from zirconia to yttria sampling the ysz phase diagram using sputter deposited thin films

From zirconia to yttria: Sampling the YSZ phase diagram using sputter-deposited
thin films
Thomas Götsch, Wolfgang Wallisch, Michael Stöger-Pollach, Bernhard Klötzer, and Simon Penner

Citation: AIP Advances 6, 025119 (2016); doi: 10.1063/1.4942818
View online: />View Table of Contents: />Published by the American Institute of Physics

,


AIP ADVANCES 6, 025119 (2016)

From zirconia to yttria: Sampling the YSZ phase diagram
using sputter-deposited thin films
Thomas Götsch,1 Wolfgang Wallisch,2 Michael Stöger-Pollach,2
Bernhard Klötzer,1 and Simon Penner1,a
1
Institute of Physical Chemistry, University of Innsbruck, Innrain 80/82,
A-6020 Innsbruck, Austria
2
University Service Center for Transmission Electron Microscopy (USTEM), Vienna
University of Technology, Wiedner Hauptstraße 8–10, A-1040 Vienna, Austria

(Received 30 November 2015; accepted 12 February 2016; published online 22 February 2016)
Yttria-stabilized zirconia (YSZ) thin films with varying composition between 3 mol%
and 40 mol% have been prepared by direct-current ion beam sputtering at a substrate
temperature of 300 ◦C, with ideal transfer of the stoichiometry from the target to
the thin film and a high degree of homogeneity, as determined by X-ray photoelectron and energy-dispersive X-ray spectroscopy. The films were analyzed using
transmission electron microscopy, revealing that, while the films with 8 mol% and
20 mol% yttria retain their crystal structure from the bulk compound (tetragonal
and cubic, respectively), those with 3 mol% and 40 mol% Y2O3 undergo a phase

transition upon sputtering (from a tetragonal/monoclinic mixture to purely tetragonal
YSZ, and from a rhombohedral structure to a cubic one, respectively). Selected
area electron diffraction shows a strong texturing for the three samples with lower
yttria-content, while the one with 40 mol% Y2O3 is fully disordered, owing to
the phase transition. Additionally, AFM topology images show somewhat similar
structures up to 20 mol% yttria, while the specimen with the highest amount of
dopant features a lower roughness. In order to facilitate the discussion of the
phases present for each sample, a thorough review of previously published phase
diagrams is presented. C 2016 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license
( [ />
I. INTRODUCTION

Yttria-stabilized zirconia (YSZ) is one of the most widely-used and studied materials due to its
high ionic conductivity at elevated temperatures,1 its high chemical inertness and thermal stability,
as well as its hardness.2 Thus, it is often used as electrolytes in solid oxide fuel cells,3 but also in
chemical sensors,4 as well as in coatings for thermal barriers,5 and to create optical devices like
switchable mirrors or filters.6
However, the influence the amount of yttria in the solid solution on the relevant properties,
such as crystal structure and conductivity is often neglected. By doping zirconium oxide, ZrO2, with
yttrium oxide, Y2O3, a tetravalent ion (Zr4+) is substituted by a trivalent one (Y3+). Due to charge
neutralization, oxygen vacancies are formed, increasing the ionic conductivity. These vacancies also
play a role in stabilizing the often desired tetragonal or cubic structures.7
YSZ has been proposed to exist in various crystal structures, some of which are shown in
Figure 1. In Figure 1(a), monoclinic ZrO2/YSZ is shown (for YSZ, the only difference is that some
of the Zr are substituted by Y and some O are missing, and that the lattice parameters will thus
differ).8 In this structure, edge-sharing polyhedra are formed by seven-fold coordinated Zr atoms.
For tetragonal zirconia,9 displayed in Figure 1(b), the coordination number of Zr increases by

a Electronic mail:

2158-3226/2016/6(2)/025119/20

6, 025119-1

© Author(s) 2016.


025119-2

Götsch et al.

AIP Advances 6, 025119 (2016)

FIG. 1. Various crystal structures of YSZ as found in several phase diagrams. Based on Refs. 8–13.

one, resulting in distorted cubes. Cubic zirconia (Figure 1(c)), exhibiting the CaF2 structure, also
features cubically coordinated zirconium atoms (undistorted).10 It is to be noted that the images
in Figures 1(a) to Figure 1(c) depict ZrO2, but are also valid for YSZ, where a certain fraction of
the Zr atoms are simply replaced by Y (and the occupancies of the O sites are reduced as well).
These three structures comprise the majority of the phase diagram at the zirconia-rich side (see
below for a discussion of the phase diagrams). There have been several other proposals in the past,
most describing ordered structures, such as Zr3Y4O12, in a rhombohedral structure,11 as depicted
in Figure 1(d). This structure is made up of a mixture of edge-sharing octahedrally and seven-fold
coordinated zirconia/yttria polyhedra. This structure results from an ordering of the vacancies in
the defective fluorite structure.11 Additionally, the existence of a cubic pyrochlore (Figure 1(e)) has
been suggested, with the composition Zr2Y2O7,12 consisting of alternating layers of edge-sharing
distorted [ZrO6] octahedra and [YO8] cubes. In Figure 1(f), the structure of pure yttria (Y2O3) is
shown. This substance crystallizes in a body-centered cubic structure and contains edge-sharing
distorted [YO6] octahedra.
In order to investigate the influence of the yttria-content on the crystal structure, the morphology

and surface topology, as well as the epitaxial growth properties, we decided to employ our model
thin film systems with the goal of establishing a thin film “phase diagram” of YSZ, forming the
required basis for future investigations regarding other properties of these thin films, and, by aiming
for epitaxially-grown films, preparing model systems for future studies as eventual catalysts. The
thin film phase diagram is especially important since there is a vast number of applications of YSZ
thin films, as outlined above. It appears that, while the number of publications dealing with zirconia
is large, many of them only focus on pure ZrO2,14,15 or present results on just one stoichiometry
of yttria-stabilized zirconia.16,17 In many cases, supported films and not free-standing ones (as in
our case) are used.16 To the best of our knowledge, there have been no systematic studies regarding
the Y-influence over a wider stoichiometry range, but only for limited compositional variations
such as in Ref. 18. For the present study, we hence restricted ourselves to the low-yttria side,
featuring sample compositions ranging from 3 mol% yttria up to 40 mol%, so as to just focus on the
technologically-relevant materials. Additionally, to facilitate the discussion of the compositions and
crystal structures obtained in our study, an overview of often-used concentration quantities, as well as


025119-3

Götsch et al.

AIP Advances 6, 025119 (2016)

a thorough discussion of existing phase diagrams, going more into detail than existing publications
such as the one by Chevalier et al.,19 are presented in the next sections.

II. YSZ STOICHIOMETRY QUANTITIES

There are various ways to present the composition of solid mixtures such as yttria-stabilized
zirconia, such as the molar percentage of one of the constituents, the atomic percentages or the
Zr/Y ratio. The practically most useful quantity probably is the molar percentage of Y2O3 (mol%

Y2O3) as it is directly related to the preparation procedure. Also often used is the atomic percentage
of yttrium (at% Y), for it is the quantity that is obtained from elemental composition analyses
such as XPS or EDX. Another composition often found, especially in phase diagrams, is the molar
percentage of YO1.5 (mol% YO1.5). It should be noted that, against common belief, this value is not
simply twice that of mol% Y2O3 because n(Y2O3) and n(YO1.5) are found in the numerator, but also
in the denominator of the equations for the molar percentages, respectively.
In Table I, the conversions between all mentioned concentration quantities are listed. Each
equation describes the calculation of the property in the left column from that in the top row.

