classical thermionic emission and quantum-mechanical tunneling.57 In region I, the increased carrier density will lead to a shift of the Fermi level to a higher energy level. Therefore, the effective barrier height and width become smaller, and consequently, the grain boundary mobility will increase at higher AFXPS. In region II, the interspacing between the adjacent AlOx layers becomes smaller than the critical value (l < lc), and the scattering at the interface of the ZnO layers and AlOx layers becomes dominant. Such a scattering limits the mean free path of free electrons during the electrical transport. Meanwhile, as discussed before, AlOx interrupts the nucleation and growth of ZnO grains during deposition. At high AFXPS, the closely spaced AlOx layers result in smaller ZnO grain sizes, and more grain boundaries between ZnO gains. As a result, the grain boundary mobility decreases significantly at higher AFXPS in region II.

Figure 4.8 (a) Effective mobility μeff, intra-grain mobility μintra-grain and

grain boundary mobility, μGB, as a function of the aluminum fraction AFXPS,

(b) Schematic representation of the energy diagram at grain boundaries.

The Al doping efficiency η of Al is the fraction of Al atoms which contribute to the carrier density by the following mechanism:


Zn ZnAl Al e+ −→ +

The doping efficiency can be calculated by the following equation:

0 100%Zn XPS

n n

N AFη

−= ×

× (Eq. 4.7)

In Eq. 4.7, n and n0 are the carrier density of AZO and intrinsic ZnO, respectively. AFXPS is the aluminum fraction as measured by XPS, and NZn is the atomic density of Zn. Therefore, the product of NZn and AFXPS yields the atomic density of Al. By Rutherford backscattering spectrometry (RBS), NZn for intrinsic ZnO was measured to be 4.0×1022 cm-3. The physical meaning of the doping efficiency is the percentage of Al atoms which effectively donate free electrons to the AZO films. The calculated result is plotted in Fig. 4.9. As can be seen, η is less than 10% for the entire AZO series. As estimated by RBS, the average distance between adjacent Al atoms within the same AlOx layer was around 0.5-1 nm. As illustrated in Fig. 4.7 (a), the overlapping of the effective electric field in the same AlOx layer inhibits the release of free electrons. Therefore, the amount of effective Al donors was limited (η < 10%). Moreover, in region I, η was relatively constant, varying from 7% to 10%, implying that the effective electric field from AlOx layers in the growth direction does not inhibit the release of free electrons from AlZn

+ when the interspacing l is large enough. In region II, as explained before, η decreases monotonically with increasing AFXPS due to the overlapping effective electric field from the AlOx layers.

Figure 4.9 Al doping efficiency as a function of aluminum fraction (AFXPS).

The dashed curve is a guide to the eye.


As is common for AZO thin films, the optical band gaps of the films were derived from the so-called Tauc plots, as depicted in Fig. 4.10 (a) which shows (ε2E

2)

2 as a function of photon energy for the AZO film series. ε2 is the imaginary part of the dielectric function ε2(ω), which was extracted from the SE data in the range of 1.2 – 6.5 eV. E is the photon energy. Since ZnO has a direct band-gap, the optical band gap or Tauc gap (Eg) is defined as the photon energy, where the extrapolation of the linear part of (ε2E

2)

2 vs. E intersects the horizontal axis.59 The resulting Eg values for each AFXPS are listed in the legend of Fig. 4.10 (a). Note that the value of ε2 can also be affected by exciton absorption in ZnO60, which might slightly affect the resulting Eg values.

Figure 4.10 (a) Tauc plots of (ԑ2E2)2 vs. photon energy for AZO films different aluminum

fractions (AFXPS) to extract the optical band gap values (Eg); (b) Comparison of the shifts of

optical band gap (Eg), the Fermi level (EF-EC), binding energy of Zn 2p3/2 as a function of AFXPS.

Note that the absolute values of these three parameters are different while the scale intervals

are equal (0.1 eV per division).

The increase of Eg for higher AFXPS values as observed in Fig. 4.10 (a) is mainly due to the shift of the Fermi level, according to the Burstein-Moss effect.61 For a more detailed comparison, the shift of the Fermi level was evaluated on the basis of the carrier density. Given that i-ZnO and AZO are degenerate semi-conductors, the following equations apply:

( ) ( , )C

CE

n D E f E T dE∞

= ∫ (Eq. 4.8)

3/2

2 3

(2 *)( )

2C C

mD E E E

π= −

(Eq. 4.9)

1( , )

exp[( ) / 1]F B

f E TE E k T

≈− +

(Eq. 4.10)


where n is the electron density, and DC(E) and f(E,T) stand for the density of states in the conduction band and the Fermi distribution function at an energy E, respectively. T is the temperature (300 K in the present case), Ec is the energy level of the conduction band tail, m* is the effective electron mass which is assumed to be 0.40me, ħ and kB are the Dirac constant and Boltzmann constant, respectively. From Eqs. 4.8 to 4.10, the relationship between the carrier density and the height of the Fermi level can be obtained. Hence, the position of the Fermi level related to the conduction band tail (EF-Ec) can be calculated as a function of the carrier density. The resulting plot is presented in Fig. 4.10 (b).

The optical band gap (Eg), and the Fermi level (EF-EC), and the binding energy of Zn 2p3/2 (averaged over the film thickness, see in Fig. 4.5) are compared in Fig. 4.10 (b). All three curves show a monotonic increase with increasing AFXPS. As explained before, the shifts of Eg and the binding energy were both attributed to the increase of the Fermi level. Consequently in principle, all of these three parameters should shift with the same trend and within the same order of magnitude with the carrier density. As presented in Fig. 4.10 (b), the difference between i-ZnO and the highest doped ZnO (AFXPS = 13.6%) was 0.48 eV for Eg, 0.33 eV for EF, and 0.44 eV for the binding energy. This supports therefore the explanation that the shift of EF with carrier density is the main reason for the increase of Eg and the binding energy. However, the shift of Eg (ΔEg = 0.48 eV) is larger than that of EF (ΔEF = 0.33 eV). The difference can be qualitatively explained by the existence of the AlOx layers within the ZnO. Since amorphous Al2O3 films have a larger Eg (Eg=6-7 eV (ref. 62)) than ZnO, the AlOx layers in the AZO films might cause an additional increase of the Eg of the entire film. The binding energy also has a larger shift (ΔBE = 0.44 eV) than EF. As discussed before, the delocalization of the core electrons from Zn may contribute a higher binding energy of Zn 2p3/2, in the order of 0.1 eV.

4.4 Conclusion

In this work we have studied the structural, electrical and optical properties of Al-doped ZnO (AZO) films with various doping levels as deposited by ALD. The resistivity of the films improved from 9.6 mΩ∙cm for intrinsic ZnO to an optimum of 2.4 mΩ∙cm for Al-doped ZnO with an Al fraction of 6.9%. By depth-profiling XPS, a nanolaminate structure of AlOx/ZnO layers could be resolved

4.5 Acknowledgements 79

for Al doping levels up to 5.9%. A nanolaminate structure was also observed by cross-sectional TEM. The AlOx layers cause interface scattering during the electrical transport in ZnO layers. This effect is the main reason for the reduced mobility at grain boundaries at high doping level. At high doping levels the carrier density also shows a soft saturation. We postulate that the effective electric field generated by AlZn

+ centres limits the doping efficiency when the interspacing of the adjacent AlOx layers becomes smaller at higher doping levels. A blue shift of the optical band-gap (ΔEg = 0.48 eV) was observed, and shown to be consistent with the shifts of both the Fermi level and the binding energy of Zn 2p3/2 photoelectrons.

4.5 Acknowledgements

The authors thank Dr. M. Blauw (Holst Centre) for his help in the Hall measurements. W.Keuning and Dr. T. Fernández Landaluce are acknowledged for doing the XPS measurements and related analysis. S. Smit is thanked for his contribution to the theoretical calculation of the Fermi level. The financial support by IMEC-NL within the Holst Centre in Eindhoven, The Netherlands, is gratefully acknowledged.

4.6 References

1 W.J. Maeng, S.-J. Kim, J.-S. Park, K.-B. Chung, and H. Kim, Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures 30, 031210 (2012).

2 M. Kudo, T. Kosaka, Y. Takahashi, H. Kokusen, N. Sotani, and S. Hasegawa, Sensors and Actuators B: Chemical 23, 173 (1995).

3 D. Barreca, D. Bekermann, E. Comini, A. Devi, R.A. Fischer, A. Gasparotto, C. Maccato, G. Sberveglieri, and E. Tondello, Sensors and Actuators B: Chemical 149, 1 (2010).

4 J. Müller and S. Weissenrieder, Fresenius’ Journal of Analytical Chemistry 349, 380 (1994).

5 S.-J. Chang, T.-J. Hsueh, I.-C. Chen, and B.-R. Huang, Nanotechnology 19, 175502 (2008).


6 S. O’Brien, M.G. Nolan, M. Çopuroglu, J.A. Hamilton, I. Povey, L. Pereira, R. Martins, E. Fortunato, and M. Pemble, Thin Solid Films 518, 4515 (2010).

7 A.E. Delahoy and S. Guo, Handbook of Photovoltaic Science and Engineering, Second Edi (John Wiley & Sons, Ltd, New York, 2011), pp. 716–796.

8 F. Ruske, M. Roczen, K. Lee, M. Wimmer, S. Gall, J. Hüpkes, D. Hrunski, and B. Rech, Journal of Applied Physics 107, 013708 (2010).

9 E. Fortunato, D. Ginley, H. Hosono, and D.C. Paine, MRS Bulletin 32, 242 (2007).

10 J.N. Duenow, T.A. Gessert, D.M. Wood, T.M. Barnes, M. Young, B. To, and T.J. Coutts, Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 25, 955 (2007).

11 J. A. van Delft, D. Garcia-Alonso, and W.M.M. Kessels, Semiconductor Science and Technology 27, 074002 (2012).

12 A. Favier, D. Muñoz, S. Martín de Nicolás, and P.-J. Ribeyron, Solar Energy Materials and Solar Cells 95, 1057 (2011).

13 P. Banerjee, W.-J. Lee, K.-R. Bae, S.B. Lee, and G.W. Rubloff, Journal of Applied Physics 108, 043504 (2010).

14 J.Y. Kim, Y.-J. Choi, H.-H. Park, S. Golledge, and D.C. Johnson, Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 28, 1111 (2010).

15 J. Elam and S. George, Chemistry of Materials 15, 1020 (2003).

16 K.T.R. Reddy, T.B.S. Reddy, I. Forbes, and R.W. Miles, Surface and Coatings Technology 151-152, 110 (2002).

17 J.F. Chang, H.H. Kuo, I.C. Leu, and M.H. Hon, Sensors and Actuators B: Chemical 84, 258 (2002).

18 M. Franke, T. Koplin, and U. Simon, Small 2, 36 (2006).

19 O. Kluth, G. Schöpe, B. Rech, R. Menner, M. Oertel, K. Orgassa, and H. Werner Schock, Thin Solid Films 502, 311 (2006).

20 F.K. Shan, G.X. Liu, W.J. Lee, and B.C. Shin, Journal of Applied Physics 101, 053106 (2007).

21 H. Zhu, H. Jia, D. Liu, Y. Feng, L. Zhang, B. Lai, T. He, Y. Ma, Y. Wang, J. Yin, Y. Huang, and Y. Mai, Applied Surface Science 258, 6018 (2012).

22 Y.-C. Cheng, Applied Surface Science 258, 604 (2011).

4.6 References 81 23 S.J. Kwon, Japanese Journal of Applied Physics 44, 1062 (2005).

24 J.S. Jur and G.N. Parsons, ACS Applied Materials & Interfaces 3, 299 (2011).

25 S. Kwon, S. Bang, S. Lee, S. Jeon, W. Jeong, H. Kim, S.C. Gong, H.J. Chang, H. Park, and H. Jeon, Semiconductor Science and Technology 24, 035015 (2009).

26 H.B. Profijt, S.E. Potts, M.C.M. van de Sanden, and W.M.M. Kessels, Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 29, 050801 (2011).

27 C.H. Ahn, H. Kim, and H.K. Cho, Thin Solid Films 519, 747 (2010).

28 V. Lujala, J. Skarp, M. Tammenmaa, and T. Suntola, Applied Surface Science 82-83, 34 (1994).

29 H. Saarenpää, T. Niemi, A. Tukiainen, H. Lemmetyinen, and N. Tkachenko, Solar Energy Materials and Solar Cells 94, 1379 (2010).

30 J.F. Moulder, W.F. Stickle, P.E. Sobol, and K.D. Bomben, Handbook of X-ray

Photoelectron Spectroscopy: a Reference Book of Standard Spectra for

Identification and Interpretation of XPS Data (Physical Electronics Inc., Eden Prairie, 2005).

31 E. Langereis, S.B.S. Heil, H.C.M. Knoops, W. Keuning, M.C.M. van de Sanden, and W.M.M. Kessels, Journal of Physics D: Applied Physics 42, 073001 (2009).

32 T. Tiwald, “PSEMI” Oscillator Model (Woollam Co. News, 2006), pp. 6–7.

33 N. Ehrmann and R. Reineke-Koch, Thin Solid Films 519, 1475 (2010).

34 A. Pflug, V. Sittinger, F. Ruske, B. Szyszka, and G. Dittmar, Thin Solid Films 455-456, 201 (2004).

35 H.C.M. Knoops, et al., to be published.

36 S. Kasap and P. Capper, Springer Handbook of Electronic and Photonic

Materials, First edition (Springer, New York, 2006), pp. 30–31.

37 S.K. Kim, C.S. Hwang, S.-H.K. Park, and S.J. Yun, Thin Solid Films 478, 103 (2005).

38 D. Kim, H. Kang, J.-M. Kim, and H. Kim, Applied Surface Science 257, 3776 (2011).

39 G. Luka, T. Krajewski, L. Wachnicki, B. Witkowski, E. Lusakowska, W. Paszkowicz, E. Guziewicz, and M. Godlewski, Physica Status Solidi (a) 207, 1568 (2010).


40 A. Yamada, B. Sang, and M. Konagai, Applied Surface Science 112, 216 (1997).

41 J.W. Elam, D. Routkevitch, and S.M. George, Journal of the Electrochemical Society 150, 339 (2003).

42 D.-J. Lee, H.-M. Kim, J.-Y. Kwon, H. Choi, S.-H. Kim, and K.-B. Kim, Advanced Functional Materials 21, 448 (2011).

43 J.-S. Na, G. Scarel, and G.N. Parsons, Journal of Physical Chemistry C 114, 383 (2010).

44 H. Tong, Z. Deng, Z. Liu, C. Huang, J. Huang, H. Lan, C. Wang, and Y. Cao, Applied Surface Science 257, 4906 (2011).

45 M. Gao, X. Wu, J. Liu, and W. Liu, Applied Surface Science 257, 6919 (2011).

46 X.-J. Yang, X.-Y. Miao, X.-L. Xu, C.-M. Xu, J. Xu, and H.-T. Liu, Optical Materials 27, 1602 (2005).

47 M. Chen, X. Wang, Y.H. Yu, Z.L. Pei, X.D. Bai, C. Sun, R.F. Huang, and L.S. Wen, Applied Surface Science 158, 134 (2000).

48 B.-Y. Oh, M.-C. Jeong, and J.-M. Myoung, Applied Surface Science 253, 7157 (2007).

49 R.W. Grant, J.R. Waldrop, S.P. Kowalczyk, and E.A. Kraut, Journal of Vacuum Science & Technology 19, 477 (1981).

50 E.A. Kraut, R.W. Grant, J.R. Waldrop, and S.P. Kowalczyk, Physical Review Letters 44, 1620 (1980).

51 T. Nagata, O. Bierwagen, M.E. White, M.Y. Tsai, Y. Yamashita, H. Yoshikawa, N. Ohashi, K. Kobayashi, T. Chikyow, and J.S. Speck, Applied Physics Letters 98, 232107 (2011).

52 J. Szuber, E. Bergignat, G. Hollinger, A. Polakowska, and P. Koscielniak, Vacuum 67, 53 (2002).

53 V.I. Nefedov, V.G. Yarzhemsky, A.V. Chuvaev, and E.M. Trishkina, Journal of Electron Spectroscopy and Related Phenomena 46, 381 (1988).

54 J. Blomquist, Journal of Electron Spectroscopy and Related Phenomena 36, 69 (1985).

55 L. Meda, C. Nicastro, F. Conte, and G.F. Cerofolini, Surface and Interface Analysis 29, 851 (2000).

4.6 References 83 56 T. Yamada, H. Makino, N. Yamamoto, and T. Yamamoto, Journal of Applied Physics 107, 123534 (2010).

57 K. Ellmer, A. Klein, and B. Rech, Transparent Conductive Zinc Oxide (Springer, New York, 2008).

58 K. Ellmer and R. Mientus, Thin Solid Films 516, 4620 (2008).

59 J. Tauc, R. Grigorovici, and A. Vancu, Physica Status Solidi (B) 15, 627 (1966).

60 H. Fujiwara and M. Kondo, Physical Review B 71, 075109 (2005).

61 B. Elias, Physical Review 93, 632 (1954).

62 G. Dingemans and W.M.M. Kessels, Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 30, 040802 (2012).

Chapter 5

Enhanced doping efficiency of Al-doped ZnO by atomic layer deposition using

dimethylaluminum isopropoxide as an alternative aluminum precursor *

Abstract

Atomic layer deposition offers the unique opportunity to control, at the atomic level, the 3D distribution of dopants in highly uniform and conformal thin films. Here, it is demonstrated that the maximum doping efficiency of Al in ZnO can be improved from ~10% to almost 60% using dimethylaluminum isopropoxide (DMAI, Al(CH3)2(OiPr)) as an alternative Al precursor instead of the conventionally-used trimethylaluminum (TMA, Al(CH3)3). Due to the steric hindrance of the isopropoxyl ligand of the precursor, the Al atoms can be deposited more widely dispersed, which enables higher active-dopant densities and hence a higher conductivity of the Al-doped films.

*Published as: Y. Wu, S.E. Potts, P.M. Hermkens, H.C.M. Knoops, F. Roozeboom, W.M.M.

Kessels, Chem. Mater. 25 4619 (2013).

86 Enhanced Al doping efficiency in ZnO

5.1 Introduction

Intrinsic and doped ZnO thin films have a growing number of prominent applications in electronic devices. In solar cells and displays they are used as transparent conducting oxides (TCO),1,2 in transparent thin-film transistors for displays they are used as semiconducting layers,3 and in low power gas sensors they serve as gas sensitive layers.4,5 One of the commonly introduced dopants is Al which is used improve the conductivity of ZnO films.6

Atomic layer deposition (ALD) is a deposition technique which allows for the preparation of high quality, highly conformal and uniform thin films with precise growth control.7,8 Due to the cyclic nature of the technique, ALD is also particularly well suited to deposit doped ZnO films when employing supercycles in which ALD cycles with Zn and dopant precursors are alternated. In principle, under carefully chosen conditions, ALD would allow for the precise, atomic scale control over the concentration, position and spacing of the dopant in the ZnO lattice. However, when the Al-doped films are deposited from the commonly applied precursors diethylzinc [DEZ, Zn(C2H5)2] and trimethylaluminum [TMA, Al(CH3)3], in combination with water, typically a nanolaminate structure is obtained. This means that ZnO layers are alternated by layers of AlOx clusters9 that have a relatively large areal density of Al atoms. The latter leads to overlapping effective electric fields from adjacent Al dopant atoms10 while the AlOx clusters also act as scattering centres for electrons limiting their mobility and, consequently, the film conductivity.11 In order to alleviate this problem, efforts have been undertaken to improve the control of the spatial distribution of Al atoms. For example, Yanguas-Gil et al. dosed organic molecules prior to dosing the TMA to reduce the saturation growth rate of the Al.12 However, this method introduces additional steps to the ALD cycles while it also did not lead to a sufficient improvement in conductivity of the films. In this work, we present another, very straightforward approach to increase the conductivity of Al-doped ZnO films. We demonstrate that an alternative Al precursor, with bulkier ligands leading to steric hindrance, affords control of the spatial distribution of Al atoms, thereby increasing the doping efficiency of the Al atoms.

