Content uploaded by Jun Jiang
Author content
All content in this area was uploaded by Jun Jiang on Jul 23, 2014
Content may be subject to copyright.
JECR SPECIAL ISSUE ON ELECTRO-CHEMO-MECHANICS
On the variability of reported ionic conductivity in nanoscale
YSZ thin films
Jun Jiang &Joshua L. Hertz
Received: 15 March 2013 /Accepted: 19 August 2013 /Published online: 11 September 2013
#Springer Science+Business Media New York 2013
Abstract Yttria-stabilized zirconia (YSZ) is the most common
material used as a solid oxide electrolyte, which is a key
component in solid oxide fuel cells, solid oxide electrolysis
cells, and certain chemical sensors. High efficiency in these
devices requires increased oxygen ion conductance at interme-
diate temperatures. Nanoscale YSZ thin films are quite prom-
ising in this regard, as the conduction path may be reduced
below what is conventionally achievable. Still, as the thickness
is decreased to nanoscale, structural properties like lattice pa-
rameter and grain morphology are typically altered. These may
affect the electrochemical properties in non-trivial ways. Recent
reports on nanoscale YSZ thin films have provided inconsistent
and, at times, controversial results. In this paper, we present a
review of reports on nanoscale YSZ thin films, focusing prin-
cipally on single component YSZ films as opposed to hetero-
geneous multilayer films. Reports of significantly increased
conductivity come from studies that use a variety of substrates,
grain morphologies, and, to some extent, film thicknesses.
Mechanical strain in the films is not typically reported but is a
suspected cause of the variability in conductivity.
Keywords Solid oxide fuel cell .Nanoscale thin film .
Variable conductivity .Strain
1 Introduction
Solid oxide fuel cells (SOFCs) are silent, efficient, modular,
and capable of operating on a wide variety of fuels [1–3].
Relative to other fuel cell technologies, the use of a solid
electrolyte provides structural stability and eliminates the
corrosion and containment issues associated with liquid elec-
trolytes. Still, there are relatively few materials with an ion
conductivity at moderate temperatures that approaches that of
liquid electrolytes [4]. Operating at high temperature (≥
800 °C) increases the expense of SOFCs. The desire for
reduced operating temperatures remains a key driver of SOFC
research [5]. Development of electrolyte materials that possess
high oxygen ion conductivity at relatively low temperature is
essential for the increased commercialization of electrochem-
ical devices including fuel cells, gas sensors, and ionic
membranes [1,6].
Yttria-stabilized zirconia (YSZ), which has the advantages
of moderately high ionic conductivity, low electron conduc-
tivity, low cost, and high chemical stability, is a traditional
solid electrolyte material. The primary requirement of the
electrolyte is to be highly conductive to oxygen ions and
highly resistive to electrons (and holes). Electrolytes in an
SOFC are exposed to both the oxidizing cathode and reducing
anode environment, so they must also be chemically and
chemomechanically stable at high temperatures in contact
with both environments. YSZ has been used or researched
for a number of applications, including thermal barrier coat-
ings, high-K dielectrics, optical coatings, oxygen sensors, and
solid oxide fuel cells. This material has satisfactory properties,
the main concern being oxygen ion conductivity that is
insufficient for a number of applications until tempera-
tures of 800 °C to 1000 °C. The fabrication of thin
films—especially nanoscale thin films—has been a promising
way to enhance its conductance at intermediate temperatures.
YSZ thin films have been studied as model solid state elec-
trochemical systems [7–11]aswellaspossibleusewithinfuel
cells [12–15].
Oxygen ion conduction in YSZ is via a vacancy
mechanism and an expression for the temperature de-
pendence of the ionic conductivity within a single
J. Jiang (*):J. L. Hertz
Department of Materials Science and Engineering,
University of Delaware, 201 DuPont Hall, Newark, DE 19716, USA
e-mail: jiangjun@udel.edu
J. L. Hertz
Department of Mechanical Engineering, University of Delaware,
126 Spencer Laboratory, Newark, DE 19716, USA
J Electroceram (2014) 32:37–46
DOI 10.1007/s10832-013-9857-1
crystal of YSZ (or within a single grain of polycrystalline
YSZ) can be derived as [16]
σ¼σ0
kT exp −ΔGAþΔGm
kT
ð1Þ
where ΔG
A
is the energy of vacancy association and ΔG
m
is
the energy of vacancy migration. Migration energies in YSZ
are thought to be larger than the association energy. Conduc-
tion at grain boundaries is typically thought to be reduced by
the presence of insulating phases or intrinsic space charge-
related defect concentration reduction [17].