III. DISCUSSION OF EXISTING PHASE DIAGRAMS

A large amount of phase diagrams concerning the zirconia-yttria system are reported in literature, but these do not all agree with each other and, in fact, contain many contradictory examples. Hence, we strive to give a brief overview of the published proposals in order to facilitate
the discussion of our findings later. The focus of this small review will thus be put on those
yttria-concentrations that were also used in our study: 3 mol%, 8 mol%, 20 mol% and 40 mol%
Y2O3. These samples will be referred to as 3YSZ, 8YSZ, 20YSZ and 40YSZ, respectively, with the
number denoting the amount of yttria (in mol% Y2O3) present. One of the first phase diagrams of
this system was published by Duwez et al.20 in 1951, featuring no ordered phases such as Zr3Y4O12.
Also, according to this diagram, no pure cubic zirconia could exist. Rather, the monoclinic polymorph would transform into its tetragonal counterpart at approximately 1000 ◦C, which then would
melt at about 2700 ◦C. The first occurrence of the cubic form of the yttria-doped variant would be
at 5 mol% Y2O3. At low temperatures, a miscibility gap between cubic and monoclinic YSZ would
exist, which would react towards the tetragonal polymorph upon reaching the respective eutectoidic
temperature between 400 ◦C and 500 ◦C. Thus, 3YSZ would be either monoclinic or tetragonal,
depending on whether the latter was frozen in a metastable state, and all other samples of interest
would be pure cubic YSZ (since that region extends from 7.5 mol% to more than 50 mol% Y2O3,
where a two-phase area between cubic YSZ and bcc yttria starts).
In 1963, Fan et al.21 published an incomplete version of the low-temperature region of the phase
diagram, differing strongly from that by Duwez et al. by containing a compound of the composition
Zr2Y2O7 at 33.3 mol% Y2O3, exhibiting a cubic pyrochlore structure. Also, the zirconia-rich region
of the diagram features dissimilarities in the progression of the phase boundaries, featuring no cubic

TABLE I. Conversion between commonly-used concentration values for yttria-stabilized zirconia. y always refers to the
quantity in the left column, x to that in the top row.
vert.: y, horiz.: x

mol% Y2O3

mol% Y2O3

—

mol% YO1.5

2x
y = 1+x
2x
y = 2x+3
2x
y = 1+x
y = 1−x
2x

at% Y (incl. O)
at% Y (excl. O)
Zr/Y Ratio

mol% YO1.5
y=

x
2−x

—
y=

2x
6−x

y=x
y=

1−x
x

at% Y (incl. O)
y=
y

3x
2−2x
6x
= x+2

at% Y (excl. O)
y=

x
2−x

y=x

—

y=

6x
y = x+2
y = 2−5x
6x

y=

2x
6−x

—
1−x
x

Zr/Y Ratio
y=

1
1+2x
1
y = 1+x
2
y = 6x+5
1
y = 1+x

—


025119-4

Götsch et al.

AIP Advances 6, 025119 (2016)

YSZ at lower temperatures, but rather suggesting the existence of solid solutions of zirconia and yttria, respectively, in the pyrochlore, as well as the stability of monoclinic solid solutions until yttria
concentrations corresponding to the pyrochlore. Also, the pyrochlore and Y2O3 show a miscibility
gap between approximately 47 mol% and 33 mol% Y2O3. On the basis of this phase diagram, each
of our samples would contain the pyrochlore structure, as all the specimens up to 20YSZ would be
a mixture between either monoclinic or tetragonal YSZ and Zr2Y2O7, and 40 YSZ would consist of
only the pyrochlore structure, which would exhibit a certain degree of flexibility regarding the exact
composition. However, no further study was able to confirm the existence of the Zr2Y2O7 phase.
At the end of the 1960s, another systematic study was performed by mixing ZrO2 and Y2O3
in 5 mol% steps and heating the mixture with a CO2 laser.22–24 The pyrochlore suggested by Fan
et al. was not found, with cubic YSZ seemingly taking its place (also including a region of immiscibility between cubic YSZ and yttria, but at higher concentrations), and the zirconia-rich region
shows similarities to Duwez’ version. In contrast to the proposal by Fan et al.,21 monoclinic YSZ
is only found up to a very small amount of yttria, with the cubic polymorph becoming the stable
phase at room temperature already below 10 mol% Y2O3. It has to be noted, though, that the data
below 500 ◦C is sparse in this study. This may be the reason for the omission of the eutectoidic
decomposition of the tetragonal phase, as seen in Duwez’ version (and many later diagrams), for
instance. At 8 mol%, 20 mol% and 40 mol% Y2O3, a cubic structure should be found and, for the
sample with the lowest yttria content, namely 3 mol%, tetragonal YSZ would be the stable form,
with small amounts of monoclinic YSZ at low temperatures.
A slightly different low-yttria region was found by Srivastava et al.,25 who, like Duwez and
co-workers, observed a eutectoidic transformation at around 4 mol% Y2O3 at 565 ◦C, below which
tetragonal YSZ would decompose into the monoclinic and cubic variants, which is where 3YSZ

would be located. Otherwise, no large differences from the previous diagram are to be found, with
the cubic YSZ/Y2O3 two-phase region starting only at 80 mol% instead of already at a bit more
than 40 mol% in the case of the previously mentioned study, but also comprising all other three
concentrations investigated in this work.
Also in the early 1970s, Rouanet26 could not validate the existence of a Zr2Y2O7 compound
either, but instead located a new ordered phase: Zr3Y4O12, with a hexagonal crystal structure.
This new phase, also called the δ phase (derived from the zirconia-scandia system),11 is located
at 40 mol% Y2O3, which would correspond directly to our 40YSZ sample, and is stable up to
1250 ◦C, where it decomposes peritectoidically. At the low-yttria side of the diagram, no eutectoid
is visible in the diagram, which is due to the omission of the low-temperature part (only data above
1000 ◦C are shown). Judging from the limited information available, 3YSZ would be tetragonal,
8YSZ a mixture between tetragonal and cubic polymorphs, and 20YSZ would be found in its cubic
structure. The eutectoid, however, was published again by Scott,27 with the two-phase region below
the eutectoid extending to 10 mol% Y2O3. The ordered phase, Zr3Y4O12, on the other hand, is not
featured in this diagram. Instead, two cubic bcc-phases are found between 80 mol% and 95 mol%
Y2O3, which, according to later work of the same author,28 are due to contaminations causing
non-equilibrium effects. Scott also examined the concentration regions where the various polymorphs could be kept in a metastable state at room temperature, which results in the information
that only tetragonal YSZ is possible between 3 mol% and 6 mol% Y2O3 — at lower concentrations,
the monoclinic analogue is obtained, and, at higher concentrations, cubic YSZ becomes more stable
(with small regions of overlap between the polymorphs). This would mean that 3YSZ would be
tetragonal, 8YSZ, 20YSZ and 40YSZ already cubic.
Gorelov investigated the low-yttria content region more closely,29 and found four ZrO2 polymorphs (and, hence, also four YSZ solid solutions) instead of the usual three. In addition to the monoclinic one, and the already known tetragonal analogue, which is stable between 1200 ◦C and 2300 ◦C,
another tetragonal phase was observed between 2300 ◦C and 2500 ◦C, where it transforms to cubic
zirconia. This new tetragonal phase is primarily distinguished by differences in lattice parameters:
the c axis is compressed and a is increased. Both tetragonal phases decompose eutectoidically upon
cooling, with the high-temperature one forming cubic and the low-temperature tetragonal compound,
and the latter yielding monoclinic and cubic YSZ. Due to the limitation on concentrations below
10 mol%, no conclusion can be drawn for 20YSZ or 40YSZ, but 8YSZ should be cubic according to

025119-5

Götsch et al.

AIP Advances 6, 025119 (2016)

this diagram, and 3YSZ could be either a mixture between monoclinic and cubic YSZ, or one of the
tetragonal polymorphs, depending on the ability to keep them metastable at low temperatures.
Stubican and coworkers found a eutectoidic decomposition of cubic YSZ into monoclinic zirconia and Zr3Y4O12 at temperatures as low as 400 ◦C,30–32 which could have an influence on 3YSZ,
8YSZ and 20YSZ. However, looking at the higher-temperature reading, the two former could also
be within the tetragonal/cubic two-phase region, and 20YSZ in the cubic area. Their version of the
phase diagram also contains the peritectoidically transforming Zr3Y4O12 phase, again at 40 mol%.
In the diagram published by Pascual and Duran,33 on the other hand, this ordered phase would
transform dystectoidically into cubic YSZ at 1375 ◦C. However, they also proposed the existence of
a second phase, for which they coined the term “1:6 phase” as it complies with the stoichiometry
of ZrY6O11, which supposedly crystallizes hexagonally, transforms dystectoidically at 1700 ◦C and
is found at 75 mol% Y2O3. What they could not verify was the miscibility gap between monoclinic
ZrO2 and Zr3Y4O12— that is, there is no eutectoid for cubic YSZ. The sample with 3 mol% Y2O3
would be tetragonal at elevated temperatures, 8YSZ would exhibit the cubic polymorph, and 20YSZ
either the cubic one too, or a mixture between the cubic and the δ phase.
In 1984, Ruh et al.34 again focused on the high-ZrO2 region of the YSZ phase diagram. Their
findings do not deviate drastically from other proposals. What is, however, interesting, is that the cubic
phase starts already for yttria-contents below 8 mol%. Also, according to this diagram, the tetragonal
phase would only be stable up to about 2 mol%, which would mean that 3YSZ would already be a
mixture between the tetragonal and cubic structures, even at relatively low temperatures. Again, as
already found in other references,25,29 the tetragonal variant undergoes a eutectoidic decomposition.
Other papers by Yoshikawa, Suto et al.35,36 also discussed the low-yttria region, and their findings
correspond well with those of Ruh et al., with the cubic YSZ region starting below 8 mol% Y2O3 too,
with the same conclusions regarding our investigated samples being drawn as for Ruh’s work. Some
of these results are confirmed by another publication from Stubican in 1988,37 although the cubic