When ALD cycles of ZnO are alternated by a cycle of Al2O3, the number of the Al precursor molecules that adsorb and react on the ZnO surface determines

5.2 Experimental Section 87

the density of the Al atoms in the Al-doped ZnO. This implies that if a larger Al precursor is employed than TMA, fewer Al atoms would be incorporated due to steric hindrance caused by the precursor ligands13. With the Al atoms wider spaced laterally, the formation of AlOx clusters will be less pronounced and this reduces the overlap of the effective electric field from adjacent Al atoms as well as the scattering of electrons by AlOx clusters. This should increase the doping efficiency and the electron mobility of the films. To test this hypothesis, we used dimethylaluminum isopropoxide [DMAI, Al(CH3)2(OiPr)] as an alternative Al precursor to the conventional TMA. Each DMAI monomer contains one isopropoxyl ligand (OiPr), which is estimated to be 4.9 Å in size and therefore 1.9 times bulkier than a methyl (CH3) ligand (2.6 Å in size)14,15. Consequently, during the Al2O3 cycles, each adsorbed -Al(CH3)(OiPr) surface group16 can shield a greater surface area than a corresponding -Al(CH3)2 surface group, thereby sterically hindering further reactions between the adjacent –OH surface groups and incoming DMAI molecules. Such a more pronounced influence of steric hindrance for DMAI when compared to TMA is also indicated by a reduced growth per cycle (in terms Al atoms deposited per cycle) obtained for Al2O3 films prepared with DMAI.16

5.2 Experimental Section

Al-doped ZnO films were deposited using an Oxford Instruments OpALTM ALD reactor at a substrate temperature of 250 °C. Si wafers with thermally-grown SiO2 layers with a thickness of 450 nm were used at substrates. DEZ, DMAI, TMA and deionized water (DI H2O) vapor were used as precursors.16 Films with a thickness of 40±3 nm were deposited, as monitored by in situ spectroscopic ellipsometry, and their doping level was varied by employing different ratios between the ZnO and Al2O3 cycles.11 The doping levels were expressed by the aluminum fraction (AF),

.%

100% .% .%

Al atAF

Al at Zn at= ×

+ (Eq. 5.1)

which was calculated from the atomic fractions of Al (Al at.%) and Zn (Zn at.%) as measured by depth-profiling X-ray photoelectron spectroscopy (XPS).11 The areal density of the Al atoms was obtained from Rutherford backscattering spectrometry (RBS) experiments. The resistivity ρ and carrier density n of the


films were obtained by four-point probe and optical measurements, respectively. The electron mobility μ was obtained by the expression11,17

1( )e nρ µ −= × × (Eq. 5.2)

This procedure for extracting μ has been validated by a comparison with data from Hall measurements.11


First, the values of the growth per cycle (GPC) for the ZnO and Al2O3 cycles were investigated by in situ spectroscopic ellipsometry (SE). In supercycles, these values can differ from the values obtained for pure ZnO and Al2O3 films. As shown in Fig. 5.1, a nucleation delay was observed for ZnO after one cycle of Al2O3. Both for DMAI and TMA, about 4 cycles of ZnO were required after an Al2O3 cycle. This effect can possibly be related to the Brønsted acidity of ZnOH* and AlOH* (asterisk denotes surface-bound species).9,18 During the transition between Al2O3 and ZnO cycles, ZnOH* and AlOH* coexist on the surface. Since ZnOH* is relatively basic compared to AlOH*, a proton-transfer surface reaction can occur:

* * *

2ZnOH + AlOH ZnOH AlO+ −→

The formation of the ZnOH2+⋯AlO-* complex consumes surface hydroxyl

groups, leaving fewer reactive sites on surface for the growth of ZnO in the subsequent cycles. Furthermore, besides the nucleation delay for ZnO, it was also observed that the GPC of the Al2O3 cycle was smaller than the value for pure Al2O3 films in the case that TMA was employed as precursor. This effect was reported before9,18 and was attributed to the etching of surface Zn atoms via ligand exchange during the Al2O3 cycles:

* *

3 3 3 3 2ZnOH Al(CH ) Al(OH)(CH ) + Zn(CH )+ →

Such “etching” was not observed in the DMAI case, where the GPC of the Al2O3 cycle was in good agreement with the value of the pure Al2O3 films, as indicated by the dashed line in Fig. 5.1 (a).


Figure. 5.1 Growth per cycle of consecutive ZnO and Al2O3 ALD cycles

using (a) DMAI and (b) TMA as the Al precursor and as measured by in

situ spectroscopic ellipsometry. Dashed lines indicate the growth per cycle of pure ZnO and Al2O3 films.

The resistivity of the Al-doped ZnO films is shown in Fig. 5.2 (a), which shows two regions for both DMAI and TMA separated by a minimum resistivity obtained at a certain AF value. In the case of DMAI, a lower resistivity (1.1 mΩ∙cm) was achieved than for TMA (2.4 mΩ∙cm) at a lower AF (4.6% for DMAI and 6.9% for TMA). Figure 5.2 (b) reveals the carrier density and mobility of the Al-doped ZnO films. At low AF values (region I), the carrier density increases significantly with increasing AF as a result of the effective Al doping. In region II, at higher AF values, the carrier density saturates. For the case of DMAI, this happens at a value of ~1.0×1021 cm-3, which is significantly higher than the maximum value of 3.7×1020 cm-3 for the TMA case.11 The mobility shows a decrease with AF over that largest part of the AF range.

The improvement in the conductivity of the films prepared with DMAI is particularly prominent at lower Al doping levels. This indicates a large difference in doping efficiency, i.e. the percentage of Al atoms which effectively donate a free electron to the ZnO films. With one Al atom donating at maximum one free electron, this doping efficiency η can be calculated from the data by


0atomic density of active dopants100% 100%

atomic density of AlZn

n n

N AFη

−= × = ×

×

(Eq. 5.5)

where n and n0 are the carrier densities of Al-doped and intrinsic ZnO, respectively. NZn is the atomic density of Zn as measured by RBS, which was 4.0×1022 cm-3. n-n0 is the increase of the carrier density due to active Al dopants or, equivalently, it is the active dopant density. As shown in Fig. 5.2 (c), the films prepared with DMAI show a significantly higher doping efficiency (up to 60%) than the films prepared with TMA (~10%).

Figure 5.2 (a) Resistivity, (b) carrier density and mobility, and (c) doping

efficiency of Al-doped ZnO films prepared with DMAI and TMA as Al precursors as a function of Al fraction (AF). In both cases two regions can


be distinguished (as indicated by dashed and dotted lines), a region in

which the resistivity decreases and a region in which the resistivity increases.

The higher doping efficiency for the case of DMAI explains the fact that the carrier density increases faster with increasing AF than for the case of TMA. It also implies that a larger active dopant density is obtained at a certain AF value when comparing the DMAI with the TMA case. In turn, a larger percentage of active dopants means also a smaller percentage of Al atoms that are ineffective, such as Al atoms located within ZnO grains in the form of neutral impurities6,19 or at grain-boundaries in form of AlOx clusters11. Such a smaller percentage of ineffective Al atoms has an effect on the mobility of the films as shown in Fig. 5.3 in which the mobility has been plotted as a function of the active dopant density. In region I, the mobility decreases with increasing active dopant density, mainly due to ionized impurity scattering by the ionized Al dopants.19 This effect is particularly pronounced for the DMAI case, where the mobility decreases monotonically for active dopant densities in the range (1-9)×1020 cm-3 as a result from the high carrier density. In this region the mobility follows a similar trend as reported by Ellmer et al. on the basis of the work by Masetti et al.19,20 However, in region II, the mobility drops much faster with the active dopant density, especially for the case of TMA. This can be understood from the relatively large density of ineffective Al atoms that are present in the films in these cases. These ineffective Al atoms act as additional scattering centers for electrons reducing the mobility of the films significantly. Figure 5.3 indicates therefore clearly the impact of the higher doping efficiency for the case of DMAI. The higher doping efficiency implies that it is possible to increase the density of Al atoms in the ZnO leading to a higher active dopant density without the adverse effect of significantly increasing the density of ineffective Al atoms which drastically reduces the mobility. In this way, much larger active dopant densities at reasonable mobility levels can be achieved with DMAI than with TMA.


Figure 5.3 Mobility of Al-doped ZnO films as a function of active dopant

density. The films have been prepared with DMAI and TMA as Al

precursors. The regions as indicated in Fig. 5.2 are shown as well.

The results presented in Fig. 5.2 confirm that the doping efficiency and hence the conductivity of Al-doped ZnO can be improved by using DMAI as alternative Al precursor. To verify the hypothesis that this is related to more steric hindrance for DMAI than for TMA (both precursor differ only by one ligand), the areal densities of Al atoms deposited in one Al2O3 cycle were determined by RBS for films of 40 nm thick (the number of Al2O3 cycles was kept the same in both cases such that the cycle ratio ZnO:Al2O3 had to be adjusted). As shown in Table 5.1, the number of Al atoms deposited per cm-2 in one Al2O3 cycle in the DMAI case was 3.8 times lower than that in the TMA case. This is in line with the aforementioned fact that the OiPr ligand is considerably larger than the CH3 ligand. This corroborates that the Al atoms deposited were more widely dispersed for ZnO films prepared by DMAI explaining the higher doping efficiency obtained by DMAI.

Table 5.1 Conditions and results from Rutherford backscattering spectrometry (RBS) for Al-doped ZnO films prepared by DMAI or TMA as

an Al precursor.

DMAI TMA Number of Al2O3 cycles 19 19 Cycle ratio ZnO:Al2O3 14:1 12:1

Al fraction by RBS: AFRBS 2.9% 13.4%

Al atoms deposited per Al2O3 cycle

(0.29±0.01)×1015 cm-2 (1.10±0.04)×1015 cm-2

5.4 Conclusions 93

5.4 Conclusions

In conclusion, it has been demonstrated that ALD is well suited to prepare Al-doped ZnO films with low resistivity values by controlling the spatial distribution of the Al dopant atoms through careful selection of the Al precursor. When using DMAI instead of TMA, the bulky isopropoxyl ligand leads to more widely dispersed Al atoms in the ZnO due to the self-limiting nature of the ALD surface reactions and the effect of steric hindrance. This significantly improves the doping efficiency of the Al for the case of DMAI compared to TMA and this, in turn, means that much higher active dopant densities can be obtained while reducing the adverse influence of ineffective Al atoms that act as scattering centers for electrons and hence reduce the mobility of the Al-doped films. Being very straightforward, it is expected that this approach can also be used to prepare other doped films by ALD for which an atomic level control of the 3D distribution of the dopant atoms is vital.

5.5 Acknowledgement

The authors thank Holst Centre/IMEC-NL, The Netherlands for financially supporting this project. The work of S.E.P. is supported by NanoNextNL, a micro and nanotechnology programme of the Dutch ministry of economic affairs, agriculture and innovation (EL&I) and 130 partners. Air Liquide is acknowledged for providing the DMAI precursor.

5.6 References

1 J.A. van Delft, D. Garcia-Alonso, and W.M.M. Kessels, Semicond. Sci. Technol. 27, 074002 (2012).

2 N. Yamamoto, H. Makino, S. Osone, A. Ujihara, T. Ito, H. Hokari, T. Maruyama, and T. Yamamoto, Thin Solid Films 520, 4131 (2012).

3 H. Yabuta, M. Sano, K. Abe, T. Aiba, T. Den, H. Kumomi, K. Nomura, T. Kamiya, and H. Hosono, Appl. Phys. Lett. 89, 112123 (2006).

4 J.F. Chang, H.H. Kuo, I.C. Leu, and M.H. Hon, Sensors Actuators B Chem. 84, 258 (2002).

5 S. O’Brien, M.G. Nolan, M. Çopuroglu, J.A. Hamilton, I. Povey, L. Pereira, R.


Martins, E. Fortunato, and M. Pemble, Thin Solid Films 518, 4515 (2010).

6 K. Ellmer, A. Klein, and B. Rech, Transparent Conductive Zinc Oxide: Basics and

Applications in Thin Film Solar Cells (Springer, New York, 2008).


8 H.B. Profijt, S.E. Potts, M.C.M. van de Sanden, and W.M.M. Kessels, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 29, 050801 (2011).

9 J.-S. Na, G. Scarel, and G.N. Parsons, J. Phys. Chem. C 114, 383 (2010).

10 D.-J. Lee, H.-M. Kim, J.-Y. Kwon, H. Choi, S.-H. Kim, and K.-B. Kim, Adv. Funct. Mater. 21, 448 (2011).

11 Y. Wu, P.M. Hermkens, B.W.H. van de Loo, H.C.M. Knoops, S.E. Potts, M.A. Verheijen, F. Roozeboom, and W.M.M. Kessels, J. Appl. Phys. 114, 024308 (2013).

12 A. Yanguas-Gil, K.E. Peterson, and J.W. Elam, Chem. Mater. 23, 4295 (2011).

13 R.L. Puurunen, Chem. Vap. Depos. 9, 249 (2003).

14 G.S. McGrady, J.F.C. Turner, R.M. Ibberson, and M. Prager, Organometallics 19, 4398 (2000).

15 N.Y. Turova, V.A. Kozunov, A.I. Yanovskii, N.G. Bokii, Y.T. Struchkov, and B.L. Tarnopol’skii, J. Inorg. Nucl. Chem. 41, 5 (1979).

16 S.E. Potts, G. Dingemans, C. Lachaud, and W.M.M. Kessels, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 30, 021505 (2012).

17 H.C.M. Knoops, B.W.H. van de Loo, and S. Smit, J. Appl. Phys. (2013).

18 J. Elam and S. George, Chem. Mater. 15, 1020 (2003).


20 G. Masetti, M. Severi, and S. Solmi, IEEE Trans. Electron Devices ED30, 764 (1983)

Chapter 6

On the factors limiting the doping efficiency in atomic layer deposited ZnO:Al thin films: a

dopant distribution study by atom probe tomography*

Abstract

The maximum conductivity that can be obtained in Al doped ZnO thin films prepared by atomic layer deposition (ALD) is limited by the low doping efficiency of Al. To better understand about the origin of this limitation, the 3-dimensional distribution of Al atoms in the ZnO host material matrix has been examined at an atomic scale using a combination of high-resolution Transmission Electron Microscopy and Atom Probe Tomography. The Al distribution in ZnO films prepared in supercycle mode is often presented as atomically flat δ doping layers. In practice, a broadening of the AlOx layers is observed, with a full width half maximum of ~2 nm. This results in a distribution where the Al atoms can be spread throughout the entire film rather than being tightly confined to the nominal deposition layers. The low doping efficiency that is obtained when the local Al density is high (> ~1 nm-3) can be ascribed to the solubility limit of Al in ZnO as well as to the disorder-induced carrier localization.

* Y. Wu, A.D. Giddings, M.A. Verheijen, T.J. Prosa, D.J. Larson, F. Roozeboom, W.M.M.

Kessels (in preparation for publication).

98 On the factors limiting the Al doping efficiency in ZnO

6.1 Introduction

Transparent conducting oxide (TCO) thin films are a class of materials that are commonly used as transparent electrodes in a wide variety of commercial optoelectronic devices, such as in displays and solar cells.1,2 The most commonly used TCO is indium tin oxide (ITO). Because of the significantly lower cost of the constituent materials, ZnO has been reported to be a promising alternative TCO.3 One major requirement for a TCO to be considered as a commercial replacement of ITO is to achieve a film resistivity in the order of 1 mΩ∙cm or lower.4 However, the resistivity of intrinsic ZnO films cannot meet this requirement. Therefore, doping is often used to decrease the resistivity of ZnO films. The most common dopant for ZnO is Al, with the resultant material known as Al-doped ZnO (ZnO:Al or AZO).

There are several methods by which ZnO:Al thin films can be prepared, each of which influences the performance of the resulting material. For example, resistivity values in the order of 0.1 mΩ∙cm have been achieved in ZnO:Al thin films prepared by pulsed laser deposition (PLD) and magnetron sputtering deposition (MSD).4 One particular deposition technique, atomic layer deposition (ALD)5–8, provides a number of potential advantages compared to PLD and MSD which are particularly relevant for emerging, demanding applications: (i) the cyclic nature of ALD allows accurate control of film thickness even for ultrathin films; (ii) ALD offers excellent uniformity, necessary for large area applications; (iii) ALD is unique in its ability to cover challenging surface topologies with a very high degree of conformality; (iv) the absence of energetic species (ions, neutrals, photons) makes it a “mild” deposition technique, which does not cause damage to the underlying surface; (v) high temperatures are not required, making the technique suitable for situations where thermal budgets are restricted; and, finally, (vi) the atomic composition can, in principle, be accurately controlled in a highly repeatable manner by tuning the ALD cycle ratio between the host matrix and dopant materials.

Despite the advantages of growing AZO thin films by ALD, the decrease of the resistivity of ZnO films by Al doping is still limited. For example, in previous work,9 it was shown that by varying the Al doping level (defined as the proportion of Zn atoms replaced by Al; see below for formal definition of AF), the resistivity could be improved from 9.8 mΩ∙cm to 2.2 mΩ∙cm. However, this

6.1 Introduction 99

is still not sufficient to meet the required performance target of <1 mΩ∙cm. Furthermore, the doping efficiency, i.e. the percentage of Al atoms that act as active dopants releasing free electrons is generally lower than 10 %. The inactive Al atoms can act as scattering centers, rather than contributing to the number of free electrons. This will have a negative effect on the carrier mobility and hence on the electrical transport in the ZnO:Al films.

The limited improvement in conductivity and the low doping efficiency of ALD-ZnO:Al have been attributed as principally being due to the inhomogeneous distribution of the Al atoms in ZnO:Al films, originating from the so-called “supercycle” growth mode.7,9,10 As shown in Fig. 6.1 (a), one supercycle consists of m ZnO cycles and one Al2O3 cycle, where m is called the cycle ratio. By modifying this cycle ratio, the doping level can be tuned. The targeted film thickness can be achieved by repeating a certain number of these supercycles, as shown in Fig. 6.1 (b). Such a growth mode can lead to a lamellar distribution of Al atoms in the ZnO:Al films. To increase the Al doping level, the cycle ratio m is lowered, which will correspondingly decrease the space between the adjacent AlOx layers, l, as indicated in Fig. 6.1 (b). Our previous work revealed two distinct trends in ZnO:Al films grown as a function of cycle ratio m. Firstly, it was shown that the doping efficiency starts to decrease when l is below a critical value of lc = 3.8 nm, which corresponds to a doping level of 6.9%.9 Secondly, the resistivity starts to increase when l > lc. Such a correlation between the doping efficiency and the spacing l of adjacent AlOx layers has also been observed by Lee et al.,7 who reported that lc ranged from 2.3 to 2.6 nm for their films.


Figure 6.1 (a) Schematic representation of one ALD supercycle for the

preparation of ZnO:Al films. One supercycle consists of m cycles of ZnO and one cycle of Al2O3, where m is called the cycle ratio. One individual

ZnO or Al2O3 ALD cycle consists of four steps: precursor dosing/purge/co-

reactant dosing/purge. Diethylzinc (DEZ), trimethylaluminum (TMA) and deionized water vapor are used as Zn precursor, Al precursor and co-

reactant, respectively. (b) Schematic representation of the sequence of the ALD cycles in supercycle mode for the growth of ZnO:Al films. The total thickness of the film is determined by the number of supercycles.

Before presenting the following detailed studies on the correlation between doping efficiency and space, two remarks should be made. Firstly, the space only describes the dopant distribution in one dimension (i.e. growth direction). It should be realized that the areal density of Al atoms within each individual AlOx layer can also play a role. There have been two experimental studies that support this, showing that by increasing the lateral distance of Al atoms the doping efficiency and the film conductivity can be improved.10,11 In one of these studies,11 dimethylaluminum isopropoxide (DMAI, [Al(CH3)2(µ-OiPr)]2) was selected as an alternative Al precursor to trimethyl aluminum (TMA, [Al(CH3)3]2). Due to steric hindrance of the larger OiPr ligand, the number of Al atoms deposited per Al2O3 cycle was significantly reduced from 11.0 to 2.9 nm-

2. As a result, a higher doping efficiency of 50-60 % compared to <10 % for TMA and a lower resistivity of 1.1 mΩ∙cm compared to 2.4 mΩ∙cm for TMA were


achieved. Another additional approach to reduce the number of Al atoms incorporated in one ALD cycle is by deactivating the surface functional groups using ethanol as a reactant before the dosing step of TMA. This approach also improved the doping efficiency by a factor of ~2.10

Secondly, the conventional view of the ALD growth is that of a digital growth mode, as illustrated in Fig. 6.1 (b): Al atoms are concentrated only at distinct film depths thus resulting in a so-called δ-doping profile. Such a view is an oversimplification compared to the true distribution. For example, the actual vertical distribution of Al atoms needs to be taken into account. It is possible that intermixing of the Al and Zn atoms may occur at the interface between the AlOx layers and the ZnO matrix, resulting in a broadened distribution of the Al-containing layers. To verify such degree of intermixing, if any, a study of the spatial distribution of Al and Zn atoms at an atomic scale is required.