YSZ thin films may exhibit different properties from the
bulk, including different microstructure, phase, stress state,
stoichiometry, lattice parameter, purity, etc. In addition, het-
erogeneous chemistry is likely to exist at the film-substrate
interface. All of these factors may lead to altered local defect
concentration or mobility. Recent studies have reported YSZ
films with thicknesses ranging from a few microns to a few
nanometers. As the thickness decreases, a variety of interest-
ing phenomena have been reported [18–20]. Multilayer YSZ
thin films have also been studied in order to achieve increased
conductivity from mechanical strain, amorphization, lattice
mismatch defects, or vacancy redistribution induced at the
interfaces [20–23]. In this paper, only single layer nanoscale
YSZ thin films will be reviewed.
A number of techniques have been reported for the pro-
duction of YSZ films, including sol–gel methods using dip
coating and spin coating [24,25]; powder methods such as
electrophoresis and spraying [26]; and physical vapor deposi-
tion methods, including evaporation, sputtering [7,27,28],
and pulsed laser deposition (PLD) [23,29]. These methods for
the production of YSZ thin films were reviewed previously
[30].
From the recently published literature (Table 1), nanoscale
YSZ thin films are reported to exhibit a very wide range of
conductivity values, as shown in Fig. 1(a). This figure presents
data from researchers that used different deposition techniques,
substrates, and processing atmospheres to prepare the nanoscale
thin films, but all represent films of nominally pure YSZ
deposited as a single layer. The dopant contents of all of these
films were between 7.5–9.5 mol%. All of the measurements
were recorded with the sample in air. YSZ is very difficult to
chemically reduce, typically only exhibiting significant electron
conduction at oxygen partial pressures well below 10
−20
atm.
Combined with the high oxygen diffusivity and the short dif-
fusion path from the air interfacetotheentirevolumeofthefilm
samples, it is difficult to imagine that any of these films, even if
slightly reduced when first deposited, remain reduced when
heated to measurement temperatures. Thus, it is strongly as-
sumed that the values presented in Fig. 1(a) represent oxygen
ion and not electron (or hole) conductivity. From Fig. 1(a),YSZ
thin films were found to exhibit conductivities as low as 10
−4
S/
cm and as high as 1 S/cm at 500 °C. In addition, the activation
energy for nanoscale YSZ thin films ranged from 0.62 eV to
1.24 eV in the presented temperature range.
The range of conductivity values reported for the films can be
compared to Fig. 1(b), which similarly plots the conductivities
of polycrystalline and single crystal bulk YSZ samples mea-
sured in a number of the same reports for calibration of the thin
film results (Table 2). As can be seen, the bulk samples fall
within a very narrow range of values, as expected for the
chemically stable and relatively impurity-insensitive YSZ. From
Tabl e 1 Literature review on
YSZ thin films with respect to
their dopant concentration, sub-
strate, processing method and
thickness, as shown in Fig. 1(a)
The symbol “*”denotes the film
that was reported to have the
maximum deviation in conduc-
tivity from expected values and
wasthuschosentobepresentedin
Fig. 1(a)
The data is ordered in the manner
in which each reference is
discussed in detail below
No. Dopant percentage Substrate Processing method Ref. Film thickness (nm)
19.5% MgO(001) PLD [31] 15*, 29, 58–2000
28.7% Al
2
O
3
Spin coating [32]400–700*
3 8.7 % MgO (110), MgO (111) Sputtering [19] 58*, 107, 194
4 9.5 % MgO (100) Electron beam
evaporation
[33] 17*, 35, 70, 210
5 8 % Si (100) Sol–gel [25]580*
69% Al
2
O
3
(0001) Sputtering [34] 6*, 12, 50, 100
78% MgO(100) PLD [35]12*,25
8 8 % Si (100) PLD [36] 20*, 55, 90
9 8 % Si (100) PLD [37] 20, 55, 90*
10 8 % Al
2
O
3
PLD [37]105*
11 8 % Al
2
O
3
(0001) PLD [38]600–1500*
12 8 % MgO (100) Sputtering [39]500–800*
13 9.1 % SiO
2
Sputtering [8]1000*
14 8 % Al
2
O
3
(0001) PLD [40] 28*, 250
15 8 % MgO (100) PLD [40]65*
38 J Electroceram (2014) 32:37–46
Fig. 1(b), YSZ bulk samples were found to exhibit conductiv-
ities all within 3×10
−4
S/cm and 10
−3
S/cm at 500 °C. In
addition, the activation energies ranged within 1.02 eV and
1.23 eV in the presented temperature range. The relatively good
repeatability among the bulk samples comes despite differences
in preparation method, dopant content, and, likely, purity. The
comparative lack of repeatability among the thin film results has
been somewhat controversial. In this paper, single layer nano-
scale YSZ thin films are reviewed, aiming at finding any prin-
ciples closely associated with variations in conductivity.