region of the phase diagram starts only well above 8 mol% Y2O3 at lower temperatures (also due to
the eutectoid already published earlier30–32), meaning that both 3YSZ and 8YSZ would be mixtures
between tetragonal and cubic structures, except for high temperatures in the case of 8 mol% Y2O3,
where it would be purely cubic. They mention that this deviation from the work of Ruh et al. could,
however, also be due their use of a hydrothermal method since it was difficult ot obtain equilibrium
at lower temperatures. This time, in contrast to their previous proposal, they found that the rhombohedral, ordered phase, Zr3Y4O12, transforms dystectoidically at 1382 ◦C to the cubic analogue. What
they couldn’t confirm was the ZrY6O11 compound proposed by Pascual and Duran.
The 1980s also brought about the advent of calculated phase diagrams. For instance, Degtyarev
and Voronin constructed such diagrams based on the calculated thermodynamic properties of all the
phases.38–40 This included the rhombohedral Zr3Y4O12 phase, which transforms dystectoidically to
cubic YSZ. Also, this diagram features an eutectoidic decomposition of tetragonal, as well as the
cubic structure (into monoclinic YSZ and the ordered phase). They also computed the diagrams
for an increased pressure, where the region of stability for the cubic polymorph is enlarged and
the monoclinic form is generally less stable.40 At ambient pressures, 3YSZ is located within the
tetragonal/cubic mixture area, and 8YSZ at intermediate temperatures as well. 20YSZ and 40YSZ
should be inside the cubic region, even though none of their diagrams actually shows 40 mol%, as
no hints of the δ phase are visible in the partial diagrams they show.
The two eutectoids were also found by the calculations of Du et al,41 who did additional
experiments that confirmed their computations. A year later, the same group published another
version,42 with the main difference being in the high Y2O3 region. They, however, employed two
different sets of model parameters, which showed drastic differences in the eutectoidic region between the monoclinic zirconia and Zr3Y4O12. In another revision of this diagram by Jin and Du,43
another miscibility gap was introduced at the low-temperature end below 479 ◦C, where the ordered
phase decomposes into the monoclinic solid solution and Y2O3. This would suggest that, at low
temperatures, a demixing of the solid solutions would be favoured. If high-temperature phases are
quenched and brought to room temperature, 3YSZ and 8YSZ would be tetragonal or cubic (if initial
temperatures exceeded 2000 ◦C), and 20YSZ cubic. 40YSZ would crystallize in the rhombohedral
Zr3Y4O12 structure — at least for a limited temperature range between 479 ◦C and 1376 ◦C.


025119-6

Götsch et al.

AIP Advances 6, 025119 (2016)

TABLE II. Summary of the YSZ phase diagrams found in literature. Abbreviations: cub. = cubic, tetr. = tetragonal,
dystect. = dystectoidic, peritect. = peritectoidic, eutect. = eutectoidic.
Year
1951
1963
1968
1971
1974
1975
1978
1982
1983
1984
1987
1988
1987
1990
1992
1995
1996
1997
2005

Citation
Duwez et al.20

Fan et al.21
Ruh, Rouanet, Skaggs22–24
Rouanet26
Srivastava et al.25
Scott27
Gorelov29
Stubican et al.30–32
Pascual, Duran33
Ruh et al.34
Suto, Yoshikawa et al.35,36
Stubican37
Degtyarev, Voronin38–40
Du et al.41,42
Jin, Du43
Suzuki44
Yashima45
Suzuki46
Fabrichnaya et al.47

Exp./Theo.

Ordered Phase(s)

Main Features

exp.
exp.
exp.
exp.
exp.

exp.
exp.
exp.
exp.
exp.
exp.
exp.
theo.
theo.
theo.
exp.
exp.
exp.
theo.

—
Zr2Y2O7
—
Zr3Y4O12
—
—
—
Zr3Y4O12
Zr3Y4O12, ZrY6O11
—
—
Zr3Y4O12
Zr3Y4O12
Zr3Y4O12
Zr3Y4O12

Zr3Y4O12
—
Zr3Y4O12
Zr3Y4O12

no cub. ZrO2, eutect. tetr. YSZ
no cub. YSZ
peritect. Zr3Y4O12
eutect. tetr. YSZ
two cub. phases
two tetr. phases, two eutect.
peritect. Zr3Y4O12, eutect. cub. YSZ
dystect. ordered phases
cub. YSZ for < 8 mol%Y2O3
cub. YSZ for < 8 mol%Y2O3
dystect. Zr3Y4O12
dystect. Zr3Y4O12
dystect. Zr3Y4O12
dystect. Zr3Y4O12
peritect. Zr3Y4O12
no equilibrium < 1200 ◦C
peritect. Zr3Y4O12
dystect. Zr3Y4O12

In an experimental phase diagram, created by Suzuki,44 only minor differences to other proposals can be discerned. For instance, Zr3Y4O12 transforms peritectoidically at 1360 ◦C. This diagram also features the eutectoids for the cubic and tetragonal YSZ species and the miscibility gap
between monoclinic zirconia and Zr3Y4O12, and the crystal structures of our samples would be the
same as for the previously discussed versions. Yashima et al.,45 on the other hand, took a closer
look at the low-yttria content region (up to around 20 mol% Y2O3). They found it impossible to
reach thermodynamic equilibrium below 1200 ◦C, and, hence, their diagram does not show equilibrium lines for those temperatures. However, like Scott,27 they managed to determine the regions
of metastability for each of the solid solutions, according to which cubic YSZ is obtainable for

concentrations above 11 mol% Y2O3 (e.g. 20YSZ), and the tetragonal structure is obtained above
1 mol% Y2O3 (hence, it being the stable phase for 3YSZ and 8YSZ).
Suzuki later published another (partial) phase diagram for YSZ,46 investigated using conductivity measurements. This version also is in good agreement with other publications. Fabrichnaya
et al.47 reported a theoretical diagram in 2005, which differs significantly from other recent proposals in our region of interest, in that, for a certain range in Y content, the cubic polymorph (for
example for 20 mol% Y2O3) would not decompose at all upon lowering the temperature (i.e. there
is no eutectoid). The remaining regions resemble previous diagrams, such as the dystectoidically
transforming δ phase, with other implications on our samples being that 3YSZ would crystallize
tetragonally and 8YSZ both, cubically and tetragonally.
Table II summarizes these diagrams, allowing for a quick overview of the major points of each
of them, where the evolutionary nature of the phase diagram proposals can be witnessed, especially
with regard to the ordered compounds — at first, none had been reported, then there were different
suggestions, and, soon it became accepted and almost every diagram contained Zr3Y4O12.

IV. EXPERIMENTAL DETAILS
A. Thin film deposition

The thin films have been deposited on NaCl(001) single crystals at elevated temperatures of
623 K to facilitate crystallization and possibly epitaxy using a custom-made sputter-gun (see below
for details) in a modular high-vacuum apparatus with a base pressure in the low 10−7 mbar regime.


025119-7

Götsch et al.

AIP Advances 6, 025119 (2016)

FIG. 2. Overview of the redesigned sputter device, featuring a modular setup with an easily removable shielding (at the top),
onto which the filament is mounted, allowing for a significant speed-up in filament changes. In addition to a higher mechanical
stability, this new design enables the sputtering of insulating targets due to the higher temperatures reached because of the

closer proximity to the glowing filament.