In this work, we investigate two research questions: (1) what is the actual distribution of Al atoms in ZnO:Al films and (2) how does the distribution affect the doping efficiency? To address these questions, we prepared ZnO:Al thin films by ALD using the conventional Al precursor TMA in the supercycle growth mode. The spatial distribution of the Al atoms was obtained at an atomic-resolution by Transmission Electron Microscopy (TEM) and Atom Probe Tomography (APT). Based on the obtained distribution, the local Al density, instead of the spacing between adjacent AlOx layers, l, is chosen as the parameter to describe the dopant distribution. When the local Al density is high, the doping efficiency is limited by the two proposed limiting factors: the solid solubility limit of Al atoms in a ZnO matrix and the disorder-induced carrier localization.


Intrinsic and Al-doped ZnO films were prepared using an Oxford Instruments OpAL ALD reactor at a substrate temperature of 250°C. Diethylzinc ()DEZ, trimethylaluminum (TMA) and deionized water vapor were used as precursors.9 From our previous work,9 we have identified the supercycle conditions to grow the three types of ZnO:Al films that can be representatively labelled as lowly-, optimally- and highly-doped films. Table 6.1 describes the growth parameters used to create the three conditions as well as the resulting


electrical properties. The number of supercycles, M, is the number of Al2O3 ALD cycles used in each film’s deposition. These values are 3 (lowly-doped), 11 (optimally-doped) and 19 (highly-doped), respectively, for the films denoted AZO-1 to AZO-3. Earlier, we reported that a minimum resistivity of 2.4 mΩ∙cm was obtained for the optimally doped case AZO-2. For comparison, the low doped in AZO-1 results in a lower carrier density, while the high doped film AZO-3 leads to lower carrier mobilities.

In Table 6.1 the doping level is presented by the aluminum fraction (AF), which represents the proportion of Zn atoms replaced by Al atoms in an ZnO:Al film. The value of AF is determined by:

Al at.%100%

Al at.% Zn at.%AF = ×

+ (Eq. 6.1)

The atomic percentages were measured either by X-ray photoelectron spectroscopy (XPS), denoted as AFglobal, or by APT, denoted as AFlocal. Note that, AFglobal represents the averaged Al doping level throughout an entire ZnO:Al film, while AFlocal represents an Al doping level of a selected nano-scale sub-volume of the AZO film, extracted from the APT data set.

Table 6.1 Properties of ZnO:Al films with film thicknesses of 40±2 nm as

studied by APT and TEM in this work.9 These films are selected as characteristic representatives of the growth conditions of ZnO:Al films

with lowly-doped (AZO-1), optimally-doped (AZO-2) and highly-doped (AZO-3) Al doping levels. The spacing l indicates the vertical distance between adjacent AlOx layers. The doping level in the ZnO:Al films is

characterized by the Al fraction, as measured by XPS.9 The electrical properties were measured for films deposited on p-Si wafers with 450 nm

thermal oxide.9

Layer

Growth parameters Doping level Electrical properties

Supercycles

M

Cycle ratio

m

Spac-ing

l (nm)

Al fraction AFglobal (%)

Resistivity ρ

(mΩ∙cm)

Carrier density

ne (1020 cm-3)

Mobility µ

(cm2/V∙s)

Doping efficiency

η (%)

AZO-1 3 85 13.7 1.9 5.1 1.1 10.8 7.1

AZO-2 11 23 3.8 6.9 2.4 3.0 8.7 8.7

AZO-3 19 12 2.0 16.4 4.9 3.7 3.5 4.7


In this work, three ZnO:Al films with a film thickness of 40±2 nm each were prepared in one deposition run under conditions identical to those used as the films listed in Table 6.1. For this experiment, the facilitate TEM and APT measurements, the films were grown sequentially in one film stack, and capped and separated by intrinsic ZnO films. A diagram of the stack is shown in Fig. 6.2 (a). This geometry allowed a single APT or TEM measurement to collect data on all three film types. The stack was grown in the same deposition run on both a Si substrate (with native oxide) for the TEM measurements and on a commercially available Si microtip array coupon with flat-top Si microtips (also with native oxide) for the APT measurements.

In addition to the three ZnO:Al films of interest, the complete stack also contained a ZnO capping film and ZnO buffer films between the ZnO:Al films. The top 100 nm ZnO cap film served two purposes: (i) during Focused Ion Beam (FIB) milling it was used as a sacrificial film to protect the ZnO:Al films from ion implantation which could damage the structure, and (ii) to have sufficient material for the apex of the APT specimens to reach an equilibrium shape during the initial stages of field evaporation, before the film of interest was reached. The two ZnO buffer films of 20 nm between the ZnO:Al films were intended: (i) to clearly separate ZnO:Al films for the TEM and APT imaging and data collection and (ii) to allow the APT specimen apex to return to an equilibrium shape between each ZnO:Al film during the APT measurement. This prevents changes of the specimen tip shape which may occur during the measurement of one film to influence the measured geometry of the next. The 60 nm i-ZnO film at the bottom was deposited to provide sufficient material so that the third ZnO:Al film could be completely evaporated before the field evaporated volume of the APT specimen reached the Si substrate.

The flat-top Si microtips upon which the deposition was made had an initial diameter of ~2 µm.12 After the ALD coating they were prepared using standard FIB annular milling protocols13,14 into APT specimen needles with an initial diameter ~50 nm and half shank angle ˜15°. By depositing directly onto the

microtip coupon the APT specimens could be prepared with reduced FIB requirements since no lamella lift-out was necessary. High resolution SEM images of the specimen needles were taken before APT measurement and, when possible, after the measurement to provide geometric information to assist with the data reconstruction. Deposition onto pre-sharpened microtips12


with an iniwal diameter ˜10 nm was also prototyped, but this proved inferior to using the flat-top type. The APT measurements were performed in CAMECA LEAP 4000X15 microscopes with the specimen at a temperature of 25 K, laser pulse energies between 0.3 pJ and 10 pJ, and laser pulse rates of 160 kHz (reflectron-type system) or 500 kHz (straight-flight path system). Cross-sectional TEM studies on FIB prepared lift-out samples were performed in high-angle annular dark-field (HAADF) STEM and in bright-field TEM modes using a probe-corrected TEM (JEOL JEM ARM 200F) equipped with a 100 mm2 Centurio SDD detector and a FEI Tecnai F30ST TEM.


6.3.1 Two-dimensional depth profiles

The morphology of the three ZnO:Al films has been imaged by both TEM and APT. The lamellar structures of the ZnO:Al films were imaged in HAADF-STEM mode as shown in Fig. 6.2 (b). The dark streaks represent the positions of the Al-rich layers, where Al-atoms were incorporated during single Al2O3 ALD cycles of the supercycle. Fig. 6.2 (c) shows the two-dimensional (2D) profiles of the Al distribution in the stack, extracted from the 3D APT data. The projected volume for the 2D APT profiles have a depth of 5 nm. The grey-scale shading in the 2D APT profiles represents the Al doping levels, and is quantified by AFlocal . Consistent with the TEM images, the grey streaks in Fig. 6.2 (c) indicate the Al-rich layers. In both cases, rather than being flat, the Al-rich layers are observed to be wrinkled. This reflects the faceting of the ZnO:Al polycrystalline9 grains obtained for the ALD-prepared films.

The equivalent estimated local atomic density of Al (nAl) is also indicated in Fig. 6.2 (c). This parameter is derived from AFlocal by:

Al -Zni APTn n AF= × (Eq. 6.2)

where ni-Zn is the bulk atomic density of Zn in our i-ZnO films grown, estimated at 4.0×1022 cm-3 from our previous work.9

In both the TEM images and the 2D APT profile, 3 and 11 individual Al-rich layers can be distinguished in the AZO-1 and AZO-2 films, respectively. These values are consistent with the number of supercycles, M, used to incorporate


Al into these two films (cf. Table 6.1). The AZO-3 film, prepared using M = 19 supercycles, has a lamellar structure that is less clearly resolved in both the TEM (Fig. 6.2 (b)) and the APT profiles (Fig. 6.2 (c)). This can be attributed to the combination of the small interspacing (~2 nm), the projection of fine-scale surface roughness on the 2D profiles and the broadening of the Al-rich layers. The last factor will be addressed in the next section.

Figure 6.2 (a) Schematic representation of the structure of the

ZnO/ZnO:Al film stack prepared on Si substrate and microtip coupons. (b) High-resolution HAADF-STEM images of the three ZnO:Al films. (c) Two-

dimensional depth profile of the Al-fraction measured by APT showing a 5 nm projections of the Al-distribution in the film stack. (d) One-

dimensional local depth profile of the Al-fraction in the three ZnO:Al films measured by APT. These profiles are extracted from a cylindrical sub-

volume with a diameter of 15 or 20 nm, where the interface between the

ZnO matrix and the Al-rich layers are approximately locally flat and orientated normally to the sampling direction.


6.3.2 One-dimensional local depth profiles

One-dimensional (1D) local depth profiles created from the 3D APT data are presented in Fig. 6.2 (d). The term “local” is used to emphasize that the profiles are not the projections from the entire 3D data set, but instead are collected from cylindrical sub-volumes with a diameter of 15 or 20 nm. The sub-volume is selected, as far as possible, from regions in which the interfaces between the ZnO matrix and the Al-rich layers were virtually locally flat for the full width of the diameter of the cylinder, for all the layers in the sub-volume. The cylindrical sub-volumes were oriented to be normal to the interface plane, which, because of the faceting, would not necessarily be parallel to the local growth direction of the film. However, given the scale of the roughness, particularly in the case of layer AZO-2, it must be accepted that it is not possible to have a perfectly positioned analysis sub-volume, a factor that will introduce apparent interface broadening into the 1D profiles. Despite this, individual peaks, which represent the Al-rich layers, are observed in all three ZnO:Al films. Both the number of peaks and the spacing between the peaks are consistent with the growth properties, as listed in Table 6.1.

These 1D profiles clearly show that the distribution of Al in the depth direction is not a δ function. Instead, from the full shape of the peaks in the upper profile in Fig. 6.2 (d), it can be estimated that the peaks have a full width at half maximum (FWHM) of ~2 nm. The Al atoms in a single Al-rich layer are incorporated into the film by a single ALD cycle which has a growth rate of 0.1 nm/cycle. Compared to this width of 0.1 nm, the FWHM of 2 nm represents a significant broadening of the Al-rich layers. Note, that in the upper profile of Fig. 6.2 (d), an asymmetric peak broadening is observed: more downwards broadening than upwards. Asymmetric interface profiles in APT data must be treated with caution because local magnification effects can occur when the layers have different evaporation fields, leading to fluctuations in the atomic density.16,17 However, in the case of the current measurements the density fluctuations in the 1D profile are such that there is an apparent higher atomic density on the interface going from AlOx layer (higher evaporation field) to ZnO (lower evaporation field). This would, therefore, tend to suppress the perceived downward broadening rather than enhance it.


There are three potential reasons for the measured broadening in the 1D profiles: (i) the roughness of the layer and misalignment from normal of the sampling volume to each of the surfaces causing projection effects (ii) the resolution limit of the technique, and iii) an actual vertical spread of the Al atoms in the films.

Option (i) reflects the issue how to extract local compositional information from a sample with a complex 3D geometry. In the samples studied here, the surface topography of all layers within the stacks varies strongly along each interface. For instance, the volume of the AZO-2 stack depicted in Fig. 6.2 (b) would not allow for extracting a 20 nm wide cylindrical volume perpendicular to the interfaces without having significant interface broadening. Taking this into account, the location of the 15 - 20 nm diameter sampling cylinder was chosen based on the shape of the layers to provide a suitable compromise between sharper profiles and sufficient compositional sensitivity. By such careful selection of the sub-volumes in the APT data sets, we tried our best to minimize the effect of projected interface broadening.

Option (ii) assumes that the limit of the spatial resolution of APT has resulted in broadening of the Al profiles. This resolution is highly specimen dependent. In some publications the depth resolution of APT is reported to be better than 0.6 nm/decade,18,19 For some materials, the reconstruction of individual atomic planes is used to demonstrate “atomic” resolution.19 However, for measurement of compound semiconductors the field evaporation behaviour is often too disordered to achieve this value. As mentioned before, the broadening of the peaks of the Al-rich layers in the AZO-1 film is estimated to have an FWHM of 2 nm. As shown in Fig. 6.3, by assuming all the peak broadening is a result of the experimental profile, described by a Gaussian broadening function, we can estimate an upper-bound to this factor. From the degree of separation observed in the 1D profile of the AZO-3 film, we can conclude that the contribution of the potential instrument resolution limitation to the peak broadening is narrower than an FWHM of ~1.5 nm. This suggests that there must be additional broadening of the peaks other than potential instrument resolution limitations.


Figure 6.3 Simulated one-dimensional Al profiles in the ZnO:Al films with

different FWHMs: (a) AZO-1 with an FWHM of 2 nm, (b) AZO-3 with an FWHM of 2 nm and (c) AZO-3 with an FWHM of 1.5 nm. As observed in

the upper profile in Fig. 6.2 (d), the measured Al profile in the AZO-1 film has a broadening with an FWHM of ~2 nm. The corresponding Al profile

simulated by a Gaussian distribution with the same FWHM is shown in (a). When this FWHM is applied to the AZO-3 film, the resulting simulated Al profile is shown in (b). In this profile, the minimum AFlocal is around 90%

of the maximum value. While in the measured data (the bottom profile in Fig. 6.2 (d)), the minimum AFlocal is around 60% of the maximum value,

which corresponds to a broadening with an FWHM of ~1.5 nm, as shown in (c). Therefore, the contribution of the potential resolution limitation to

the peak broadening is narrower than an FWHM of 1.5 nm. Compared to this value, the FWHM of 2 nm in the measured Al profile in the AZO-1 film

indicates that besides the potential resolution limitation, there must be

other factors that contribute to the peak broadening.


Having rejected measurement and instrumental artifacts to be the sole cause of the broad Al peaks, we need to consider option (iii), i.e. a physical reason for a vertical distribution of Al dopants. Thermally driven atomic diffusion of Al into a ZnO matrix has been observed in other experiments, but appears to be relevant only when the samples are annealed at temperatures higher than 800°C because of the high activation energy for diffusion of Al in ZnO (Ea = ~2.8 eV).20,21 Therefore, such atomic diffusion is unlikely to be the reason for the observation of a significant broadening in our case, where the substrate heater temperature is only 250°C. A more likely explanation is the intermixing of Al and Zn atoms that can occur easily via surface reactions during ALD growth.22 An etching effect has been observed when TMA reacts with a ZnO surface,11,23,24 i.e. TMA can remove a surface Zn atom and consequently, an Al atom can occupy the created surface site. This etching effect, might introduce the downward spreading of Al atoms, accounting for the asymmetry of the profile.

We now apply the information we have obtained about the actual vertical concentration profiles of Al in ZnO to reconsider the classical representation of the compositional distribution of a film deposited using a supercycle ALD recipe. In the conventional (nominal) description, a δ-function profile is assumed, as shown in Fig. 6.4 (a), and the electrical parameters are considered to be dependent on the space between the adjacent AlOx layers. Based on the observations of the broadening of Al-rich layers, we propose the refined Al profile as shown in Fig. 6.4 (b). Here, the broadening of Al-rich layers is taken into account. The Al peaks are represented with a shape similar to a Gaussian distribution. For smaller spacings, occurring at higher doping levels, there is overlapping of Al profiles. This results in a non-zero local Al density throughout the entire stack. The exact shape of the resulting profile will affect both the conductivity9 as well as the doping efficiency. The latter will be discussed in more detail below.


Figure 6.4 Schematic representation of models of (a) the δ dopant

distribution7,9 and (b) the actual distribution of Al atoms in our ZnO:Al films. In schematics for high doping levels, the Al atoms are indicated by either red or blue colored balls to differentiate Al atoms from adjacent

AlOx layers. In the refined distribution model, the broadening of AlOx layers is considered. The overlapping of the Al profiles of the individual

Al-rich layers occurs high doping levels with short spacings between layers.

6.3.3 Factors limiting the doping efficiency of Al

Based on the obtained Al distribution, two factors limiting the doping efficiency are discussed in this section: the solubility limit and the disorder-induced carrier localization.

The solid solubility of Al in the ZnO matrix is relatively low, because Zn and Al elements have different oxidation states, ionic radii and coordination preferences.25 The values of the solubility limit reported in the literature vary. For example, a value of AFglobal = 0.3% is reported for ZnO:Al powders synthesized via the Pechini route, and is based on the observation that an additional phase ZnAl2O4 was detected by X-ray diffraction (XRD) after annealing.25 For ZnO:Al thin films prepared by the sol-gel method26 or by the spray pyrolysis technique,27 a degradation of the crystallinity of the ZnO:Al


films was observed when AFglobal > 2%. In the case of the ZnO:Al thin films grown by ALD, a decrease of crystallinity was observed when AFglobal > 3%.28 Due to the facts that AFlocal was not measured in these literature, and that the aforementioned deposition methods (except for ALD) possibly resulted in a homogeneous Al distribution, here we assume the same values of AFlocal as the reported AFglobal in these literature. When the AFlocal is at the solubility limit (AFsol.), similar to Eq. 6.2, the Al density at the solubility limit (nAl.sol.) can be calculated by:

Al.sol. -Zn .i soln n AF= × (Eq. 6.3)

where ni-Zn is the atomic density of Zn in i-ZnO as in Eq. 6.2. The corresponding Al-Al interatomic distance (dsol.) at the solubility limit can be approximated by:

1/3

sol Al.sol.d n

−≈ (Eq. 6.4)

Based on the solubility limit values mentioned above (0.3% ≤ AFlocal ≤ 3%), the corresponding value of dsol. will range from 0.9 to 2.0 nm. When the local Al density is below the solubility limit (nAl < nAl.sol. and dAl-Al > dsol.), all Al atoms can be substitutionally incorporated on the Zn positions in the ZnO matrix, and effectively act as electron donors. When the local Al density is above the solubility limit (nAl > nAl.sol. and dAl-Al < dsol.), a fraction of the Al atoms may form AlOx clusters, ZnAl2O4 compounds or segregate to the grain boundaries. These Al atoms are unlikely to act as electron donors. Therefore, the doping efficiency is lower when the local Al density is high.

Besides the aforementioned factor of the solubility limit, the reduced doping efficiency at high doping levels can also result from the disorder-induced carrier localization. In literature, it has been reported that intrinsic defects, such as oxygen vacancies (VO

2+),29,30 and extrinsic defects, such as substitutional dopant atoms at zinc sites (e.g. GaZn

+),31 can cause disorder in the ZnO lattice. During electrical transport, electrons can be backscattered by such disorder, leading to the self-interference of coherent electron wave functions.29 As a result, the scattered electrons are weakly localized, which limits the contribution of the electrons to the conductivity.31 The inelastic diffusion length (Lin) can be expressed by:31

/in in BL D D k Tτ= = (Eq. 6.5)


where D is the diffusion constant, τin is the inelastic relaxation time, and ħ, kB and T are the reduced Planck constant, Boltzmann constant and absolute temperature, respectively. With increasing doping levels, the lattice disorder increases. When the interatomic distance between dopant impurities is smaller than Lin, electrons start to interfere constructively.31 Taking this theory into our ZnO:Al case, the value of D is ~ 0.213 cm2/s in the ZnO matrix31, resulting in Lin =~ 0.74 nm at room temperature (T = 298 K). It is reasonable to propose that when the interatomic distance of Al is smaller than Lin, i.e. dAl-Al<~ 0.74 nm, the weak localization of electrons will become significant. These weakly localized electros may not contribute to the carrier density measured by the Hall measurements, and hence not to the calculated doping efficiency.

Summarizing, the Al-Al interatomic distance at the solubility limit (dsol.) and the inelastic diffusion length are both in the scale of ~ 1 nm. Therefore, in order to obtain a high doping efficiency, an Al-Al interatomic distance dAl-Al > ~1 nm is desired, which corresponds to a local Al density nAl < ~1 nm-3, i.e. AFlocal < ~2.5%. The measured Al concentrations in the three profiles displayed in Fig. 2 (d) appear to exceed this doping level, either over the full thickness of the stack (AZO-3) or locally (AZO-1 and AZO-2). This indicates that at the depths where ALD Al2O3 cycles are inserted during film growth, the amount of incorporated Al atoms is too high to obtain an optimal conductivity and doping efficiency. On the other hand, the broadening of the Al profiles, most likely due to intermixing during deposition, actually has a beneficial effect on the doping efficiency in the AZO-1 and AZO-2 stacks, as a smaller fraction of the dopants contributes to cluster formation than in the case of a delta-doped layer. In order to achieve a higher doping efficiency, the amount of Al atoms incorporated during one ALD Al2O3 cycle should be reduced. This confirms why the approaches of reducing incorporated Al atoms as described in refs.10,11 give rise to a significant enhancement of doping efficiency.

