Magnetron Sputtered Scandium Doped Aln Thin Films With High SC Content

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 73

Enni Hartikainen

MAGNETRON SPUTTERED SCANDIUM DOPED


ALUMINIUM NITRIDE THIN FILMS WITH HIGH
SCANDIUM CONTENT
Structure and Piezoelectric Properties

Master of Science Thesis


Faculty of Engineering and Natural Sciences
Examiners: Prof. Petri Vuoristo
Prof. Jukka Vanhala
Dr. Stefan Mertin
May 2021
i

ABSTRACT

Enni Hartikainen: Magnetron Sputtered Scandium Doped Aluminium Nitride Thin Films with High
Scandium Content − Structure and Piezoelectric Properties
Master of Science Thesis
Tampere University
Materials Science
May 2021

The topic of this master’s thesis is to investigate how the piezoelectric properties of magnetron
co-sputtered ScAlN thin films are affected by the Sc content and sputtering process. Piezoelectric
thin-film materials have become an important part of everyday life as they are used in several
components e.g., in mobile phones, smart watches and in steering systems in the automotive
industry. Aluminium nitride (AlN) is one of the most common piezoelectric materials used in sur-
face, bulk acoustic wave devices such as filters and resonators and piezo-driven MEMS (micro-
electro-mechanical systems). During recent years scandium (Sc) doped AlN thin films (ScAlN)
have gained much interest due to their enhanced piezoelectric properties as compared to pure
AlN. The piezoelectric coefficients increase until the polar wurtzite crystal structure turns into a
non-polar cubic structure above around 42% Sc content.
ScAlN is typically deposited by a magnetron sputtering process, which is close to the well-
established AlN process. In this thesis work reactive magnetron co-sputtering from Sc and Al
targets was used in order to grow Scx Al(1-x) N thin films with different Sc content. It was inves-
tigated and confirmed that the Sc content depends linearly only on the target power ratio, inde-
pendent from the total power. However, adding Sc into columnar, c-axis oriented AlN can lead
to abnormally oriented grain (AOG) growth, which disturbs the piezoelectric properties. Those
grains exhibit the correct wurtzite microstructure but their polar c-axis is slightly misoriented. The
amount of AOGs is more pronounced when the Sc content increases and the film stress increases
towards tensile. However, it was found that at a specific Sc content, the amount of AOG growth
depends also on the absolute Sc sputtering power which can be adjusted.
To evaluate the effect of the Molybdenum (Mo) bottom electrode (BE) quality on the co-
sputtered ScAlN film growth, the ScAlN films were grown on Mo BEs of different thickness. The
ScAlN films had less AOGs on thinner Mo BE, and thus 100–150 nm BE thicknesses were se-
lected for following experiments. The highest Sc content of the ScAlN thin films deposited was
measured as 33.5%. The transversal piezoelectric coefficient e31,f of co-sputtered ScAlN films
were measured with a piezo-cantilever test set-up. This experiment showed that during piezoelec-
tric thin-film process development it is important to measure not only the thin-film structure, but
also the piezoelectric response of the film across the wafer. At some wafer positions no piezoelec-
tric response was detected, which was not evident when looking only at the film microstructure.
The highest value was determined to −2.4 C/m2 for a ScAlN film with 33.5% Sc content. This
is 2.6 times higher than e31,f for measured reference AlN, and well-aligned with theoretical curve
and literature values. The result of this work is a good foundation for future improvement and
process development of piezo-MEMS devices containing such ScAlN thin films.

Keywords: scandium aluminium nitride, reactive magnetron sputtering, co-sputtering, thin-film


microstructure, piezoelectric properties

The originality of this thesis has been checked using the Turnitin OriginalityCheck service.
ii

TIIVISTELMÄ

Enni Hartikainen: Magnetronisputteroidut korkealla skandiumpitoisuudella seostetut alumiininitridi-


ohutkalvot − Rakenne ja pietsosähköiset ominaisuudet
Diplomityö
Tampereen yliopisto
Materiaalitekniikka
Toukokuu 2021

Tämän diplomityön aiheena on pietsosähköisten alumiininitridiohutkalvojen seostaminen kor-


kealla skandiumpitoisuudella. Mobiili- ja älylaitteiden yleistyessä pietsosähköiset materiaalit ovat
merkittävä osa monen jokapäiväisestä elämää. Pietsosähköisiä komponentteja käytetään esi-
merkiksi mobiililaitteissa, älykelloissa ja autojen ohjausjärjestelmissä. Alumiininitridi (AlN) on esi-
merkki materiaalista, jota käytetään yleisesti mm. radiotaajuusvastaanottimissa, mikromekaanisis-
sa antureissa, mikrofoneissa ja ajastinpiirien resonaattoreissa. AlN-ohutkalvojen skandiumseos-
tus on viime vuosina herättänyt laajaa kiinnostusta, sillä niiden pietsosähköiset ominaisuudet ovat
paremmat kuin alumiininitridillä. On havaittu, että pietsosähköiset ominaisuudet paranevat kun-
nes skandiumalumiininitridin (ScAlN) wurtziitti-kiderakenne muuttuu kuutiolliseksi noin 42 % skan-
diumpitoisuudella, jolla tarkoitetaan metallisuhdetta Sc/(Sc+Al).
AlN-ohutkalvot kasvatetaan yleensä magnetronisputteroinnilla, joka soveltuu myös ScAlN-ohut-
kalvojen kasvatukseen. Tämän diplomityön seostetut ohutkalvot sputteroitiin kahdesta kohtiosta
samanaikaisesti. Kohtioiden tehosuhdetta säätämällä kalvon skandiumpitoisuutta voitiin muuttaa.
Diplomityössä todetaan, että skandiumpitoisuus riippuu lineaarisesti kohtioiden tehosuhteesta riip-
pumatta kokonaistehosta. Seostus aiheuttaa kuitenkin lisääntynyttä epänormaalia rakeen muo-
dostumista, jonka seurauksena osa kiteistä kasvaa vinosti muun kalvon pylväsmäiseen kasvuun
nähden. Näitä vikakiteitä ilmenee jo matalalla skandiumpitoisuudelle ja niiden määrä kasvaa pitoi-
suuden kasvaessa. Vikakiteet vaikuttavat ohutkalvon kokonaispietsosähköisyyteen ja siksi niiden
määrä on tärkeää minimoida. Vikakiteiden määrän todettiin olevan sitä pienempi, mitä pienempi
oli kohtioiden yhteenlaskettu kokonaisteho ja mitä enemmän kalvojännitys oli puristuksen puolella.
ScAlN-ohutkalvot kasvatettiin pohjaelektrodille, jonka materiaalina käytettiin sputteroitua mo-
lybdeenia. Pohjaelektrodin paksuuden vaikutusta tutkittiin kasvattamalla ScAlN-ohutkalvo eri pak-
suisille pohjaeletrodeille. Koska ohuimmat, 100–150 nm paksut pohjaelektrodit tuottivat vähiten
vikakiteitä, valittiin nämä tulevien pohjaelektrodin paksuudeksi. Korkein saavutettu skandiumpitoi-
suus oli tässä työssä 33,5 %. Ohutkalvojen pietsosähköisiä ominaisuuksia määritettiin mittaamalla
elektrodi-skandiumalumiininitridi-elektrodiaktuaattorilla päällystetyn piipalkin taipumaa aktuaatto-
rin jännitteen funktiona. Tämän avulla voidaan laskea taipumavoiman määrittävä pietsosähköinen
jännityskerroin e31,f . Lisäksi huomattiin, että pietsosähköistä vastetta ei aina ollut koko ohutkal-
von laajuudella hyvästä mikrorakenteesta huolimatta. Tästä pääteltiin, että vaste on syytä mitata
ohutkalvon kehitysvaiheessa sen toimivuuden tarkistamiseksi. Tässä työssä mitattu korkein arvo
oli −2,4 C/m2 , joka mitattiin 33,5 % skandiumpitoisuudella seostetuille alumiininitridiohutkalvoille.
Tämä arvo on 2,6 kertaa korkeampi kuin verrokkinäytteinä olleille seostamattomille alumiininitridi-
näytteille. Tulos on hyvä pohja jatkotutkimuksille.

Avainsanat: skandiumalumiininitridi, magnetronisputterointi, ohutkalvon mikrorakenne, pietsosäh-


köiset ominaisuudet, pietsosähköinen jännityskerroin

Tämän julkaisun alkuperäisyys on tarkastettu Turnitin OriginalityCheck -ohjelmalla.


iii

PREFACE

This Master’s Thesis was conducted at VTT Technology Research Centre of Finland
(VTT) as a part of the New Materials task of Business Finland funded project Beyond
SOI. One goal of this project task is to develop a co-sputtering process for scandium alu-
minium nitride (ScAlN) with high Sc content on the Evatec Clusterline 200 II sputtering
equipment. I am grateful for the opportunity to contribute in this project.

I am grateful to VTT and Dr. Heini Saloniemi for giving me the chance to work in her team
of Process Engineering. Furthermore, I would like to thank my thesis advisor Dr. Stefan
Mertin for giving me excellent advises and valuable feedback during the work. Thank you
for the trust and support from day one. Dr. Abhilash Thanniyil Sebastian and Konsta
Airola deserve a special acknoledgement for helping me with tasks around microfabri-
cation. My thanks go also to all the operators who helped me with measurements, to
the maintenance team who made the sputtering system work, and to the other trainees,
who reminded me of the importance of coffee breaks. Furthermore I acknowledge the
access to measurement equipment of Aalto University. Especially I would like to thank
Professor Mervi Paulasto-Kröckel for giving me access to her laboratory, and Kristina Be-
spalova for her great help with the piezo-cantilever test set-up. I would also like to send
my thanks to Volker Röbisch from Evatec for the discussions regarding ScAlN sputtering.
To my examiners from Tampere University, Professor Petri Vuoristo and Professor Jukka
Vanhala, I am grateful for the academic advices and guidance, which were always highly
appreciated.

If someone had asked me 11 years ago, how my career as a university student would
end, I would not have imagined a worldwide pandemic and a switch the field I studied.
I would not be the person I am now without all the friends I made during these years.
Thank you, kiitos, takk fyrir, danke, tack, merci, and grazie! Also, a little special thank to
my goddaughter Matilda and her twin Matleena. Thank you for all the laughter and joy!

The biggest thanks go to my family, my parents and brothers, who have given me their
unconditional love and support. Thank you for always being there. Finally, I want to thank
my partner Tuomas. You have seen this journey with all its highs and lows, and without
you and Iita I would not have made this.

Tampere, 17th May 2021

Enni Hartikainen
iv

CONTENTS

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Piezoelectric Principle and Thin-Film Deposition Technology . . . . . . . . . . . 4
2.1 Piezoelectricity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1 Piezoelectric Materials and Piezoelectric Effect . . . . . . . . . . 4
2.1.2 Piezoelectric Coefficients of a Clamped Film . . . . . . . . . . . . 7
2.2 Aluminium Nitride Thin Films . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Electric and Piezoelectric Properties . . . . . . . . . . . . . . . . 8
2.2.2 Scandium doped AlN Thin Films . . . . . . . . . . . . . . . . . . 10
2.3 Thin-Film Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.1 Physical Vapour Deposition . . . . . . . . . . . . . . . . . . . . . 13
2.3.2 Chemical Vapour Deposition . . . . . . . . . . . . . . . . . . . . 14
2.3.3 Other Thin-Film Deposition Methods . . . . . . . . . . . . . . . . 15
2.4 Magnetron Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.1 Basic Principle of Magnetron Sputtering . . . . . . . . . . . . . . 16
2.4.2 Reactive Sputtering of AlN . . . . . . . . . . . . . . . . . . . . . 18
2.4.3 Reactive Co-Sputtering of Scx Al(1-x) N . . . . . . . . . . . . . . . . 21
3 Experimental Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1 Thin-Film Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.1 Piezoelectric Layer Deposition . . . . . . . . . . . . . . . . . . . 23
3.1.2 Electrode Deposition . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Film Structure and Thickness Determination . . . . . . . . . . . . . . . . 25
3.2.1 X-ray Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.2 Atomic Force Microscopy . . . . . . . . . . . . . . . . . . . . . . 27
3.2.3 Scanning Electron Microscopy . . . . . . . . . . . . . . . . . . . 28
3.2.4 Spectroscopic Ellipsometry and Reflectometry . . . . . . . . . . . 30
3.2.5 Sheet Resistance . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3 Piezoelectric characterisation . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.1 Microfabrication of Piezo-Cantilevers . . . . . . . . . . . . . . . . 33
3.3.2 Determination of Transversal Piezoelectric Coefficient e31,f . . . . 34
4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.1 ScAlN Thin Films by Reactive Magnetron Co-Sputtering . . . . . . . . . . 36
4.1.1 Influence of Process Chamber Conditioning . . . . . . . . . . . . 36
4.1.2 Influence of Target Power on Sc content . . . . . . . . . . . . . . 38
4.1.3 Influence of Process Gas Flows and Process Pressure . . . . . . 40
v

4.1.4 Influence of Deposition Time and Film Thickness . . . . . . . . . 41


4.2 Magnetron Sputtering Process for Mo Electrodes . . . . . . . . . . . . . 43
4.2.1 Influence of Process Pressure . . . . . . . . . . . . . . . . . . . 43
4.2.2 Influence of Substrate Temperature . . . . . . . . . . . . . . . . . 45
4.2.3 Influence of Underlaying Material . . . . . . . . . . . . . . . . . . 46
4.3 Influence of Mo Bottom Electrode on ScAlN Thin-Film Growth . . . . . . . 47
4.4 Piezoelectric Properties of ScAlN Thin Films . . . . . . . . . . . . . . . . 48
4.4.1 Influence of Sc Content on Piezoelectric Coefficient e31,f . . . . . 49
4.4.2 Influence of Stress on Piezoelectric Coefficient e31,f . . . . . . . . 50
4.4.3 Piezoelectric Performance Compared to Film Structure . . . . . . 51
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
vi

LIST OF FIGURES

1.1 Applications of AlN thin films . . . . . . . . . . . . . . . . . . . . . . . . 2

2.1 Principal of piezoelectric coefficients . . . . . . . . . . . . . . . . . . . . 7


2.2 Wurtzite AlN crystal structure . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 AlN piezoelectric performance vs. RF bias potential . . . . . . . . . . . . 10
2.4 Piezoelectric response vs. Sc-content . . . . . . . . . . . . . . . . . . . 11
2.5 Growth of abnormally oriented grains . . . . . . . . . . . . . . . . . . . . 12
2.6 Amount of abnormally oriented grains with film stress . . . . . . . . . . . 12
2.7 Schemata of Magnetron Sputtering . . . . . . . . . . . . . . . . . . . . . 17
2.8 Schemata of balanced and unbalanced magnetron sputtering . . . . . . . 17
2.9 Schematic illustration of reactive sputtering . . . . . . . . . . . . . . . . . 18
2.10 Schematic illustration of process hysteresis curve . . . . . . . . . . . . . 19
2.11 Hysteresis of AlN sputtering process . . . . . . . . . . . . . . . . . . . . 20
2.12 Thornton diagram of thin-film growth . . . . . . . . . . . . . . . . . . . . 20
2.13 Schematic illustration of reactive co-sputtering process . . . . . . . . . . 22

3.1 AlN film stress vs. substrate RF bias . . . . . . . . . . . . . . . . . . . . 24


3.2 Ar flow vs. BE Mo film stress . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Schematic illustration of XRD system . . . . . . . . . . . . . . . . . . . . 26
3.4 XRD example of a θ–2θ scan . . . . . . . . . . . . . . . . . . . . . . . . 27
3.5 Example of AFM scanning . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.6 Interaction volume in SEM . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.7 Example of EDX spectrum . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.8 SEM cross section of Scx Al(1-x) N thin film . . . . . . . . . . . . . . . . . . 31
3.9 Schematic of spectroscopic ellipsometry . . . . . . . . . . . . . . . . . . 31
3.10 Mask design for cantilever fabrication . . . . . . . . . . . . . . . . . . . . 33
3.11 Principal of piezo-cantilever test . . . . . . . . . . . . . . . . . . . . . . . 34
3.12 Linear displacement vs. voltage curve . . . . . . . . . . . . . . . . . . . 35

4.1 Influence of conditioning . . . . . . . . . . . . . . . . . . . . . . . . . . . 37


4.2 Sc-Al power ratio vs. Sc content . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Sc-Al power ratio vs. abnormally oriented grains . . . . . . . . . . . . . . 40
4.4 Film stress vs. N2 flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.5 Total gas effect on abnormally oriented grains . . . . . . . . . . . . . . . 42
4.6 Influence of deposition time . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.7 Mo BE film stress vs. Ar flow . . . . . . . . . . . . . . . . . . . . . . . . 44
vii