The sputtering was performed in 5 × 10−5 mbar Ar, and as targets, pellets of the different YSZ
samples (commercial powders with 3 mol%, 8 mol%, 20 mol% and 40 mol% Y2O3, by Sigma
Aldrich) were used. These targets were prepared by pressing the powders onto a spirally-shaped
Ta wire at approximately 20 kN in a KBr pellet press used for infrared spectroscopy studies. By
submerging the coated sodium chloride crystals in water, the thin films can be floated off, and
afterwards collected using TEM gold grids to yield unsupported thin films with nominal thicknesses
of 25 nm.
While there were successes in sputtering YSZ with 8 mol% yttria using our previously published direct-current ion beam sputter device,48,49 a redesign was required in order to suit more
insulating targets, such as some of the oxides used in this study, without having to resort to
radio-frequency power supplies, as is often the case for commercial sputtering systems. The main
changes, displayed in Figure 2, encompass a new, modular head for the device, where the target can
be brought in closer proximity to the hot filament, causing the target to reach higher temperatures,
decreasing the resistivity. Also, not visible in the illustration, to limit the heat conductance away
from the oxide, the target mounting was altered to split up the structural connection (now using
ceramics) from the electric connection (via a very thin Ta wire). The new design also brings about
a higher mechanical stability compared to previous versions, resulting in a better stability of the
deposition process due to less vibrations being transmitted to target and the gun head, allowing
for the growth of more homogeneous films. Additionally, the revised shielding (needed to block
the line of sight between the target and the substrate in order to avoid evaporated or sputtered
tungsten (oxide) contaminations) now features a conical center part instead of a cylindrical one,
and is lower in height, maximizing the substrate area that is coated. Another benefit relating to the
shielding stems from the new filament mountings, which are now directly attached to the removable
shielding, meaning that the whole filament can now be swapped much quicker than before.
B. Characterization of the films

The unsupported films were investigated with respect to their crystallographic properties, structure and homogeneity using a FEI Tecnai F20 S-TWIN (high-resolution) analytical (scanning)
transmission electron microscope (200 kV), equipped with an EDAX Apollo XLT2 silicon-drift
detector for energy-dispersive X-ray spectrometry (EDX) and a GIF Tridiem electron energy-loss

spectrometer.
For the X-ray photoelectron spectroscopy (XPS) compositional analyses, a Thermo Scientific
MultiLab 2000 spectrometer (with a base pressure in the high 10−11 mbar to low 10−10 mbar range),
fitted with a monochromated Al-Kα X-ray source, an Alpha 110 hemispehrical sector analyzer
and an ion gun for sputter-depth profiling (operated at 3 kV using argon), was utilized. For these
investigations, additional samples had been prepared by depositing the oxides on silicon wafers.


025119-8

Götsch et al.

AIP Advances 6, 025119 (2016)

The atomic force microscopy (AFM) surface topology images of the unsupported specimens,
from which the surface roughness values were calculated, were recorded using a Veeco Digital
Instruments Dimension 3100 in tapping mode. For this, Veeco RTESPW silicon cantilevers with
force constants between 20 N m-1 and 80 N m-1 as well as resonance frequencies from 256 kHz to
317 kHz were employed.
C. Characterization of the target materials

In order to gain an understanding about the change of crystallographic structure from the
target to the thin film, X-ray diffraction studies have been performed on the target powders using
a Siemens D5000 diffractometer under ambient conditions, while recording a 2θ range from 20◦ to
70◦ with a step size of 0.02◦.
V. RESULTS AND DISCUSSION
A. Composition

To check the purity and composition of the thin films, various methods were employed: using
X-ray photoelectron spectroscopy, sputter depth profiles were recorded, energy-dispersive X-ray

(EDX) spectra were taken both in TEM, as well as in STEM mode (spectrum imaging), and electron
energy-loss spectrometry was utilized as well. Figure 3 displays the surface-sensitive XP spectra
(Figure 3(a)) and EDX spectra (Figure 3(b)), which, due to the larger mean free path of X-rays in
contrast to electrons of the same energy, corresponds to an integrated composition over the whole
thickness range.
These two sets of plots show that the samples are contamination-free. In Figure 3(a), the main
peaks visible are the O 1s (at 532 eV), the corresponding O KLL auger peak at around 1000 eV
binding energy, the Zr 3d and 3p (180 eV and 331 eV, respectively), as well as the Y 3d (158 eV)
and 3p peaks (301 eV). Going from 3YSZ to 40YSZ, the relative change of Zr/Y peak intensities
can be observed nicely. For the thinner films, namely 3YSZ and 40YSZ (and, to a smaller extent,
20YSZ), the Si 2p peak at 99 eV and the Si 2s one (149 eV) from the underlying substrate (XPS
samples were deposited directly on silicon wafers) begin to be visible at the surface spectrum
already. At 245 eV, the Ar 2p peak can be seen, which is due to incorporated argon in the thin

FIG. 3. Surface-sensitive X-ray photoelectron spectra (a) and energy-dispersive X-ray spectra recorded in the TEM (b) both
show that the films are impurity-free (note that the NaCl does not originate from the deposition process, but rather from the
substrate, and that it can be removed by rinsing with water) and that the stoichiometry is the same as in the target.


025119-9

Götsch et al.

AIP Advances 6, 025119 (2016)

films, most likely due to implanted Ar+ ions into the target during the sputter-depositing process.
For 40YSZ, two very small peaks are found at 1070 eV and 508 eV, which are Na 1s and Na
KLL peaks, originating from the sample preparation process when transferring it into the XPS
chamber. This can be confirmed by sputtering the film for a few seconds, after which the sodium
contamination vanishes. Hence, it was only on the top of the surface and does not arise from the thin

film deposition process.
In the EDX spectra (Figure 3(b)), the peaks from other elements than Zr, Y or O can readily be
explained as well: the very small Na and Cl signals are remnants of the sodium chloride substrates
(and, hence, not from the deposition process itself) and can be minimized by introducing additional
cleaning steps simply using water. Au stems from the TEM gold grid, upon which the films are
placed, Cu from the specimen holder, and since the pole piece of the lens contains Fe and Co, these
consequently yield X-ray fluorescence peaks. This fluorescence is due to the measurement positions
being chosen close to the grid bars in order to minimize charging effects. The proximity to the
gold grid causes the high-energy Au X-ray lines to be emitted, in turn resulting in the emission of
X-ray fluorescence from the pole pieces for two of the samples (3YSZ and 40YSZ). Thus, it can be
concluded that the films are of high purity, and the composition is the same as in the target as well,
as the data in Table III show.
There, the stoichiometries, as obtained from various methods, are shown: the mean of an XPS
depth profile, the quantification results from the EDX spectra shown in Figure 3(b), the integrated
EDX spectra from spectrum images, and the quantification of the Zr and Y L3 edges. The EDX
spectra have been quantified using the Cliff-Lorimer method with the software Digital Micrograph
by Gatan, using the k-factors contained therein. Calculating the mean of all these values, one
can see that the yttria content in the thin film correlates well with that in the target: for 3YSZ,
3.3(7) mol% Y2O3 is measured, for 8YSZ 8(2) mol%, for 20YSZ 17(2) mol%, and 40YSZ contains
38(3) mol% yttria. Looking at the separate values, it immediately comes to mind that the EELS
quantification is off the most, for example only yielding 13.8 mol% Y2O3 for 20YSZ. This is due to
the difficulties arising when quantifying EELS edges, which, in contrast to for instance the Gauss
peaks in EDX spectroscopy, is not as straightforward. Also, the L3 edges are found at 2080 eV
energy-loss (yttrium) and 2222 eV (zirconium), respectively, where the intensity in the spectrum is
already extremely low. All other methods are in good agreement with the target values, with the
EDX results in TEM mode corresponding better than those in STEM, because a larger area on the
specimen was sampled, allowing us to use a higher beam current without destroying the thin films
due to charging and, thus, obtain a better signal-to-noise ratio.
Because not only the composition is relevant, but also its depth and spacial homogeneity, XPS
depth profiling and EDX spectrum imaging were employed (Figure 4). The XPS depth profile

(Figure 4(a)), which was normalized to the film thickness (determined by taking the inflection point
of a sigmoidal fit of the Si 2p peak intensity) in order to allow for easier comparison of the profiles if
the films do not have the same thicknesses, shows that the composition does not change drastically.
The slight increase in yttria-content at higher etch depths comes from the increased silicon content,
which results in the other peak intensities diminishing, making the quantification less reliable. For
40YSZ, this plot shows that the XPS measurements yield too low Y2O3 concentrations, as already
seen in Table III. For the other specimens, the values correspond well with the compositions from
the targets, indicated by the dashed lines.
TABLE III. Composition of the thin films, as determined by various methods. The mean of these values correlates with the
target compositions.

target

XPS depth profile

3
8
20
40

3.8(12)
6.8(3)
17.7(30)
33.7(9)

mol% Y2O3
EDX TEM
3.2
8.3
17.8

41.0

EDX STEM

EELS

Mean

4.0(1)
9.5(2)
18.5(2)
39.0(5)

2.4
5.8
13.8
38.5

3.3(7)
8(2)
17(2)
38(3)


025119-10

Götsch et al.