6.3.4 Al distribution at grain boundaries

So far, the discussion focused on the distribution of Al within the ZnO:Al grains, and the discussion presented above holds under the assumption of an infinite ZnO:Al grain size, i.e. effects introduced by grain boundaries are not yet taken into account. The effect of grain boundaries on the electrical transport, such as grain boundary scattering, can also play an important role. In fact, in order to


decrease the resistivity of ZnO:Al films, the improvement of the carrier mobility is as important as that of the carrier density. In the case of ITO, the carrier mobility is in the range of 20-40 cm2 V-1 s-1,32 while the mobility of ZnO:Al is only ~10 cm2 V-1 s-1.4,9 One significant contribution to the differences between carrier mobility in ZnO:Al and ITO is the limit from grain boundary scattering.33 This scattering has a minor effect for (Sn-doped) indium oxide due to a low trap density at the grain boundaries (~1.5×1012 cm-2), but plays an important role for ZnO:Al due to a high trap density (~3×1013 cm-2).34 Previous work9 indicates that the Hall mobility of ZnO:Al prepared by ALD is typically 80% of the intra-grain mobility. The remaining 20% difference is ascribed to grain boundary scattering. Therefore, it is also relevant to investigate the Al distribution at grain boundaries.

In order to study the Al distribution at grain boundaries, first the morphology of the entire ZnO/ZnO:Al stack was imaged by bright-field TEM, as shown in Fig. 6.5 (a). Grains with a columnar-growth shape throughout the entire stack are observed. Based on this image, the width of the grains is estimated to be ~40 nm. The 3D morphology of the three Al-rich layers in the AZO-1 film is presented via the APT-derived isosurfaces of AFlocal = 1.7% and AFlocal = 3.0%, as shown in Fig. 6.5 (b) and (c). These three Al-rich layers are indicated by different colors for visual clarity. The 1.7% isosurface provides a 3D representation of the faceted Al-rich layers shown in the TEM and APT images (Fig. 6.2 (b) and (c)). The 3.0% isosurface highlights some regions inside the Al-rich layer that are found to have an enhanced Al-concentration. The 3.0% isosurface has a cross-shaped pattern. Especially in the top-most Al-rich layer, the cross-shaped pattern is more obvious, and has a dimension of around 20-30 nm.


Figure 6.5 (a) Cross-sectional bright-field TEM image of the entire film stack. Iso-concentration surfaces created from the APT data set at (b)

AFlocal = 1.7% and (c) AFlocal = 3.0% in the AZO-1 film. This layer was prepared with 3 supercycles, i.e. there are 3 Al-rich layers in this film. For

clarity, each of the three Al-rich layers in this AZO-1 film has been assigned a different color. (d) Schematic representation illustrating

different AFlocal within grains and at grain boundaries in the AZO-1 film. The grey rectangular boxes indicate ZnO:Al grains in a columnar shape, as

observed in (a). The gaps between the boxes indicate grain boundaries.

The colored planes indicate the Al-rich layers, and correspond to the isosurfaces in (b). The cross-shape patterns within the planes indicate the

enrichment of Al atoms at grain boundaries, and correspond to the isosurfaces in (c). The AFlocal is above 3 % in the cross-shape patterns,

and between 1.7% and 3% at the transparent parts of the planes. In rest parts of the rectangular boxes, the AFlocal is below 1.7 %.

We propose that the correlation between the microstructure of ZnO grains imaged by TEM and the lateral Al distribution measured by APT can be schematically represented as in Fig. 6.5 (d). The transparent parts of the colored planes in Fig. 6.5 (d) indicate Al-rich layers in ZnO grains with 1.7% < AFlocal < 3.0%, while the cross-shape patterns within the planes follow the shape and locations of grain boundaries in the Al-rich layers, and indicate Al-rich layers at grain boundaries with AFlocal > 3.0%. Based on the dimension of the cross-shape pattern (~ 20-30 nm), the grains can be estimated to have a similar size as the pattern, which is consistent with the grain size observed in

6.4 Conclusions 115

the TEM image (~40 nm). The columnar shape of the ZnO:Al grains is also consistent with the elongated grains of the entire film stack as shown in Fig. 6.5 (a).

The formation of the enrichment of Al at grain boundaries can be explained as following. In literature, it has been observed that grain boundary can act as sinks for dopants.35 Moreover, the defects at grain boundaries may also act as trapping centers and immobilize these limited amounts of Al-atoms. In the previous section, it is proposed that the broadening of Al-rich layers in the growth direction results from the intermixing between Al and Zn atoms. Here we further propose that during the film growth, such intermixing occurs mainly within the ZnO:Al grains, leaving a relatively lower AF in the original Al-rich layers (1.7% < AFlocal < 3.0%) within the ZnO:Al grains. Conversely, the Al atoms may be less mobile at the grain boundaries, resulting in a lower degree of intermixing, and thus a relatively high AFlocal (> 3.0%). Thus, an enrichment of Al is observed at the cross-section between the grain boundaries and at the depths where Al atoms are incorporated from the Al2O3 ALD cycles. On the other hand, no enrichment of Al is observed in those areas of the grain boundaries in between the Al-rich layers, suggesting no significant segregation out of the Al-rich layers. In terms of effect on electrical properties, this means the contribution of Al atoms to the relatively high grain boundary scattering in ZnO:Al is limited, at least in cases where there is a large spacing between the Al-rich layers.

6.4 Conclusions

In this work, the Al distribution in ZnO:Al films and its effect on doping efficiency are studied. The 3D spatial distribution of Al in the ZnO:Al thin films is studied at an atomic scale using a combination of TEM and APT. The ZnO:Al films are characterized by layers with a high local Al concentration. These layers are not δ-layers, but rather exhibit a vertical concentration profile, which causes overlapping of the Al-distributions of adjacent Al-rich layers. Because of the solubility limit of Al in ZnO and the disorder-induced carrier localization, a high local Al density will result in a low doping efficiency. In addition, an enrichment of Al at grain boundaries is observed at the depths where Al atoms are incorporated, while its contribution to grain boundary scattering can be limited.


6.5 Acknowledgments

The authors thank Holst Centre/IMEC-NL, The Netherlands, for financially supporting this project, as well as Solliance, a solar energy R&D initiative of ECN, TNO, Holst Centre, Eindhoven University of Technology, Imec and Forschungszentrum Jülich. The funding of the atomic resolution TEM facility by the Dutch province of Noord-Brabant as well as the fruitful discussions with Dr. A. Illiberi (TNO) and Prof. dr. P. Koenraad (Eindhoven University of Technology) are gratefully acknowledged.

6.6 References

1 K. Ellmer, Nature Photonics 6, 809 (2012). 2 J.A. van Delft, D. Garcia-Alonso, and W.M.M. Kessels, Semicond. Sci. Technol. 27, 074002 (2012). 3 E. Fortunato, D. Ginley, H. Hosono, and D.C. Paine, MRS Bull. 32, 242 (2007). 4 T. Minami, Semicond. Sci. Technol. 20, S35 (2005). 5 S.M. George, Chem. Rev. 110, 111 (2010). 6 H.B. Profijt, S.E. Potts, M.C.M. van de Sanden, and W.M.M. Kessels, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 29, 050801 (2011). 7 D.-J. Lee, H.-M. Kim, J.-Y. Kwon, H. Choi, S.-H. Kim, and K.-B. Kim, Adv. Funct. Mater. 21, 448 (2011). 8 N.P. Dasgupta, S. Neubert, W. Lee, O. Trejo, J.-R. Lee, and F.B. Prinz, Chem. Mater. 22, 4769 (2010). 9 Y. Wu, P.M. Hermkens, B.W.H. van de Loo, H.C.M. Knoops, S.E. Potts, M.A. Verheijen, F. Roozeboom, and W.M.M. Kessels, J. Appl. Phys. 114, 024308 (2013). 10 A. Yanguas-Gil, K.E. Peterson, and J.W. Elam, Chem. Mater. 23, 4295 (2011). 11 Y. Wu, S.E. Potts, P.M. Hermkens, H.C.M. Knoops, F. Roozeboom, and W.M.M. Kessels, Chem. Mater. 25, 4619 (2013). 12 K. Thompson, D.J. Larson, and R. Ulfig, Microsc. Microanal. 11, 2004 (2005). 13 D.J. Larson, D.T. Foord, A.K. Petford-Long, H. Liew, M.G. Blamire, A. Cerezo, and G.D.W. Smith, Ultramicroscopy 79, 287 (1999). 14 M.K. Miller, K.F. Russell, K. Thompson, R. Alvis, and D.J. Larson, Microsc.

6.6 References 117

Microanal. 13, 428 (2007). 15 T.F. Kelly and D.J. Larson, Annu. Rev. Mater. Res. 42, 1 (2012). 16 F. Vurpillot, A. Cerezo, D. Blavette, and D.J. Larson, Microsc. Microanal. 10, 384 (2004). 17 D.J. Larson, B.P. Geiser, T.J. Prosa, and T.F. Kelly, Microsc. Microanal. 18, 953 (2012). 18 S. Koelling, M. Gilbert, J. Goossens, A. Hikavyy, O. Richard, and W. Vandervorst, Appl. Phys. Lett. 95, 144106 (2009). 19 F. Vurpillot, G.D.A. Costa, A. Menand, and D. Blavette, 203, 295 (2001). 20 T. Nakagawa, I. Sakaguchi, K. Matsumoto, M. Uematsu, H. Haneda, and N. Ohashi, Key Eng. Mater. 421-422, 197 (2009). 21 K.M. Johansen, L. Vines, T.S. Bjørheim, R. Schifano, and B.G. Svensson, Phys. Rev. Appl. 3, 024003 (2015). 22 M. Knez, Semicond. Sci. Technol. 27, 074001 (2012). 23 J.-S. Na, G. Scarel, and G.N. Parsons, J. Phys. Chem. C 114, 383 (2010). 24 J. Elam and S. George, Chem. Mater. 15, 1020 (2003). 25 H. Serier, M. Gaudon, and M. Ménétrier, Solid State Sci. 11, 1192 (2009). 26 X. Zi-qiang, D. Hong, L. Yan, and C. Hang, Mater. Sci. Semicond. Process. 9, 132 (2006). 27 A. El Manouni, F.J. Manjón, M. Perales, M. Mollar, B. Marí, M.C. Lopez, and J.R. Ramos Barrado, Superlattices Microstruct. 42, 134 (2007). 28 P. Banerjee, W.-J. Lee, K.-R. Bae, S.B. Lee, and G.W. Rubloff, J. Appl. Phys. 108, 043504 (2010). 29 D. Saha, A.K. Das, R.S. Ajimsha, P. Misra, and L.M. Kukreja, (n.d.). 30 A. Tiwari, C. Jin, J. Narayan, and M. Park, J. Appl. Phys. 96, 3827 (2004). 31 V. Bhosle, a. Tiwari, and J. Narayan, J. Appl. Phys. 100, (2006). 32 S. De Wolf, A. Descoeudres, Z.C. Holman, and C. Ballif, Green 2, 7 (2012). 33 K. Ellmer and R. Mientus, Thin Solid Films 516, 4620 (2008). 34 K. Ellmer and R. Mientus, Thin Solid Films 516, 5829 (2008). 35 T. Suzuoka, Trans. Japan Inst. Met. 2, 25 (1961).

Chapter 7

Compositional, structural and electrical properties of In2O3 thin films prepared by

atomic layer deposition from cyclopentadienyl indium and O2/H2O*

Abstract Recent studies on In2O3 thin films prepared by atomic layer deposition (ALD) primarily focused on identifying appropriate indium precursors. Libera et al. (Chem. Mater. 23, 2150 (2011)) reported that cyclopentadienyl indium (InCp), combined with H2O and O2 can yield fairly high growth per cycle values (1.0 nm) at temperatures as low as 100 oC. Recently, Macco et al. (ACS Appl. Mat. Interfaces 7, 30 (2015)) demonstrated that In2O3 films with record optoelectronic quality can be obtained by post-crystallizing the films grown by this process and identified atomic H as the main donor in such films. In this work, the crystal growth and the incorporation of hydrogen during the ALD process have been investigated by using transmission electron microscopy (TEM) and atom probe tomography (APT) in combination with deuterium isotope labelling. From the TEM results it can be concluded that an amorphous-to-crystalline transition occurs upon increasing the temperature (from 100 to 150 °C). At higher deposition temperatures (>200 oC), enhanced nucleation of crystals and the incorporation of carbon impurities in the film lead to a reduced grain size and even amorphization of the film, respectively, resulting in a strong reduction in carrier mobility. From the APT studies using D2O as co-reactant, it was found that the incorporated hydrogen (deuterium) in the film mainly originates from the co-reactant and not from the precursor. In addition, it was established that the incorporation of hydrogen during film growth is higher in the amorphous phase than in the crystalline phase.

* Y. Wu, D. Vanhemel, B. Macco, S. Koelling, M.A. Verheijen, P.M. Koenraad, F. Roozeboom,

W.M.M. Kessels, (in preparation for publication).

120 Compositional, structural and electrical properties of In2O3

7.1 Introduction

In2O3 can exhibit a high optical transparency and a low electrical resistivity (<1 mΩ∙cm).1 For this reason, In2O3-based thin films find applications as transparent conducting oxide (TCO) in e.g. solar cells2–4, (organic) light emitting diodes5,6 and flat panel displays.7,8 In many of these applications, a precise thickness control at the nanometer scale, a high uniformity on large areas, a good conformality over three-dimensional surface topologies and a low processing temperature are becoming increasingly important. Today, sputtering9 and chemical vapor deposition (CVD)10 are the mainstay in commercial deposition of In2O3 thin films. However, ultimately these methods face limitations with respect to the aforementioned requirements. For example, it is challenging to prepare thin films on three-dimensional surface topologies with a high conformality for sputtering, whereas a high substrate temperature is often required for CVD. With atomic layer deposition (ALD) it is possible to deposit thin films over large-area substrates with superior uniformity and conformality and with precise thickness control.11,12 It has also been demonstrated that the substrate temperature of In2O3 films deposited by ALD can be as low as 50 °C.13

In2O3 has been prepared by ALD with different In precursors. For instance, ALD of In2O3 films was first realized using InCl3 and H2O as precursor and co-reactant, respectively.14 However, this process requires high substrate temperatures (400 °C-500 °C) and features a fairly low growth per cycle (GPC) of 0.02-0.03 nm. Moreover, it was shown that the deposited In2O3 can be etched by the InCl3 precursor. Trimethyl indium [In(CH3)3] combined with H2O can yield In2O3 films as well.15 However, for this process, the values of GPC obtained at relatively low substrate temperatures (<250 °C) are lower than 0.04 nm. Meanwhile, the resistivity was observed to be relatively high (2-6 mΩ∙cm) due to the low carrier density (2-8×1019 cm-3). Indium acetylacetonate [In(acac)3] can be used to prepare In2O3 films with either H2O or O3, but the maximum GPC is only 0.03 nm or 0.06 nm, respectively.16 Recently, diethyl[1,1,1-trimethyl-N-(trimethylsilyl)silanaminato]-indium (INCA), [3-(dimethylamino-kN)propyl-kC]dimethyl-indium (DADI), and triethyl indium [In(C2H5)2] were used as In precursors with O3 as a co-reactant 13 Even though the GPC values of 0.04-0.12 nm are relatively high, the film resistivities obtained at 100 °C exceed 1 mΩ∙cm.

7.1 Introduction 121

It has also been reported that In2O3 can be prepared at low substrate temperatures and with fairly high GPC values by using cyclopentadienyl indium (InCp) as In-precursor.17,18 Together with O3 as co-reactant, GPC values as high as 0.13-0.20 nm were obtained at substrate temperatures ranging from 200 °C to 450 °C.17 In contrast to the aforementioned In-precursors where the In atom is in an oxidation state of +3, the oxidation state of the indium atom in InCp is +1. Therefore, O3 acts as an active oxidant to oxidize the In atom from the +1 oxidation state in InCp to the +3 state in In2O3. However, during the ALD process, the thermal decomposition of O3 is catalyzed by the In2O3 film surface, resulting in a significant non-uniformity in film thickness, i.e. a lower film thickness is observed at the downstream region of the O3 injection.17,18 Moreover, the minimum substrate temperature of 200 °C is not sufficiently low to allow the growth of In2O3 films on organic and other temperature sensitive substrates.17 In their follow-up work, Libera et al. reported that the co-reactant O3 can be replaced by the combination of H2O and O2. With their recipe using InCp, H2O and O2, a high GPC of 0.1 nm at temperatures as low as 100 °C could be realized, while also a high uniformity was obtained.18 Their in-situ study using a quartz crystal microbalance (QCM) and a quadruple mass spectrometer (QMS) revealed that H2O serves to release the -Cp ligands from the film surface, and that O2 oxidizes the indium atoms from the +1 to the +3 oxidation state. No significant growth was observed when using only H2O or O2 as a co-reactant. Moreover, a higher GPC and a lower film resistivity were obtained when H2O and O2 were fed into reactor during growth simultaneously, instead of exposing the surface to H2O and O2 sequentially (in either order).

In addition to yielding high GPC values at low deposition temperatures, the process using InCp and simultaneous dosing of H2O and O2 also results in a low film resistivity (<1 mΩ∙cm) over a range of substrate temperatures (100-200 °C).18 Interestingly, the carrier density and mobility were found to change drastically within this temperature range. These changes were attributed to the change of crystallinity as evidenced by X-ray diffraction (XRD) data. However, other properties, such as the grain morphology and the elemental film composition, can also have important effects on the electrical properties. However, the nature of the species acting as electron donors in In2O3 is still under debate. In the work of Libera et al., as well as in other literature reports,15 the free-electron donors were attributed to oxygen vacancies (VO

2+). However, it has been pointed out that an oxygen vacancy acts as a deep donor


in crystalline In2O3 films,19 and that hydrogen can act as a shallow electron donor.20 Hydrogen can lead to unintentional doping of In2O3 films and can be easily overlooked by compositional studies by e.g. X-ray photoelectron spectroscopy (XPS). When hydrogen atoms act as donors, they can either be present as interstitial hydrogen at the antibonding centre of In-O bonds (Hi

+) or as substitutional hydrogen at the location of oxygen vacancies (HO

+).19 Recently, we have also provided evidence that HO

+ and Hi+, instead of VO

2+, are the main electron donors in In2O3 films which are crystallized by post-annealing after deposition by ALD.21

The high growth per cycle, low substrate temperature and low film resistivity of the In2O3 films obtained when using InCp, H2O and O2 make the process a very interesting route to prepare In2O3 thin films by ALD. Using the same precursor and co-reactants, we obtained In2O3 films with comparable electrical properties as Libera et al. To complement their work, we investigated the correlation between the compositional, structural and electrical properties. In addition, we performed an explorative study on the origin and distribution of hydrogen in the films, considering its possible role as a donor in In2O3. After the experimental details, we will first describe the optimization of our ALD recipe. Next, studies on the composition, crystallinity, microstructure and electrical properties will be presented and discussed for films prepared at different substrate temperatures. Next, elastic recoil detection (ERD) and atom probe tomography (APT) studies of the distribution of hydrogen in In2O3 will be presented. Finally, the correlation between the crystallinity, elemental composition and electrical properties will be discussed.


The In2O3 films were deposited using an Oxford Instruments OpAL ALD reactor. InCp was used as the indium precursor supplied by SAFC Hitech. Unless specified otherwise, deionized water vapor (DI H2O) and oxygen gas (O2) were both used as co-reactants in the second half-cycle. In specific samples, deionized deuterated water (DI D2O) was used. InCp was held in a stainless steel bubbler heated to 40°C, and was delivered to the deposition chamber by a 50 sccm flow of Ar carrier gas. H2O was vapor drawn and O2 was delivered through a mass flow controller at a flow of 50 sccm. Both Si (100) wafers with ~ 450 nm thermally grown SiO2 and Si (100) wafers with 1-2 nm native oxide


were used as substrates. The substrate temperature during ALD refers to the temperature of the substrate stage. The chamber walls were also set to the substrate temperature, with a maximum of 180 °C.