4.8 Deposition rate of Mo vs. Ar flow . . . . . . . . . . . . . . . . . . . . . . 44


4.9 Sheet resistance vs. TE deposition temperature . . . . . . . . . . . . . . 45
4.10 Sheet resistance vs. TE thickness . . . . . . . . . . . . . . . . . . . . . 46
4.11 AFM scans of Mo on SiO2 and AlN . . . . . . . . . . . . . . . . . . . . . 47
4.12 Scx Al(1-x) N on different BE Mo . . . . . . . . . . . . . . . . . . . . . . . . 48
4.13 e31,f vs. Sc content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.14 e31,f vs. film stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.15 Scx Al(1-x) N film at different radius . . . . . . . . . . . . . . . . . . . . . . 52
4.16 Displacement vs. voltage . . . . . . . . . . . . . . . . . . . . . . . . . . 54
viii

LIST OF SYMBOLS AND ABBREVIATIONS

AFM Atomic force microscopy

ALD Atomic layer deposition

AlN Aluminium nitride

AlNMo Molybdenum bottom electrode grown on AlN seed layer

AOG Abnormal oriented grain

BAW Bulk acoustic wave

BE Bottom electrode

CVD Chemical vapour deposition

d33,f Longitudinal piezoelectric coefficient of a thin film

e31,f Transversal piezoelectric coefficient of a thin film

EBPVD Electron beam physical vapour deposition

EDX Energy dispersive X-ray spectroscpy

FWHM Full width at half maximum

IC Integrated circuit

ICP Inductively coupled plasma etching

kt2 Electro-mechanical coupling coefficient

LIDAR Laser imaging, detection and ranging

MEMS Micro-electro-mechanical system

MOCVD Metalorganic chemical vapour deposition

PECVD Plasma enhanced chemical vapour deposition

PLD Pulsed layer deposition

pMUT Piezo-drive micromachined ultrasonic trasnducers

PVD Physical vapour deposition

PZT Lead zirconate titanate

RC XRD rocking curve

RF Radio frequency

SAW Surface acoustic wave


ix

ScAlN Scandium doped aluminium nitride

SE Secondary electrons

SEM Scanning electron microscopy

SiO2 Silicon dioxide

TE Top electrode

XRD X-ray Diffraction

ZnO Zinc oxide


1

1 INTRODUCTION

Piezoelectric thin-film devices have become an important part of our everyday life. They
are used, for instance, in radiofrequency (RF) filters in mobile phones [1], in gyroscopes [2],
in Laser Imaging, Detection, and Ranging (LIDAR) systems [3], and ultrasonic transduc-
ers [4]. As mobile devices become more and more common, there is a constant interest
on making them more performal and cost efficient while their size gets smaller. These
technologies are often based on silicon (Si) processing. Common piezoelectric materials
used in the aforementioned applications are for instance lead zirconate titanate (PZT),
zinc oxide (ZnO) and aluminium nitride (AlN) [5].

AlN is common and widely spread piezoelectric thin-film material, as it is used in surface
acoustic wave (SAW) and bulk acoustic wave (BAW) filters and resonators [1, 6, 7], energy
harvesters [8], and in micro-electro-mechanical systems (MEMS) [9]. AlN is a suitable
material for those applications due to its stability, compatibility with Si processing, high
sound velocity and excellent performance at high frequencies over 2 GHz [1]. In addition,
due to its low dielectric losses, AlN is an interesting material for microphones and piezo-
driven Micromachined Ultrasonic Transducers (pMUT) [4]. Figure 1.1 illustrates examples
of some of the applications mentioned above. In Fig. 1.1a and c it can be seen how the
AlN thin film would either produce mechanical movement when electric voltage is applied
to the BAW filter or a LIDAR mirror. The electric signal generated in the AlN film by moving
a mass plate is exploited in gyroscopes shown in Fig. 1.1b. The pMUT in Fig. 1.1d can
exploit both piezoelectric phenomena, actuating and sensing.

Scandium aluminium nitride has gained much interest in piezo-MEMS industry since
2008, when Akiyama et al. [11] discovered that the piezoelectric properties of AlN are
enhanced when it is doped with Sc. Since then, Sc doped AlN thin films (ScAlN) have
been an emerging topic of active research. So far it is known that the higher the Sc doping
is, the more the piezoelectric coefficients of ScAlN thin films increase. For a high piezo-
electric performance, the ScAlN films must grow in the desired AlN piezoelectric wurtzite
phase in c-axis film texture. The piezoelectric effect is enhanced until the crystal structure
turns from wurtzite to cubic above around 42% Sc content [12, 13]. However, adding Sc
in wurtzite AlN can lead to abnormal grain growth in the thin film already at low Sc con-
centrations. Those grains exhibit the correct wurtzite phase but the piezoelectric c-axis is
slightly misoriented [14]. The higher the amount of Sc, the more pronounced the growth
2

a) b) c) d)

BAW filter Gyroscope LIDAR mirror pMUT

Figure 1.1. Applications for piezoelectric AlN thin films. Figure adapted from [2–4, 10].

of abnormally oriented grains (AOGs) is. This disturbs the piezoelectric performance of
the thin film [15]. Therefore, it is important to reduce the number of such AOGs in the
ScAlN thin film.

Thin films can be deposited with several different techniques among physical vapour de-
position (PVD) and chemical vapour deposition (CVD). Reactive magnetron sputtering
has become a reliable and state-of-the-art way to deposit wurtzite AlN [1]. It is also a
suitable method for depositing ScAlN thin films either from ScAl alloy target or by co-
sputtering from pure Sc and Al targets [15]. By co-sputtering, the amount of Sc can be
modified, which is useful for studying and characterising films with different Sc concentra-
tions.

In this work ScAlN thin films are deposited by reactive magnetron co-sputtering. The Sc
content of the films is varied, up to 33.5%. The influence of different process parameters
in the co-sputtering process for Scx Al(1-x) N thin-film process are inspected together with
structural characterization. Film crystallinity is characterised by X-ray diffraction (XRD)
and the surfaces of the films are inspected with scanning electron microscopy (SEM). Fur-
thermore, the transversal piezoelectric coefficient e31,f is measured with a piezo-cantilever
set-up in order to inspect, how the e31,f of co-sputtered thin films changes with Sc content
and across a single wafer. The highest achieved value for piezoelectric coefficient e31,f
was −2.4 Cm−2 for a film with 33.5% Sc content.

The main research questions of this work are:

1. Is it possible to control the Sc content of ScAlN films precisely by adjusting only the
target power ratio between Sc and Al target?
2. Can the bottom-electrode process be tuned in such a way that piezoelectric AlN
wurtzite c-axis textured growth of ScAlN films is better supported?
3

3. Can the process parameters of the reactive co-sputtering process of ScAlN thin
films with Sc content above 30% and grown on Mo electrodes be adjusted and
tuned in order to better support a high piezoelectric response, which is well in line
or higher than the theoretical and literature values?

The thesis starts with an introduction to piezoelectricity and piezoelectric materials in Sec-
tion 2. In the second part of the theory, the state-of-the-art of AlN thin films and the doping
of it with Sc is discussed in detail. Then, different thin-film deposition methods are de-
scribed before concentrating on magnetron sputtering. The experimental methods used
are described in Section 3. Results are presented and discussed in Section 4. Finally,
in Section 5 the main results are concluded and an outlook on future research topics is
given.
4

2 PIEZOELECTRIC PRINCIPLE AND THIN-FILM


DEPOSITION TECHNOLOGY

In this chapter the theoretical background and the state-of-art of the thesis topic is in-
troduced. First, in Section 2.1 piezoelectric materials and the piezoelectric effect are de-
scribed. Their special properties as thin films are also introduced. Aluminium nitride
(AlN) is one of the most used piezoelectric thin-film materials in piezo-MEMS (micro-
electro-mechanical system) devices. The properties of AlN thin films are described in
Section 2.2. The piezoelectric principle of AlN films and how scandium (Sc) doping en-
hances the piezoelectric response is explained. In Section 3.1 thin-film deposition meth-
ods are introduced and Section 2.4 deals with details about magnetron sputtering, which
is the deposition technique used in this work.

2.1 Piezoelectricity

Piezoelectricity is a phenomenon that was discovered already at the end of the 19th cen-
tury and which is nowadays used widely in electronics [16]. This section concentrates on
the general properties of piezoelectric materials and theory about the piezoelectric effect
and finally describes piezoelectric thin films.

2.1.1 Piezoelectric Materials and Piezoelectric Effect

Piezoelectricity means that mechanical and electrical properties of a material are cou-
pled. Applying mechanical stress or strain to piezoelectric material induces an electric
field. In other words, it produces charges. This phenomenon is called direct piezoelectric
effect and was discovered in 1880 by Pierre and Jacques Curie. Later, in the year 1881,
Gabriel Lippman predicted mathematically the converse piezoelectric effect. Converse
piezoelectricity means that applying an electric field on the material generates mechani-
cal stress. [16]

The direct piezoelectric effect can be described through an electromechanical coupling


5

coefficient k t 2 , which is defined as

Output electrical energy


kt2 = (2.1)
Input mechanical energy

The converse piezoelectric effect can be described by using the same coefficient as

Output mechanical energy


kt2 = (2.2)
Input electrical energy

Piezoelectric materials require an asymmetric structure, because only then they can be
polarised. Polarisation means separation of negative and positive charges in the mate-
rial which causes electric fields. In case of centrosymmetric materials, applied stress or
electric field leads to reversed polarisation as the inversion eventually leads to the same
crystal-stress state. [16]

Piezoelectric materials can be pyroelectric or ferroelectric or either of these. If a material is


pyroelectric, it has an electric response to change in temperature. Pyroelectricity occurs
in all crystalline materials that allow polar directions [17]. Some pyroelectric materials
are also ferroelectric. Ferroelectric materials can be semiconductors, but often they are
dielectrics with low electric conductivity. In ferroelectrics, the direction of the polarisation
can be changed by applying an external high electric field. The polarisation remains after
the electric field is switched off. Ferroelectrics lose their polarisation if they are heated to
their Curie temperature TC [18].

Piezoelectric materials are nowadays used widely in electric applications, such as energy-
harvesters, transducers, sensors and filters [1, 4, 6–8]. They can be single-crystals,
ceramics, composites, or polymers that have an asymmetric structure. Materials with
perovskite and wurtzite crystal structures are most commonly used in piezoelectric appli-
cations. For example, lead zirconate titanate Pb(Zrx Ti(1-x) )O3 , commonly known as PZT,
exhibits a ferroelectric perovskite structure. The piezoelectric coefficients are generally
higher for ferroelectric perovskites than for non-ferroelectric wurtzites. However, since
the permittivity of wurtzites is lower than the permittivity of perovskites, they can be more
suitable for some applications, for example for sensing applications [5]. Aluminium nitride
(AlN) and zirconium oxide (ZnO) are examples for wurtzite crystal structure.

For non-ferroelectric piezoelectric materials, such as AlN, it is required that they have
a macroscopic oriented microstructure, because they cannot be polarized by applying
external electric field. If some of the grains are misoriented or grow in the wrong polar
axis direction, they decrease the total net polarity of the material because the piezoelectric
effect is evaluated over all grains that are present [19]. Due to this, non-ferroelectric
materials need to be deposited so that the polarisation axis is correctly aligned during the
film growth and the resulting film is well textured [20].
6

To describe piezoelectricity more precisely, few basic equations need to be introduced.


The electric polarisation P can be defined as

Q
P = = eS (2.3)
A

where Q is charge, A is the area, S refers to the dimensionless strain, and e is the
piezoelectric stress constant. For piezoelectric materials the electric displacement D can
be written as

D = εS E + P = εS E + eS (2.4)

where εS is the permittivity of the material at a constant strain S and E the applied electric
field.

Depending on the application, the longitudinal, transversal or shear modes are exploited.
The longitudinal piezoelectric mode is utilized in, for example, pressure sensors, as the
mechanical stress is applied perpendicular to the thin film. The transversal mode is uti-
lized in, for example, cantilevers and membranes, as bending of the film causes stresses
in that direction [21]. The shear mode is not described in this work.

For defining longitudinal and transversal coefficient, an understanding of strain S and


electric displacement D in these directions is needed, as shown in Fig. 2.1. In this figure
a schematic of piezoelectric material is represented. By convention, index 3 means lon-
gitudinal direction along the polar axis and transversal direction is perpendicular to this.
Using the piezoelectric strain constant d, the longitudinal component of the piezoelectric
effect can be written as

S3 = d33 E3 (2.5)

where S3 is the strain and d33 the longitudinal piezoelectric coefficient along the polar axis
and E3 the applied electric field. Likewise, using the piezoelectric stress constant e the
electric displacement D3 can be written as

D3 = e31 S1 (2.6)

where e31 is the transversal piezoelectric coefficient and S1 the strain along the material,
perpendicular to the polar axis and the applied electric field.

In general, piezoelectric constants dij and eik are described as tensor, which lead to
multiple different terms and complex equations. Indexes i, j , and k refer to different
7

Figure 2.1. Schematic presentation of a) longitudinal and b) transversal piezoelectric


coefficients of piezoelectric material. Figure adapted from [22].

tensor components. For simplicity, in this work we focus only on the longitudinal and
transversal components, which can be directly measured in thin films.

2.1.2 Piezoelectric Coefficients of a Clamped Film

Piezoelectric thin films behave differently than bulk materials. They are normally de-
posited on silicon wafers by different physical vapour deposition (PVD), chemical vapour
deposition (CVD), atomic layer deposition (ALD), or sol–gel deposition processes. The
aim is to grow thin films of desired crystal structure. As mentioned before, perovskite
and wurtzite structures are mostly used in the field. In addition, thin films having wurtzite
crystal structure need to be grown directly with correct polar axis orientation [5]. Those
films are called textured thin films.

For most applications, the piezoelectric thin films are grown on electrodes. The electrodes
can be, for example platinum (Pt) or molybdenum (Mo), and they are often grown on a
seed layer, which is needed for good adhesion between the silicon substrate and the
electrode and to achieve good crystallinity of the thin film [5].

For thin films two previously introduced piezoelectric coefficients can be measured di-
rectly, the longitudinal d33,f and the transversal e31,f piezoelectric coefficient. The suffix f
implicates that the measure is given to a material which is clamped on a substrate. Lon-
gitudinal and transversal piezoelectric coefficients are defined as

S3 e33
d33,f = = E (2.7)
E3 c33

cE
e31,f = e31 − 13 e33 (2.8)
cE
33

The electro-mechanical coupling of a film in thickness extension mode (c-axis or polar


axis) can be derived with the help of these directions. Then, electro-mechanical coupling
8

coefficient kt2 can be calculated as

e233
kt2 = (2.9)
cD S
33 ε33

where c33 D is the stiffness constant at constant D field, εS is the permittivity of the material
at a constant strain S , and indexes 33 mark the longitudinal direction [23].

To characterise the piezoelectric properties of the clamped thin film, in this work the
transversal piezoelectric coefficient e31,f is measured. It can be measured without tak-
ing into account the elastic constants of the piezoelectric film itself and its value is greater
than e31 for the free-standing material [23].

2.2 Aluminium Nitride Thin Films

Aluminium nitride (AlN) is a ceramic material with a wurtzite structure which is schemat-
ically illustrated in Fig. 2.2. As a bulk, AlN can be used as an insulator. It has a wide
band-gap of 6.2 V at room temperature and a high thermal conductivity, 321 Wm−1 K−1 .
AlN is thermally stable up to 2200°C. This is useful in applications in which heat needs to
be conducted away without electrical short cut. [24]

However, if AlN crystal growth is well controlled, AlN also becomes an interesting material
for piezoelectric devices. This means that AlN is grown as a thin film in the polar c-
axis direction of the wurtzite crystal. As AlN has a high surface acoustic wave velocity
of 6000 ms−1 and is compatible with Si manufacturing technologies, it has become an
important thin-film material for a variety of applications, such as surface acoustic wave
(SAW) and bulk acoustic wave (BAW) filters and resonators [6, 7, 9], and in energy
harvesters [8]. In recent years, there has been a growing interest in using AlN thin films
in piezoelectric MEMS devices [19, 25]. This is due to the stability of AlN: in a wurtzite
AlN crystal the polarity remains always the same and its maximum usage temperature is
high, 1150°C [11].

Aluminium nitride thin films can be produced, for example, by metalorganic chemical
vapour deposition (MOCVD) [26], pulse laser deposition (PLD) [24], ALD [27], and mag-
netron sputtering [28]. In each different process method the goal is to reach film growth
along the polar c-axis by controlling the process temperature, the process gas flow, and
the substrate bias.