AIP Advances 6, 025119 (2016)

FIG. 4. (a) Results from the XPS depth profiles, showing the yttria-concentration as a function of sample depth. In (b), an
extracted EDX spectrum from the spectrum image of 3YSZ is shown exemplarily, featuring the Y and Zr Kα peaks. Prior
to quantification of the respective map, the principal component has been selected by principal component analysis, thus
removing the noise. The mean concentration value as calculated from the EDX map for this compound is 4.0(1) mol% yttria
(see Table III). The very small standard deviation confirms the high homogeneity of our samples.

In Figure 4(b), the results from a corresponding EDX analysis are shown for the case of 3YSZ.
The spectrum was derived from a mapping generated by a principal component analysis, removing
the noise sufficiently to allow for a reliable quantification. The spectrum image, recorded over a
35 nm × 35 nm region reveals a very high homogeneity despite the short dwell times, with the yttria
content being 4.0(1) mol% Y2O3,and a negligibly small standard deviation (see also Table III). For
the other three specimens, the same procedure has been applied, yielding similarly high degrees of
homogeneity.
Figure 5(a) displays an exemplary chemical state depth profile for the Zr 3d peak of the 20YSZ
thin film. It can be observed that, at the surface, the film is fully oxidized, as is typical for sputtered thin films prepared from oxidic targets. With increasing depth, small amounts of “suboxide
A” (blue) and another state, denoted as “suboxide B”, arise. The latter state is just another more
reduced type of suboxide, closer to the metallic state, for the occurrence of pure metallic zirconium
in these samples is unlikely. These reduced states arise from the depth profiling process itself
because the oxygen is preferentially sputtered. Thus, even though the fully oxidic signal drops
to about 80 %, the film most likely still is completely oxidized. Closer to the substrate interface
(Si), a larger suboxide portion is observed, which stems from the fact that the silicon signal is
becoming dominant, reducing the intensity of the zirconium peak, rendering the fitting procedure
less reliable. An example of such a fit is shown in Figure 5(b). This region was obtained for a
depth corresponding to the point that is marked by an arrow in the plot in Figure 5(a). For the fit,
products of Gaussian/Lorentzian peaks were fitted to the spectrum, with restraints on the peak ratios
(determined from the fully oxidized surface) and widths. It can be seen that the model of three
distinct Zr states describes the experimental data very well.

B. Crystallographic properties

Before starting any discussion regarding the crystal structure of the thin films, a determination
of the phases present in the sputtering target materials is appropriate. Figure 6 shows the X-ray
diffractograms of the respective source materials. What can immediately be seen is that 3YSZ,
i.e. the sample with the lowest yttria-content, features more peaks than any of the other samples.
In fact, the reflexes visible can be attributed to both monoclinic ZrO2/YSZ as well as tetragonal
YSZ.8,51 Thus, the starting material for 3 mol% thin films would agree with all those phase diagrams
that predict a corresponding two-phase region at this concentration.22–24
For 8YSZ, the diffractogram looks a lot simpler, with peaks assignable to either tetragonal or
cubic YSZ,51 which already shows a drawback when determining the phases via diffraction techniques: tetragonal and cubic diffractograms are almost impossible to distinguish as both structures
feature lattice planes with the same spacing. For instance, the tetragonal (101) spacings are the same


025119-11

Götsch et al.

AIP Advances 6, 025119 (2016)

FIG. 5. (a) An exemplary chemical state depth profile for the Zr 3d peak of 20YSZ. That small amounts of suboxides, in
the form of two distinct, differently reduced states (suboxide A and B), are visible, stems from the sputtering process during
depth profiling. (b) Zr 3d region for the point indicated with an arrow in (a) including the fitted peaks.

as those of the cubic (111) planes, and the t(002) and c(200) reflexes also are located at the same
position in the diffractogram. However, during previous investigations, the crystal structure of this
powder has been determined to be tetragonal.50 The plot corresponding to 20YSZ thus strongly
resembles that of 8YSZ, making it difficult to determine whether they feature different crystal
structures. That the cubic phase was assigned to 20YSZ was based on the published phase diagrams
discussed above, for none of them feature a tetragonal phase at 20 mol% Y2O3. It has to be noted
that the two peaks between 40◦ and 50◦ that are pronounced for 20YSZ actually arise from the
sample holder (marked with “SH” in the plot), and are visible to a lesser extent in all diffractograms.

The sample with 40 mol% Y2O3 gives rise to a slightly different-looking diffractogram, which
is characteristic for Zr3Y4O12 (i.e. the ordered δ phase).11 The existence of this phase would invalidate all phase diagrams not containing it.20–25,27 It has to be considered, however, that the ordered
compound could either not be thermodynamically stable, or the temperatures required during the
sample pretreatment in these studies could have exceeded the limit of stability for this phase (at
approximately 1300 ◦C), hence causing it to be overlooked. Such could have been the case, for
example, for the early studies by Ruh, Rouanet and Skaggs,22–24 where the samples were heated
with a CO2 laser.
In Figure 7, TEM bright field images are shown to give an overview of the morphology of the
thin films. From Figure 7(a) it can be deduced that the thin film prepared by sputtering the 3YSZ
target consists of small nanocrystallites (about 10 nm in diameter). Strong diffraction contrast arises
from those particles in ideal Bragg orientation, and, as Moiré patterns are visible on some of them,
the grains also seem to overlap. The crystallites do not feature a common, regular shape: while there


025119-12

Götsch et al.

AIP Advances 6, 025119 (2016)

FIG. 6. X-ray diffractograms of the powders, from which the targets were produced. Some of the major peaks are indexed,
with SH denoting reflexes originating from the sample holder. The assignment of the tetragonal phase to 8YSZ was done
according to earlier investigations,50 and 20YSZ is assumed to be cubic based on the published phase diagrams.

are some that look approximately rectangular, most of them are in the form of rounded, irregular
particles.
For 8YSZ (Figure 7(b))), the morphology resembles that of 3YSZ, with some of the crystallites
being larger. Again, crystallinity is confirmed by the presence of Bragg contrast within the grains.
20YSZ in Figure 7(c), too, looks very similar to 3YSZ, albeit the particles are even a bit smaller.
Like for the 3 mol% specimen, Moiré patterns indicate that there are multiple crystallites stacked on

each other, from which information regarding the growth process of the films can be gathered since
a columnar growth (i.e. single grains stretching from one surface to the other through the whole
film), as is often obtained by magnetron sputtering,52 can be excluded. This could be advantageous
for several applications, such as electrolytes, because the non-columnar microstructure would allow
for the preparation of pinhole-free films much easier since, as long as the film is thick enough, the
probability that the pinholes are covered by new particles, is much higher than for adjacent columns.
In Figure 7, some thin areas are still visible between the grains, but the films are also a lot thinner
than what would usually be the case for electrolyte applications.
The overview micrograph of 40YSZ, seen in Figure 7(d), looks vastly different from the other
ones as no defined particles are visible. Rather, the whole film looks rather corrugated and irregular.
Also, while there is strong contrast in places, no particles with either Bragg contrast or Moiré
patterns are discernible.


025119-13

Götsch et al.

AIP Advances 6, 025119 (2016)

FIG. 7. Bright field overview images of the specimens with varying Y2O3 content. While the first three micrographs look
similar, the one for 40 mol% yttria is distinctly different.

The selected area electron diffraction (SAED) patterns of the various samples, as deposited
at 300 ◦C, are displayed in Figure 8 (preparations were carried out at room temperature as well,
but the films were all amorphous and, thus, are not shown here). At first glance, a strong ordering
for all films except for that with 40 mol% becomes apparent from the pronounced texturing in

FIG. 8. Selected area electron diffraction patterns for the samples show a large degree of ordering except for 40YSZ. For
the assignment of tetragonal/cubic structures, see Table IV. Note that the t(200)/t(110) peaks could also be assigned to

t(002)/t(112).


025119-14

Götsch et al.