Film growth was monitored by in-situ spectroscopy ellipsometry (SE). The SE set-up used for these measurements is a J.A.Woollam Co. Inc. M-2000D spectrometer with an XLS-100 light source (1.2-6.5 eV of photon energy). Cauchy modeling was applied to determine the film thickness. After growth, the film thickness was verified by ex-situ SE at room temperature. The resistivity of the films was measured using a Signatone four-point probe (FPP) station, while the resistivity, Hall mobility and Hall carrier density were also measured with a Europa HMS-5300 Hall effect measurement system. XPS set-up used in this work was a Thermo Scientific K-Alpha KA1066 spectrometer, using monochromatic Al Kα (hv = 1486.6 eV) X-ray radiation. The XPS sensitivity factors used for oxygen and indium are 2.88 and 37.95, respectively, which were calibrated from the stoichiometry values of O/In obtained by Rutherford backscattering spectrometry (RBS).22

The XRD measurements were performed using an X’Pert Pro MRD, manufactured by Panalytical. ω-2θ scans were performed on In2O3 films prepared on Si wafers with native oxide. To avoid a strong Si(400) reflection in the XRD measurements, the symmetry axis was given a 3° offset from the normal of the sample, i.e. ω- θ=3°. The areal densiwes of indium and oxygen atoms were measured by RBS, while the areal densities of hydrogen (1H) and deuterium (2H) were measured by elastic recoil detection (ERD). The RBS/ERD measurements in this work were performed through Detect99 at the Singletron at AccTec BV in Eindhoven, The Netherlands. The samples were first investigated with ERD and subsequently with RBS using a 1.8 MeV He+ beam. The RBS spectra were registered simultaneously by two detectors at scattering angles of 105° and 170°. TEM studies were performed using a JEOL JEM ARM 200F transmission electron microscope. The cross-sectional TEM sample was prepared using a standard FIB lift-out procedure. APT was carried out using a LEAP 4000X-HR from Cameca utilizing picosecond laser pulses with a wavelength of 355 nm. For the analyses the In2O3 thin films were prepared into tips using a FEI Nova 600 FIB/SEM dual beam system.23 The analyses were carried out at a base temperature of 20 K and with a laser power of 10-25 pJ per pulse. In order to assure that the contribution to the hydrogen signal from


the vacuum background is minimal, measurements were carried out at different laser pulse frequencies (125-250kHz) and evaporations rates (1.5-3%). No significant changes of the measured hydrogen concentrations were observed over this range.

7.3 Results and discussions

7.3.1 ALD growth

As a first step, the self-limiting nature of the surface reactions for the ALD process with InCp and simultaneous dosing of O2 and H2O was verified for our ALD reactor. So-called saturation curves were measured at a substrate temperature of 200 °C. Similar to the recipe of Libera et al., our recipe consists of InCp dose ― InCp purge ― O2 flow stabilizawon ― O2/H2O dose ― O2/H2O purge. The saturation curves of the dosing and purging steps are shown in Fig. 7.1. Self-limiting growth was observed for the InCp dosing and the O2/H2O dosing steps. In the InCp purge step, no CVD-component was apparent from the curve even when the purge time was set to 0 s (note that also the O2 flow stabilization step acts as a kind of purge step). In the last step, ~7 s of O2/H2O purging is required to avoid CVD-type reactions between H2O, O2 and the InCp dosed in the next cycle. Both our work and the aforementioned work by Libera et al. 18 proved that almost no growth occurs with only InCp and O2 reactant. Therefore, the step of 5 s O2 flow stabilization is not considered to affect the growth of In2O3 during one ALD cycle, and is thus not considered in the measurement of the saturation curves. Based on the saturation curves, the optimized recipe has been set at 5 s InCp dose ―2 s InCp purge ― 5 s O2 flow stabilizawon ― 0.25 s O2/H2O dose ― 10 s O2/H2O purge. This cycle is quite similar to the one from Libera et al., which is 3 s InCp dose ― 5 s InCp purge ― 4 s O2/H2O dose ― 5 s O2/H2O purge. This similarity suggests that the ALD process for In2O3 using InCp, O2 and H2O can be quite easily transferred between ALD setups.

7.3 Results and discussions 125

Figure 7.1 Saturation curves for ALD of In2O3 films using InCp as precursor and a mixture of O2/H2O as co-reactant. The substrate temperature was

200 °C. Lines serve as a guide to the eye. The optimized recipe consists of 5 s InCp dose ― 2 s InCp purge ― 5 s O2 flow stabilizawon ― 0.25 s

O2/H2O dose ― 10 s O2/H2O purge.

Based on this recipe, the GPC at different substrate temperatures has been studied. As shown in Fig. 7.2, a relatively constant value of GPC between 0.11 and 0.13 nm is observed for the temperature range of 100 °C-300 °C. In the temperature range of 100 °C to 200 °C, this value corresponds to (3.0-3.8)×1014 cm-2 In atoms deposited per cycle, as deduced from the RBS data, as shown in Fig. 7.3. The GPC starts dropping around 100 °C, which can likely be attributed to incomplete surface reactions due to insufficient thermal energy.11 For reference, Fig. 7.2 also represents an overview of GPC values for other ALD processes for In2O3 reported in the literature. Compared to these processes, the ALD process with InCp and H2O/O2 yields relatively constant and high GPC values over a wide temperature range. This underlines the fact that the ALD process based on InCp and O2/H2O is well-suited for ALD of In2O3.


Figure 7.2 Growth per cycle as a function of substrate temperature for

In2O3 films prepared by ALD. The results obtained in this work are compared to those reported in the literature by Libera et al.,18 Nilson et

al.,16 Lee et al.,15 Ozasa et al.,24 Gebhard et al.,25 Maeng et al.,13 Asikainen

et al.14 and Ritala et al.

26.

Figure 7.3 Growth per cycle in terms of areal atomic densities (atoms/cm2)

of indium and oxygen as a function of substrate temperature as

measured by RBS.


7.3.2 Film properties

Compositional, structural and electrical properties have been determined for the In2O3 films prepared between 100 °C and 350 °C. Some of the results are summarized, as listed in Table 7.1.

Table 7.1 Overview of the growth per cycle, the compositional properties

and the electrical properties of the In2O3 films prepared at different substrate temperatures. The films had a thickness of ~40 nm. The growth

per cycle was measured by in-situ spectroscopy ellipsometry. The O/In ratio and the carbon content were measured by X-ray photoelectron

spectroscopy. The atomic percentage of hydrogen was calculated by combining the hydrogen content measured by elastic recoil detection

with the indium and oxygen contents measured by Rutherford

backscattering spectrometry. The resistivity, carrier density and mobility were measured by Hall measurements. The typical experimental errors

are shown in the first entry of each column. A dash means “not measured”. The hydrogen content in the In2O3 films was only determined

for the films deposited at low substrate temperatures as demarcated by the dashed line.

Sub-strate temp.

(°C)

Growth per cycle

(nm)

Compositional properties Electrical properties O/In ratio

Carbon (at. %)

Hydrogen (at. %)

Resistivity (mΩ∙cm)

Carrier density

(1020 cm-3)

Hall mobility (cm2/V∙s)

100 0.11±0.01 1.56±0.07 0.0 4.9±0.5 0.43±0.02 4.0±0.2 37±2 130 0.12 1.59 0.0 3.3 0.72 1.7 51 150 0.12 1.58 0.0 3.6 0.79 1.2 68 170 0.12 1.57 0.3±0.1 - 0.85 1.8 40 200 0.12 1.58 0.5 - 0.91 2.0 35 250 0.12 1.57 3.7 - 1.3 2.1 23 300 0.13 1.55 6.4 - 2.14 1.8 16 350 0.12 1.51 10.0 - 3.93 1.5 10

Elemental composition. The atomic ratio of oxygen over indium appeared to be constant (~ 1.57) regardless the substrate temperature, indicating that the films are not stoichiometric, but instead, oxygen-rich. Carbon could only be detected at substrate temperatures above 150 °C, its concentration significantly increasing with increasing temperature, as shown in Fig. 7.4. These carbon impurities most likely originate from thermal decomposition of the -Cp ligand of the InCp precursor during growth. The hydrogen content was investigated for the films prepared at the temperatures of 100 °C, 130 °C and


150 °C, yielding atomic percentages of around 3-5 at.%. The origin and distribution of hydrogen will be discussed in more details in the next section.

Figure 7.4 Atomic percentage of carbon in In2O3 films with a thickness of

40 nm as a function of substrate temperature as measured by X-ray photoelectron spectroscopy.

Electrical properties. The resistivity, carrier density and mobility at different substrate temperatures were obtained by Hall measurements on In2O3 films with a thickness of ~ 40 nm, as shown in Fig. 7.5. A resistivity lower than 1 mΩ∙cm can be obtained at substrate temperatures below 200 °C, while the lowest resistivity of 0.43 mΩ∙cm is achieved at 100 °C for the set of films studied. Note that, the resistivity stays relatively constant when the film thickness varies in the range of 10 to 70 nm, but increases drastically when the thickness decreases below 10 nm. For example, the resistivity of the film with a thickness of 6 nm prepared at 100 °C is 6.5 mΩ∙cm, as shown in Fig. 7.6. Changing the substrate temperature from 100 °C to 150 °C, yields a significant drop of carrier density from 4.0×1020 to 1.2×1020 cm-3 and a rise of carrier mobility from 37 to 68 cm2/(V∙s). This rise in carrier mobility can (partially) be attributed to the drop in carrier density, as a lower carrier density indicates fewer electron donors and hence less ionized impurity scattering,21,27 as also mentioned by Libera et al.18Upon increasing the substrate temperature from 150 °C to 350 °C, no significant change of the carrier density was observed, while the carrier mobility drops from 68 to 10 cm2/(V∙s). The drop of the carrier mobility may be correlated to the reduced grain size and the presence of


carbon impurities, as will be discussed later. The trends of the resistivity, carrier density and mobility as a function of substrate temperature are similar with those in the work of Libera et al.,18 as also shown in Fig. 7.5.

Figure 7.5 (a) Hall resistivity and (b) carrier density and mobility as a

function of substrate temperature for ~40 nm-thick In2O3 films. Solid markers are results obtained in this work (on Si wafers covered with ~450

nm thermal oxidized SiO2), whereas open markers refer to the data reported by Libera et al.

18 (on fused silica substrates).


Figure 7.6 Resistivity as a function of film thickness of In2O3 thin films

prepared at various substrate temperatures. The resistivity was measured by the four-point probe (FPP) technique and the In2O3 films were

deposited on Si wafers with thermally grown SiO2 with a thickness of ~

450 nm.

Crystallinity. The crystallinity at different substrate temperatures was studied by XRD, as shown in Fig. 7.7. For the film prepared at 100 °C, only a weak (222) peak can be observed, indicating a predominantly amorphous film. Above 100 °C, various diffraction peaks belonging to the cubic bixbyite structure of In2O3 can be observed in the XRD spectra, indicating the presence of a polycrystalline phase in the films. The intensity ratios of the peaks within each spectrum, except for the 100 °C case, are similar to those in the standard powder diffraction spectrum (JCPDS, No. 65-3170). This implies a random orientation of the grains in these films. Compared to the case of 150 °C, the peak intensities in the spectra above 150 °C are lower, indicating a slightly reduced crystallinity for increased substrate temperatures. The change of crystallinity from amorphous to polycrystalline upon the increase of substrate temperature from 100 °C to 150 °C was also observed by Libera et al.

18 However, the random grain orientation observed in this work is different from the observations in their work, revealing a <400> or <222> texture, depending on the growth conditions.


Figure 7.7 X-ray diffraction (XRD) ω-2θ spectra for In2O3 films with a thickness of 40 nm prepared on Si(100) wafers at different substrate

temperatures. XRD peak intensities from a standard powder diffraction

spectrum of cubic In2O3 are given as bar graph (JCPDS, No. 65-3170).

Microstructure. The microstructure at different substrate temperatures was studied using bright field transmission electron microscopy (BFTEM). Top-view BFTEM images of the In2O3 films with a thickness of ~40 nm deposited on SiO2 coated TEM windows are shown in Figs. 7.8 (a) to (g). The corresponding selected area electron diffraction (SAED) patterns of the films are shown in Fig. 7.9. The film prepared at 100 °C is not fully amorphous as was already indicated by the XRD spectrum. Instead, crystals with a diameter of ~ 50 nm within an amorphous matrix are observed (Fig. 7.8 (a)). For the film prepared at 150 °C, as shown in Fig. 7.8 (c), the top-view image shows a surface fully covered by crystalline grains. Compared to Fig. 7.8 (a), a higher density of grains with a similar lateral diameter of ~ 50 nm is present in Fig. 7.8 (c). In the SAED pattern of the film prepared at 150 °C, discrete diffraction rings representing the crystalline part of the film are present, while diffuse rings reflecting the presence of an amorphous fraction cannot be distinguished, implying that this film is virtually fully crystalline. The cross-sectional image of the film prepared at 150 °C (Fig. 7.10) shows that the grains have columnar shapes, that extend throughout the entire film. Note that, the exact lateral size of the grains cannot be extracted unambiguously from this image due to partial


overlap of grains within the projected thickness of the TEM sample. Based on the top-view BFTEM images, the SAED pattern and the cross-sectional image, it is reasonable to believe that the In2O3 films prepared at 150 °C are completely crystalline. Above 150 °C, with the increase of the substrate temperature, higher grain densities and smaller grain sizes are observed from the top-view images (Figs. 7.8 (d) to (g)). The formation of crystalline nuclei is known to be thermally activated.28 Therefore, an increase of substrate temperature leads to a higher nucleation rate and consequently to higher grain densities. Moreover, in the corresponding SAED patterns, discrete rings and diffuse rings co-exist, indicating that these films are partially amorphous. This can possibly again be attributed to the enhanced carbon levels in these films as will be discussed later.

Figure 7.8 Top-view bright-field transmission electron microscopy

(BFTEM) images of In2O3 films with a thickness of 40 nm as prepared on Si3N4 TEM windows covered with 3-4 nm ALD SiO2. The In2O3 was

deposited at various substrate temperatures: (a) 100 °C, (b) 130 °C, (c)

150 °C, (d) 200 °C, (e) 250 °C, (f) 300 °C, (g) 350 °C. The corresponding selected area electron diffraction (SAED) patterns are presented in Fig.

7.9.


Figure 7.9 Selected area electron diffraction (SAED) patterns of In2O3

films with a thickness of 40 nm as prepared on Si3N4 TEM windows

covered with 3-4 nm SiO2. Patterns are given for various substrate temperatures: (a) 100 °C, (b) 130 °C, (c) 150 °C, (d) 200 °C, (e) 250 °C, (f)

300 °C, (g) 350 °C. Most of the SAED patterns show a superposition of

diffuse rings (from the amorphous fraction of the film) and sets of (interrupted), more discrete rings (from the crystalline fraction of the

film). The interrupted nature of the diffraction rings reflects the limited number of grains contributing to the SAED patterns. The contribution of

the diffuse rings decreases from (a) to (b), in line with the observed increase in crystallinity in this temperature regime. In (c) and (d), the

diffuse rings are absent, indicating the fully crystalline nature of these

films. For (e) to (g), a weak diffuse ring can be recognized, in line with the decrease in crystallinity observed by XRD (Fig. 7.7). For all SAED patterns

a selected area aperture with a physical diameter of 1.3 µm was used. All patterns can be indexed with the cubic bixbyite structure of In2O3.


Figure 7.10 Cross-sectional high resolution transmission electron

microscopy image of an In2O3 film with a thickness of 40 nm prepared at

the substrate temperature of 150 °C on Si wafer with 1-2 nm native oxide.

7.3.3 Origin and distribution of H in In2O3 films

The over-stoichiometric O/In ratio might be the first indication that oxygen vacancies may not be the main electron donors in the In2O3 films. As hydrogen is also a possible donor in In2O3 films, the hydrogen distribution was investigated. Two aspects were considered: the source of hydrogen in the film, and the correlation between hydrogen content and microstructure. In order to study the contribution of the various possible sources of hydrogen, heavy water (D2O) was used as a co-reactant in these investigations instead of H2O. To address the impact of the microstructure, we studied the films prepared on Si(100) wafers at substrate temperatures of 100 °C, 130 °C and 150 °C. These films have a low carbon content, i.e. below the detection limits of XPS and RBS, and show relatively low film resistivity values compared to the films prepared at higher temperatures (>150 °C). Moreover, the microstructures of the films prepared at low temperatures are clearly different: the films are mostly amorphous at 100 °C, while fully crystalline films are obtained at 150 °C. The concentration and distribution of hydrogen in these films were investigated by ERD and APT.

As listed in Table 7.2, the In2O3 films prepared with O2/D2O contain both isotopes of hydrogen: hydrogen 1H and deuterium 2H. More than half of the hydrogen present is deuterium, which obviously comes from the D2O during


growth. The hydrogen 1H isotope can come from -Cp ligands, ambient water present in the deposition chamber, water adsorbed at the substrates or water adsorbed at the film when exposed to air. The total amount of hydrogen in the In2O3 films, including both hydrogen and deuterium, is 4.9 ± 0.5 at.% in the film prepared at 100 °C, as listed in Table 7.1. Note that this atomic percentage is averaged over the entire film, while the actual distribution of the isotopes is not homogeneous throughout the film, as will be discussed later. This atomic percentage corresponds to an averaged atomic density of (3.9 ± 0.4) × 1021 cm-

3. The concentration is smaller in the films prepared at 150 °C (3.6 at.%, as listed in Table 7.1). This difference can be mainly assigned to a change in the 2H concentration, because the averaged atomic percentages of 2H and 1H are 1.1 at.% and 0.3 at.% lower at 150 °C, respectively.

Table 7.2 Areal atomic density of hydrogen (1H) and deuterium (2H) in 40

nm thick In2O3 films prepared by InCp and O2/D2O as measured by elastic recoil detection (ERD). 1H in the In2O3 films is observed to be mainly

concentrated within ~ 10 nm of the surface and within ~10 nm from the substrate. No 1H is observed in the intermediate centre part of the films.

2H is distributed homogeneously throughout the 40 nm film. The

systematic errors in 1H and 2H are 7% and 10%, respectively.

Substrate Temp.

(°C)

Areal atomic density (1015 cm-2) 1H 2H

near surface

in the centre

near interface

throughout the film

100 4.0 0 1.1 8.8 130 2.5 0 1.0 6.2 150 2.5 0 1.8 6.2

Despite its limited depth resolution (~ 10 nm), ERD measurements can still provide a rough indication of the distribution of 1H and 2H in the growth direction. The distribution of the 2H atoms is found to be almost homogeneous throughout the entire film within the depth resolution. Such a distribution is consistent with the fact that the 2H atoms are incorporated in the films by surface reactions during every ALD cycle. In the case of the 1H isotope, as listed in Table 7.2, it is observed that more than 50 % of the 1H atoms are localized within ~10 nm of the surface. We propose that this part of 1H originates from water adsorbed on the film after growth. The remaining 1H atoms are mainly concentrated within ~10 nm from the substrate. This part of the 1H atoms may


originate from the hydroxyl groups and water molecules present on the surface of the substrates before ALD.

ERD is a quantitative analysis technique. However, it has only a limited spatial resolution. In order to obtain a three-dimensional image of the distribution of hydrogen, and to investigate the correlation between the crystallinity and hydrogen content, the spatial distributions of 1H and 2H in In2O3 films were measured by APT. Two films were selected for the APT study: one was prepared at 100 °C, and the other at 150 °C. Various APT samples were prepared for films prepared under these conditions by transferring different areas of the films on the Si substrates to the APT tips using a focused ion beam. Figure 7.11 (a) represents the layout of the APT samples. The reconstructions of the distribution of the elements from three representative areas (two from the 100 °C film and one from the 150 °C film) are shown in Figs. 7.11 (b), (c) and (d). Figure 7.11 (b) is representative for the major part of the film prepared at 100 °C, for which a homogeneous distribution of 2H is observed. Figure 7.11 (c) represents a distinct area of the same film, for which a relatively low density of 2H can be observed. Figure 7.11 (d) is representative for the entire area of the film prepared at 150 °C. At this temperature, 2H is homogeneously distributed. The density of 2H is higher in Fig. 7.11 (b) than in Fig. 7.11 (d), while the density in Fig. 7.11 (c) lies in between.

Based on the TEM images in Figs. 7.8 and 7.10, the microstructures of the In2O3 films prepared at 100 °C and 150 °C are proposed as in shown Fig. 7.11 (e) and (f), respectively. In Fig. 7.11 (e), the major fraction of the In2O3 film is amorphous, with only a few crystal grains being present. Such a microstructure is supported by the top-view BFTEM image shown in Fig. 7.8 (a). Here we further assume the shape of grains to be as indicated by the red patterned area in Fig. 7.11 (e): grains extending from the interface with the substrate to the top film surface and which are wider at the top than at the bottom. Such assumption is based on a growth model that nuclei form on the film surface during growth and extend upwards during subsequent film growth and - at a slower rate - also downwards due to solid-phase crystallization.26 For the film prepared at 150 °C, as shown in Fig. 7.11 (f), grains have a columnar shape, and extend from the bottom to the top of the films. Such a microstructure is supported by the cross-sectional TEM image shown in Fig. 7.10.