2.2.1 Electric and Piezoelectric Properties

AlN thin films show the largest piezoelectric performance when the wurtzite AlN crystals
are grown very close to c-axis, [0001] direction, perpendicular to the wafer surface. Fig-
9

Figure 2.2. Schematic illustration of wurtzite AlN polyhedra from [23]. a) The AlN unit
cell indicating lattice parameters a and c. Yellow atoms implicate nitrogen and grey atoms
aluminium in the crystal lattice. b) Polarity of an AlN unit cell.

ure 2.2b shows how the polarity of the AlN lattice is arranged in the lattice. If this well
textured growth is not reached, it may lead to misoriented grains which may lead to a lack
of piezoelectric performance [19]. The piezoelectric properties change for example by in-
version of the wurtzite crystal from Al polarized to N polarized [29] and enhanced by dop-
ing the AlN with other elements, such as magnesium and zirconium [30], chromium [31],
titanium [32] and tantalum [9]. One of the doping elements is scandium (Sc) which al-
lows the enhancing of the piezoelectric performance of AlN by increasing the transversal
piezoelectric coefficient e31,f by a factor of 2.5 as compared to pure AlN when the Sc
content was 42% [14]. This is discussed in more detail in the next section.

Table 2.1. Properties of selected piezoelectric materials. All values from [5] except if
mentioned differently.

Material Max T d33 d33,f e31 e31,f


calculated calculated
[°C] [pCN−1 ] [pCN−1 ] [Cm−2 ] [Cm−2 ]
AlN 1150 [11] 5.5 3.9 −0.58 −1.05
PZT 250 [11] 268 98 −3.60 −14.7
ZnO 554 [11] 12.4 7.4 −0.37 −0.68

As mentioned previously, compared to other piezoelectric materials, such as PZT and


ZnO, AlN has an excellent stability at elevated temperatures while its piezoelectric co-
efficients remain moderate. Comparing the values presented in Table 2.1, it can be con-
cluded that AlN is an excellent materials for high temperature applications, while its piezo-
electric coefficient remains moderate. PZT is one of the most used piezoelectric materials
with excellent d33 value, but it can only be used in relatively low temperatures. ZnO is a
piezoelectric ceramic used in cell phones as a filter material, but it is not suitable for high
temperature applications [11]. In addition to the high utilisation temperature, as the band-
10

gap of AlN is wide, it has a high electric resistance and a low dielectric loss, which is
useful for high frequency applications [23]. AlN is often used in MEMS resonators and
bulk acoustic wave (BAW) and surface acoustic wave (SAW) filters [25].

The piezoelectric properties of AlN thin films depend strongly on the thin-film crystallinity,
and they get better when the rocking curve (RC) measured by X-ray diffraction (XRD) is
below 2° [33, 34]. It has been shown that the c-axis orientation is mainly controlled by the
process gas flow and the process temperature [35]. Also the substrate potential has an
effect on the piezoelectric properties, as can be seen in the graph in Fig. 2.3. If substrate
bias is not high enough, the longitudinal piezoelectric coefficient d33,f decreases. The
substrate potential is influenced by process pressure, total gas flow and RF-induced bias
voltage [28].

2.2.2 Scandium doped AlN Thin Films

Scandium (Sc) is a rare-earth metal discovered in Scandinavia at the end of the 19th
century. Its density and melting temperature are higher than that of aluminium (Al) [36,
37]. In 2008 it was discovered by Akiyama et al. that the piezoelectric performance of AlN
thin films can be increased by doping AlN with scandium [11]. The e31,f value for 42% Sc
doped AlN can be more than 2.5 times higher than for AlN thin film [13, 15, 38, 39], and the
increase of d33,f was reported to be over 4 times higher for 43% Sc content [40]. Another
advantage is that Scx Al(1-x) N can be grown with similar processes as AlN thin films. Sc
doped AlN thin films can be used to e.g., enhance the piezoelectric performance of bulk
acoustic wave (BAW) resonators and radio frequency (RF) filters [14].

Scandium has a valence of 3, which is the same as the valence of aluminium. The Sc
atom is slightly larger than Al atom, and thus substituting Al with Sc increases lattice
parameters of the unit cell especially in a-axis which in turn softens the structure. In

Figure 2.3. Longitudinal piezoelectric coefficient d31,f versus substrate bias potential of
AlN thin film deposition on Pt electrode [28].
11

Figure 2.4. Piezoelectric response versus Sc concentration of Scx Al(x-1) N thin films with
varying Sc content x grown at 400°C [41].

addition, electronegativity of Sc is stronger than for Al. The soft structure and the higher
electronegativity contribute to the enhanced piezoelectric properties, which explain the
improved piezoelectric effect [13]. In bulk acoustic wave (BAW) resonators, the enhanced
piezoelectric properties are observed as a lowering of the sound velocity along c-axis [14].

Akiyama’s graph (see Fig. 2.4) shows an increase of piezoelectric response until Sc con-
centration of 42–45%. After that, the piezoelectric performance decreases rapidly. The
piezoelectric response of the Scx Al(1-x) N depends on the Sc-concentration, where x refers
to the metal ratio Sc/(Sc+Al). This can be explained by looking at the crystal structure of
the film: the piezoelectric response is enhanced until the wurtzite structure is lost [13].
Below 41% Sc content Scx Al(1-x) N grows in hexagonal wurtzite phase. When the Sc con-
tent increases to 42–45% also a non-polar cubic rock-salt crystal structure appears while
the hexagonal wurtzite crystals are still present. At higher Sc concentrations above 45%,
cubic rock-salt crystal structure gets dominating, as the length of the a-axis has grown
longer and become as long as the c-axis [11].

Growing Scx Al(1-x) N thin films is challenging because ScN has a cubic rock-salt structure
which is non-polar and thus not piezoelectric. Sc atoms have to be forced into the wurtzite
AlN structure which is relatively straightforward at low Sc contents. However, the more
Sc there is, the more difficult this becomes [14]. Also, the Scx Al(1-x) N should grow in the
correct polar orientation along the c-axis for the highest piezoelectric performance [33].
12

Another challenge of Scx Al(1-x) N thin films are abnormally oriented grains (AOG). They
are wurtzite grains that do not grow in the desired c-axis orientation [14]. This makes
the film surface rougher and disturbs the piezoelectric response of the films, which in turn
affects the device’s performance [15]. For example, the devices based on surface acoustic
waves (SAW), the increased surface roughness due to AOGs leads to wave scattering
and attenuation [14]. AOGs can appear in Scx Al(1-x) N already at low Sc contents, but
their number increases as the Sc content gets higher [14, 40]. The cone shape shown
in Fig. 2.5 of AOGs is due to faster growth in width than the surrounding (0001) oriented
columnar film grains [14].

The AOG growth may start on the substrate, but it is reported that they can nucleate also

Figure 2.5. Growth of abnormally oriented grain on Pt electrode. The AOG cone starts
growing from a nucleus indicated by the red arrow [14].

Figure 2.6. Stress dependency of number of abnormally oriented grains [15]. The Sc
concentration of the films is 27%.
13

on defects and on the grain boundaries between the c-axis oriented grains [14]. This
was found to be related with some accumulation of Sc on the grain boundaries which
leads to grains growth in the wrong orientation. Their tilt changes the growth direction
and ultimately the c-axis of the wurtzite phase AOGs is misoriented. Previous research
has shown that the number of abnormal grains increases as the film stress increases
towards tensile stress [13, 15]. Consequently, for more compressive films, fewer of those
grains can be expected as can be seen in Fig. 2.6. Therefore, compressive stress range
of approximately 0 to −500 MPa was chosen as the target film stress for the Scx Al(1-x) N
deposition development in this work. Also the quality of the substrate surface seems
to decrease the number of AOGs, as their number was reduced when the surface was
plasma etched before Scx Al(1-x) N deposition on Si wafer [40]. This strategy was followed
for substrates used in this work.

2.3 Thin-Film Deposition

Thin films are material layers of solid materials that have different properties than the
same material as bulk [42]. Thickness of thin films can vary from one atomic layer to about
10 µm. They have applications in improving material surface properties e.g., absorption,
reflection, corrosion and abrasion resistance, and electrical behaviour. In micro- and
nanotechnology thin films are used for manufacturing electric and mechanical devices
such as sensors, transistors, and microelectromechanical systems (MEMS) [42].

Thin films can be deposited via different methods depending on the application. Chemical
vapour deposition (CVD) and physical vapour deposition (PVD) are both used in thin-film
technology. Different PVD methods are introduced in Section 2.3.1 and CVD methods
in Section 2.3.2. Atomic layer deposition (ALD) and sol−gel deposition are described
shortly in Section 2.3.3.

2.3.1 Physical Vapour Deposition

Physical vapour deposition is a category for a group of thin-film deposition methods in


which the films are deposited on a substrate surface in a vacuum from a source which
can be in a solid or a crucible. There are several different methods on transporting the
source material to the substrate [43]. The source material can be for example heated by
resistive heating or by an electron beam, or bombarded with an ion or a laser beam.

Thermal evaporation is a vacuum process, in which the material source is placed in a


vacuum below 1×10−6 mbar. The source material can be a pure atomic element or com-
pounds of metals or non-metals. The material source is usually placed at the bottom of the
process chamber as it is heated to its melting or sublimation temperature. Then, vapour
of the evaporated material is transferred towards the substrate which gets coated [44].
14

Deposition rate in thermal evaporation is moderate, 10–100 nm/min [45].

In electron-beam evaporation, also known as electron-beam physical vapour deposition


(EBPVD), a beam of high-energy electrons are accelerated towards the source material.
This causes local heating and evaporation of the source material [46]. Typically, the base
pressure is in the range of 1 ×10−7 to 1×10−9 mbar. Deposition rate of electron-beam
evaporation can be as high as 25 µm/min [47].

Pulsed laser deposition (PLD) is based on locally evaporating the source material with the
help of a high-power pulsed laser. The source and the substrate are placed in a vacuum
or in a reactive gas environment depending on the required final film composition. As the
laser beam interacts with the target, a small region gets rapidly heated and cooled and
the target material is evaporated or sublimed. [48]

Focused ion-beam sputtering is a high-vacuum deposition process which results in high-


purity thin films [42]. In ion-beam sputtering, an ion beam is focused on the surface
material source. As Ar is ionized at a high electric field, an ion beam is formed which
sputters material from the target to the substrate [49]. Another sputtering technique is
magnetron sputtering. As the name indicates, this method exploits permanent magnets
in the sputtering source leading to much higher deposition rates than in ion-beam sputter-
ing [50]. Magnetron sputtering, the method used in this work, is introduced more closely
in Section 2.4.

2.3.2 Chemical Vapour Deposition

Chemical vapour deposition (CVD) is a widely used thin-film technology in micro- and
nano-electronics. CVD is based on a chemical reaction which occurs on the substrate
surface. A thin film is formed, as the reactants in the vapour phase react heterogeneously
at the substrate surface and form a solid product. The rate of film growth can be con-
trolled with the temperature and the flow of the process gas brought to the process by a
carrier gas. At a low temperature, the CVD process is controlled by the kinetics of the
reaction and at a high temperature the reaction is controlled by diffusion of the reactive
gas. Depending on the desired thin films, the process can take place in a horizontal or
a vertical process reactor. In the horizontal reactors, the gas flow is horizontal and the
film quality may be affected by the position of the wafer in the chamber. In the vertical
reactors, the reactive gas is introduced either from the bottom or from the top of the re-
action chamber or both at the same time, which results in more uniform films. Depending
on if the reaction is exothermic or endothermic, hot-wall or cold-wall chambers are used,
respectively. With hot-wall chambers there is also a risk, that a film forms on the chamber
walls. Due to this, hot-wall chambers need more frequent cleaning than cold-wall cham-
bers. CVD processes are used for example for pyrolysis, oxidation, reduction, nitride and
carbide formation. [51]
15

Low pressure chemical vapour deposition (LPCVD) occurs at 0.13–13 mbar pressure in
a hot-wall reactor. In LPCVD, a carrier gas is not needed. LPCVD results in uniform
films with a low risk of particle contamination, but residual stresses generally remain
high [51]. Plasma enhanced chemical vapour deposition (PECVD) is used for depositing
oxides, nitrides and carbides. For example diamond like carbon (DLC) and carbon nano-
tubes (CNT) can be produced by PECVD processes. In PECVD, radio frequency (RF)
generated plasma is present to accelerate the process gas decomposition, which leads
to better reactivity. One advantage of this is that the process temperature of PECVD is
lower than in thermal CVD, which is desired in some oxide processes [43].

Metalorganic chemical vapour deposition (MOCVD) is used when the desired application
requires exceptional step coverage [52]. The applications of MOCVD are based on ele-
ments in groups II-V of periodic table, which are transition metals. Application include for
example gallium nitride (GaN) and gallium arsenide (GaAs) based devices, for example
laser diodes, light-emitting diodes (LED) and solar cells. MOCVD can also be used to
deposit AlN thin films [26].

2.3.3 Other Thin-Film Deposition Methods

Other methods used to deposit piezoelectric thin films are atomic layer deposition (ALD)
as well as sol−gel deposition. ALD is developed from a variety of CVD processes. It is,
however, significantly distinguished from the conventional CVD. It is a deposition method
in which thin films are grown by layer-by-layer. The single atomic layers are grown by
introducing chemical precursors in the process chamber. These precursors react with the
sample surface with a self-limiting reaction which forms an atomic monolayer at the sur-
face. After the reaction the chamber is purged with an inert carrier gas before introducing
a next pulse of a different precursor gas for the next layer [43, 53]. ALD is a very slow
deposition method but it has a perfect side-wall coverage. It is also possible to grow AlN
with ALD [20].

In sol−gel deposition, thin films are formed through gelation of a colloidal metal alkoxide
solution, which is deposited on the wafer by dip- or spin-coating followed by hydrolysis and
condensation. After that the wafer is taken to thermal treatment to sinter the film which
controls the grain growth. The properties of the film are controlled for example by pH of
the solution, the temperature, the reaction time and the concentration of the reagents.
The advantages of sol−gel deposition are that the films are chemically homogeneous
with high purity and the grain size distribution is narrow [54]. However, the shrinkage
during drying may lead to fractures in the film. The sol−gel method can be used, for
example, to deposit PZT [55] and ZnO [56].
16

2.4 Magnetron Sputtering

Magnetron sputtering is widely used in integrated circuits (IC) and in microfabrication [43],
but it can be also used for protective coatings and hard coatings for tools [50] or even
functional coating on architectural glazing [57]. In this section, magnetron sputtering is
discussed as a thin-film deposition method in detail. It is chosen as the deposition method
for the piezo-layer and the electrode deposition in this thesis. One of the process types,
reactive magnetron sputtering, is discussed in Section 2.4.2 and co-sputtering set up is
described in Section 2.4.3.

2.4.1 Basic Principle of Magnetron Sputtering

Magnetron sputtering is a low pressure thin-film deposition process with a vacuum base
pressure of 1× 10−7 to 1 × 10−8 mbar. In magnetron sputtering, material is sputtered
from the target by bombarding it with energetic ions. The target can be a pure element,
an alloy or a compound. The plasma is often created with a noble gas which is ionised.
Chemical reactions are not always desired in the sputtering process and that is why noble
gases are mainly used in sputtering. Argon (Ar) is widely used, as it is cheap compared to
other noble gases which could be used for plasma [58]. In addition, argon is also a heavy
enough element for most of the target materials. As Ar is ionised, there are positively
charged Ar+ ions and negatively charged electrons e− present in the plasma. The Ar+
ions are accelerated towards the negatively charged target (cathode). Figure 2.7 shows a
schematic drawing of the magnetron sputtering process. The Ar+ ions sputter off atoms or
compounds from the target surface to form the thin film on the anode, where the substrate
is also placed [43, 59].

One of the advantages compared to other PVD and CVD methods is that sputtering can
be used for depositing large surfaces. Therefore, it can be used for example on depositing
a UV photocatalytic thin films on large windows, which can be as large asa few meters
long and wide [60]. Another example is the functional coating in architectural glazing that
are up to 3.2 × 6 m large [57].

In magnetron sputtering, a set of permanent magnets are placed behind the target. One
set of magnets is placed in the centre, and another set with opposite orientation of the
magnetic poles is placed on the outer ring as shown in Fig. 2.7. This creates closed mag-
netic field lines parallel to the target surface. This increases the probability of ionization
and collisions of electrons and keeps the plasma density high close to the target as elec-
trons are trapped there [43, 58]. Applying a negative potential on the target attracts the
positively charged Ar+ ions from the plasma which then causes bombardment on the tar-
get and sputters material off. As the plasma density close to the target is increased, using
magnetrons as sputtering source increase the sputter yield compared to other sputtering
17

methods [50].

Depending on the set of magnets in the magnetron sputtering system, the system can
be balanced or unbalanced. In the balanced magnetron sputtering, which is also known
as conventional magnetron sputtering, all magnets in the system are equally strong and
the magnetic field lines form closed loops. If one set of the magnets is stronger than the
other, the magnetron system is called unbalanced. Depending on which magnets are
stronger, the unbalanced magnetron systems can be either unbalanced type I or type II
(see Fig. 2.8). This opens some of the magnetic field lines and the plasma can spread
towards the substrate surface. This can be beneficial in increasing the ion bombardment
on the substrate surface and increase the film density [50].