1.1 Reports indicating increased conductivity with nanoscale
thin films
Kosacki et al. [31,41] reported in-plane ionic conductivity as
a function of film thickness for films that ranged from 15 nm
to 2 μm for YSZ films with (001) texture grown epitaxially on
MgO (001) substrates via PLD at 500 °C. The in-plane con-
ductivity was measured by two-probe impedance using silver
electrodes. A 1–2 orders of magnitude increase in the conduc-
tivity was observed for films less than 60 nm thick, with a
maximum conductivity at 800 °C of 0.6 S/cm observed for a
15 nm thick film. The author concluded that parallel paths for
ion conduction existed from a YSZ bulk path and a YSZ/
substrate interfacial path. Both reports suggested this as the
cause of an observed increasing conductivity with decreasing
film thickness. In later work [32], the authors also prepared
8.7 mol% nanocrystalline YSZ thin films in the thickness
range of 0.4 μm–0.7 μmonAl
2
O
3
substrates using a polymer
precursor spin coating technique. The conductivity of nano-
crystalline YSZ thin films was about one order of magnitude
higher compared with bulk polycrystalline YSZ.
Sillassen et al. [19](seealso[42]) prepared 8.7 mol% YSZ
thin films by reactive DC magnetron sputtering with thickness
varied from 58nm to 420 nm using MgO (110), SrTiO
3
(STO)
(100), and MgO (111) substrates with a deposition tempera-
ture of 800 °C. Pole figures were used to demonstrate epitaxial
growth of YSZ (220) on MgO (110), YSZ (200) on STO
(100), and YSZ (111) on MgO (111), however epitaxy was
lost for films deposited on MgO when above an orientation-
dependent critical thickness. The in-plane conductivity was
measured using a four-terminal configuration with Ag paste
electrodes. A greater than three order of magnitude increase in
the lateral ionic conductivity of epitaxial YSZ films relative to
bulk ceramics was reported. A superposition of two parallel
contributions was suggested: one due to the bulk and one
attributable to the film-substrate interface. At temperatures
below≈350 °C, the contribution from the film-substrate inter-
face dominated with a decrease in activation energy, whereas
at higher temperatures a more “bulk-like”conductivity was
Tabl e 2 Literature review on bulk YSZ with respect to their microstruc-
ture and dopant concentration, as shown in Fig. 1(b)
No. Sample Dopant concentration Ref.
1 Polycrystalline 8.7 % [32]
2 Polycrystalline 8 % [19]
3 Single crystal 10 % [33]
4 Single crystal 8 % [34]
5 Polycrystalline 8 % [35]
6 Polycrystalline Not reported [36]
7 Bulk grain 8 % [38]
8Bulk 8% [39]
9 Polycrystalline 8 % [40]
Fig. 1 Arrhenius plots of the electrical conductivities of YSZ taken from
a number of literature sources. In both plots, the line represents typical
values reported for bulk YSZ, calculated as a linear regression of all of the
data presented in plot (b). The samples are: (a) thin films (data references
are listed in Table 1); and (b) bulk samples, both single crystal and
polycrystalline (data references are listed in Table 2). Many of the same
references supply data to both plots
J Electroceram (2014) 32:37–46 39
observed with activation energies close to the activation ener-
gy for bulk YSZ ceramic. The strain model proposed by
Schichteletal.[20] was used to explain the increase in
conductivity for the films deposited on STO. The high density
of misfit dislocations at the semicoherent interface was be-
lieved to have led to even larger conductivity increases in the
YSZ/MgO system.