Figure 7.11 (a) Schematic representation of the layout of the samples

analysed by atom probe tomography (APT). Reconstructions of the three-

dimensional atomic distributions of each element in the films prepared at 100 °C and 150 °C: (b) a typical part of the film prepared at 100 °C in

which the deuterium distribution is homogeneous within the In2O3 films;

(c) a distinct area of the same film prepared at 100 °C in which an inhomogeneous distribution of deuterium is observed; (d) a typical part

of the film prepared at 150 °C in which deuterium is distributed homogeneously. Different elements are indicated by different colors as

listed on the right side of the figure. Schematic representations of the microstructure for the In2O3 films prepared at (e) 100 °C and (f) 150 °C.

The grey area represents an amorphous matrix in the film, while the color

line patterns represent crystalline grains with different grain orientations and shapes, based on the information extracted from the TEM images in

Figs. 7.8 (a) and Fig. 7.10. The dashed blue rectangles indicate the proposed locations of the three areas analysed in the APT studies in b, c

and d.

Based on the fact that most APT samples prepared from the 100 °C film show similar results as in Fig. 7.11 (b), and the fact that the In2O3 films prepared at 100 °C are mostly amorphous, it is reasonable to believe that the area studied in Fig. 7.11 (b) represents an amorphous matrix. Since the In2O3 films prepared at 150 °C are fully crystalline, Fig. 7.11 (d) represents the hydrogen distribution in a crystalline grain. Since the distribution of 2H in Fig. 7.11 (c) is a mix of Figs. 7.11 (b) and (e), we assume that the In2O3 film in Figure 7.11 (c) represents a


crystalline grain within an amorphous matrix. The proposed locations of the three representative areas in the films are illustrated by blue dashed rectangles in Figs. 7.11 (e) and (f), respectively.

To further quantify the distribution of hydrogen, four volumes from the three representative areas were selected to create one-dimensional local depth profiles of 1H and 2H. The term “local” is used to underline that the profiles are not the projections from the entire three-dimensional data set of the three areas, but instead are collected from cylindrical volumes with a diameter of 10 nm. The depth profiles are shown in Fig. 7.12. The locations of the selected volumes are indicated by the black rectangles in the schematics included in the figures. The four volumes are chosen to be representative of (a) a mix of a crystalline grain and an amorphous matrix, (b) a crystalline grain, (c) only an amorphous matrix in the film prepared at 100 °C, and (d) a crystalline grain in the film prepared at 150 °C. Note that, the atomic percentages of 1H and 2H obtained by APT are slightly different from the percentages determined by ERD/RBS, which is expected to yield more accurate quantitative results, at least averaged over the full film thickness. An enrichment of 1H at the interface between the In2O3 film and the Si wafer is observed in all four profiles, which is consistent with the ERD data. However, in contrast to the ERD measurements, no enrichment of 1H is observed at the film surfaces. This might result from the removal of adsorbed hydrogen during the FIB fabrication for the APT samples. Moreover, around 1-2 at.% of 1H is observed throughout the film, which is not observed in the ERD measurements. The observed 1H in the bulk of the films might actually be present in the In2O3 films, but it might also be due to hydrogen present in the background during the APT measurements. In the film prepared at 100 °C, the 2H concentration is ~ 4 at.% in the amorphous matrix (Fig. 7.12 (a)), ~ 2 at.% in the crystalline grains (Fig. 7.12 (b)), while at the location where simultaneously a crystalline grain and an amorphous matrix are measured (Fig. 7.12 (c)), the percentages are ~2 at.% and ~ 4 at.% in the crystalline grain and in the amorphous matrix, respectively. In the fully crystalline film prepared at 150 °C, the atomic percentage of 2H is ~2 at.% (Fig. 7.12 (d)). Summarizing, irrespective of the substrate temperature between 100 °C and 150 °C, the 2H content is consistently higher in the amorphous phase (~4 at.%) than in the crystalline phase (~ 2 at.%).


Figure 7.12 Local one-dimensional depth profiles of hydrogen (1H) and

deuterium (2D). The embedded schematics present the microstructure of the In2O3 films and correspond to those in Fig. 7.11 (e) and (f). The black

rectangles indicate the locations of the analysed volumes, which are cylindrical in shape with a diameter of 10 nm: (a) an amorphous matrix,

(b) a crystalline grain and (c) a mix of an amorphous matrix and a crystalline grain in the film prepared at 100 °C, and (d) a crystalline grain

in the film prepared at 150 °C.

7.3.4 Correlation between crystallinity, hydrogen content and electrical properties

On the basis of the obtained information about the structural, compositional and electrical properties, possible correlations between these properties will be discussed. At high substrate temperatures (> 150 °C), the increased grain density and the reduced grain size can enhance the grain boundary scattering, while the increase of carbon content may enhance the neutral impurity


scattering. Both effects can result in a decreased mobility and consequently an increased resistivity. Moreover, the carbon content will probably also hinder the crystallization of the films, resulting in the co-existence of amorphous and crystalline phases at temperatures above 150 °C.

Concerning the films prepared in the low temperature range (100 – 150 °C), there are two observations about crystallinity and hydrogen content that are of special interest. First of all, as shown in the APT study in this work, the deuterium content is ~ 2 at.% in the crystalline part and ~ 4 at.% in the amorphous part, regardless the substrate temperature. Secondly, in our previous work, we observed that the films changed from mostly amorphous to fully crystalline during post-deposition annealing, while the hydrogen content only dropped from 4.2 at.% to 3.9 at.%.21 Therefore, it might be that the difference in deuterium content in the amorphous and crystalline regions as found by APT is due to a difference in deuterium incorporation from D2O during growth on these surfaces.

Interestingly, besides the difference in the hydrogen content, the values of the carrier mobility are also different between the as-deposited films from this work and the post-deposition annealed films from our previous work. The as-deposited crystalline film shows a maximum carrier mobility of 68 cm2/(V∙s) in the temperature range studied. In contrast, the film, which was crystallized after post-deposition annealing, shows a higher value of 138 cm2/(V∙s). The electron scattering in the In2O3 films prepared by thermal crystallization of amorphous films is dominated by phonon scattering and ionized impurity scattering.21 Since the as-deposited and post-deposition annealed films are both crystalline, the level of phonon scattering can be expected to be similar for these two films. The carrier density in the as-deposited film (1.2×1020 cm-3) is smaller than that in the post-deposition annealed film (1.8×1020 cm-3), indicating less ionized impurities and thus less impact of ionized impurity scattering in the as-deposited film. Therefore, it can be concluded that the lower carrier mobility in the as-deposited film should result from other scattering mechanisms, such as grain boundary scattering and neutral impurity scattering. However, the influence of hydrogen on neutral impurity scattering is expected to be small, as reported in our previous work.21 Even though the carbon content is below the detection limit in the as-deposited films prepared at 100 - 150 °C, from the trend observed in Fig. 7.4, the traces of carbon might

7.4 Conclusions 141

still be present in the film prepared at 150 °C. If so, the carbon present can contribute to the neutral impurity scattering in this film. The effect of grain boundary scattering on the two films can be different due to the difference in grain sizes. Since the lateral grain size of the as-deposition deposited film (~50 nm) is smaller than that of the post-annealed one (>400 nm), the higher areal density of grain boundaries in the as-deposited film is likely to cause stronger grain boundary scattering.

As mentioned in the introduction, hydrogen is proposed to be the main electron donor in crystalline In2O3 films. On the basis of this work, two interesting observations related to the hydrogen content and the carrier density can be made. (i) From the measured atomic densities of hydrogen and the carrier densities and assuming that all free electrons are generated by hydrogen donors (such as HO

+ and Hi), the doping efficiency of hydrogen would be 10.2%, 6.9% and 4.5% for films prepared at 100 °C, 130 °C and 150 °C, respectively. Considering other potential donors, such as oxygen vacancies, the actual doping efficiency could be even lower. (ii) With the increase of substrate temperature from 100 °C to 150 °C, the carrier density drops significantly from 4.0×1020 to 1.2×1020 cm-3, while the hydrogen content only decreases from 4.9 to 3.6 at.%. From these two observations, the hydrogen content and the carrier density do not seem to be strongly correlated. Therefore, this work does not provide strong evidence that hydrogen is the main electron donor in the In2O3 films, and further studies will be necessary. For example, investigation of the chemical environments of hydrogen atoms in In2O3 lattice, present as, for example, -OH, H+, HO

+, H-H, etc., can be helpful to identify whether hydrogen acts as the main electron donor in In2O3.

7.4 Conclusions

In this work, we have investigated In2O3 films prepared by ALD using InCp as indium precursor and a combination of O2/H2O as co-reactant. Following the earlier report of Libera et al., the ALD recipe was established and it was shown that the process yields a high growth per cycle of ~0.12 nm in a wide temperature range (100 °C - 300 °C). Consistent with the earlier results also low film resistivities (<1 mΩ∙cm) and an increasing degree of crystallinity were obtained in the temperature range of 100 °C to 150 °C. The composition and microstructure of the films were studied more extensively in this work and


these properties were correlated to the electrical properties. In particular, TEM studies revealed more details about the microstructure of the In2O3 films: crystalline grains in an amorphous matrix were observed in the film prepared at 100 °C whereas a columnar grain structure was observed in the fully crystalline film prepared at 150 °C. Carbon impurities were present in the films prepared at higher temperatures (>150 °C), which likely explains the reduced crystallinity and carrier mobility for these elevated temperatures when compared to the films prepared at 150 °C. Furthermore, since hydrogen has been proposed as the main electron donor in In2O3 films in the literature, we studied its origin and distribution in the In2O3 films by the combination of RBS/ERD and APT. Deuterium isotope labeling experiments revealed that the hydrogen in the films mainly originates from the co-reactant H2O, while from the APT study it was concluded that more hydrogen is incorporated from the co-reactant H2O during ALD growth in an amorphous In2O3 phase than in a crystalline phase. However, no strong correlation between the hydrogen content and the carrier density was observed. A more detailed understanding of the doping of the In2O3 films will require additional research, e.g. an investigation of the chemical environments of hydrogen atoms in the In2O3 lattice.

7.5 Acknowledgments

The authors thank Holst Centre/IMEC-NL, The Netherlands, for financially supporting this project, as well as Solliance, a solar energy R&D initiative of ECN, TNO, Holst Centre, Eindhoven University of Technology, Imec and Forschungszentrum Jülich. Solliance is gratefully acknowledged for funding the TEM facility. This work was partly financed by the Netherlands Organization for Scientific Research (NWO).

7.6 References

1 O. Bierwagen, Semicond. Sci. Technol. 30, 024001 (2015).

2 T. Koida, H. Fujiwara, and M. Kondo, Sol. Energy Mater. Sol. Cells 93, 851 (2009).

3 L. Barraud, Z.C. Holman, N. Badel, P. Reiss, A. Descoeudres, C. Battaglia, S. De Wolf, and C. Ballif, Sol. Energy Mater. Sol. Cells 115, 151 (2013).

7.6 References 143

4 H. Schmidt, H. Flügge, T. Winkler, T. Bülow, T. Riedl, and W. Kowalsky, Appl. Phys. Lett. 94, 243302 (2009).

5 Y. Tak, K. Kim, H. Park, K. Lee, and J. Lee, Thin Solid Films 411, 12 (2002).

6 H. Kim, J.S. Horwitz, G.P. Kushto, Z.H. Kafafi, and D.B. Chrisey, Appl. Phys. Lett. 79, 284 (2001).

7 B.H. Lee, I.G. Kim, S.W. Cho, and S. Lee, Thin Solid Films 302, 25 (1997).

8 U. Betz, M. Kharrazi Olsson, J. Marthy, M.F. Escolá, and F. Atamny, Surf. Coatings Technol. 200, 5751 (2006).

9 T. Sasabayashi, N. Ito, E. Nishimura, M. Kon, P.K. Song, K. Utsumi, a Kaijo, and Y. Shigesato, Thin Solid Films 445, 219 (2003).

10 C. Wang, V. Cimalla, G. Cherkashinin, H. Romanus, M. Ali, and O. Ambacher, Thin Solid Films 515, 2921 (2007).


12 H.B. Profijt, S.E. Potts, M.C.M. van de Sanden, and W.M.M. Kessels, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 29, 050801 (2011).

13 W.J. Maeng, D. Choi, J. Park, and J. Park, J. Alloys Compd. 649, 216 (2015).

14 T. Asikainen, M. Ritala, and M. Leskelä, J. Electrochemical Soc. 141, 3210 (1994).

15 D.-J. Lee, J.-Y. Kwon, J. Il Lee, and K.-B. Kim, J. Phys. Chem. C 115, 15384 (2011).

16 O. Nilsen, R. Balasundaraprabhu, E.V. Monakhov, N. Muthukumarasamy, H. Fjellvåg, and B.G. Svensson, Thin Solid Films 517, 6320 (2009).

17 J.W. Elam, A.B.F. Martinson, M.J. Pellin, and J.T. Hupp, Chem. Mater. 18, 3571 (2006).

18 J.A. Libera, J.N. Hryn, and J.W. Elam, Chem. Mater. 23, 2150 (2011).

19 S. Limpijumnong, P. Reunchan, A. Janotti, and C. Van de Walle, Phys. Rev. B 80, 193202 (2009).

20 P. King, R. Lichti, Y. Celebi, J. Gil, R. Vilão, H. Alberto, J. Piroto Duarte, D. Payne, R. Egdell, I. McKenzie, C. McConville, S. Cox, and T. Veal, Phys. Rev. B 80, 081201 (2009).


21 B. Macco, H.C.M. Knoops, and W.M.M. Kessels, ACS Appl. Mater. Interfaces 7, 16723 (2015).

22 D. Vanhemel, Atomic Layer Deposition of In2O3: Growth, Morphology and Hydrogen Doping, Master thesis, Eindhoven University of Technology, The Netherlands, 2015.

23 D.J. Larson, D.T. Foord, T.C. Anthony, I.M. Rozdilsky, A. Cerezo, and G.W.D. Smith, Ultramicroscopy 75, 147 (1998).

24 K. Ozasa, T. Ye, and Y. Aoyagi, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 12, 120 (1994).

25 M. Gebhard, M. Hellwig, H. Parala, K. Xu, M. Winter, and A. Devi, Dalton Trans. 43, 937 (2014).

26 M. Ritala, T. Asikainen, and M. Leskelä, Electrochem. Solid-State Lett. 1, 156 (1998).


28 D.C. Paine, T. Whitson, D. Janiac, R. Beresford, C.O. Yang, and B. Lewis, J. Appl. Phys. 85, 8445 (1999).

29 B. Macco, M.A. Verheijen, L.E. Black, B. Barcones, and W.M.M. Kessels, ACS Appl. Mater. Interfaces, submitted (2016).

Chapter 8

Amorphous-to-crystalline transition of In2O3 thin films during atomic layer deposition*

Abstract

It has been reported that In2O3 thin films can be prepared by atomic layer deposition (ALD) using InCp as indium precursor and H2O/O2 as co-reactants in a wide range of substrate temperatures (100 °C to 350 °C). The films prepared at 100 °C are mostly amorphous, while the films prepared at 150 °C are virtually fully crystalline. This chapter investigates the amorphous-to-crystalline transition of In2O3 thin films during ALD, and is split into two main parts related to: 1) the formation of crystalline nuclei and 2) the subsequent crystalline growth of these nuclei. The first part of the study focuses on the effects of substrate material and substrate temperature on the formation of nuclei. A higher substrate temperature can result in increased areal grain densities, as observed from top-view bright field transmission electron microscopy. In the second part of the study, the roles of the thermal budget and film thickness on the growth of nuclei will be addressed. The formation of nuclei appears to occur on the film surface, in the bulk of the films as well as the interface with substrates, the probabilities of these processes depending on substrate nature, deposition temperature and growth time.

* Y. Wu, D. Vanhemel, B. Macco, M.A. Verheijen, F. Roozeboom, W.M.M. Kessels, (in

preparation for publication).

148 Amorphous-to-crystalline transition of In2O3

8.1 Introduction

As reported in Chapter 7, In2O3 films can be prepared by atomic layer deposition (ALD) using recipes with InCp as indium precursor and a combination of H2O and O2 as co-reactants. Such recipes yield a fairly high growth per cycle in a wide range of substrate temperatures (100 °C to 350 °C). The films prepared at 100 °C are mostly amorphous, while the films prepared at 150 °C are virtually fully crystalline. Both amorphous and crystalline In2O3 films find their applications. As introduced in Chapter 3, the amorphous phase is favored for applications of the films as amorphous oxide semiconductors (AOSs), since the absence of grain boundaries can provide adequate short-range uniformity and surface smoothness,1 while the carrier mobility is also not affected by grain boundary scattering.2 In contrast, the crystalline phase is favored for applications of the films as transparent conducting oxides (TCOs). Crystalline In2O3 can yield a relatively high carrier mobility (68 cm2/Vs), which mitigates the requirement of a high carrier density while still yielding a high conductivity. In case of a relatively low carrier density, the optical losses in the infrared due to free carrier effects can be small. Interestingly, for In2O3, the change in crystallinity induced by increasing the substrate temperature from 100 °C to 150 °C, is accompanied by a significant decrease in carrier density and an increase of carrier mobility, implying a correlation between the crystallinity and the electrical properties. The relatively low crystallization temperature implies that the thermal budget of processing of devices containing an In2O3 layer should be carefully chosen. Thus, it is of fundamental interest to understand the details of the amorphous-to-crystalline transition of In2O3 thin films. Such study can potentially be helpful when trying to further tune the electrical properties of In2O3 films by controlling the crystallinity and the microstructure.

According to literature, the amorphous-to-crystalline transition of (tin-doped) In2O3 films can be controlled during film growth3 or during post-annealing4,5 by the thermal budget, i.e. the total of thermal energy provided over a certain time. The transition consists of two stages: the formation of crystalline nuclei and their subsequent growth.4,6,7 In case of post-annealing of amorphous films, the formation of nuclei in the bulk of the material is composed of a gradual structural relaxation and a subsequent local transition from a relaxed amorphous state into a crystalline state.4,6 During the structural relaxation, the

8.1 Introduction 149

realignment of In-O bonds occurs. This creates a locally ordered structure with a volume smaller than ~ 1 nm. The growth of nuclei within a bulk amorphous matrix is a progressive displacement of the interface between the crystalline grains and the amorphous matrix.7

Alternatively, nuclei can already be formed on the In2O3 film surface during growth, as reported in the work of Macco et al.,8 where In2O3 films with a thickness of 75 nm were prepared using the same ALD recipe on Al2O3 substrates at a substrate temperature of 100 °C and annealed at a range of annealing temperatures. In their work, crystalline nuclei in the near surface region of an amorphous matrix were observed by cross-sectional TEM images, top-view scanning electron microscope (SEM) images, and atomic force microscopy (AFM) images. On the film surface, surface diffusion processes9 can facilitate the formation of nuclei, and may result in a lower activation energy than in the bulk film where significant rearrangements are needed for nucleation.

A third region for nucleation is at the interface with the substrate. The impact of the substrate surface on the nucleation is not yet fully understood, and in the case of In2O3 not yet even specifically addressed in literature. Nucleation during thin film growth might be affected by a range of substrate surface parameters, such as surface energy (or interface energy), density of impurities and chemical reactivity. For example, it has been reported that in the case of Si/SiO2 multilayers, the interface tension between the substrate and the thin film can be different when different substrate materials are used, and that the interface tension can significantly influence the wetting of the film and hence the formation of nuclei.10

This chapter aims at a better understanding of the amorphous-to-crystalline transition of In2O3 thin films during ALD growth and it builds on the work of Macco et al. Transmission electron microscopy (TEM) was used as the analysis method. Areal grain densities and grain sizes have been determined from top-view bright field transmission electron microscopy (BFTEM) images, and insight into the crystallinity was extracted from the corresponding selected area electron diffraction (SAED) patterns. The study of the amorphous-to-crystalline transition consists of two parts: the formation of crystalline nuclei and the subsequent growth. In the first part, it will be shown that the substrate material can significantly influence the grain density, and that a higher


substrate temperature can lead to a higher areal grain density. In the second part of the results, the roles of thermal budget and film thickness will be addressed. A high thermal budget can facilitate the solid-phase crystallization and the formation of new nuclei. In parallel, during growth additional nuclei can be formed on the film surface, leading to a higher areal density for thicker films. At the end of the second section, a comparison between the results from Macco et al. and our data is made and a schematic representation is proposed to summarize the crystallization process of In2O3 films prepared by ALD.