AIP Advances 6, 025119 (2016)

TABLE IV. The lattice parameter c, as calculated from the diffraction patterns. This parameter is the same, regardless
whether the crystal structure is cubic or tetragonal, since the (002) reflexes in both cases are located at the same positions in
the pattern.
mol% Y2O3

c / nm

3
8
20
40

0.51
0.52
0.52
0.53

the ring-patterns. Only 40YSZ features a Debye-Scherrer like ring pattern typical of disordered
polycrystalline materials. Taking a closer look at the diffraction patterns, the same problem as for
the XRD analysis occurs: it is basically impossible to distinguish between tetragonal and cubic YSZ
simply by looking at the SAED patterns. As mentioned, this stems from the fact that the tetragonal

and cubic structures both have lattice planes with the same spacings, even though the unit cell
dimensions are different.
Before assigning any lattice spacings to each diffraction pattern, an evaluation of the crystal
structure is required. For this, intensity profiles were generated by angular integration of each
diffraction pattern using the PASAD software.53 From the peak positions of four spots, the lattice
constant c (which is the same as a in the case of the cubic system) was calculated. This was chosen
because the tetragonal (002) and cubic (200) planes feature the same spacings and, hence, yield
spots at the same distance from the direct beam.51 The calculation of c was done by a simple
crystallographic relation between the lattice spacing, d hk l , the miller indices (h, k and l), as well as
the the lattice parameter, c (denoted by this letter to better illustrate the connection to the tetragonal
c parameter),

2
c = d hk
· (h2 + k 2 + l 2) .
(1)
l
The lattice constants computed this way are listed in Table IV. It can be observed that there
is a strong increase of the unit cell dimensions upon increasing the yttria-content from 3 mol% to
8 mol%, namely from 0.51 nm to 0.52 nm. This is expected behavior since the ionic radius of Y
is larger than that of Zr. However, by substituting even more Zr with Y (i.e. reaching 20 mol%
Y2O3), c remains approximately constant (at now 0.52 nm) within the margin of error, before
increasing again towards 40 mol% yttria (to 0.53 nm — this specimen is here assumed to be cubic
instead of rhombohedral as well, for details regarding this, see the discussion below). This would be
counter-intuitive, as an increased fraction of larger ionic species will inevitably cause the unit cell
to be scaled up appropriately. Hence, this retention of the lattice parameter can only be explained
by a phase transformation to the cubic polymorph, upon which the unit cell volume would increase
due to the larger a and b dimensions. And, judging from the existing phase diagrams discussed in
Section III, the samples with lower Y2O3 content will be tetragonal, and 20YSZ as well as 40YSZ
cubic.

The texturing in the 3YSZ pattern in Figure 8(a) can be used to assist in determining the lattice
planes that cause the ordered spots to appear: the innermost ring (0.29 nm) can be ascribed to the
tetragonal (101) plane (these planes and spots will from now on be called t(101), where the letter
in front indicates whether the cubic (c) or tetragonal (t) structure is referenced), which stems from
crystallites that are not grown epitaxially, and, thus, exhibit a rather large disorder, causing the more
ring-like appearance of the signals from this plane. The next spots at 0.26 nm can, in principle,
derive from the t(110) or t(002) planes, as they both feature the same spacings. The assignment,
again, is difficult, but can be done by looking at the next diffraction spots (0.18 nm), which, too,
can originate from two lattice planes, namely t(200) and t(112). Looking at the expected angles
between each pair of spots, for t(002) and t(200), one would expect 90◦, which is not the case as the
measured angle between these spots is 45(1)◦. The distinction between either t(002)/t(112) (45.67◦),
t(110)/t(200) (45◦) and t(110)/t(112) (44.33◦) is not unambiguous since all three angles fall within
the standard deviation of the measured angle. However, the zone axis would differ between the


025119-15

Götsch et al.

AIP Advances 6, 025119 (2016)

¯
¯ and t(110)/t(200) would
assignments: for t(002)/t(112), it would be [110],
for t(110)/t(112) [110],
yield a [001] zone axis. A <110> zone axis, resulting from an YSZ{110}/NaCl(001) interface,
could show epitaxy, depending on whether the YSZ surface is O- or Zr-terminated, for the latter
would form a square surface unit cell (with a = 0.359 nm), and the oxygen atoms a rectangular one
(0.503 nm × 0.206 nm), with a certain corrugation in the O positions. For the (001) face, both atom
types arrange in the same square lattice with a periodicity constant of 0.355 nm. Even though the

assignment is still not absolutely clear, the spots are labeled with t(110) and t(200), respectively, for,
in addition to the surface arrangement, an epitaxial growth along one of the principal directions of
the crystal structure seems more likely.
A very similar diffraction pattern is obtained for the specimen with 8 mol% of yttria in
Figure 8(b), with slightly increased lattice spacings for each set of planes: 0.30 nm for the t(101),
0.26 nm and 0.19 nm for t(110) and t(200), respectively. Thus, the same discussion as for 3YSZ
also applies for 8YSZ. The degree of texturing in the SAED pattern is approximately the same,
meaning that this thin film was grown with a high amount of epitaxy as well. The spots in 20YSZ
(Figure 8(c)) can be assigned to the cubic polymorph, as already discussed. Here, the innermost,
disordered ring-like arrangement of spots at 0.30 nm stems from the c(111) planes. The c(200) signals are found at 0.26 nm and show a pronounced ordering. In fact, two sets of spots are discernible,
suggesting two preferred orientations in terms of rotation around the growth direction axis. The
angle to the c(220) spots, which are found at 0.18 nm, is measured to be 45.2(5)◦, correlating nicely
with the theoretical value of 45◦. From these two planes, the zone axis of this film can be determined
to be along the [001] direction. This ordered growth along the unit cell axis corroborates the theory
that, for the tetragonal analogues, the same is more likely than the growth along one of the <110>
directions.
Taking a closer look at the 40YSZ diffraction pattern (Figure 8(d)), a much worse ordering
than for the other samples can be identified; however, there is a very faint texturing visible within
the rings, indicating a not complete disordering. The spacings of these rings can all be attributed to
cubic YSZ, in contrast to the target’s rhombohedral structure (Zr3Y4O12): at 0.30 nm, there is the
c(111) signal, at 0.27 nm the c(200) one, and the c(220) rings are located at 0.19 nm. Still further to
the outside, the ring stemming from the c(311) planes is visible at 0.16 nm. While the rhombohedral
ordered phase would have similar spacings (e.g. 0.30 nm for the (003) planes), the δ phase would
have a lot more rings in the diffraction pattern.54 This suggests that a phase transformation has
taken place during the sputtering process. This could be the result of the sputtered clusters traveling
in close proximity to the hot filament when leaving the sputtering device (see Figure 2) since the
glowing filament is at a temperature well above the stability limit of the δ phase, which, according
to most phase diagrams, is around 1300 ◦C, where the compound transforms either peritectoidically
or dystectoidically. Another reason could be the effect of the substrate: if depositing a thin film
on the cubic sodium chloride (100) facets, the growth in a cubic structure will most likely be

favored due to a template-effect. However, if this was the case, one would assume a larger degree
of ordering since epitaxy would be forced due to this effect. Regarding the substrate, it thus seems
more likely that small amounts of NaCl have been incorporated into the crystal structure of the thin
film, as these ions also have the ability to stabilize cubic zirconia and YSZ.50 It is, of course, also
possible that, during the sputtering process, some argon ions have been implanted in the lattice of
the target, which will stabilize the cubic polymorph.55 Furthermore, the grain size (which, judging
from the diffraction rings and the overview image is very small in the case of this sample) has
a big influence on the crystal structure as well, as demonstrated by Drazin et al.,56 even though
their assessment of the extent of the cubic region clashes with our results regarding the tetragonal
structure for 8YSZ. It is challenging to determine which of these processes is the cause for the
phase transformation. Nevertheless, some conclusions can be drawn that give an indication as to
which reasons are likely and which are not. First, as already mentioned, if this transformation was
the cause of a substrate-template effect, a higher degree of ordering would be expected, which is
not the case; hence, this effect can be excluded. Similarily, if implanted Ar+ were the reason, the
transformation would already occur in the target — then, a cubic target would be sputtered, due to
which one would again expect similar results to, for example, 20YSZ. The disorder, thus, cannot be
explained by a template effect or an argon-induced target transformation. If the phase transition was


025119-16

Götsch et al.