Figure 2.7. Schemata of conventional magnetron sputtering with magnetron, target, and
substrate. The plasma formed by Ar and N2 is indicated in purple.

Figure 2.8. Schemata of targets and magnets in balanced (a) and unbalanced (b, c)
magnetron sputtering configuration [50].
18

One of the disadvantages of magnetron sputtering is that the plasma density is highest in
the regions where the magnetic field is the strongest. Due to this the ion bombardment on
the target surface increases on this region and the target is not eroded evenly. This leaves
a so called electron race track on the target surface [50]. The uneven erosion of the target
is problematic as it causes non-uniform deposition of the thin film. Also the target material
cannot be fully utilized. This issue can be improved by using a magnetic system in which
the distance between the different magnets can be modified or by rotating an off-centred
magnet system around the target [61].

2.4.2 Reactive Sputtering of AlN

Reactive sputtering is a sputtering method in which the thin film is formed through a
chemical reaction and thus it has a different composition than the original target material.
The principal of reactive sputtering is presented in Fig. 2.9. The compound in the film is
formed between a metal or a metalloid, and a non-metal. This is achieved by adding the
non-metal element as a reactive gas to the process. Reactive gas is typically O2 or N2
depending on the desired resulting film. Thus the sputter deposited compounds are often
oxides and nitrides [62].

One advantage of reactive sputtering is that the thin-film composition can be controlled.
The formation of the film compound takes mostly place at the target surface and on the
substrate surface. When the reactive gas and the target material form a compound on

Figure 2.9. Schematic drawing of reactive sputtering process with pulsed DC power
and RF substrate bias. The plasma is shaped with a moving magnetron system. Target
substrate distance is adjustable.
19

the target surface, the target is called poisoned. Consequently, as the compound forms
also at the target surface, hysteresis is a phenomenon that has to be taken into account
in the reactive sputtering process [58, 63]. A schematic illustration of the hysteresis in
Figure 2.10 shows the different process regions. In the metal mode, the sputter rate is
higher than in the compound mode, also known as poisoned mode. The reason for this
is that the sputtering rate is lower than in elemental sputtering of a pure metal, because
compounds have higher binding energy than pure elements [63].

In addition, the change of the sputtering rate depends on the process history. This can be
explained as following [62]: As the flow of the reactive gas is gradually increased starting
from the metal mode, it is first fully consumed on the target surface before reaching the
poisoned mode. As the transition region is reached, the sputter rate remains stable until
it starts to decrease. In this region the sputter rate is sensitive to small changes in the
process gas flow. Once the poisoned mode is reached, the process rate is low and does
not significantly change as the reactive gas flow is increased. When the reactive gas flow
is decreased after the poisoned mode, the sputter rate remains lower longer than when
the reactive gas flow was initially increased in the metal mode region. As the reactive gas
flow is further decreased, the metal mode of sputtering is reached again through a similar
transition region.

In practise it is easier to evaluate the hysteresis of a reactive sputtering process using the
target voltage. This was determined for an AlN process. The corresponding target voltage
versus N2 flow hysteresis curve is shown in Figure 2.11. From the graph it can be seen
that at least a N2 -Ar ratio of 1:1 is needed to work in the poisoned mode. Following the
standard process for AlN, in this work hysteresis does not need to be taken into account,
as the N2 flow is kept higher compared to the Ar flow. Due to this, the AlN process occurs
always in the poisoned mode.

When the target is in the poisoned mode, it becomes less conductive and charge trans-

Figure 2.10. Schematic example of the hysteresis of a reactive sputtering process [58].
20

450
Increasing N2 flow
Decreasing N2 flow
400

350
Voltage [V]

300

250

200
4 6 8 10 12 14 16
N2 flow [sccm]

Figure 2.11. Hysteresis curve of AlN reactive sputtering process. Ar flow is kept constant
at 10 sccm. Target power was 1000 W.

Figure 2.12. Thornton diagram of thin-film growth [64].


21

port is significantly reduced. Due to this, strong electric fields can build up which can
discharge with an electrical breakdown, which is known as arcing. Arcing makes the
plasma unstable and can release macro particles in the process. Moreover, it can be
even harmful to the target or the system. To prevent the charge build-up during reactive
sputtering, pulsed direct current (DC), radio frequency (RF) or alternating current (AC)
power are used. Pulsed DC means that a short pulse of positive voltage of the power
supply is applied. The positive pulse attracts electrons to the target surface which neu-
tralizes the positive charges that are built up on the cathode during the negative cycle [58].
When AC power is used, voltage changes constantly in cycles. For RF-power the cycle
is very rapid and attracts only electrons to the target surface as Ar+ ions are too heavy to
react.

Thornton diagram in Fig. 2.12 illustrates how the thin-film growth changes when the sub-
strate temperature and the Ar pressure are changed. Thornton describes different zones
of similar film growth that depend on the Ar pressure and the ratio of the substrate tem-
perature T and the film material melting temperature TM in Kelvin [64]. The illustration
shows that as the substrate temperature is increased, grains grow in a more columnar
way. The column size increases as substrate temperature is further increased until about
the ratio T/TM =0.8. After that grains start to recrystallize. The desired zone of grain growth
for AlN is Zone T. Melting temperature TM of AlN is 2200°C (2474 K) and the substrate
temperature T in magnetron sputtering is 300°C (574 K) which gives a ratio T/TM of 0.25.
To remain in the Zone T, this requires that the Ar pressure is kept below approximately
1 mTorr (1 × 10−3 mbar).

2.4.3 Reactive Co-Sputtering of Scx Al(1-x) N

Co-sputtering is a sputtering method in which the film material is sputtered from more
than one target as shown in Fig. 2.13. It can be used to grow both pure metal (alloy) films
and ceramic or dielectric films through reactive sputtering. If co-sputtering is used for
sputtering an alloy or a compound, the amount of each component can be controlled by
controlling the power of each target [40]. This is the biggest advantage of co-sputtering.
In case of a nitride such as Scx Al(1-x) N the co-sputtering process need to be also reactive
(see Section 2.4.2).

Since the targets are not directly above the substrate, the film uniformity may cause chal-
lenges. Using two targets at the same time changes the process environment, and thus
the process conditions are not the same as when only one source is used. In the case
of Scx Al(1-x) N there are two targets which both consume the reactive gas N2 instead of
one. This needs to be compensated by adjusting the N2 flow. The thin-film uniformity and
the Sc distribution across the wafer can be optimized by changing the target-to-substrate
distance and using a rotating chuck (substrate holder). In addition, changing the ratio of
22

Figure 2.13. Schematic illustration of reactive co-sputtering process. Plasma is created


with Ar and N2 gas mixture. The chuck rotates during deposition. Target-substrate dis-
tance is adjustable.

the process gases may have an effect on the film uniformity.

Scx Al(1-x) N thin films can be produced also by reactive sputtering from an ScAl alloy target.
The advantage of a ScAl alloy target is that the Sc content of the film will be very close
to the content of the alloy target. In addition, the Sc concentration is the same across the
whole wafer and does not depend on the used target power. This is useful if properties
of a certain Sc content are already known and the film quality and the deposition process
are well established. But at the same time, not being able to tune the Sc content is a big
disadvantage for research, since each different Sc content in the film would require a new
target.

Furthermore, the growth of abnormally oriented grains as discussed in Section 2.2.2


causes issues in the thin film quality. As Sc needs to be forced in the wurtzite AlN,
the sputtering parameters need to be optimized. The process temperature for Scx Al(1-x) N
sputtering was chosen as the same as for AlN process, approx. 300°C according to the
Thornton diagram. Finding suitable process parameters for the target power and for Ar
and N2 gas flows for Scx Al(1-x) N sputtering process is one of the great challenges of this
thesis.
23

3 EXPERIMENTAL METHODS

The experimental set-up and methods for the thin-film deposition and characterisation
are described in this chapter. First, the deposition of piezoelectric thin films and metal
electrodes is described in Section 3.1. Second, the structural characterisation of thin-film
crystallinity, electron microscopy imaging, surface roughness and film thickness deter-
mination is explained in Section 3.2. Finally, electrode patterning and the piezoelectric
characterisation of AlN and Scx Al(1-x) N films are described in Section 3.3.

3.1 Thin-Film Deposition

Piezoelectric and metal thin films of this work were deposited with an Evatec Cluster-
line 200 II cluster sputtering system, which has five chambers for different process pur-
poses. One of the chambers is an inductively coupled plasma (ICP) etch chamber. ICP
etching is used before each sputtering process to remove native oxides from the Si wafers
(p-type (100) silicon) or the previously deposited thin-film layers, i.e. Mo electrodes. After
etching, the robot handler takes the wafers to the sputtering process module for thin-film
deposition without breaking the vacuum.

3.1.1 Piezoelectric Layer Deposition

The AlN and Scx Al(1-x) N thin films were deposited by pulsed DC reactive magnetron sput-
tering. AlN films were deposited by reactive sputtering from a pure Al target. For 20%
and 30% Sc content reference films ScAl alloy targets were used with respective Sc con-
tent. The diameter of targets used in the single-target process is 304 mm. The pulsed
DC power used during the sputtering was between 7000–7500 W and the base pressure
2.0 × 10−8 to 1.8 × 10−7 mbar. Ar is used as the sputter gas and N2 as the reactive
gas. In AlN process, the Ar flow was 20 sccm and N2 flow 60 sccm, as the ratio between
the gases was kept constant at 1:3. The same ratio was used also in the ScAlN pro-
cess, but the Ar flow was decreased to 12 sccm and N2 to 36 sccm. The thickness of the
piezoelectric layers deposited from single target was 1000 nm.

The piezoelectric films with varying Sc content were deposited in another multi-target
process chamber by co-sputtering using pure Al and Sc targets of 100 mm diameter.
24

Depending on the desired Sc content, the pulsed DC power on each target was varied
between 250–655 W for the Sc target and 600–1000 W for the Al target. The vacuum base
pressure of the process chamber was between 8.2 × 10−8 and 1.3 × 10−7 mbar. The
effective Ar flow was varied between 2.5–10.5 sccm and N2 flow between 21–25 sccm.
The thickness of co-sputtered Scx Al(1-x) N piezoelectric layer was approximately 600 nm.

In order to find good quality film produced by the single target several calibration runs
were performed. On the first deposition the effect of the substrate RF bias power was
investigated by changing the substrate RF bias. It has great effect on film stress, which
should be adjusted to around 0±50 MPa. A typical optimisation of the film stress by
substrate RF bias is presented in Figure 3.1.

3.1.2 Electrode Deposition

Molybdenum (Mo) was applied as bottom electrode (BE) and top electrode (TE) of the
piezo-cantilevers used in this work. The Mo electrodes were deposited from one of the
targets in the multi-target process chamber which is also used for co-sputtering. As Mo is
sputtered as a pure element from 100 mm Mo target, only Ar was used as process gas.

For the bottom electrode good crystallinity is crucial for textured film growth of the next
layer. Therefore, a thin AlN film, called AlN seed layer, was deposited on a 200 nm thick
SiO2 thermal oxide, which improves not only the adhesion but also the crystallinity of the

300
250 Stress X
200 Stress Y
150
100
50
Stress [MPa]

0
-50
-100
-150
-200
-250
-300
14.5 15.5 16.5 17.5 18.5 19.5
RF power [W]

Figure 3.1. Film stress versus substrate bias RF power of piezo-AlN by single target
sputtering process deposition.
25

Mo film. The thickness of the seed layer was 30–35 nm. In this process the Mo film
stress is optimized by adjusting the Ar flow as shown in Fig. 3.2 with the target of near
zero film stress (0±50 MPa). The thickness-dependant stress optimization is discussed
more closely in the results in Section 4.2.1. The temperature of the substrate holder, also
known as chuck, was set to 300°C for Mo BE process. Depending on the thickness of the
bottom electrode, the Ar flow was in the range 13–17 sccm.

The Mo top electrode was deposited with a similar process on top of the piezoelectric
layer. The thickness of the TE was approximately 200 nm. The Ar flow was 16 sccm and
temperature of the chuck was set to 200°C. As with the previously deposited films, the
target for the TE film stress was also near 0±50 MPa.

3.2 Film Structure and Thickness Determination

The thin-film growth was evaluated by structural characterisation. The crystallinity was
evaluated by X-ray diffraction (XRD) as described in Section 3.2.1. In Section 3.2.2 it is
explained how the surface roughness of bottom and top electrode Mo was evaluated by
atomic force microscopy (AFM). Scanning electron microscopy (SEM) was used to evalu-
ate the film morphology in SEM plane view and film thickness in SEM cross sections (see
Section 3.2.3). Film thickness of AlN and Scx Al(1-x) N was determined optically by spec-
troscopic ellipsometry and reflectometry (see Section 3.2.4) or by measuring the sheet

50

-50
Stress [MPa]

-100

-150

Stress X
-200
Stress Y

-250
11.0 12.0 13.0 14.0 15.0 16.0 17.0
Ar flow [sccm]

Figure 3.2. Film stress versus Ar flow of bottom electrode Mo film stress with an AlN
seed layer on thermal oxide.
26

resistance of metallic Mo (see Sec. 3.2.5).

3.2.1 X-ray Diffraction

X-ray diffraction (XRD) is based on the scattering of X-rays from atoms parallel to atomic
plane in crystal structure as shown in Fig. 3.3. The structure of crystalline materials and
textured films is periodic, which means that the atoms form a long-range order which is
repeated in each grain. Each crystal orientation has a specific X-ray diffraction pattern
and a spacing d between the atomic planes. The crystal planes cause diffraction pattern
according to Braggs’ law

2d sin θ = nλ (3.1)

In the equation θ is the angle between the sample surface and the incident X-ray beam
and λ is the wavelength of the X-rays. 2θ is the angle between the diffracted and the
undiffracted beam as shown in Fig. 3.3.

The X-rays that are scattered with the same angle form diffraction peaks. The intensity
and width of the diffraction peaks can be used in analysing the crystal structure and the
thin-film texture. The more textured the film and oriented the crystals are, the narrower

Figure 3.3. Schematic illustration of XRD indicating the 2θ diffration condition. Adapted
from [65].
27

Figure 3.4. Example of a θ–2θ scan of Scx Al(1-x) N thin film on Mo bottom electrode.

the peaks diffracted from crystal planes. The 2θ scan in Fig. 3.4 is an example of the
diffraction pattern in which peaks from the orientated planes of each element can be
found.

To analyse the distribution of the crystallites close to a crystal plane orientation, the scan-
ning is done at a fixed 2θ angle which is different for each compound or element, for
example 36° for AlN and 40° for Mo. The sample is scanned around the ω axis. From
this scan the full width at half maximum (FWHM) is also known as rocking curve (RC)
[34]. The smaller the RC value, the better the crystallites are oriented along a certain
axis. The rocking curve value normally decreases as the film grows thicker because at
the beginning of film growth film crystallites are still slightly misoriented and may grow in
different directions [33].

The equipment used in this work (XRD Panalytical X’Pert Pro) exploits two wavelengths
of Cu, Kα1 =1.541 Å and Kα2 =1.544 Å. The RC value for different Scx Al(1-x) N and Mo
films were measured during the process development and after the whole stack has been
deposited.

3.2.2 Atomic Force Microscopy

Surface roughness of the films was measured by atomic force microscopy (AFM) with
Bruker Dimension Icon AFM. In AFM a very fine tip, with a size of less than 10 nm,
is attached to a cantilever. As the surface is scanned, the cantilever displacement is
measured with a laser. AFM scanning can be done in contact mode, in which the repulsive
forces between the tip and the surface are detected, or in non-contact mode, in which
the tip is kept 50–150 Å from the surface and attractive interatomic forces are detected.
Tapping mode is a combination of the two, as the cantilever is oscillating and only gently
touching the surface. [66]
28

(b) Surface roughness analysis.

(a) Image of an AMF scan.

Figure 3.5. Typical AFM image of a Scx Al(1-x) N thin film at wafer centre.

In this work AFM measurements were conducted in tapping mode and the scanned area
was 1 × 1 µm2 . The measurements were done at the wafer centre and at the radius of
R = 60 mm towards the right wafer edge when the wafer flat was facing the operator. The
scan frequency was 0.50 kHz with 512 scan lines and the peak force amplitude was set
to 120 nm. The lift height of the probe was adjusted automatically.

Once the scanning is done, image analysis is still needed. The obtained data files were
analysed with Gwyddion which is an open source software for image data analysis. To
analyse the AFM image, the raw-data image is first flattened and scan defects, such as
line defects, are corrected if needed. An example of the final AFM image is shown in
Figure 3.5a. After that it is possible to obtain the surface roughness value of the scanned
area, as shown in Figure 3.5b.