The In2O3 films were deposited in an Oxford Instruments OpALTM ALD reactor. InCp was used as indium precursor, and a combination of both H2O and O2 was used as co-reactant (for more details see Chapter 7). Silicon nitride (Si3N4) TEM window grids were used as substrates, either bare or covered by SiO2 or Al2O3. Both the SiO2 and the Al2O3 layers were prepared by ALD and had a film thickness of 3-4 nm. TEM studies were performed using a JEOL JEM ARM 200F transmission electron microscope. The areal grain density was obtained by directly counting the amount of grains from the top-view BFTEM images.


8.3.1 The formation of nuclei

a) The effect of substrate material

It is important to note that the substrate materials can affect the formation of nuclei in the In2O3 films. This is demonstrated in the top-view TEM images in Fig. 8.1, which show the In2O3 films prepared at 150 °C on the substrates of Si3N4, Al2O3 and SiO2. The film prepared on the Al2O3 substrates shows larger grain sizes and a smaller grain density than those prepared on the Si3N4 and SiO2 substrates. As discussed above, the impact of substrate materials on the nucleation is not yet fully understood. The present results clearly show that the nucleation probability is different for the three substrate materials used, suggesting that at least for the samples with a high grain density interface nucleation contributes to the crystallization of the film. In many applications, the nature of the underlying layer on which the In2O3 has to be deposited is


governed by the specifications of the device. Below, we will focus on substrate independent parameters that allow for tuning the crystalline morphology, while keeping the substrate constant, i.e. SiO2. In the final discussion, we compare our new results with those obtained by Macco et al. for deposition on Al2O3.

Figure 8.1 Top-view TEM images of In2O3 films with a thickness of 40 nm prepared at 150 °C on various substrate materials: (a) Si3N4, (b) Al2O3 and

(c) SiO2.

b) The effect of substrate temperature

To study the effect of the deposition temperature on the nucleation probability, In2O3 films with a thickness of 40 nm were grown on SiO2 substrates at a range of substrate temperatures. BFTEM top view images of these samples are shown in Fig. 7.8. The areal grain density increases with increasing substrate temperature, as shown in Fig. 8.2. This increase can be attributed to a larger thermal budget due to a higher substrate temperature, since the formation of nuclei is a thermally activated process.4 The activation energies for the formation of nuclei has been reported to be 1.3±0.2 eV in the case of post-deposition annealing of Sn-doped In2O3 films prepared on glass substrates.4 This activation energy was determined from a so-called Arrhenius plot, presenting the relation between the crystallization rate and the substrate temperature. From the data presented in Fig. 7.8, the activation energy for the nucleation cannot be determined, since the films were already crystallizing during growth. Post-deposition imaging of fully crystallized films does not allow for determining the nucleation frequency per unit time. Moreover, these experiments do not reveal whether crystallization of these films is a bulk phenomenon or rather occurs at the substrate interface or the top surface


during growth. Below, we will describe dedicated experiments to address this aspect.

Figure 8.2 Areal grain density as a function of substrate temperature. The In2O3 films have a film thickness of ~ 40 nm, and were prepared on SiO2

covered TEM windows at various substrate temperatures. The areal grain

densities were counted from top-view BFTEM images shown in Fig. 7.8.

8.3.2 The growth of nuclei

In this section the effects of two parameters on the growth of nuclei will be discussed: the thermal budget applied during the deposition and the thickness of the In2O3 films. The In2O3 films for this study were all prepared on SiO2 substrates at a substrate temperature of 150 °C. At a fixed substrate temperature, the thermal budget applied during the deposition is directly related to the total time that a sample resides in the deposition chamber. It is essential to realize that the effect of thermal budget interferes to a certain extent with the study of the effect of film thickness, because a longer deposition time is required for the deposition of thicker films, thus involving a larger thermal budget. Therefore, two comparative series of samples were made, having equal thermal budgets, the one series being realized by continuous film growth and the other one by a combination of a short growth time and varying anneal times.


In the annealing series, the effect of annealing was studied. As shown in Fig. 8.3, samples with a thickness of 10 nm were prepared within a deposition time of 0.5 hour, and then kept in the deposition chamber at the same substrate temperature for different annealing times. The annealing times of samples A, B, C and D were 0, 0.5, 1.5 and 17.5 hours, respectively. Thus, with an additional 0.5 hour of the deposition time, the total times that sample A, B, C and D resided in the deposition chamber were 0.5, 1, 2 and 18 hours, respectively. The long annealing time of sample D was chosen to represent the maximum impact of thermal budget on grain growth.

In the thickness series, the effect of film thickness was studied , wherein samples A, B’ and C’ with thicknesses of 10, 20 and 40 nm, respectively were prepared. As shown in Fig. 8.3, the deposition times of sample B’ and C’ (1 and 2 hours, respectively) are therefore equivalent to the times that sample B and C reside in the chamber. In this way, the differences in microstructures between sample B and B’ and between C and C’ can be attributed to the effect of film thickness, i.e. nucleation during growth rather than post-growth nucleation.

Figure 8.3 Schematic illustration showing the growth and annealing time

and the film thickness of the samples in the annealing series and the

thickness series. The annealing series was prepared with a constant film thickness of 10 nm, and right after, annealed for various times in the

deposition chamber. The thickness series was prepared in a series with

variable deposition times, giving rise to different film thicknesses, without subsequent annealing.


a) The effect of thermal budget

As shown in Fig. 8.4 (a), crystalline nuclei in an amorphous matrix were observed in sample A, which was not annealed. In the SAED pattern, diffuse rings representing the presence of an amorphous phase can be observed. After 0.5 hour of annealing, the grains become larger, and cover a larger area in the top-view image (Sample B, Fig. 8.4 (b)), indicating the growth of the grains due to the thermal budget provided by annealing. By comparing the top-view images in Figs. 8.4 (a) and (b), the rate of the grain growth is estimated to be ~ 1 nm/min, which is smaller than the value of 2 nm/min reported in the work of Macco et al. The lower rate in our case may be attributed to interface strain, reducing the growth rate. This effect is expected to be more pronounced in case of a lower film thickness (10 nm in our case compared to 75 nm in the aforementioned work of Macco et al.). Meanwhile, diffuse rings can still be observed in the SAED pattern of sample B, indicating that this film is not fully crystalline yet. After 1.5 hours of annealing, the top-view image (Sample C, Fig. 8.4 (c)) is fully covered by crystalline grains, and diffuse rings can no longer be distinguished in the SAED pattern, indicating a fully crystalline film. Sample D, which was annealed for 17.5 hours, shows a similar top-view image and a similar SAED pattern (Fig. 8.4 (d)) as the sample annealed for 1.5 hours, indicating that no significant grain coarsening occurs during prolonged annealing.

The areal grain densities determined from the top-view images are plotted as a function of annealing time in Fig. 8.5. Most of the nuclei are formed during the first 0.5 hour of deposition. During annealing, the areal grain density increases, indicating the formation of new nuclei. The areal grain density saturates at a value of ~400 µm-2 after around 1.5 hours of annealing. This saturation corresponds to the situation that the In2O3 films finally become fully crystalline, leaving no more amorphous matrix for the formation of new nuclei.


Figure 8.4 Top-view bright field transmission electron microscopy images and the corresponding selected area electron diffraction (SAED) patterns of the annealing series schematically illustrated in Fig. 8.3: (a) Sample A,

(b) Sample B, (C) Sample C and (D) Sample D. These samples were prepared on SiO2 covered TEM windows at a substrate temperature of

150 °C for 0.5 hour, yielding a film thickness of 10 nm. Right after the deposition, Sample A, B, C and D have been annealed in the deposition

chamber for 0, 0.5, 1.5 and 17.5 hours, respectively.

Figure 8.5 Areal grain density as a function of the residence time of the

samples in the deposition chamber, i.e. deposition time plus annealing time. The areal grain densities were counted from the corresponding top-

view BFTEM images of the thickness series and the annealing series. Sample labels are given for the data points, whereas the colors of the

data points correspond to the illustration in Fig. 8.3.


b) The effect of film thickness

The top-view BFTEM images and the SAED patterns of the thickness series are shown in Figs. 8.6 (a) to (c). Diffuse rings can still be observed in the SAED pattern of sample B’ (Fig. 8.6 (b)), but not in that of sample C’ (Fig. 8.6 (c)). i.e. similar to sample B, sample B’ is composed of crystalline grains in an amorphous matrix; while similar to sample C, sample C’ is fully crystalline. Such similarities indicate that during additional film growth, the solid-phase crystallization takes places, i.e. a longer deposition time results in a higher thermal budget and hence a higher degree of crystallinity.

Figure 8.6 Top-view BFTEM images and the corresponding SAED patterns

of the thickness series indicated in Fig. 8.3: (a) Sample A, (b) Sample B’ and (C) Sample C’. These samples were prepared on SiO2 covered TEM

windows at a substrate temperature of 150 °C. The thicknesses of Sample

A, B’ and C’ are 10, 20 and 40 nm, which correspond to the deposition times of 0.5, 1 and 2 hours, respectively.

As shown in Fig. 8.5, the areal grain density increases with increasing film thickness. This increase can again be partially attributed to the effect of thermal budget, i.e. additional nuclei can form due to a longer time that the samples reside in the deposition chamber. The areal grain density of Sample C’ is ~570 µm-2, which is higher than the value of Sample C (~392 µm-2). Such a difference indicates that in addition to the effect of the thermal budget, the

8.4 Summary and discussion 157

kinetics at the growth front during thickness increase of the film is also important in terms of formation of additional nuclei. It appears that the formation of crystalline nuclei is easier during film growth than during annealing which is possibly due to the surface rearrangement.11

8.4 Summary and discussion

For a thorough understanding of the nucleation mechanisms playing a role in the series discussed in this paper, it is essential to compare the results with those of Macco et al. In that study, 75 nm In2O3 layers were deposited at 100 °C on Al2O3 substrates. These layers were characterized by: i) Amorphous growth of the film up to several tens of nm’s thickness (nucleation at the Al2O3 surface appeared to be negligible); ii) Formation of nuclei during further film growth, yielding a final grain density of ~6 µm-2; iii) Grain growth without further nucleation upon annealing at temperatures between 150 °C and 200 °C.

Our series of samples differ in two ways from those of the study of Macco et

al.: a higher deposition temperature of 150 °C was used and SiO2 was used as an alternative substrate material. A higher substrate temperature can result in a higher areal grain density. This phenomenon has been presented above (Fig. 8.2) for SiO2 substrates as is also observed for Al2O3 substrates: an areal grain density of ~25 µm-2 is present at the substrate temperature of 150 °C (Fig. 8.1 (b)), compared to the value of ~6.4 µm-2 at 100 °C (see Macco et al.). The substrate material appears to have a much larger effect. For the same deposition temperature of 150 °C, densities of several hundreds µm-2 were measured on SiO2 substrates (Fig. 8.1 (c)), compared to the value of ~25 µm-2 on the Al2O3 substrate (Fig. 8.1 (b)). This implies that at least part of the nucleation occurs at the interface to the substrate.

In order to be able to discuss the other contributions to nucleation (i.e. bulk and surface nucleation), two aspects need to be addressed.

Firstly, it is interesting to observe that for our 10 nm layers the areal grain density increases during annealing (Fig. 8.5). Contrary, the areal grain density stays relatively constant during post-annealing of as-deposited amorphous films in the aforementioned work of Macco et al. Here, we propose two possible explanations for the different trends in areal grain density with annealing in these two cases: 1) Some nuclei might already exist in the sample


without annealing in our case, but are too small to be discernable in the BFTEM images acquired (e.g. Fig. 8.4 (a)). During annealing, these small nuclei can grow, and become large enough to be recognized from the TEM images, resulting in an apparent increase of areal grain density. 2) The films were prepared at different substrate temperatures, i.e. 150 °C in the present case and 100 °C in the case of Macco et al. Even though the nucleation starts in an amorphous matrix in both cases, the extent of structural disorder of the amorphous phase and hence the number of available nucleation sites can be different, which for example has been observed in the crystallization of amorphous silicon.12 The film prepared at 150 °C in the present case may have more available nucleation sites, resulting in the formation of new nuclei during annealing in the bulk part of the film. From the present studies, we cannot confirm or exclude either of the two explanations.

Secondly, as discussed before, by comparing the areal grain densities of the annealing and thickness series in Fig. 8.5, we can conclude that nuclei are formed during film growth, most likely occurring on the film surface. The formation of nuclei on the surface was also evidenced in the work of Macco et

al. Thus, this process appears to be important both at 100 °C as well as at 150 °C.

As a summary, Fig. 8.7 schematically represents the formation and growth of nuclei in ALD-grown In2O3 films on SiO2 and Al2O3. At the early stage of film growth, the formation of nuclei occurs on the interface with the substrates, in the bulk of the film or on the film surface, as indicated by the colors of green, red and yellow in Fig. 8.7, respectively. If the film is prepared at 150 °C on an Al2O3 substrate, as the case of Fig. 8.1 (b), the formation of nuclei on the interface with the substrate has a low probability. If the film is deposited at 100 °C on an Al2O3 substrate, as reported in the work of Macco et al., the formation of nuclei occurs on the film surface only. After the early stage of film growth, the following can take place depending on the process.

If the film growth is interrupted, and the film is subsequently kept at the same substrate temperature for annealing, the thermal energy leads to the further growth of already existing nuclei by solid-phase crystallization (dashed areas in Fig. 8.7). In the meantime, new nuclei can form in the bulk of the film by solid-phase crystallization. Once the annealing time becomes sufficiently long, i.e. sufficient thermal budget is provided, the film becomes fully crystalline.

8.5 References 159

Alternatively, if the deposition process continues, the same processes of solid-phase crystallization and the formation of new nuclei also occur. Moreover, the existing crystals will also grow upwards at the same pace as the increase in film thickness during film growth. Also, the formation of additional nuclei can take place, probably mostly on the film surface. As a result, with continued growth, the film ends up with a similar microstructure as a film after a long annealing step, but with a higher final grain density.

Figure 8.7 Schematic representation illustrating the formation and

growth of nuclei during deposition or annealing of an In2O3 film. The grey

areas represent the amorphous matrix. The colored areas represent the crystalline grains that are formed in different regions: (orange) on the

film surface; (red) solid-phase crystallization in the bulk of the films; (green) on the interface with the substrate. The dashed, colored areas represent the growth of grains via a progressive displacement of the

amorphous/crystalline interface. In the case of a film prepared on a SiO2 substrate at a substrate temperature of 150 °C, nuclei can form in all the

three regions; in the case of a film prepared on an Al2O3 substrate at 150 °C, nuclei are likely to form in the bulk by solid-phase crystallization

and on the film surface; in the case of a film prepared on an Al2O3

substrate at 100 °C, only the formation of nuclei on the film surface can be observed, as reported by Macco et al.

8


8.5 References

1 T. Kamiya and H. Hosono, NPG Asia Mater. 2, 15 (2010).

2 E. Fortunato, P. Barquinha, and R. Martins, Adv. Mater. 24, 2945 (2012).

3 S. Muranaka, Jpn. J. Appl. Phys. 2062, L 2062 (1991).

4 D.C. Paine, T. Whitson, D. Janiac, R. Beresford, C.O. Yang, and B. Lewis, J. Appl. Phys. 85, 8445 (1999).

5 B. Macco, Y. Wu, D. Vanhemel, and W.M.M. Kessels, Phys. Status Solidi - Rapid Res. Lett. 8, 987 (2014).

6 A.S.A.C. Diniz and C.J. Kiely, Renew. Energy 29, 2037 (2004).

7 I.A. Rauf and L.M. Brown, Acta Metall. Mater. 42, 57 (1994).

8 B. Macco, M.A. Verheijen, L.E. Black, B. Barcones, and W.M.M. Kessels, ACS Appl. Mater. Interfaces submitted (2016).

9 M. Knez, Semicond. Sci. Technol. 27, 074001 (2012).

10 M. Zacharias and P. Streitenberger, Phys. Rev. B - Condens. Matter Mater. Phys. 62, 8391 (2000).

11 H. Morikawa, H. Sumi, and M. Kohyama, Thin Solid Films 281-282, 202 (1996).

12 J.N. Lee, B.J. Lee, D.G. Moon, and B.T. Ahn, Jpn. J. Appl. Phys. 36, 6862 (1997).

Chapter 9

Concluding remarks and outlook

Both ZnO and In2O3 are popular semiconducting metal oxides, and find many applications as transparent conducting oxides (TCO) in solar cells, amorphous oxide semiconductors (AOS) in flat panel display and active layers in gas sensors. In this dissertation work, Al-doped ZnO and In2O3 have been selected as two representative cases of semiconducting metal oxides to investigate the properties of these materials prepared by atomic layer deposition (ALD). ALD is an emerging technique becoming highly popular because of its potential to produce ultrathin films with superior material quality and precisely tunable layer thickness and composition control over large substrate areas.

The focus of this study was in particular on the correlation between the film composition, electrical properties and microstructure. We demonstrated that the film properties can be modified, improved and even tailored during ALD growth. From this work, several conclusions and future perspectives can be drawn.

Al-doped ZnO (ZnO:Al) is generally considered as the most promising alternative for Sn-doped indium oxide (ITO). The main function of the Al doping is to improve the electrical conductivity of ZnO thin films. In this respect, the supercycle growth mode of ALD is a powerful method to precisely control the dopant content. By varying the ratio of the ZnO and Al2O3 ALD cycles, a series of ZnO films with different Al contents were prepared, and an optimized doping level with respect to the minimum film resistivity was obtained. Yet, we found that the supercycle growth mode still suffers from poor levels of Al doping efficiency (<10 %), whereas the fundamental understanding necessary to improve for the low doping efficiency remained limited. For example, the Al distribution was often simply presented as atomically flat δ-doping layers in the literature. However, our atom probe tomography (APT) study revealed a broadening of the dopant layers rather than a δ doping distribution. These findings of the Al distribution led us to propose two factors to explain the low

164 Concluding remarks and outlook

doping efficiency: the solid solubility limit and the disorder-induced carrier localization. Therefore, one of the directions towards a better doping efficiency is to reduce the inhomogeneity of the dopant distribution. Note that, the Al distribution in the growth direction can be controlled by tuning the cycle ratio, while the number of incorporated Al atoms during individual doping cycles is determined by the surface reactions. Therefore, in addition to the tuning of the cycle ratio, one very effective way to realize a more refined control of dopant distribution is by directly modifying the surface reactions during the doping cycle. Our work demonstrated that by using a bulkier Al precursor, steric hindrance can give rise to a reduced amount of incorporated Al dopants and hence, to a significantly increase of the doping efficiency. To visualize the effect of steric hindrance on the doping efficiency, we recommend to measure and compare the dopant distributions when using different dopant precursors by using APT, transmission electron microscopy (TEM), etc. Besides the steric hindrance of the dopant precursor proposed in this work, the inhomogeneity of dopant profiles can also be reduced by manipulating the structure of the doping cycle. For example, multistep doping cycles, consisting of multiple ALD half-reactions, have been proposed to reduce the amount of incorporated dopants during the doping cycles.

In2O3 exhibits excellent properties (e.g. high carrier mobility (100 cm2/Vs)), and is still widely used for electronics, despite the scarcity of indium element. Moreover, very recently H-doped In2O3 was demonstrated to exhibit higher carrier mobility than conventional Sn-doped In2O3, thus reaffirming the potential of (H-doped) In2O3. Yet, ALD of In2O3 is still at an infancy stage: so far most research efforts have been dedicated to the optimization of the growth recipes. In this respect, in this work first a reliable recipe for ALD In2O3 was established based on cyclopentadienyl indium and H2O/O2, yielding high values of growth per cycle and high film conductivity in a wide substrate temperature window. This recipe can potentially become a widely accepted recipe for the ALD growth of In2O3. Interestingly, by varying the substrate temperatures, In2O3 films with different degrees of crystallinity (from mostly amorphous to fully crystalline) can be obtained. This outcome opens up the possibility to carry out a more detailed study of the crystallization process of In2O3. We found that the control of the microstructure can be realized by varying the growth conditions, such as deposition time, substrate material and temperature. With the change of the crystallinity, changes in the electrical

165

properties of the In2O3 films were also observed which can be used for further optimization. For example, the carrier density is higher in the amorphous phase than in the crystalline phase. Therefore, control of the microstructure of In2O3 films can potentially be utilized to control other film properties. For example, by tuning the film crystallinity, one can influence the grain morphology and grain boundary structure, and thus the carrier mobility as this mobility is influenced by charge scattering at grain boundaries.