AIP Advances 6, 025119 (2016)

to occur during the time of flight of the sputtered clusters due to thermal reasons, one could again
argue that it’s the same case as for 20YSZ, with cubic particles arriving at the substrate. However, it
is not inconceivable to imagine that, if clusters that are already cubic are heated up by the filament,
they arrive in a “hot” state at the surface. If a rhombohedral-cubic transition has to occur first, on the
other hand, these clusters could lose some heat due to the transformation. Thus, maybe the effective

temperature of the material deposited at the substrate plays a role in the ordering since the particles
have to possess a certain mobility to rearrange on the surface, which will be higher at elevated
temperatures. The effect of Na/Cl impurities in the lattice would only affect the first layer of clusters
arriving at the surface, which could also cause an ordering to happen because to incorporate these
ions into the lattice, a large degree of mobility is required, which also allows for the rearrangement
of the YSZ-atoms itself, although all further grains that have no contact with the substrate could
only be cubic due to epitaxy of the already present crystallites, which would then lead to a high
ordering again. As for the crystallite size effect, this could be possible, but one has to ask why these
crystallites are smaller than for the other targets in the first place; this sounds more like a secondary
effect instead of the cause of this transformation.
To shed more light on the crystallography of these specimens, high-resolution TEM (HRTEM)
images containing lattice fringes are given in Figure 9. In Figure 9(a), a micrograph of 3YSZ
containing multiple crystallites is shown. In the grain on the left, lattice spacings of 0.29 nm can be
¯ lattice planes. The angle between
measured in two directions. These correspond to t(101) and t(101)
◦
them was measured to be 70.0 . From theoretical calculations, a reference value of 69.3◦ is obtained
for these two planes, which is in good agreement with the experimental value. It is to be noted that
this particle is not viewed in a [001] zone axis for, otherwise, these fringes would not be visible.
The high-resolution image of 8YSZ (Figure 9(b)) contains one particle (bottom) also exhibits
t(101) planes (0.29 nm), while, above, there are t(110) fringes with d = 0.25 nm observable. Since
both fringes originate from different grains, no conclusion can be drawn regarding the zone axis,
and no angle-analysis is possible.
In the case of 20YSZ, in Figure 9(c), a particle in [001] zone axis could be imaged (in the
center). This contains both c(200) and c(020) fringes, with a spacing of 0.26 nm each. The angle
between them is measured 90◦, as expected from theory. For 40YSZ, the disorder and nanocrystallinity can immediately be seen in the image (Figure 9(d)): the Fourier transform shows a ring-like

FIG. 9. HRTEM images of the specimens. The scale bar is 10 nm in all cases.

025119-17

Götsch et al.

AIP Advances 6, 025119 (2016)

pattern, even for this small (20 nm × 20 nm) image, and the crystallites visible are smaller than for
the other specimens. In the grain with the markings, two spacings of 0.30 nm can be measured, with
¯ fringes that should span
an angle of 70.4◦ between them. This corresponds to the c(111) and c(111)
◦
70.5 , which correlates nicely with the measured values, confirming the assignment of the cubic
structure to this thin film.
C. Surface topography

Figure 10 shows topography and phase images (insets, 1 àm ì 1 àm each) of the unsupported
thin films, as taken from AFM measurements. In all images, the vacuum side (i.e. not the side
exposed to the NaCl during the growth) is displayed. In Figure 10(a), the surface of 3YSZ is shown
to contain very regular-looking, rectangular structures. These are also visible in the phase image.
With the exception of the one on the top right, all of them feature 90◦ angles (the distorted-looking
one is probably skewed due to the film hanging through between the grid bars). These most likely
are crystallites, like small single crystals grown on top of the surface. In fact, the same kind of
structure can also be seen within the surface (see also the phase image), like platelets layered above
one another. It seems that the crystallites visible on top are just the beginnings of a new layer being
deposited.
The surface of 8YSZ (Figure 10(b)), in principle, looks similar to that of the 3 mol% Y2O3
specimen, in that it contains the same plates, at least to a certain extent. The surface generally is
more irregular, also featuring more round disk-like grains stacked on top of each other. And 20YSZ
(Figure 10(c)) looks even more distinct from the other two surfaces, for it showcases flake-like, very
irregular grains, in a layered structure. So there is a trend in surface topography when going from

3YSZ to 20YSZ, in that the microstructure at the surface becomes less ideal and more irregular and
seemingly disordered (even though the electron diffraction experiments showed that the ordering of
the films is comparable).
40YSZ (Figure 10(d)) looks drastically different, with the surface appearing much smoother,
which would be in line with the observed disordering and nanocrystallinity (i.e. the crystallites are
too small to be imaged using the AFM tip). While no platelets are visible, what can be seen are
the pits already observed previously for 8YSZ thin films when either depositing them at higher

FIG. 10. 1 àm ì 1 àm AFM surface topography (main panels) and phase images (insets) of the unsupported thin films.


025119-18

Götsch et al.

AIP Advances 6, 025119 (2016)

TABLE V. The RMS surface roughness values calculated from the images in Figure 10.
mol% Y2O3

Sq / nm

3
8
20
40

6.3
11.2
6.3

3.7

substrate temperatures or after post-annealing.48 They have been attributed to a high mobility in the
vicinity of vacancies, which could very well be the case for 40YSZ here, too, as this stoichiometry
brings about the highest concentration of oxygen vacancies. This also suggests that the atoms in the
growing thin film have enough mobility to rearrange in order to form such pits.
Table V lists the RMS surface roughness values as calculated from the 1 àm ì 1 àm images
in Figure 10. 40YSZ has a drastically lower roughness than the other samples, as could already
be concluded by looking at the surface topography, with 3YSZ and 20YSZ being comparable
(again, see the similar surface structure), and 8YSZ has a higher roughness, most likely due to the
big trenches in the film. In summary, the films are very similar, and of low roughness, with the
exception of 40YSZ, which features a completely different morphology with a higher degree of
smoothness.

VI. CONCLUSION

Impurity-free thin films of various YSZ samples with different yttria-content could be prepared
with excellent transfer of stoichiometry to the thin film by several modifications to our home-built
ion beam sputtering device. Prior to the thin film deposition, the targets have been analyzed regarding their crystal structures using XRD, and the results showed that the commercially-available
3YSZ sample was a mixture of monoclinic YSZ and tetragonal YSZ, while 8YSZ was cubic.
20YSZ showed a cubic crystal structure and 40YSZ was found to be the rhombohedral δ phase
corresponding to the compound Zr3Y4O12.
In the thin film, different phases are found for some of the compositions, as obtained by
analysis of the lattice parameters calculated from the electron diffraction patterns: 3YSZ is not
heterogeneous any longer, but a single phase, namely tetragonal YSZ, while 8YSZ retains its tetragonal structure. 20YSZ is cubic, just like in the target, and 40YSZ is also cubic. This means that for
two stoichiometries, phase transitions have occurred during the deposition. This could be due to an
effect from the substrate acting as a deposition template, which is plausible in the case of 3YSZ,
because a tetragonal structure will be favoured over the monoclinic one on a cubic substrate, provided similar lattice spacings are present. For 40YSZ, the diffraction pattern shows strong disorder,
3YSZ grows epitaxially (as do the other two samples). There seems to be a high mobility within the
40YSZ film, however, as the surface, which features smaller crystallites than the other specimens,

shows pits that have previously been seen on 8YSZ thin films,48 as well as on single crystals of YSZ
(with 7 mol% yttria).57 They are thought to originate from high mobility around point defects in the
lattice. This would favour theories regarding a rearrangement after the deposition, while, at the same
time, one would expect a higher ordering then. There is also the possibility of a phase transition
during the time of flight of the sputtered clusters due to the limited thermal stability of the Zr3Y4O12
phase and the close proximity to the hot filament.
Figure 11 summarizes the correlations between target and thin film crystal structures in a
schematic way. In this graphic, the symbols relate to the crystal structure as follows: square means
cubic, skewed square monoclinic, the oblong rectangle corresponds to the tetragonal phase, and
the hexagon to the rhombohedral one. The squiggly arrows denote a phase transition, which occur
for 3YSZ (monoclinic/tetragonal → tetragonal) and 40YSZ (rhombohedral → cubic), and, for the
ordered thin films, the directions at the top indicate the zone axes.


025119-19

Götsch et al.

AIP Advances 6, 025119 (2016)

FIG. 11. Schematic thin film phase diagram for YSZ. The symbols denote the crystal structure: the square means cubic,
the skewed square monoclinic, the oblong rectangle tetragonal, and the hexagon depicts the rhombohedral structure. Straight
arrows mean a retention of crystal structure has occurred, while sinusoidal arrows mark a phase transition during the thin film
deposition process. The directions above the thin film phases label the zone axes.