Karthikeyan et al. [33] prepared 9.5 mol% YSZ thin films
with thickness range from 17 nm to 210 nm on MgO (100)
substrates by electron beam evaporation. Despite the relatively
high dopant levels, the films were found to be a mixture of cubic
and tetragonal phases. The in-plane conductivity was measured
by two-probe impedance using platinum electrodes. As with the
reports discussed above, the conductivity of the films increased
with decreasing film thickness. The increases in conductivity
were attributed to space charge effects at the film-substrate
interface and at the grain boundaries. As evidence, the authors
noted a high activation energy for the conductivity relaxation
time as determined by impedance measurements. The increased
conductivity was also observed in samples deposited on Al
2
O
3
and suggested that it was not primarily an effect specific to one
particular kind of substrate. In later work, Karthikeyan et al. [43]
reported microstructural studies carried out on YSZ thin films
grown on MgO (100) and Ge (100) single crystal substrates
using in-situ transmission electron microscopy. A one order of
magnitude increase in total conductivity was reported for films
less than 20 nm thick grown on an MgO substrate. For 17 nm
films, the authors again noted an increased activation energy of
the electrical relaxation time of 1.7 eV. This activation energy
was ~1.1 eV for 933 nm thick films, comparable to bulk YSZ.
Similar to the previous reports, increasing conductivity with
decreasing film thickness gave evidence of enhanced interfacial
conductivity.
Zhang et al. [25] fabricated dense, crack-free, and homo-
geneous (RE
2
O
3
)
0.08
(ZrO
2
)
0.92
(RE = Sc, Y) nanocrystalline
thin films on Si (100) substrates using a sol–gel method. At
temperatures above 600 °C, the electrical conductivity of
(Sc
2
O
3
)
0.08
(ZrO
2
)
0.92
and (Y
2
O
3
)
0.08
(ZrO
2
)
0.92
nanocrystaline
thin films in pure cubic phase was found to be one order of
magnitude higher compared with that of the corresponding
bulk materials. The in-plane conductivity was measured by
two-probe impedance using platinum electrodes. The authors
attributed the conductivity increase to reductions in the grain
boundary resistance; this was related to nanometric grain sizes
causing increased grain boundary purity (as previously dem-
onstrated in traditionally processed YSZ by Aoki et al. [44]).
Jiang et al. [34] used sputtering to create epitaxial YSZ thin
films on (0001) Al
2
O
3
substrates at a deposition temperature
of 650 °C. Growth of the films was in the (111) direction.
Samples with thickness of 6 nm, 15 nm, 25 nm, and 100 nm
were characterized structurally and by impedance spectrosco-
py. The in-plane conductivity was measured by two-probe
impedance using interdigitated platinum electrodes. Low-
angle grain boundaries were found by high resolution trans-
mission electron microscopy, but they were not disruptive to
the lattice structure. The thinnest films were found to have a
moderate increase in conductivity upon annealing at 650 °C in
air, which was believed to be due to recrystallization around
the low angle grain boundaries. After the conductivity values
stabilized, it was found that the conductivity of the 100 nm
thick films was similar in both magnitude and activation
energy to bulk samples. On the other hand, the 6 nm thick
samples exhibited an activation energy of 0.79 eV leading to a
conductivity that was slightly improved from bulk values at
temperatures ≥500 °C but improved from bulk values by
about 1.5 orders of magnitude at 300 °C. From the structural
characterization, significant in-plane compressive strain and
out-of-plane dilatative strain was found in amounts that in-
creased as the film thickness decreased.