3.2.3 Scanning Electron Microscopy

Scanning electron microscopy (SEM) was used for verifying the thickness of the deposited
metal films and the amount of abnormally oriented grains (AOG) in the case of Scx Al(1-x) N
films. Generally SEM is used to gain more information about the sample than is possible
by using optical microscopy. Optical microscopy is limited to the wavelength of light, and
thus very small details in the sub-micrometer scale cannot be observed [67].

Scanning electron microscopes are based on the interaction of the sample surface and
the beam electrons. The surface is scanned with an electron beam with high energy.
The beam electrons interact with the sample material in different ways. The interaction
volume and the different ways of interaction are shown in Fig. 3.6. If the electron beam
penetrates the sample, electrons with low energy escape from the sample, they are called
secondary electrons (SE). SEs from close to the sample surface are collected with sec-
ondary electron detector, which in this work was in-lens detector. Because the origin of
29

Figure 3.6. Schematic of interaction volume and different signals in electron mi-
croscopy [67].

SEs is only from the depth of a few nanometres, they form good topographical images
with good details. [67]

Characteristic X-rays are used for elemental analysis in energy dispersive X-ray spec-
troscopy (EDX). Generally EDX can be used to gain qualitative and quantitative data of
the elemental composition of the sample [67]. Qualitative analysis means that for ex-
ample the elements and their location on the sample surface is detected. Quantitative
analysis gives also the ratio of the elements. An example of EDX spectrum gained from
the samples of this work is shown in Fig. 3.7. The peaks that are visible in the spectrum
are for Sc, Al and N. The spectrum is analysed by Inca software, which automatically
finds elements corresponding to the peaks. The user still has to confirm which elements
are present in the sample. During dead time data is processed in the system and new
data is not collected. The dead time increases when acquisition time is increased. The
acquisition time is the total time that signals are collected [67].

Two different electron microscopes were used in this work: Zeiss Supra 35 and Zeiss
LEO 1560. Both SEMs are placed at VTT Micronova cleanroom facilities. Zeiss Supra 35
was used mainly for cross-sectional and plane-view imaging. The used acceleration volt-
age was in the range of 3.5–6.0 keV and the working distance was around 4 mm. The
Zeiss LEO 1560 was used for energy dispersive X-ray diffraction (EDX) analysis of the
30

Figure 3.7. Example of an EDX spectrum in which Sc, Al, and N peaks are visible.

scandium concentration of the Scx Al(1-x) N films. EDX scanning was done with electron
voltage of 8.0 keV and working distance of 7 mm. The dead time in EDX measurement
was 20–25% and acquisition time 40 s.

Figure 3.8 shows a cross-section image of a Scx Al(1-x) N film on Mo bottom electrode on
AlN seed layer. In order to be able to measure the thickness of the different layers, it
is important to see clearly the interfaces of different layers when taking cross-sectional
images. Also, it is important to get a full understanding and a good overview of the whole
stack in which all layers are visible. From the plane-view images, the number of abnormal
oriented grains (AOG) of Scx Al(1-x) N films is inspected. For analysis of the number of
AOGs an open source software ImageJ was used. For this, it was important to have
a good contrast in the SEM images, as the software calculates the pixels in a certain
grey-scale of a given threshold.

3.2.4 Spectroscopic Ellipsometry and Reflectometry

Spectroscopic ellipsometer can be used to determine thickness and optical properties


such as refractive index n and absorption coefficient k of an optically transparent material.
The principle of ellipsometry is presented in Fig. 3.9. After light is linearly polarized, it
31

Figure 3.8. SEM cross-section image of a Scx Al(1-x) N thin film on AlNMo.

interacts with all thin-film layers and the wafer. Due to this interaction, the reflected light
becomes elliptically polarized. This signal is detected and can be analysed by finding a
good data fit of the measured data with optical models [68]. Advantages of ellipsometry
are that it can be used to measure the film thickness even if n and k are unknown, and
that it can measure multiple layers if their optical models are known.

Figure 3.9. Schematic illustration of spectroscopic ellipsometry [68].

The refractive index changes as Sc content varies [69]. Due to this ellipsometry was
used to determine the thickness of Scx Al(1-x) N thin films with different Sc content x. The
32

equipment used in this work was Semilab SE-2000 Spectroscopic ellipsometer in the
Micronova cleanroom facilities. As films were measured, first the refractive index model
was created by scanning the film at the centre and at the edge with four different angles,
60°, 65°, 70°, and 75°. When a good optical model was found by fitting existing models
to the measured data, the thickness measurement could be done on the whole wafer
at 49 measurement points across the wafer in order to determine the film thickness and
uniformity. These measurements were done at only one angle, 75°.

Reflectometry is another method that can be used to measure the layer thickness and
the refractive index of optically transparent materials. It is also based on the interaction
of light and its reflection at the thin film and the wafer interface, which is measured di-
rectly above the sample. However, it can be challenging to measure the thickness of
multiple layers with a reflectometer [70]. It is also challenging or even impossible when
n and k are not well known and when the films are slightly absorbing (k ̸= 0). In this
work FilmTekTM 2000M is used for measuring the thickness of the reference AlN and
Sc0.20 Al0.80 N films because their optical constants are well known.

3.2.5 Sheet Resistance

Sheet resistance R S is used to evaluate the thickness of Mo bottom and top electrodes
as they cannot be optically measured. The measurement is done with a probe with four
contact points. When all four points are in contact with the surface and current is applied
between the two outer contact points, the generated voltage between the two points in
the middle can be measured [71].

Sheet resistance is directly proportional to the specific resistivity of the material and in-
versely proportional to the thickness of the film, and thus they can be calculated if one is
known [71]. However, the resistivity of a thin film is not exactly the same as the resistivity
of a bulk material due to possibly different material density and defects in the thin film.
Thus the absolute thickness need to be verified by other methods e.g., by SEM cross
section. Nevertheless, sheet resistance measurement gives a good estimation of the film
thickness once the absolute thickness is verified.

3.3 Piezoelectric characterisation

For measuring the piezoelectric properties of the selected thin films, wafers were pat-
terned with a standard photolithography process described in Section 3.3.1. After the
cantilevers were diced, the transversial piezoelectric coefficient e31,f is measured with a
piezo-cantilever set-up presented in Section 3.3.2.
33

3.3.1 Microfabrication of Piezo-Cantilevers

For measuring the piezoelectric properties of Scx Al(x-1) N thin films, piezo-cantilevers were
fabricated. The deposition of metal electrodes and piezo-layer was described in Sec-
tion 3.1. A mask design (see Fig. 3.10) with 25 cantilevers was used for the cantilever
fabrication. There were three top electrode designs with different lengths and widths. In
each cantilever, there were also electrodes for measuring the capacitance and the dielec-
tric constant of the piezolayer in future measurements. The cantilevers were fabricated
using two mask process with three different etch steps: top-electrode layer patterning
and etching to open bottom-electrode contact window, wet etching of the piezo-layer in
the above opened window and finally patterning and etching of the top-electrode layer.

In the first step, the goal was to open contacts to the bottom electrode. To do this, a
standard photolithography step was carried out. Here, the wafers were spin coated with
a positive photoresist and soft-baked. Then, the wafers were aligned with the first mask
and exposed with ultraviolet (UV) light in the contact mode and developed. This opens
a window in the photoresist layer where the top Mo layer was etched by an inductively
coupled plasma (ICP) etcher using SF6 plasma. The remaining photoresist layer was
stripped using an oxygen plasma.

During the next step, the piezo-layer in the removed Mo window was wet etched using
tetramethylammonium hydroxide (TMAH) which was heated to 80°C. The unetched parts
of the top Mo layer was used as a protection mask during wet etching of the piezo-layer.
Etch durations depended on the thickness and Sc content of the piezo-layer. Finally, the
top-electrode layer was patterned using another lithography process as described above
using a second mask. The top-electrode layer was dry etched, and the remaining pho-
toresist layer was removed. Finally the wafers were cleaned in acetone and isopropanol
solvents.

Figure 3.10. Mask design for cantilever fabrication showing the different electrode de-
signs. The scale bar in the image is 20,000 µm.
34

After patterning the wafer, the cantilevers were diced with 50 and 100 µm thick dicing
saws. The width of each cantilever was 5 mm and length 72 mm. These dimensions are
determined by the piezo-cantilever test system.

3.3.2 Determination of Transversal Piezoelectric Coefficient e 31,f

The transversal piezoelectric coefficients e31,f can be measured with a four-point-bending


set-up [17, 21] or with a piezo-cantilever set-up [72]. The piezo-cantilever testing used
in this work is based on the converse piezoelectric effect. The test set-up is described
in Figure 3.11. The image shows the clamping, from which voltage can be applied to the
cantilever. The clamping also fixes the cantilever firmly to the system. As electric voltage
of ±30 V is applied to the top and bottom electrodes (TE and BE) of the cantilever, the
mechanical strain of ScAlN film deflects the cantilever system. This is due to the converse
piezoelectric effect as the film expands or shortens according to the voltage, and the
caused strain S1 bends the cantilever. The displacement is measured with an optical
displacement sensor. As described by Mazzalai et al. [72], the transversal piezoelectric
coefficient e31,f can then be calculated as

1 YCant (tCant )2 z(x2 )


e31,f =− , (3.2)
3 (1 − ν)cf x1 (2x2 − x1 ) V

where YCant is the Young’s modulus of the cantilever (169 MPa for Si [20]), tCant is the
thickness of the cantilever, and ν is the Poisson’s ratio of the cantilever material (0.064

Clamping Displacement
sensor
± 30 V

0
x1
x2

z(x2)
TE Mo x
ScAlN
BE Mo
SiO2
Si

Figure 3.11. Schematic presentation of piezo-cantilever test. Piezoelectric thin film is pre-
sented in green and top and bottom electrodes in grey. Blue represents the Si substrate
and orange is the SiO2 thermal oxide.
35

4.0E-07

3.0E-07

2.0E-07
Displacement [µm]

1.0E-07

0.0E+00

-1.0E-07

-2.0E-07

-3.0E-07

-4.0E-07
-30 -20 -10 0 10 20 30
Voltage [V]

Figure 3.12. Linear displacement versus voltage curve for a Sc0.23 Al0.77 N thin film.

for Si in [110] direction [20]). The ratio between cantilever and electrode width is cf , x1
is the length of the top electrode and x2 is the distance from the clamping to the mirror
surface where displacement is measured. The displacement z(x2 ) and the voltage V
are measured values. A typical displacement versus voltage curve for AlN and Scx Al(1-x) N
films is presented in Fig. 3.12. The value used in calculation is the slope z/V of this curve.
36

4 RESULTS AND DISCUSSION

In this section the most important experiments and the results are presented and dis-
cussed. First, in Section 4.1, the effect of the different process parameters on Scx Al(x-1) N
thin films are resented. In Section, 4.2 the process parameters for bottom and top Mo
electrodes are discussed. Finally, the results of the piezoelectricity measurements of
different Scx Al(1-x) N films are presented and discussed in the context of other results in
Section 4.4.

4.1 ScAlN Thin Films by Reactive Magnetron Co-Sputtering

The influence of the different process parameters on Scx Al(1-x) N film growth was studied
with several deposition processes, with varying scandium content. During optimisation,
Scx Al(1-x) N thin films were deposited on p-type (100) silicon wafers on both bare Si and
Si wafers with thermal oxide and Mo bottom electrode. The goal is to have well textured
columnar Scx Al(1-x) N films without abnormally oriented grains on Mo bottom electrode with
Sc content above 30%.

4.1.1 Influence of Process Chamber Conditioning

Condition of the process chamber remains stable as long as it is kept at vacuum base
pressure while there are no processes running. However, the condition changes when
the chamber is opened, the sputtering target is changed or in the case of multi-source
sputtering chamber, other materials have been deposited in the same chamber. Other
processes may also create contamination on the chamber walls which need to be re-
moved. For this reason extra dummy wafers are needed to condition the chamber at
the beginning of each process run before the process wafers are deposited. This condi-
tioning removes oxides and other possible contaminants from the target and the process
conditions become more stable for the process wafers, also thermally. The number of
conditioning wafers depends on the process. For standard metal processes typically two
conditioning wafers are enough, but for Scx Al(1-x) N thin-film deposition more dummies are
needed.

The conditioning effect of two different co-sputtered Scx Al(1-x) N process runs on different
37

Figure 4.1. SEM images of co-sputtered ScAlN thin films at wafer centre in order: a, d)
first wafer, b, e) third wafer and c, f) fifth wafer. The stress value indicates local film stress.

days are presented in Fig. 4.1. Run 1 is presented in Fig. 4.1a–c, and Run 2 in Fig. 4.1d–f
in which the first, the third and the fifth conditioning wafers of the Scx Al(1-x) N are shown.
The Sc content of the films in Run 1 is 24.0% and in Run 2 it is 33.5%. As can be seen
on the images, the conditioning effect is more pronounced during Run 1. The film of the
first wafer in Fig. 4.1a looks already rather good. However, the film stress value indicated
in the Figure 4.1a, is rather compressive (−365 MPa) which has lowered the number of
abnormally oriented (AOGs). On the third wafer in the process in Fig. 4.1b, AOGs cover
the whole wafer surface but on the fifth wafer, presented in Fig. 4.1c, the number of AOGs
has decreased again. Also, the film stress during the first five wafers has not stabilized to
a certain value.

For Run 2, the SEM images in Fig. 4.1d-f are the films of the first, third, and fifth wafer
which were processed after the co-sputtering process module was opened for mainte-
nance. Then the module was pumped down to vacuum base pressure. The effect of
38

process conditioning is more challenging to see in this run, which may be due to lower to-
tal power than in the first example. The biggest difference can be seen in local film stress.
On the first wafer in Fig. 4.1d, the local film stress is −410 MPa, while the film stress of
the third and the fifth wafer in Figures 4.1e and f, has stabilised at around −670 MPa. The
film stress is very compressive on all of the wafers, which in this case is more related to
other process parameters. However, during the first five wafers, the stress has stabilized
around a certain value.

Even if the first wafers of both process runs look reasonable regarding the number of
AOG, they are not generally of desired film quality. The target may still be contaminated
due to other processes or chamber maintenance. Even if the film would look good, the
lattice parameters and crystal structure are not necessarily correct. This was seen for ex-
ample when the film thickness was measured optically as the data fitting of the measured
curve is not good for the first wafer. To confirm this, the first conditioning wafers could be
inspected in addition by XRD.

In general, this experiment would need to be repeated in order to better understand,


how many wafers are needed for the conditioning. However based on this experiment,
five conditioning dummy wafers were used in this work. The co-sputtering process is
rather time consuming, as five conditioning dummy wafers corresponds approximately to
approximately five hours of process time.

4.1.2 Influence of Target Power on Sc content

The dependency of the film composition and the Sc/(Sc+Al) target power ratio is pre-
sented in the graph in Figure 4.2. The scandium concentrations presented in the graph
are measured by EDX on an SEM at VTT Micronova cleanroom facilities. Samples that
are deposited with a total target power of 1000 W are presented in the graph with circles.
In this case Sc target power was varied between 250–400 W. Square markers represent
target power ratios for which Sc target power is varied between 300–655 W and Al target
power is kept constant at 1000 W. Consequently the total power varies. The graph con-
firms that the Sc content in the co-sputtering process depends linearly on the Sc/(Sc+Al)
target power ratio of the used targets independently from the total power. As power of the
scandium target is increased, the scandium content increases.

However, greater power on Sc target also affects the film growth and therefore the amount
of abnormally oriented grains as seen in Figure 4.3. The film presented in Fig. 4.3a (De-
position 1) is deposited with high total power of 1450 W, 450 W on Sc-target and 1000 W
on Al-target. The film in Fig. 4.3b (Deposition 2) is deposited at approximately the same
power ratio of 0.3, but with total power of 1000 W, with Sc-target power 300 W and Al-
target power 700 W. The Sc concentrations were measured to be 24.0% and 23.1%,
respectively. The films presented in Fig. 4.3c and d (Depositions 3 and 4) are deposited
39

similarly with the power ratio of approximately 0.4. The film in Fig. 4.3c is deposited with
Sc target power 655 W and Al target 1000 W, and Fig. 4.3d with Sc target power 400 W
and Al target power 600 W. The Sc concentrations were measured as 33.7% and 33.5%,
respectively.

As the films deposited with high total power (Depositions 1 and 4) are compared to low
total power processes with corresponding power ratios (Depositions 2 and 3), it seems
that especially the size of AOGs is larger in films deposited with high total power. The
size difference could be due to increased Sc power. Interestingly, the film stress of the
film from Depositions 2 is in the target range of 0±50 MPa. Normally, more tensile stress
leads to more AOGs, which is not the case in this experiment. It seems that the number of
AOGs can be controlled by lowering the total power, while keeping the target Sc/(Sc+Al)
power ratio the same and thus getting the desired Sc content.