These data will, thus, help with the choice of the composition of YSZ for various applications where a certain crystal structure is desired, since the ionic conductivity is dependent on the
crystal structure. Also, this lays a foundation for further work in this area, as we have been able
to prepare model thin films with a high degree of ordering (at least for stoichiometries between
3 mol% and 20 mol% yttria), being ideally suitable e.g. for further studies using impedance spectrosopy to further elucidate the suitability as electrolytes, catalytical testing (for methane steam
reforming, as used within fuel cells, also with nickel or copper particles embedded in the film),

and optical as well as electronic spectroscopies to gather a deeper understanding of the role of the
yttrium-concentration on their physico-chemical properties.

ACKNOWLEDGMENTS

This work was financially supported by the Austrian Science Fund (FWF) through grant
F4503-N16 and has been performed within the framework of the Forschungsplattform Materials
and Nanoscience.
1

T. H. Etsell and S. N. Flengas, Chem. Rev. 70, 339 (1970).
P. Yashar, J. Rechner, M. S. Wong, W. D. Sproul, and S. A. Barnett, Surf. Coat. Technol. 94-95, 333 (1997).
3 Brian C. H. Steele and A. Heinzel, Nature 414, 345 (2001).
4 N. Miura, T. Sato, S. A. Anggraini, H. Ikeda, and S. Zhuiykov, Ionics 20, 901 (2014).
5 Ceramics Science and Technology, edited by R. Riedel and I.-W. Chen (Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim,
Germany, 2013)
6 P. Amézaga-Madrid, A. Hurtado-Macías, W. Antúnez-Flores, F. Estrada-Ortiz, P. Pizá-Ruiz, and M. Miki-Yoshida, J. Alloys
Compd. 536, S412 (2012).
7 Y. Shen, S. Shao, H. Yu, Z. Fan, H. He, and J. Shao, Appl. Surf. Sci. 254, 552 (2007).
8 D. K. Smith and W. Newkirk, Acta Crystallogr. 18, 983 (1965).
9 B. Bondars, G. Heidemane, J. Grabis, K. Laschke, H. Boysen, J. Schneider, and F. Frey, J. Mater. Sci. 30, 1621 (1995).
10 D.-N. Wang, Y.-Q. Guo, K.-M. Liang, and K. Tao, Science in China 42, 80 (1999).
11 H. G. Scott, Acta Crystallogr., Sect. B: Struct. Crystallogr. Cryst. Chem. 33, 281 (1977).
12 J. M. Pruneda and E. Artacho, Phys. Rev. B 72 (2005).
13 M. G. Paton and E. N. Maslen, Acta Crystallogr. 19, 307 (1965).
14 A. Torres-Huerta, M. Domnguez-Crespo, E. Ramrez-Meneses, and J. Vargas-Garca, Appl. Surf. Sci. 255, 4792 (2009).
15 J. Lian, J. Zhang, F. Namavar, Y. Zhang, F. Lu, H. Haider, K. Garvin, W. J. Weber, and R. C. Ewing, Nanotechnology 20,
245303 (2009).
16 M. Tsuchiya, A. M. Minor, and S. Ramanathan, Philos. Mag. 87, 5673 (2007).
17 M. Tsuchiya, S. K. Sankaranarayanan, and S. Ramanathan, Prog. Mater. Sci. 54, 981 (2009).

18 W. Jung, J. L. Hertz, and H. L. Tuller, Acta Mater. 57, 1399 (2009).
19 J. Chevalier, L. Gremillard, A. V. Virkar, and D. R. Clarke, J. Am. Ceram. Soc. 92, 1901 (2009).
20 P. Duwez, F. H. J. Brown, and F. Odell, J. Electrochem. Soc. 98, 356 (1951).
2


025119-20
21

Götsch et al.

AIP Advances 6, 025119 (2016)

F.-K. Fan, A. K. Kuznetsov, and É. K. Keler, Bull. Acad. Sci. USSR, Div. Chem. Sci. (Engl. Transl.) 12, 542 (1963).
R. Ruh, K. S. Mazdiyasni, and H. O. Bielstein, Am. Ceram. Soc. Bull. 47, 366 (1968).
23 A. Rouanet, C. R. Seances Acad. Sci., Ser. C 267, 1581 (1968).
24 S. R. Skaggs, U.S.A.E.C., Report No. SC-RR-720031, 1 (1972).
25 K. K. Srivastava, R. N. Patil, C. B. Choudhary, K. V. G. K. Gokhale, and E. C. Subbarao, Trans. J. Br. Ceram. Soc. 73, 85
(1974).
26 A. Rouanet, Rev. Int. Hautes Temp. Refract. 8, 161 (1971).
27 H. G. Scott, J. Mater. Sci. 10, 1527 (1975).
28 H. G. Scott, J. Aust. Ceram. Soc. 17, 16 (1981).
29 V. P. Gorelov, Tr. Inst. Elektrochim., Ural. Nauchn. Tsentr, Akad. Nauk SSSR 26, 69 (1978).
30 V. S. Stubican, R. C. Hink, and S. P. Ray, J. Am. Ceram. Soc. 61, 17 (1978).
31 V. S. Stubican and J. R. Hellmann, Adv. Ceram. 3, 25 (1981).
32 V. S. Stubican, J. R. Hellmann, and S. P. Ray, Mater. Sci. Monogr. 10, 257 (1982).
33 C. Pascual and P. Duran, J. Am. Ceram. Soc. 66, 23 (1983).
34 R. Ruh, K. S. Mazdiyasni, P. G. Valentine, and H. O. Bielstein, J. Am. Ceram. Soc. 67, C (1984).
35 N. Yoshikawa and H. Suto, Nippon Kinzoku Gakkaishi 50, 1101 (1986).
36 H. Suto, T. Sakuma, and N. Yoshikawa, Trans. Jpn. Inst. Met. 28, 623 (1987).

37 V. S. Stubican, Adv. Ceram. 24A, 71 (1988).
38 S. A. Degtyarev and G. F. Voronin, Russ. J. Phys. Chem. (Engl. Transl.) 61, 320 (1987).
39 S. A. Degtyarev and G. F. Voronin, CALPHAD: Comput. Coupling Phase Diagrams Thermochem. 12, 73 (1988).
40 S. A. Degtyarev and G. F. Voronin, Russ. J. Inorg. Chem. (Engl. Transl.) 62, 1320 (1988).
41 Y. Du, Z. P. Jin, and P. Y. Huang, Chin. J. Met. Sci. Technol. 6, 13 (1990).
42 Y. Du, Z. P. Jin, and P. Y. Huang, J. Am. Ceram. Soc. 74, 1569 (1991).
43 Z. P. Jin and Y. Du, CALPHAD: Comput. Coupling Phase Diagrams Thermochem. 16, 355 (1992).
44 Y. Suzuki, Solid State Ionics 81, 211 (1995).
45 M. Yashima, M. Kakihana, and M. Yoshimura, Solid State Ionics 86-88, 1131 (1996).
46 Y. Suzuki, Solid State Ionics 95, 227 (1997).
47 O. Fabrichnaya, C. Wang, M. Zinkevich, C. G. Levi, and F. Aldinger, J. Phase Equilib. Diffus. 26, 591 (2005).
48 T. Götsch, L. Mayr, M. Stöger-Pollach, B. Klötzer, and S. Penner, Appl. Surf. Sci. 331, 427 (2015).
49 L. Mayr, N. Köpfle, A. Auer, B. Klötzer, and S. Penner, Rev. Sci. Instrum. 84, 094103 (2013).
50 M. Kogler, E.-M. Köck, S. Vanicek, D. Schmidmair, T. Götsch, M. Stöger-Pollach, C. Hejny, B. Klötzer, and S. Penner,
Inorg. Chem. 53, 13247 (2014).
51 D. G. Lamas and N. E. Walsöe de Reca, J. Mater. Sci. 35, 5563 (2000).
52 J. S. Lamas, W. P. Leroy, and D. Depla, Thin Solid Films 525, 6 (2012).
53 C. Gammer, C. Mangler, C. Rentenberger, and H. P. Karnthaler, Scr. Mater. 63, 312 (2010).
54 S. P. Ray and V. S. Stubican, Mater. Res. Bull. 12, 549 (1977).
55 J. Holgado, R. Escobar Galindo, A. van Veen, H. Schut, J. d. Hosson, and A. González-Elipe, Nucl. Instrum. Methods
Phys. Res., Sect. B 194, 333 (2002).
56 J. W. Drazin, Ricardo H. R. Castro, and D. J. Green, J. Am. Ceram. Soc. 98, 1377 (2015).
57 R. G. Green, L. Barré, and J. B. Giorgi, Surf. Sci. 601, 792 (2007).
22