It is important to note that hydrogen can be unintentionally doped into In2O3 thin films during the growth. The incorporation of hydrogen atoms can be beneficial in terms of improving the film conductivity, since in the literature hydrogen is proposed to be an electron donor in crystalline In2O3. To understand the origin of hydrogen in the ALD In2O3 films, we used hydrogen isotope labeling experiments to differentiate hydrogen incorporation from the different possible sources. Elastic recoil detection (ERD) revealed that the hydrogen atoms mainly originate from the co-reactant H2O. In addition, the distributions of the hydrogen isotopes were studied by atom probe tomography (APT). By correlating the distributions with the microstructure obtained by transmission electron microscopy (TEM), it could be concluded that the hydrogen content is higher in the amorphous phase than in the crystalline phase. Currently, it is not yet fully clear what the main electron donor in In2O3 is, oxygen vacancies (VO

2+) or hydrogen impurities (interstitial Hi

+ and substitutional HO+), as no strong correlation between the hydrogen

content and the carrier density was observed. However, our early results on the origin and distribution of hydrogen can be helpful for more detailed studies on the hydrogen doping of In2O3 films. We propose that investigations of the chemical environments of H atoms in In2O3 may reveal how efficiently H atoms can act as donors. For example, infrared spectroscopy can be used to investigate the bonding states of hydrogen in In2O3. In addition, it is also important to note that compared to the case of intentional doping, such as Al doping in ZnO, the hydrogen content in In2O3 remains more challenging to control, and may be unstable when the films are exposed to the ambient condition. Therefore, a further study on the control of hydrogen content is of fundamental importance, and can be used to control the electrical properties of In2O3 films.

166 Concluding remarks and outlook

Based on the two study cases of ZnO:Al and In2O3, it can be concluded that the microstructure, the composition and the electrical properties of these metal oxide thin films are strongly correlated. These correlations enable us to control one property by modifying the others. Especially, doping can have a strong impact on the electrical properties of the films. The understanding of the ALD growth and the control of film properties can be further extended to multicomponent metal oxides, such as the tertiary oxide In-Sn-Zn-O system for TCOs and the In-Ga-Zn-O system for AOSs. For example, the insights of the supercycle growth mode can be applied to control the composition and distribution of constituents in such tertiary oxides, while the understanding of hydrogen doping can be used to further enhance the conductivity of other TCOs.

Summary

Growth, phase and doping control in ZnO and In2O3 thin films prepared by atomic layer

deposition

Electronic devices are playing an increasingly important role in our society. The further development of modern electronics mainly takes place into two directions: miniaturization and diversification. Such a development will require advanced materials with novel or superior properties beyond those of the conventional materials. Semiconducting metal oxides such as In2O3 and ZnO represent an important class of materials showing sufficient transparency in the visible part of the electromagnetic spectrum, and simultaneously showing semiconductor behavior. They find many applications as thin films, for example, as amorphous oxide semiconductors (AOS) in displays, as transparent conducting oxides (TCOs) in solar cells and display devices and as active layers in gas sensors. In order to fulfil current and future requirements set by these applications, there is a growing need for accurate tunability of the properties of these films (e.g. transparency, conductivity and morphology) during thin film preparation.

Atomic layer deposition (ALD) is an emerging technique in commercial thin-film technology and is becoming a powerful tool to prepare thin films for future electronics. The self-limiting nature of the surface reactions during ALD growth allows for the preparation of thin films with a high uniformity, excellent conformality and atomic-precision control of the film thickness over large substrate areas. Especially, the so-called supercycle growth mode enables the precise control of the composition and distribution of constituents in multicomponent metal oxides.

This dissertation work was aimed at manipulating the properties of semiconducting metal oxide thin films during ALD through the understanding

170 Summary

of the film properties at an atomic scale. In an attempt to obtain a generic understanding of the underlying growth mechanisms, we selected In2O3 and ZnO as representative cases. These materials are both important components in multi-component oxides for AOS and TCO applications, and are often used as active layers of gas sensors. H-doping in In2O3 and Al-doping in ZnO represent the cases of unintentional and intentional doping, respectively.

Al-doped ZnO (ZnO:Al, AZO) thin films can be prepared using the so-called supercycle growth mode: a wide doping range could be obtained by varying the ratio between ZnO and Al2O3 ALD cycles. The structural and electrical properties of the ZnO:Al films were investigated to obtain insight into the doping and electrical transport mechanisms in the films. An optimum doping level with respect to the minimum film resistivity was obtained, while the doping efficiency was found to be low (<10 %). The doping profile measured with X-ray photoelectron spectroscopy (XPS) revealed that the Al distribution in the films is not homogeneous, but, instead, concentrated around certain film depths. Cross-sectional transmission electron microscopy (TEM) imaging shows that the AlOx layers can interrupt the growth of ZnO:Al crystal grains. With the increasing doping levels, the carrier mobility decreases partially due to the scattering from the AlOx layers. Therefore, a further reduction of the film resistivity, requires a higher doping efficiency, and investigations of the factors limiting the low doping efficiency can be helpful.

Our subsequent study demonstrated that the maximum doping efficiency can be improved from ~10 % to ~ 60 % by using dimethylaluminum isopropoxide (DMAI, Al(CH3)2(OiPr)) as an alternative Al precursor instead of the conventionally-used trimethylaluminum (TMA, Al(CH3)3). The isopropoxide ligand is larger than the methyl ligands. Consequently the steric hindrance caused by the bulkier DMAI precursor leads to fewer incorporated Al atoms per doping cycle. As a result, the inhomogeneity of the Al distribution is reduced, which improves the film conductivity.

To investigate the factors limiting the doping efficiency in more detail, the three-dimensional (3D) distribution of Al dopants in ZnO:Al films was measured at the atomic scale using the combination of high-resolution TEM and atom probe tomography (APT). In the literature, the dopant distribution in ZnO:Al is often described as atomically abrupt δ doping layers. Our study revealed that actually a broadening of the Al-rich layers is observed, which at high doping

171

levels results in the overlap of the individual Al-rich layers and consequently, a high local Al density. Based on the 3D dopant distribution, two liming factors have been proposed to explain the low doping efficiency: the solubility limit of Al in the ZnO lattice and the disorder-induced carrier localization. The latter means that electrons are weakly localized when scattered by the disorder induced by impurities.

Regarding the ALD growth of In2O3, recent studies mainly focused on testing the appropriated indium precursors. Our work demonstrated that cyclopentadienyl indium (InCp), combined with H2O and O2 can yield high values for the growth per cycle in a wide range of substrate temperatures (100 °C to 350 °C). Film resistivity values lower than 1 mΩ∙cm could be obtained at substrate temperatures lower than 200 °C. The degree of crystallinity increases from mostly amorphous to fully crystalline with substrate temperature increasing from 100 °C to 150 °C. Higher temperatures (>150 °C) result in the incorporation of carbon impurities in the films, which can further lead to reduced crystallinity and carrier mobility. Since hydrogen can be unintentionally doped in In2O3 films, and may act electron donors in In2O3 films as reported in the literature, the origin and distribution of hydrogen in In2O3 films were investigated. Using isotopic labeling, we concluded that the hydrogen mainly originates from the co-reactant H2O. Using elastic recoil detection (ERD) and APT, the distribution of hydrogen in the In2O3 films was studied. From the combined results on the hydrogen distribution and the microstructure, we concluded that more hydrogen atoms are incorporated into the ALD grown In2O3 films in the amorphous phase than those in the crystalline phase.

On the basis of the observed differences in the crystallinity of the In2O3 films obtained at different substrate temperatures the amorphous-to-crystalline transition of In2O3 films was investigated. This transition consists of two parts: the early formation of the crystalline nuclei and their subsequent growth. In the first part of the study, the effects of substrate material (Al2O3, SiO2 and Si3N4) and substrate temperature (100 °C to 350 °C) on the formation of nuclei was addressed. Top-view bright-field TEM images revealed that the areal grain density increases at higher substrate temperatures. This can be explained by the fact that the formation of nuclei is thermally activated. In the second part, the effects of the thermal budget and film thickness on the growth of nuclei

172 Summary

were addressed. It was observed that both effects can lead to the formation of additional nuclei.

In conclusion, using Al-doped ZnO and H-doped In2O3 as representative study cases, extensive knowledge has been obtained related to the control of the film properties of semiconducting metal oxides. This knowledge can be applied to the further improvement of these films and it can also be extended to other, related multi-component films such as In-Ga-Zn-O for AOS and In-Sn-Zn-O for TCO applications.

List of publications related to this work

1. Y. Wu, P.M. Hermkens, B.W.H. van de Loo, H.C.M. Knoops, S.E. Potts, M.A. Verheijen, F. Roozeboom and W.M.M. Kessels, ‘Electrical transport and Al

doping efficiency in nanoscale ZnO films prepared by atomic layer deposition’, J. Appl. Phys., 114 024308 (2013). Chapter 4 of this thesis.

2. Y. Wu, S.E. Potts, P.M. Hermkens, H.C.M. Knoops, F. Roozeboom, W.M.M. Kessels, ‘Enhanced doping efficiency of Al-doped ZnO by atomic layer

deposition using dimethylaluminum isopropoxide as an alternative aluminum

precursor’, Chem. Mater. 25 4619 (2013). Chapter 5 of this thesis.

3. Y. Wu, A.D. Giddings, M.A. Verheijen, T.J. Prosa, D.J. Larson, F. Roozeboom, W.M.M. Kessels, ‘On the factors limiting the doping efficiency in atomic layer

deposited ZnO:Al thin films: a dopant distribution study by atom probe

tomography’, (in preparation). Chapter 6 of this thesis.

4. Y. Wu, D. Vanhemel, B. Macco, S. Koelling, M.A. Verheijen, P.M. Koenraad, F. Roozeboom, W.M.M. Kessels, ‘Compositional, structural and electrical

properties of In2O3 thin films prepared by atomic layer deposition from InCp

and O2/H2O’, (in preparation). Chapter 7 of this thesis.

5. Y. Wu, D. Vanhemel, B. Macco, M.A. Verheijen, F. Roozeboom, W.M.M. Kessels, ‘Amorphous-to-crystalline transition of In2O3 thin films during atomic

layer deposition’, (in preparation). Chapter 8 of this thesis.

6. A. Illiberi, R. Scherpenborg, Y. Wu, F. Roozeboom, and P. Poodt, ‘Spatial

Atmospheric Atomic Layer Deposition of AlxZn1-xO’, ACS Applied Materials & Interfaces 5 (24), 13124 (2013).

7. D.J. Larson, A.D. Giddings, Y. Wu, M.A. Verheijen, T.J. Prosa, F. Roozeboom, K.P. Rice, W.M.M. Kessels, B. P. Geiser and T.F. Kelly, ‘Encapsulation method

for atom probe tomography analysis of nanoparticles’, Ultramicroscopy, 159, 420 (2015).

174

8. B. Macco, Y. Wu, D. Vanhemel, W.M.M. Kessels, ‘High mobility In2O3

transparent conductive oxides prepared by atomic layer deposition and solid

phase crystallization’, Phys. Status Solidi RRL, 8, 987 (2014).

9. Y. Kuang, B. Macco, C. K. Ande, P.C.P. Bronsveld, M. A. Verheijen, Y. Wu, W. M. M. Kessels, and R. E. I. Schropp, ‘Towards the implementation of atomic

layer deposited In2O3:H in silicon heterojunction solar cells’, (in preparation).

Dankwoord

Nu het schrijven van de hoofstukken van dit proefschrift erop zit, is eindelijk het rustpunt voor mij aangebroken om een dankwoord te schrijven. Omdat dit minder wetenschappelijke onderwerpen bevat weet ik ook zeker dat deze sectie het aantrekkelijkst zal zijn voor de meeste lezers. Daar gaan we dan.

Ik wil eerst mijn begeleider Erwin bedanken. Je bood me de kans om in onze PMP groep mijn promotieonderzoek te doen. Ik herinner me nog dat je je zeer open opstelde toen ik naar deze promotieplaats solliciteerde. Nu hoop ik dat je tevreden bent met mijn prestatie in de afgelopen jaren. Jouw kritische commentaar op mijn schrijven van manuscripten en hoofdstukken hebben me geholpen bij het verbeteren van mijn schrijfvaardigheid. Tegelijkertijd heb je mij geduldig en vriendelijk begeleid om mijn doelen stap voor stap te bereiken. Ik wil ook mijn andere begeleider Fred bedanken. Jij bent altijd bereid om mij te helpen wanneer het nodig is, zowel in werk-gerelateerde , als andere, meer persoonlijk zaken. Je bood me de geestelijke begeleiding, vooral in de moeilijkste periode van mijn onderzoek. Hiermee ben je niet alleen mijn begeleider, maar ook mijn betrouwbare vriend geworden. Ik wil ook mijn co-promotor Marcel bedanken. Ondanks dat je officieel maar een halve dag per week in onze PMP-groep werkt, besteedde je extra tijd om mijn TEM samples te meten en te analyseren. Bovendien bood je me veel hulp aan in de moeilijkste tijded.

Mijn volgende dankwoorden gaan naar de technici, in het bijzonder naar Cristian en Wytze. Cristian, je hielp mij om de problemen van het OpAL systeem op te lossen. ZnO and In2O3 warenweliswaar interessante materialen in mijn onderzoek, maar ze zijn vervelend voor jou. Ze vervuilen het systeem, en het kostte je veel tijd om het systeem telkens weer schoon te maken. Om voldoende energie voor het schoonmaken te verzamelen, ging je extra tijd besteden aan zwemmen om zo jouw lichaamconditie te onderhouden. Wytze, ik bedank je voor de uitvoerige uitleg van alle analyse-apparatuur. Toen wij in het TNO-gebouw gehuisvest waren, bezocht ik je vaak op kantoor. Onze besprekingen begonnen meestal met vragen over karakterisastie van dunne lagen, maar eindigden met vele andere interessante onderwerpen. Geleidelijk zijn we vrienden geworden. Ik genoot veel van de avondetentjes bij jou huis.

176 Dankwoord

Ik wil Patryk bedanken. Ondanks dat je geen mijn master student van mij was toen ik nog nieuw was in PMP, werkten we vaak samen aan ZnO. Buiten PMP gingen we vaak uit. Geleidelijk werd je een van mijn beste vrienden. Wat ik heel belangrijk vind is dat je op mijn persoonlijk leven veel invloed hebt gehad, dat zal ik altijd blijven waarderen. Ik bedank ook mijn studenten Wouter en Dries. We werkten samen aan de ontwikkeling van het In2O3 proces. Al snel werd dit proces zeer populair binnen PMP, en het was ook zeer belangrijk voor mijn PhD project. Verder, bedank ik Dries voor zijn hulp bij het zoeken van een baan.

Ik ben zeer dankbaar omdat ik gedurende mijn gehele promotieonderzoek altijd heel fijne zaalgenoten heb gehad, zoals Stepen, Jurgen, Jan-Pieter, Claire, Valerio, Roy, Sjoerd, Jan-Willem, Henriette, Dibyyashree en Mark. Jurgen, we zijn bijna tegelijk aan de slag gegaan met onze PhD projecten. We hebben zeer vaak in de koffiepauze gepraat (hoewel dat ik nooit koffie drink). Je was altijd bereid om naar mijn gekke verhalen en klachten te luisteren. Stephen, de eerste paar jaren was ik altijd gespannen wanneer wij praatten omdat ik jouw Britse uitspraak en grapjes nog niet helemaal kon snappen. Maar geleidelijk vond ik dat je een prettige persoon bent om mee om te gaan. Bovendien vond ik je ook betrouwbaar. Wij zullen elkaar zeker in de nabije toekomst in Londen ontmoeten.

Bas, Harm en “Macco”, ik wil jullie bedanken voor jullie bijdragen aan mijn werk. Bovendien, Bas, af en toe controleerde je mijn werkplek, vooral de monitor. Je gaf me zeer positieve commentaren, zoals “Fraai paper” en “Mooie

grafiek!”, die mij motiveerden om nog harder te werken. Harm, het was altijd aangenaam om met jou op te trekken tijdens de donderdagavond-borrel, in het zwembad en samen uit te gaan eten. “Macco”, omdat je uit een zeer afgelegen provincie komt, is het verrassend te zien dat je uiterst slim bent. Eigenlijk lijkt het alsof je in staat bent alles te kunnen zoals Captain America, variërend van simulatie, en artikelen schrijven tot het maken van een schild. Ik bedank je voor je bijdrage aan mijn papers en de uitnodigingen voor de activiteiten in Sittard.

Mijn speciale dank gaat ook uit naar het Studentsportcentrum, vooral naar de mensen in en rond het zwembad. Een dergelijk mooie faciliteit met zo’n voordelige prijs is zeer uitzonderlijk en dierbaar. Zonder het zwembad zou mijn leven een stuk saaier zijn. Het zwembad heeft me geholpen om mijn

177

uithoudingsvermogen op te bouwen en om mijn geest te verversen, zodat ik opgewassen kon blijven tegen het veeleisende schrijven van mijn proefschrift. Wat zwemmen betreft, wil ik Harm en Matthieu bedanken. Al in het begin overtuigden jullie om mee te gaan zwemmen. In de loop der tijd werd dit een verslavende gewoonte. Ik moet hier zeker ook mijn trainer en goede vriend Andre bedanken. Jouw zwemtechniek en snelheid zijn indrukwekkend. Je bent niet alleen een sportman, maar je bent ook heel slim, en je kunt anderen heel goed instrueren. Bovendien ben je geestelijk volwassen en positief ingesteld. Je bent altijd bereidwillig om naar allerlei onderwerpen van mij te luisteren, inclusief de onderwerpen waarin je geen zin hebt. Je helpt me met zwemmen, maar ook met veel andere aspecten in mijn leven.

This paragraph is dedicated to my Chinese friends. Zhang Zhe and Cui Wei, we together had nice experiences in Groningen for two years. We began with exploring the Western world as young foreign students. In the end we found ourselves, and we all like The Netherlands. Now we live at different places, but the memories will be treasured forever. Xu Chencheng and Zhang Hehe, bei jedem meiner Besuche bei Euch und bei unseren gemeinsamen Reisen habe ich mich immer wie zu Hause gefühlt. Ihr seid wie Familienmitglieder für mich, mit denen ich alles teilen kann. Aachen ist mein zweites Zuhause in Europa durch Euch zwei. Wir alle wissen, dass nicht alles einfach ist im Leben als Ausländer, daher wünsche ich Euch das Beste für Eure Zukunft.

Nederland wil ik als land hartelijk bedanken voor alles wat het mij biedt. De overheid, de cultuur, de taal en al die gekke Nederlanders zijn heel aardig en vriendelijk tegen mij. Daarom doe ik altijd mijn uiterste best om in te burgeren. Het slagen voor het taalexamen en de naturalisatie strekken mij voor altijd tot eer. Ik voel me een stukje van dit land. In het verleden was ik nooit zo gelukkig en blij als nu in dit land. Niks is volmaakt. Maar als je echt liefde voor iets voelt, neem je alle zwakke plekkenop de koop toe.

最后一个段落我必然要献给我父母。从我出生起，你们就给予了我你们的所

有。由于特有的中国的独生子女政策，你们一直承受着孤独因为我多年来都不

在你们身边。你们总是隐藏你们的悲伤，并且全力支持我的决定。一方面，我

是你们的骄傲；而另一方面，我却越来越远离你们。谢谢你们的支持。我对你

们的感激之情溢于言表。

Curriculum Vitae

Personal information

Family name Wu

Given name Yizhi

Date of birth 15 July 1985

Place of birth Xiamen, China

Education

2005-2009 Bachelor of Engineering Degree in Material Science and Engineering,

Tsinghua University, Beijing, China

2009-2011 Master of Science Degree in Nanoscience (cum laude)

University of Groningen, Groningen, The Netherlands.

Master thesis project in the group Nanostructures of Functional Oxides, Zernike Institute for Advanced Materials, University of Groningen, Groningen, The Netherlands.

2011-2015 Ph.D. student in the group Plasma and Materials Processing, Department of Applied Physics, Eindhoven University of Technology, Eindhoven, The Netherlands.

Student Paper Award winner at the 14th International Conference on Atomic Layer Deposition (ALD 2014), Kyoto.

Growth, phase and doping control in ZnO and In2O3 thin ... · Growth, phase and doping control in...

Documents

Transcript of Growth, phase and doping control in ZnO and In2O3 thin ... · Growth, phase and doping control in...