For the films Depositions 1–3 in Fig. 4.3a, b, and d, the process gas flows were 2.5 sccm
for Ar and 30 sccm for N2 . The film from Deposition 4 shown in Fig. 4.3c is deposited
with Ar flow of 7.8 sccm and N2 flow of 31.5 sccm. The intention was to use approxi-
mately the same process gas flows, but unfortunately a malfunction of the Ar mass flow
controller was detected after Deposition 3. Thus Ar flow was lower than intended for De-
positions 1–3. Normally increasing total power would lead to decreased film stress, but
in the case of Deposition 4 this was compensated by increased process gas flow. Also, it
was discovered before Deposition 4 in Fig. 4.3c that adding a ramping up of the Sc target
power during the first 180 s decreases the number of AOGs as the crystallinity of the film

50
45
40
35
Sc content [%]

30
25
20
15
10 Total power 1000W
5 Al-target power 1000W
0
0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Sc/(Sc+Al) power ratio

Figure 4.2. Scandium content versus Sc/(Sc+Al) target power ratio in co-sputtering.
40

Figure 4.3. SEM plane views of ScAlN thin films deposited with similar Sc/(Sc+Al) target
power ratios of approx. 0.3 in a–b) and 0.4 in c–d). The target powers are a) Sc target
power 450 W, Al target power 1000 W, b) total power of 1000 W: Sc-target power 300 W,
Al-target power 700 W, c) Sc target power 655 W, Al 1000 W, and d) total power of 1000 W:
Sc target power 400 W, Al target power 600 W.

is better, which is used for this process. The Sc target power ramp-up does not affect the
film stress. Nevertheless, it seems that the surface area covered by AOGs increases as
the total power increases. To verify the effect of the total power, this experiment should
be repeated with the same process gas flows.

4.1.3 Influence of Process Gas Flows and Process Pressure

To evaluate the effect of the total amount of process gas and process pressure, Scx Al(1-x) N
thin films were deposited with a constant Ar flow of 9.5 sccm while N2 flow was increased.
In Figure 4.4 can be seen how the film stress changes when N2 flow is varied. The film
stress depends linearly on the process gas flow. Figure 4.5 shows SEM plane views of
41

150

100

50

0
Stress [MPa]

-50

-100

-150

-200 Stress X

-250 Stress Y

-300
20.0 22.0 24.0 26.0 28.0 30.0
N2 flow [sccm]

Figure 4.4. Sc0.24 Al0.76 N film stress versus N2 flow. Ar flow was kept constant at 9.5 sccm.

wafers from the same Sc0.24 Al0.76 N process. The N2 flows of Fig. 4.5a–c are 21 sccm,
24 sccm, and 28 sccm, respectively. It can be seen how the amount of AOGs and stress
at the wafer centre changes when N2 flow is increased.

As the total gas flow and thus the process pressure increases, the film stress changes
towards tensile stress. This increases the number of AOGs as well, which is align well
with literature [15]. In this experiment, only N2 flow was changed, which affected not only
the total pressure but also the process gas ratio. To study the effect of gas mixture more
in detail in future, it would be needed to deposit Scx Al(1-x) N thin films with a same total
process gas, but varying the ratio of Ar and N2 in the gas mixture.

4.1.4 Influence of Deposition Time and Film Thickness

To evaluate the effect of film thickness on the number of AOGs, films were deposited
with the same parameters but different process time. Samples shown in Figure 4.6 are
of co-sputtered Scx Al(1-x) N thin films with a 24.0% Sc-content sputtered on a 150 nm Mo
bottom electrode (BE). The thickness of the film in Fig. 4.6a is 580 nm and in Fig. 4.6b it
is 980 nm. The deposition times were 3000 s and 5130 s, respectively.

By comparing Fig. 4.6a and b, it can be seen that there are more abnormal grains on the
980 nm thick film than on the 580 nm thick film. The grain size is also different. The
greatest dimensions of the largest grains of the 580 nm thick film are 100–200 nm and
the smallest grains are approximately 50–70 nm. The grains in 980 nm thick film seem to
42

Figure 4.5. SEM plane views of Scx Al(1-x) N thin films with 24.0% Sc content deposited
with different gas flows. The effective Ar flow was kept constant at 9.5 sccm while N2 flow
was varied: a) 21 sccm, b) 24 sccm, and c) 28 sccm.

Figure 4.6. SEM images of co-sputtered Sc0.24 Al0.76 N thin films of different thickness on
AlNMo bottom electrode. a) 580 nm Sc0.24 Al0.76 N film and b) 980 nm Sc0.24 Al0.76 N film
co-sputtered with the same process parameters with different time.
43

grow larger. Their greatest lengths are in the range of 250–400 nm for the big grains and
110–140 nm for the small grains. Due to their size, AOGs also cover greater area as the
film grows thicker. The area coverage of the 580 nm thick film is approximately 6%, while
in the 980 nm thick film AOGs cover 35% of the surface area.

The size difference of the grains in the two examples could be explained by assuming
that the grains continue growing as the film grows thicker. It was presented in Sec-
tion 2.2.2 that AOGs grow faster than the rest of the film, which explains the clearly bigger
AOGs [14]. It is likely that some of the grains would start growing on defects at the inter-
face of substrate and film. However, there are also more small abnormally oriented grains
in the film in Fig. 4.6b as well. This would indicate that at least some of the AOGs start
growing inside the film. The grains that start growing in the film itself could start growing
at grain boundaries where Sc is likely to accumulate [14].

This could be confirmed for example by depositing the thin film in two different layers.
Once parameters for a thin film with no grains are found, the film could be used as the
first layer. Then, during a new process run starting with the first layer deposition, the
second layer would be deposited on top without stopping the process. For the second
layer, one of the parameters that is known to increase the number of AOGs could be
changed. By comparing the resulting film and the film without AOGs, it would be possible
to see if AOGs started growing in the film due to process parameter change mid-process.

4.2 Magnetron Sputtering Process for Mo Electrodes

As the final goal of work is to deposit the Scx Al(1-x) N thin films on Mo bottom electrodes, it
was important to optimize Mo sputtering process. Different aspects of process optimiza-
tion for bottom electrode (BE) and top electrode (TE) Mo layers are investigated in this
section.

4.2.1 Influence of Process Pressure

As explained in Section 3.1.2, Mo is grown on a 30–35 nm thick AlN seed layer. The AlN
seed layer process was kept constant for all the different bottom electrode Mo processes.
For bottom electrode (BE) different Mo thicknesses were evaluated. Therefore, it was
necessary to calibrate the process separately for each thickness in order to reach a film
stress close to zero. As discussed in Section 4.1.3, the increase in process gas flow
shifts film stress towards tensile stress. The graph in Figure 4.7 shows how the film stress
depends on Ar flow for different Mo bottom electrode thickness. The thinner the film is,
the more Ar is needed to reach zero stress. However, for a given thickness the film stress
depends linearly on Ar flow, and there is a parallel shift for the film stress versus Ar flow
curves at different film thicknesses. Thus, it is possible to evaluate the Ar flow for zero
44

film stress for a new film thickness with a simple calculation. For the top electrode (TE)
Mo, which is deposited on top of AlN or Scx Al(1-x) N films, the film stress can be adjusted
the same way.

50

-50
Stress [MPa]

-100

-150

100nm Mo Stress
-200
180nm Mo Stress

-250
11.0 12.0 13.0 14.0 15.0 16.0 17.0 18.0 19.0
Ar flow [sccm]

Figure 4.7. Film stress versus Ar flow of different bottom electrode Mo thickness. The
filled markers are measured film stress in x-direction, and open markers film stress in
y-direction. Mo is grown on AlN seed layer.

0.86

0.85
Deposition rate [nms-1]

0.84

0.83

0.82

0.81

0.8
12 13 14 15 16 17 18
Ar flow [sccm]

Figure 4.8. Deposition rate versus Ar flow of Mo in electrode deposition.


45

As Ar flow is changed for different bottom electrode thickness, the deposition rate changes
slightly as shown in Figure 4.8. This can be explained by increased Ar+ ion bombardment
on the target and thus sputtered material. However, deposition rate does not increase
infinitely; eventually there is so much Ar present that the Ar+ ions start colliding with each
other and with sputtered particles, and eventually less sputtered material arrives on the
substrate.

To better understand the influence of the Ar flow on the deposition rate, films should be
deposited with the same process time while varying the Ar flow. After this, the thickness
of the deposited film is measured in SEM cross section and the deposition rate for each
Ar flow can calculated.

4.2.2 Influence of Substrate Temperature

Sheet resistance R S can be used in evaluating the thickness and uniformity of a metal
film which cannot be measured optically. The effect of chuck temperature on the Mo top
electrode film growth is shown in the graph presented in Figure 4.9. Films were deposited
with a constant Ar flow of 12 sccm. The thickness of the films was approximately 280 nm.

As can be seen in the graph, as temperature increases, sheet resistance decreases. This
would indicate that there are fewer defects, such as grain boundaries, in the film which
would increase the resistance. As temperature is increased, Mo grains grow larger which

0.385

0.380

0.375
Sheet Resistance [Ω/sq]

0.370

0.365

0.360

0.355

0.350

0.345

0.340
250 300 350 400 450
Temperature [°C]

Figure 4.9. Sheet resistance versus substrate temperature of Mo top electrode deposi-
tion. Ar flow is kept constant at 12 sccm. The film thickness was approximately 280 nm.
46

leads to fewer grain boundaries. This also leads to better oriented grain growth and thus
better film texture, which should be confirmed by XRD in a future experiment.

Electrode thickness has a similar effect on the sheet resistance R S as can be seen in the
Figure 4.10. As the film grows thicker, grains grow in height which leads to decreased
sheet resistance R S . At fixed deposition temperature, R S depends linearly on the film
thickness. This was confirmed by depositing Mo thin films with different thickness in the
same process temperature of 300°C, and measuring the sheet resistance.

4.2.3 Influence of Underlaying Material

As was explained in Section 3.1.2, good crystallinity is crucial for Mo film quality, especially
as another layer is to be deposited on top of the film. In case of the bottom electrode, the
piezoelectric layer is grown on top of it, and thus good crystallinity is desired. Piezoelectric
devices are often grown on an insulating thermal SiO2 . Because SiO2 is amorphous, Mo
cannot grow well textured directly on this oxide, which can be seen in the AFM scan
presented in Figure 4.11a. The grains look more randomly oriented and elongated and
the mean roughness S a is 1.3 nm. Figure 4.11b shows an AFM scan of Mo grown on
well textured 1000 nm thick AlN film. In this case, the grains have grown bigger and their
shape has become rounder. The roughness S a of this film is 2.9 nm.

As a result, it can be concluded that the smooth roughness of the film surface does not
necessarily mean good crystallinity or that the next layer would grow well on the bottom

0.850

0.800

0.750
Sheet resistance [Ω/sq]

0.700

0.650

0.600

0.550

0.500

0.450

0.400
120.0 140.0 160.0 180.0 200.0 220.0
Thickness [nm]

Figure 4.10. Sheet resistance versus thickness of Mo top electrode deposited at 300°C.
47

(a) Mo on SiO2 , Sa = 1.3 nm (b) Mo on 1000 nm AlN, Sa = 2.9 nm

Figure 4.11. AFM scans of Mo top electrode on different substrates.

layer. The bottom Mo layer works as a seed for the next layer, and for good crystallinity of
the piezo-layer, the crystallinity of the Mo bottom electrode should be good as well.

To verify if Mo grows with better crystal orientation on AlN than on SiO2 , films should
be measured with XRD. However, it can be assumed that the 1000 nm thick AlN is well
oriented also due to its thickness. To study the Mo growth better, an AFM scan and XRD
measurements should also be done for Mo grown on 30–35 nm AlN seed layer.

4.3 Influence of Mo Bottom Electrode on ScAlN Thin-Film Growth

In order to investigate how bottom-electrode thickness affects the piezolayer quality, co-
sputtered Scx Al(1-x) N films were deposited on Mo bottom electrodes with four different
thickness. Before piezolayer deposition, bottom electrodes were measured with AFM and
XRD. Results are presented in Table 4.1. Rocking curve value was at lowest, at 1.74°, for
the 225 nm thick Mo, which is the thickest of the selected bottom electrode layers. This
is logical, as the crystallites of the film get more aligned as the film grows thicker and as
result RC gets better. However, at the same time roughness of this film was also the high-
est with S a value of 1.05 nm. The thinnest film had the lowest roughness (S a =0.80 nm),
but its crystallinity was the worst of this series (RC = 2.27°). The results of 150 nm and
190 nm films were approximately the same, and consequently less conclusive.

After the measurements of different Mo BEs, a 580 nm thick Scx Al(1-x) N thin film with
24.0% Sc content was deposited on Mo electrodes which were deposited with the same
process parameters. The SEM images at wafer centre of Scx Al(1-x) N thin film plane views
are shown in Figure 4.12. The BE thicknesses in the images are a) 100 nm, b) 150 nm,
c) 180 nm, and d) 225 nm. The average film stress of all BEs was around +60 MPa. As
can be seen in the images, there are fewest abnormally oriented grains on the Scx Al(1-x) N
film which is deposited on the thinnest bottom electrode and most AOGs on the thickest
48

Table 4.1. Properties of Mo bottom electrodes of different thickness.

Thickness Power Process time Ar flow Film stress Rocking curve Roughness S a
[nm] [W] [s] [sccm] [MPa] [deg] [nm]
100 1000 128 17 −55 2.27 0.80
150 1000 191 15 −50 1.91 0.95
190 1000 233 14 −60 1.89 0.93
225 1000 274 13 −60 1.74 1.05

Figure 4.12. SEM images of different bottom electrode Mo thickness of a) 100 nm,
b) 150 nm, c) 180 nm, and d) 225 nm. The local film stress indicated in the image is
stress of the Sc0.24 Al0.76 N film.

bottom electrode. The local stress of the Scx Al(1-x) N film on 150 nm thick Mo BE is higher
than expected but the number of AOGs seem lower than for the films on thicker BE.

This experiment series shows that there are less abnormally oriented grains on thinner BE
Mo than on thicker electrodes. In addition, the local film stress of the Scx Al(1-x) N film also
seems to increase as the bottom electrode is thicker. However, the experiment should be
repeated to confirm this. The effect of bottom electrode roughness could be evaluated
by optimizing the Scx Al(1-x) N process parameters so that the Scx Al(1-x) N film stress would
be around the same value for all films and then comparing the number of abnormally
oriented grains. Nevertheless, so far it can be concluded that thinner Mo BE is beneficial
for the Scx Al(1-x) N film growth.

4.4 Piezoelectric Properties of ScAlN Thin Films

One of the purposes of this thesis was to evaluate whether the co-sputtered Scx Al(1-x) N
thin films are piezoelectric or not. For this purpose, the transversal piezoelectric coeffi-
cient e31,f of Scx Al(1-x) N thin films with varying Sc content was measured. In this section
49

the result of piezo-cantilever measurements are presented and evaluated together with
other results. The selected co-sputtered Scx Al(1-x) N thin films were approximately 600 nm
thick and their Sc concentrations were determined to 23.1%, and 33.5%. The new ref-
erence samples were a 1000 nm AlN thin film, a 1000 nm thick Sc0.20 Al0.80 N film, and
a Sc0.30 Al0.70 N. All three films were sputtered from pure Al target or ScAl alloy target
with 20% and 30% Sc content, respectively. The sputter process module was one of
the single-target modules of the Evatec Clusterline 200 II. For the co-sputtering of the
Scx Al(1-x) N films, a multi-target chamber of the Evatec Clusterline 200 II was used. The
other two reference wafers for AlN and Sc0.20 Al0.80 N were deposited earlier with other
sputtering systems at Micronova cleanroom facilities.

4.4.1 Influence of Sc Content on Piezoelectric Coefficient e 31,f

Transversal piezoelectric coefficient e31,f was measured with a piezoelectric cantilever


set-up discussed in Section 3.3. The measured results are presented in Figure 4.13. The
red line represents the theoretical curve of e31,f by Caro et al. [38]. When the error of
the measurement is taken into account, all the AlN and Sc0.20 Al0.80 N references and co-
sputtered Scx Al(1-x) N thin films follow the theoretical curve closely. The literature value of
e31,f for AlN is −1.05Cm−2 [5]. In this thesis, the measured value of both AlN reference
films deposited with different sputtering equipment was slightly less. An explanation could
be that there is a small systematic measurement error on the used set-up. If the AlN
result were to be corrected to the literature value, it would mean that results for the co-

-3

-2.5

-2
e31,f [Cm-2]

-1.5

-1 Old reference wafers


Single target references
-0.5 Co-sputtered films
Poly. (Caro)
0
0 5 10 15 20 25 30 35 40
Sc content [%]

Figure 4.13. Transversal piezoelectric coefficient e31,f versus Sc content.


50

sputtered Scx Al(1-x) N could also be at the higher end of the error bars presented in the
graph. However, already as measured, the presented e31,f values of the co-sputtered
films are well-aligned with literature.

The only exception in this batch of samples is the film with 30% Sc content reference
film deposited from a single alloy target. This film did not behave as expected. The
result was about 15% lower than the theoretical value and approximately the same as the
reference samples with 20% Sc. This is one of the first wafers deposited from this 30%
ScAl alloy target. One explanation for the low value might be that there is a problem with
the sputtering process for the alloy target, meaning, for example, that it was still not fully
optimised. AOGs covered approximately 40% of the wafer surface at the wafer centre,
and the number decreased to 20% coverage towards the wafer’s edge (see Fig. 4.15d and
h, respectively). However, the only working cantilevers were at the centre of the wafer,
where the highest number of AOGs were situated. Another issue could be a problem with
the target quality, a matter that cannot be excluded either.

4.4.2 Influence of Stress on Piezoelectric Coefficient e 31,f

The stress dependency of piezoelectric coefficient e31,f is presented in the graph in Fig-
ure 4.14. The three films compared in the graph were deposited with slightly varied pa-
rameters. For the film of wafer W1 the total gas flow is higher, 31 sccm, than for the

-2.80
-2.70
-2.60
-2.50
-2.40
e31,f [Cm-2]

-2.30
-2.20
-2.10
-2.00
W1, 31sccm, 180s Sc Ramp Up
-1.90
W2, 28sccm, 240s Sc Ramp Up
-1.80 W3, 28sccm, 180s Sc Ramp Up
-1.70
-850 -750 -650 -550 -450 -350 -250
Local film stress [MPa]

Figure 4.14. Transversal piezoelectric coefficient e31,f versus local film stress. The Sc
content of the films is approximately 33.5%.
51

Table 4.2. Properties of Scx Al(1-x) N thin films with varying Sc-content.

Radius [mm] Parameter Unit 33.5% 33.5% 23.1% 30.0%


R00 Film stress MPa −280 −500 −520 −130
RC [deg] 1.38 1.39 1.67 1.74
AOG fraction [%] 2.5 2.9 2.5 38.2
−2
e31,f [Cm ] −2.40 −2.36 −1.72 −1.55
R60 Film stress MPa −540 −750 −800 −290
RC [deg] 1.85 1.90 1.84 1.70
AOG fraction [%] 0.6 1.3 1.5 19.9
−2
e31,f [Cm ] −2.31 Dysfunctional Dysfunctional Dysfunctional
cantilever cantilever cantilever

other wafers, and scandium target ramp-up time was 180 s. Ramp-up time means that
instead of immediately switching on the full power on the Sc target, the power was slowly
increased to full during the given time frame. This way the Sc content increases gradually
during initial film growth, and a better crystallinity of the film can be achieved. For wafers
W2 and W3 the total gas flow was the same, 28 sccm, but for wafer W2 the Sc target
ramp-up time was increased to 240 s.

When the film stress and the piezoelectric coefficient e31,f for each wafer are compared,
it looks like there is a trend within a single wafer. However, the measured e31,f is approx-
imately the same on different wafers even at the different stress ranges. Within a single
wafer the e31,f value could increase as the film stress goes towards tensile at the wafer
centre. This is well-aligned with the previous research [15]. This further trend should be
confirmed by optimizing the film stress closer to 0±50 MPa. However, the Sc content
has a stronger effect on the e31,f value, and thus optimized stress might help only a little.
This is an important note that should be taken into account when devices are designed
across the wafer surface. Thin-film quality can change drastically at different positions of
the wafer, and thus devices do not necessarily work the same way. This effect can be
reduced by optimizing the film across the wafer. However, it is necessary to measure the
piezoelectric properties during this optimization, as discussed in detail in the next section.

4.4.3 Piezoelectric Performance Compared to Film Structure

Thin-film crystallinity is one of the most important measurements when it comes to eval-
uating the general thin-film quality. As discussed previously in Section 2.2.2, the number
of abnormally oriented grains also affect the total piezoelectricity of the film. However,
sometimes a piezoelectric device still does not work even if all the structural measure-
ments indicated a adequate film quality.
52

This is the case with some of the films presented in Fig. 4.15. The local film stress, the
XRD rocking curve, the surface coverage by the AOGs, and the transversal piezoelectric
coefficient e31,f at the wafer centre and 60 mm towards the wafer flat are summarized in
Table 4.2. As can be seen in the SEM images in Fig. 4.15, the abnormal grains are few in
number and they are mostly not connected to each other for the first three co-sputtered
films. The surface area covered by AOGs is less than 10% for the films presented in
Fig. 4.15a–c and e–g. This is assumed to be a threshold, where the e31,f is expected to
drop [15]. For the fourth wafer in Fig. 4.15d and h the 40% and 20% of the surface area
is covered by AOGs, respectively.

For all the presented co-sputtered samples, cantilevers from the wafer centre worked as
expected. The crystallinity of these thin films is excellent (RC<1.7°) and piezoelectric
coefficient e31,f is as expected. At the wafer edge, crystallinity is not as good, but RC
at the wafer edge is still below 2° for each wafer, which normally is a sign of a well-
textured film [40]. However, from the thin films at R = 60 mm presented in Fig. 4.15e-
g, cantilevers worked as expected only on the film presented in Fig. 4.15e. Also the
cantilevers which were deposited from ScAl alloy target with 30% Sc content did not work
as expected. Possible reasons for this were discussed previously in Section 4.4.1. But
even for this film the c-axis texture determined by XRD RC = 1.74° shows a good value

Figure 4.15. SEM plane views at different radius of selected ScAlN films Mo bottom
electrode. Wafers taken to SEM before Mo top electrode deposition.
53

below 2°. Evaluating the film only by the SEM images, XRD RC and film stress, it would
not be possible to tell if the film show a high piezoelectricity and therefore, if the final
device would work.

One of the reasons that a cantilever does not work as expected might be that there are
defects in the Scx Al(1-x) N film which lead to leakage current between the two electrodes.
This is more likely to occur in films with a higher Sc content. As discussed before in
Section 2.2.2, Sc is likely to accumulate at grain boundaries. This may create a channel
between the electrodes for the leakage current. The area covered by the top electrode
was approximately the same, but length and width varied. Also, on longer cantilever
electrodes there is greater risk that the quality of the film changes underneath.

The leakage current was visible especially in the case of film presented in Fig. 4.15g.
For an as-expected working cantilever, the displacement versus voltage curve is linear
without any hysteresis and follows the same path both when voltage is increased and
decreased, as was shown in Fig. 3.12 in Section 3.3.2. For a cantilever which does not
work as desired, the curve becomes non-linear (see Fig. 4.16a). In the case of this film, a
small hysteresis effect could be seen. When the voltage was ramped up from 0 to +30 V,
a linear increase in displacement was observed. As voltage was ramped down from 0 to
-30 V, an initial change in displacement was observed, after which the effect plateaued
and the displacement remained constant from approximately −10 V to −30V. When only
positive voltage was applied, a hysteresis was detected (see Fig. 4.16b) which is a sign
of dielectric loss in the film in this case. The displacement and calculated e31,f of this type
of cantilever was approximately half of the value of a cantilever with linear displacement
curve. However, for all of such cantilevers from the same wafer exactly the same value
was measured, and even below the e31,f value of pure AlN. As a next step to confirm this,
dielectric losses of this type of a film should be measured.
54

1.0E-07

5.0E-08
Displacement [µm]

0.0E+00

-5.0E-08

-1.0E-07
-30 -20 -10 0 10 20 30
Voltage [V]

(a) Non-linear behaviour of a cantilever.

1.0E-07

5.0E-08
Displacement [µm]

0.0E+00

-5.0E-08

-1.0E-07
0 5 10 15 20 25 30
Voltage [V]

(b) Non-linear cantilever measured with positive voltage only.

Figure 4.16. Examples of non-linear displacement vs. voltage curve of a ScAlN film with
23.1% Sc content at R = 60 mm.
55

5 CONCLUSIONS

The goal of this thesis was to deposit piezoelectric Sc doped aluminium nitride (Scx Al(1-x) N)
thin films with over 30% Sc content on Mo bottom electrodes. The films were deposited
by reactive magnetron co-sputtering from pure Sc and pure Al targets. The amount of
Sc in the film was controlled by varying the target power ratio. The crystallinity, amount
of abnormally oriented grains (AOGs) and Sc content of the thin films were evaluated by
X-ray diffraction, scanning electron microscopy (SEM), and energy dispersive X-ray spec-
troscopy (EDX). Finally, the piezoelectric performance was evaluated by determining the
transversal piezoelectric coefficient e31,f with a piezo-cantilever set up.

One of the challenges of this research work was to find the suitable sputter process pa-
rameters for Scx Al(1-x) N thin films with high piezoelectric response. For 23% and 33% Sc
containing films columnar grown films with c-axis texture and a reasonable small amount
of abnormally oriented grains (surface area coverage less than 5%) were obtained. Ab-
normally oriented grains (AOGs) start to appear in Scx Al(1-x) N thin films already at low Sc
contents, and their number increases as the Sc content increases. AOGs are grains in the
desired wurtzite crystal structure, but their polar c-axis is slightly tilted. This disturbs the
total piezoelectric performance of the deposited thin film. They also appear more likely as
film stress turns towards tensile. Therefore, the sputter process and thin-film growth was
tuned to deposit Scx Al(1-x) N with a reasonable compressive film stress in the order of the
−500 MPa to 0 MPa. This was controlled by varying the target power ratio and process
pressure.

Another topic of this thesis was to investigate the influence of Mo bottom electrode (BE)
on the Scx Al(1-x) N thin-film properties. For this, the thickness of Mo BE was varied and
characterised by XRD and AFM. It seems that the thicker the Mo BE is, the better its
crystallinity is. However, at the same time the film surface becomes rougher. Scx Al(1-x) N
thin films that were grown on thinner Mo BE had less AOGs than films grown on thicker
Mo BE. From this result it can be concluded that thinner bottom electrodes are prefable
for the Scx Al(1-x) N thin-film growth. How this is related to the film roughness needs to be
investigated separately.

In co-sputtering, the most important factor regarding the Sc content of the Scx Al(1-x) N
thin films is the ratio Sc/(Sc + Al) of the target powers. It was confirmed in this thesis
56

that the Sc content follows linearly the target power ratio, independent of the total, re-
spectively absolute power. This can be used to select rather precisely the needed target
powers to deposit Scx Al(1-x) N films with a certain Sc content independent from other pro-
cess parameters. Thickness and Sc content uniformity across the wafer was verified by
spectroscopic ellipsometry and EDX, respectively. The Sc content varied less than 0.5%
across the wafer.

The highest Sc content of Scx Al(1-x) N thin films which was deposited during this work
with reasonable compressive film stress and crystallinity was determined to be at 33.5%.
The piezoelectric response of those films was measured with piezo-cantilever test set-up.
The transversal piezoelectric coefficient e31,f at the wafer centre was approx. −2.4 Cm−2 ,
which is well-aligned with the theory and literature. This is over 2.3 times higher than the
literature value for pure AlN, and more than 2.6 times higher than the measured AlN refer-
ence films. However, for some of the measured wafers a low number of AOGs and good
crystallinity with XRD rocking curve below 2° indicate high quality films as well, but at the
same time reasonable piezoelectric response was not detected across the whole wafer
surface. Hence, it is evident that one should measure not only the thin-film microstructure,
but also the piezoelectric response for process development and monitoring of Scx Al(1-x) N
thin films with high Sc content.

Future research could be e.g. to deposit Scx Al(1-x) N with Sc content between 33-42% with
good crystallinity and high piezoelectric response. This can be done by continuing the
work based on the results of this thesis. Also the already obtained processes for 23%
and 33% Sc concentrations could be further improved so that the film stress would be
near 0±50 MPa. Another interesting aspect could be to investigate more the effect of Mo
bottom electrode thickness. It is still unclear whether the Mo roughness or crystallinity is
more dominant factor for good Scx Al(1-x) N film growth.

As high Sc containing Scx Al(1-x) N thin films have higher piezoelectric coefficients than
AlN, they can potentially improve the performance of AlN based filters and resonators, as
well as energy harvesters and piezo-MEMS devices. In order to evaluate this, it should
be analysed which Sc content is suitable for different types of devices and applications,
leading to better efficiency and possibly to even smaller device size. The result of this work
is a good foundation for future device design improvements and process development of
ScAlN piezoelectric-driven thin-film devices.
57

REFERENCES

[1] Muralt, P. Recent Progress in Materials Issues for Piezoelectric MEMS. Journal of
the American Ceramic Society 91.5 (May 2008), 1385–1396. DOI: 10.1111/j.
1551-2916.2008.02421.x.
[2] Gorelick, S., Dekker, J., Guo, B. and Rimminen, H. Piezoelectrically Actuated Lin-
ear Resonators on Ring-shaped Suspensions for Applications in MEMS Phase-
sensitive Gyroscope. Procedia Engineering 87 (2014). EUROSENSORS 2014, the
28th European Conference on Solid-State Transducers, 1597–1600. DOI: https:
//doi.org/10.1016/j.proeng.2014.11.532.
[3] Pensala, T., Kyynäräinen, J., Dekker, J., Gorelick, S., Pekko, P., Pernu, T., Ylivaara,
O., Gao, F., Morits, D. and Kiihamäki, J. Wobbling Mode AlN-Piezo-MEMS Mirror
Enabling 360-Degree Field of View LIDAR for Automotive Applications. 2019 IEEE
International Ultrasonics Symposium (IUS). 2019, 1977–1980. DOI: 10 . 1109 /
ULTSYM.2019.8925660.
[4] Karuthedath, C. B., Sebastian, A. T., Saarilahti, J., Sillanpaa, T. and Pensala, T. De-
sign and Fabrication of Aluminum Nitride Piezoelectric Micromachined Ultrasonic
Transducers for Air Flow Measurements. 2019 IEEE International Ultrasonics Sym-
posium (IUS). 2019, 2489–2492. DOI: 10.1109/ULTSYM.2019.8925544.
[5] Muralt, P. Piezoelectric thin films for mems. Integrated Ferroelectrics 17.1-4 (Sept.
1997), 297–307. DOI: 10.1080/10584589708013004.
[6] Ruby, R., Bradley, P., Larson, J., Oshmyansky, Y. and Figueredo, D. Ultra-miniature
high-Q filters and duplexers using FBAR technology. 2001 IEEE International Solid-
State Circuits Conference. Digest of Technical Papers. ISSCC (Cat. No.01CH37177).
2001, 120–121. DOI: 10.1109/ISSCC.2001.912569.
[7] Rinaldi, M., Zuniga, C., Zuo, C. and Piazza, G. Super-high-frequency two-port AlN
contour-mode resonators for RF applications. IEEE Transactions on Ultrasonics,
Ferroelectrics, and Frequency Control 57.1 (2010), 38–45. DOI: 10.1109/TUFFC.
2010.1376.
[8] Elfrink, R., Renaud, M., Kamel, T. M., Nooijer, C. de, Jambunathan, M., Goedbloed,
M., Hohlfeld, D., Matova, S., Pop, V., Caballero, L. and Schaijk, R. van. Vacuum-
packaged piezoelectric vibration energy harvesters: damping contributions and au-
tonomy for a wireless sensor system. Journal of Micromechanics and Microengi-
neering 20.10 (Sept. 2010), 104001. DOI: 10.1088/0960-1317/20/10/104001.
[9] Liu, H., Zeng, F., Tang, G. and Pan, F. Enhancement of piezoelectric response of
diluted Ta doped AlN. Applied surface science 270 (2013), 225–230.
58

[10] Pensala, T., Thalhammer, R., Dekker, J. and Kaitila, J. Experimental investigation of
acoustic substrate losses in 1850-MHz thin film BAW resonators. IEEE transactions
on ultrasonics, ferroelectrics, and frequency control 56.11 (2009), 2544–2552.
[11] Akiyama, M., Kamohara, T., Kano, K., Teshigahara, A., Takeuchi, Y. and Kawa-
hara, N. Enhancement of Piezoelectric Response in Scandium Aluminum Nitride
Alloy Thin Films Prepared by Dual Reactive Cosputtering. Advanced Materials 21.5
(Dec. 2008), 593–596. DOI: 10.1002/adma.200802611.
[12] Akiyama, M., Tabaru, T., Nishikubo, K., Teshigahara, A. and Kano, K. Preparation
of scandium aluminum nitride thin films by using scandium aluminum alloy sput-
tering target and design of experiments. Journal of the Ceramic Society of Japan
118.1384 (2010), 1166–1169.
[13] Mertin, S., Heinz, B., Rattunde, O., Christmann, G., Dubois, M.-A., Nicolay, S. and
Muralt, P. Piezoelectric and structural properties of c-axis textured aluminium scan-
dium nitride thin films up to high scandium content. Surface and Coatings Technol-
ogy 343 (2018), 2–6.
[14] Sandu, C. S., Parsapour, F., Mertin, S., Pashchenko, V., Matloub, R., LaGrange, T.,
Heinz, B. and Muralt, P. Abnormal Grain Growth in AlScN Thin Films Induced by
Complexion Formation at Crystallite Interfaces. physica status solidi (a) 216.2 (Oct.
2018), 1800569. DOI: 10.1002/pssa.201800569.
[15] Mertin, S., Pashchenko, V., Parsapour, F., Nyffeler, C., Sandu, C. S., Heinz, B.,
Rattunde, O., Christmann, G., Dubois, M.-A. and Muralt, P. Enhanced piezoelectric
properties of c-axis textured aluminium scandium nitride thin films with high scan-
dium content: Influence of intrinsic stress and sputtering parameters. 2017 IEEE
International Ultrasonics Symposium (IUS). Ieee. 2017, 1–4.
[16] Damjanovic, D. Piezoelectricity. Encyclopedia of Condensed Matter Physics. Ed.
by F. Bassani, G. L. Liedl and P. Wyder. Oxford: Elsevier, 2005, 300–309. DOI:
https://doi.org/10.1016/B0-12-369401-9/00433-2.
[17] Muralt, P. Pyroelectricity. Encyclopedia of Condensed Matter Physics. Ed. by F.
Bassani, G. L. Liedl and P. Wyder. Oxford: Elsevier, 2005, 441–448. DOI: https:
//doi.org/10.1016/B0-12-369401-9/00434-4.
[18] Levanyuk, A. and Strukov, B. Ferroelectricity. Encyclopedia of Condensed Matter
Physics. Ed. by F. Bassani, G. L. Liedl and P. Wyder. Oxford: Elsevier, 2005, 192–
201. DOI: https://doi.org/10.1016/B0-12-369401-9/00435-6.
[19] Tonisch, K., Cimalla, V., Foerster, C., Dontsov, D. and Ambacher, O. Piezoelectric
properties of thin AlN layers for MEMS application determined by piezoresponse
force microscopy. physica status solidi c 3.6 (2006), 2274–2277.
[20] Österlund, E. Deposition and characterization of aluminum nitride thin films for
piezoelectric MEMS. PhD thesis. Aalto University, 2020.
[21] Prume, K., Muralt, P., Calame, F., Schmitz-Kempen, T. and Tiedke, S. Piezoelectric
thin films: Evaluation of electrical and electromechanical characteristics for MEMS
59

devices. IEEE transactions on ultrasonics, ferroelectrics, and frequency control


54.1 (2006), 8–14.
[22] Kaajakari, V. Practical MEMS: Design of microsystems, accelerometers, gyroscopes,
RF MEMS, optical MEMS, and microfluidic systems. Las Vegas, Nev, 2009.
[23] Muralt, P. AlN Thin Film Processing and Basic Properties. Microsystems and Nanosys-
tems. Springer International Publishing, 2017, 3–37. DOI: 10.1007/978-3-319-
28688-4_1.
[24] Vispute, R. D., Narayan, J., Wu, H. and Jagannadham, K. Epitaxial growth of AlN
thin films on silicon (111) substrates by pulsed laser deposition. Journal of Applied
Physics 77.9 (1995), 4724–4728. DOI: 10.1063/1.359441.
[25] Dubois, M.-A. and Muralt, P. Properties of aluminum nitride thin films for piezoelec-
tric transducers and microwave filter applications. Applied Physics Letters 74.20
(1999), 3032–3034.
[26] Chen, Y., Song, H., Li, D., Sun, X., Jiang, H., Li, Z., Miao, G., Zhang, Z. and Zhou, Y.
Influence of the growth temperature of AlN nucleation layer on AlN template grown
by high-temperature MOCVD. Materials Letters 114 (2014), 26–28. DOI: https:
//doi.org/10.1016/j.matlet.2013.09.096.
[27] Motamedi, P. and Cadien, K. Structural and optical characterization of low-temperature
ALD crystalline AlN. Journal of Crystal Growth 421 (2015), 45–52.
[28] Dubois, M.-A. and Muralt, P. Stress and piezoelectric properties of aluminum ni-
tride thin films deposited onto metal electrodes by pulsed direct current reactive
sputtering. Journal of Applied Physics 89.11 (2001), 6389–6395. DOI: 10.1063/1.
1359162.
[29] Anggraini, S. A., Uehara, M., Hirata, K., Yamada, H. and Akiyama, M. Polarity In-
version of Aluminum Nitride Thin Films by using Si and MgSi Dopants. Scientific
Reports 10.1 (Mar. 2020). DOI: 10.1038/s41598-020-61285-8.
[30] Yokoyama, T., Iwazaki, Y., Onda, Y., Nishihara, T., Sasajima, Y. and Ueda, M. Effect
of Mg and Zr co-doping on piezoelectric AlN thin films for bulk acoustic wave res-
onators. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
61.8 (Aug. 2014), 1322–1328. DOI: 10.1109/tuffc.2014.3039.
[31] Felmetsger, V. and Mikhov, M. Reactive magnetron sputtering of piezoelectric Cr-
doped AlN thin films. 2011 IEEE International Ultrasonics Symposium. IEEE. 2011,
835–839.
[32] Felmetsger, V. and Mikhov, M. Reactive sputtering of highly c-axis textured Ti-doped
AlN thin films. 2012 IEEE International Ultrasonics Symposium. IEEE. 2012, 782–
785.
[33] Martin, F., Muralt, P., Dubois, M.-A. and Pezous, A. Thickness dependence of the
properties of highly c-axis textured AlN thin films. Journal of Vacuum Science &
Technology A 22.2 (2004), 361–365. DOI: 10.1116/1.1649343.
60

[34] Bjurstrom, J., Rosen, D., Katardjiev, I., Yanchev, V. and Petrov, I. Dependence of the
electromechanical coupling on the degree of orientation of c-textured thin AlN films.
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 51.10
(2004), 1347–1353. DOI: 10.1109/TUFFC.2004.1350963.
[35] Iriarte, G. F., Engelmark, F. and Katardjiev, I. V. Reactive sputter deposition of highly
oriented AlN films at room temperature. Journal of materials research 17.6 (2002),
1469–1475.
[36] Nie, Z., Fu, J., Zou, J., Jin, T., Yang, J., Xu, G., Ruan, H. and Zuo, T. Advanced
aluminum alloys containing rare-earth erbium. Materials forum. Vol. 28. 2004, 197–
201.
[37] Toropova, L., Eskin, D. G., Kharakterova, M. and Dobatkina, T. Advanced aluminum
alloys containing scandium: structure and properties. Routledge, 2017.
[38] Caro, M. A., Zhang, S., Riekkinen, T., Ylilammi, M., Moram, M. A., Lopez-Acevedo,
O., Molarius, J. and Laurila, T. Piezoelectric coefficients and spontaneous polariza-
tion of ScAlN. Journal of Physics: Condensed Matter 27.24 (2015), 245901.
[39] Heinz, B., Mertin, S., Rattunde, O., Dubois, M. A., Nicolay, S., Christmann, G.,
Tschirky, M. and Muralt, P. Sputter deposition technology for Al (1- x) Sc x N films
with high Sc concentration. 2017 China Semiconductor Technology International
Conference (CSTIC). IEEE. 2017, 1–3.
[40] Fichtner, S., Wolff, N., Krishnamurthy, G., Petraru, A., Bohse, S., Lofink, F., Chem-
nitz, S., Kohlstedt, H., Kienle, L. and Wagner, B. Identifying and overcoming the
interface originating c-axis instability in highly Sc enhanced AlN for piezoelectric
micro-electromechanical systems. Journal of Applied Physics 122.3 (2017), 035301.
[41] Akiyama, M., Kano, K. and Teshigahara, A. Influence of growth temperature and
scandium concentration on piezoelectric response of scandium aluminum nitride
alloy thin films. Applied Physics Letters 95.16 (Oct. 2009), 162107. DOI: 10.1063/
1.3251072.
[42] Frey, H. and Khan, H. R. Handbook of thin film technology. Springer, 2015.
[43] Thin-Film Materials and Processes. Introduction to Microfabrication. John Wiley
Sons, Ltd, 2010. Chap. 5, 47–67. DOI: https://doi.org/10.1002/9781119990413.
ch5.
[44] Wasa, K., Kitabatake, M. and Adachi, H. 2 - Thin Film Processes. Thin Film Materi-
als Technology. Ed. by K. Wasa, M. Kitabatake and H. Adachi. Norwich, NY: William
Andrew Publishing, 2004, 17–69. DOI: https : / / doi . org / 10 . 1016 / B978 -
081551483-1.50003-4.
[45] Luttge, R. Chapter 2 - Basic Technologies for Microsystems. Microfabrication for
Industrial Applications. Ed. by R. Luttge. Micro and Nano Technologies. Boston:
William Andrew Publishing, 2011, 13–54. DOI: https : / / doi . org / 10 . 1016 /
B978-0-8155-1582-1.00002-2.
61

[46] Bashir, A., Awan, T. I., Tehseen, A., Tahir, M. B. and Ijaz, M. Chapter 3 - Interfaces
and surfaces. Chemistry of Nanomaterials. Ed. by T. I. Awan, A. Bashir and A.
Tehseen. Elsevier, 2020, 51–87. DOI: https://doi.org/10.1016/B978-0-12-
818908-5.00003-2.
[47] Saunders, S. and Nicholls, J. Chapter 14 - Oxidation, hot corrosion and protection
of metallic materials. Physical Metallurgy (Fourth Edition). Ed. by R. W. Cahn and
P. Haasen. Fourth Edition. Oxford: North-Holland, 1996, 1291–1361. DOI: https:
//doi.org/10.1016/B978-044489875-3/50019-3.
[48] Ahmad, F., Al-Douri, Y., Kumar, D. and Ahmad, S. Metal-oxide powder technology in
biomedicine. Jan. 2020, 121–168. DOI: 10.1016/B978-0-12-817505-7.00007-
5.
[49] Teixeira, V., Carneiro, J., Carvalho, P., Silva, E., Azevedo, S. and Batista, C. 11 -
High barrier plastics using nanoscale inorganic films. Multifunctional and Nanorein-
forced Polymers for Food Packaging. Ed. by J.-M. Lagarón. Woodhead Publishing,
2011, 285–315. DOI: https://doi.org/10.1533/9780857092786.1.285.
[50] Gudmundsson, J. T. Physics and technology of magnetron sputtering discharges.
Plasma Sources Science and Technology 29.11 (Nov. 2020), 113001. DOI: 10 .
1088/1361-6595/abb7bd.
[51] Seshan, K. Handbook of thin film deposition processes and techniques. William
Andrew, 2001.
[52] Yang, F. 2 - Modern metal-organic chemical vapor deposition (MOCVD) reactors
and growing nitride-based materials. Nitride Semiconductor Light-Emitting Diodes
(LEDs). Ed. by J. Huang, H.-C. Kuo and S.-C. Shen. Woodhead Publishing, 2014,
27–65. DOI: https://doi.org/10.1533/9780857099303.1.27.
[53] Johnson, R. W., Hultqvist, A. and Bent, S. F. A brief review of atomic layer deposi-
tion: from fundamentals to applications. Materials today 17.5 (2014), 236–246.
[54] Athar, T. Chapter 17 - Smart precursors for smart nanoparticles. Emerging Nan-
otechnologies for Manufacturing (Second Edition). Ed. by W. Ahmed and M. J.
Jackson. Second Edition. Micro and Nano Technologies. Boston: William Andrew
Publishing, 2015, 444–538. DOI: https://doi.org/10.1016/B978- 0- 323-
28990-0.00017-8.
[55] Dubois, M.-A. and Muralt, P. Measurement of the effective transverse piezoelectric
coefficient e31, f of AlN and Pb (Zrx, Ti1- x) O3 thin films. Sensors and Actuators
A: Physical 77.2 (1999), 106–112.
[56] Hong, H.-S. and Chung, G.-S. Surface acoustic wave humidity sensor based on
polycrystalline AlN thin film coated with sol-gel derived nanocrystalline zinc oxide
film. Sensors and Actuators B: Chemical 148.2 (2010), 347–352. DOI: https://
doi.org/10.1016/j.snb.2010.05.002.
62

[57] Mertin, S., Hody-Le Caer, V., Joly, M., Mack, I., Oelhafen, P., Scartezzini, J.-L. and
Schüler, A. Reactively sputtered coatings on architectural glazing for coloured ac-
tive solar thermal façades. Energy and Buildings 68 (2014), 764–770.
[58] Liljeholm, L. Reactive sputter deposition of functional thin films. PhD thesis. Acta
Universitatis Upsaliensis, 2012.
[59] Yap, Y. K. and Zhang, D. Physical Vapor Deposition. Encyclopedia of Nanotechnol-
ogy. Springer Netherlands, 2015, 1–8. DOI: 10.1007/978-94-007-6178-0_362-
3.
[60] Takeda, S., Suzuki, S., Odaka, H. and Hosono, H. Photocatalytic TiO2 thin film
deposited onto glass by DC magnetron sputtering. Thin solid films 392.2 (2001),
338–344.
[61] Iseki, T. Completely flat erosion magnetron sputtering using a rotating asymmetrical
yoke magnet. Vacuum 84.12 (2010), 1372–1376.
[62] Strijckmans, K., Schelfhout, R. and Depla, D. Tutorial: Hysteresis during the reac-
tive magnetron sputtering process. Journal of Applied Physics 124.24 (Dec. 2018),
241101. DOI: 10.1063/1.5042084.
[63] Aiempanakit, M., Kubart, T., Larsson, P., Sarakinos, K., Jensen, J. and Helmers-
son, U. Hysteresis and process stability in reactive high power impulse magnetron
sputtering of metal oxides. Thin Solid Films 519.22 (2011), 7779–7784.
[64] Thornton, J. A. Influence of substrate temperature and deposition rate on structure
of thick sputtered Cu coatings. Journal of Vacuum Science and Technology 12.4
(1975), 830–835.
[65] Charpentier, C. Investigation of deposition conditions and annealing treatments on
sputtered ZnO: Al thin films: Material properties and application to microcristalline
silicon solar cells. PhD thesis. Ecole Polytechnique X, 2012.
[66] Boussu, K., Van der Bruggen, B., Volodin, A., Snauwaert, J., Van Haesendonck,
C. and Vandecasteele, C. Roughness and hydrophobicity studies of nanofiltration
membranes using different modes of AFM. Journal of Colloid and Interface Science
286.2 (2005), 632–638. DOI: https://doi.org/10.1016/j.jcis.2005.01.
095.
[67] W. Zhou and Z. L. Wang, eds. Scanning Microscopy for Nanotechnology. Springer
New York, 2007. DOI: 10.1007/978-0-387-39620-0.
[68] Rosenberg, A., Hu, D., Chouaib, H., Tan, Z. and Malkova, N. Tracking the defects of
ultra-thin HfO2 using a Cody-Lorentz multiple oscillator model. Metrology, Inspec-
tion, and Process Control for Microlithography XXXII. Ed. by O. Adan and V. A.
Ukraintsev. SPIE, Mar. 2018. DOI: 10.1117/12.2296980.
[69] Baeumler, M., Lu, Y., Kurz, N., Kirste, L., Prescher, M., Christoph, T., Wagner, J.,
̇ A. and Ambacher, O. Optical constants and band gap of wurtzite Al1-
Žukauskaite,
xScxN/Al2O3 prepared by magnetron sputter epitaxy for scandium concentrations
63

up to x = 0.41. Journal of Applied Physics 126.4 (July 2019), 045715. DOI: 10 .


1063/1.5101043.
[70] Kim, K., Kwon, S. and Pahk, H. J. Fast analysis of film thickness in spectroscopic re-
flectometry using direct phase extraction. Current Optics and Photonics 1.1 (2017),
29–33.
[71] Bishop, C. A. Chapter 5 - Process Diagnostics and Coating Characteristics. Vac-
uum Deposition Onto Webs, Films and Foils (Third Edition). Ed. by C. A. Bishop.
Third Edition. Boston: William Andrew Publishing, 2015, 85–128. DOI: https://
doi.org/10.1016/B978-0-323-29644-1.00005-0.
[72] Mazzalai, A., Balma, D., Chidambaram, N., Jin, L. and Muralt, P. Simultaneous
piezoelectric and ferroelectric characterization of thin films for MEMS actuators.
2013 Joint IEEE International Symposium on Applications of Ferroelectric and Work-
shop on Piezoresponse Force Microscopy (ISAF/PFM). IEEE. 2013, 363–366.

You might also like