1.2 Reports indicating decreased conductivity with nanoscale
thin films
Guo et al. [35] deposited polycrystalline 8 mol% YSZ thin
films by PLD with thicknesses of 12 nm and 25 nm on MgO
(100) substrates and measured the conductivity in dry and
humid O
2
. Impedance spectra were measured with platinum
electrodes. The ionic conductivity of the nanostructured films
was lower by about a factor of 4 compared with microcrystal-
line bulk ceramics. The authors attributed this decrease to
lower bulk and grain-boundary conductivities, and thus the
influence of the ZrO
2
/MgO interface on ionic conduction was
considered to be negligible. The authors further concluded
that proton conduction was insignificant in the nanostructured
films, even when measured in a humidified environment.
Navickas et al. [36] prepared 8 mol% YSZ thin films
of 20 nm, 55 nm, and 90 nm thickness by PLD onto Si
(100) substrates with a native silica layer. The across-
plane conductivity was measured with circular gold
electrodes. The conductivity of the YSZ films did not signif-
icantly depend on the layer thickness and was 3–4 times lower
than a YSZ polycrystalline sample. In addition, Navickas et al.
[37] suggested that the blocking effect of a native silica
interlayer on a Si substrate allowed anisotropy studies of ion
conduction in thin films and that simultaneous in- and across-
plane conductivity measurements on YSZ layers in a single
impedance spectrum were possible in a limited temperature
range. It was found that the across-plane conductivity of YSZ
thin films was similar to the bulk values of a macroscopic
polycrystalline sample and approximately one order of mag-
nitude higher than the in-plane values. The activation energy
of across-plane ionic conductivity (~0.8 eV) was lower than
that of in-plane conduction (~1 eV). These results were attrib-
uted to the large number of grain boundaries surrounding
columnar grains impeding the in-plane (but not across-plane)
ion transport.
40 J Electroceram (2014) 32:37–46
1.3 Reports indicating unaltered conductivity with nanoscale
thin films
Joo et al. [38] deposited YSZ thinfilms with thickness ranging
between 0.6 μm and 1.5 μm on Pt (111), Pt (200) and Al
2
O
3
(0001) using PLD. The conductivity was measured using both
impedance and two-probe DC methods using platinum elec-
trodes. The across-plane and in-plane conductivity values of
the films were similar. In addition, the ionic conductivity and
activation energy of the films were similar to those of bulk
YSZ. Similar conductivities were found regardless of thick-
ness (including both film and bulk samples) and mode of
measurement (in-plane or across-plane conductivity).
Rivera et al. [39] grew textured YSZ thin films with thick-
ness in the range 500 nm –800 nm on MgO (100) by RF
sputtering at 650 °C. Films were strongly oriented along the
YSZ (200) direction, but with additional YSZ (111) peaks
visible. The confirmed column-like structure had no grain
boundaries perpendicular to the growth direction. Admittance
spectroscopy was measured with gold electrodes. The DC
across-plane conductivity in the thin films was found to be
highly similar to a bulk sample.
Jung et al. [8] focused on the ability of nanometric grain
sizes to stabilize the cubic fluorite phase at reduced yttria
concentration. Using sputtering onto unheated fused silica
substrates, 1 μm thick polycrystalline films were deposited.
Impedance spectra were measured with interdigitated plati-
num electrodes. Conductivity for a 9.1 mol% YSZ film was
similar (though perhaps slightly larger than) bulk samples.
Gerstl et al. [40] prepared 8 mol% YSZ thin films with
thicknesses of 65 nm on MgO (100) and 28 nm and 250 nm on
Al
2
O
3
(0001) using PLD at 500 °C –650 °C. The film on
MgO was further annealed for 16 h at 1000 °C. Impedance
spectra were measured with interdigitated Au/Cr electrodes.
The bulk conductivity of all three samples were found to be
close to that of the macroscopic YSZ polycrystal and the
activation energy was around 1 eV, matching reference
measurements.
2 Discussion
The ionic conductivity increases reported in nanoscale YSZ
thin films remain controversial, as the reports have been
difficult to repeat. A number of experimental difficulties bur-
den these measurements, including accurate measurement of
the conductivity of relatively resistive thin films. In a highly
revealing and informative paper, Kim et al. [45] found that the
current flowing through myriad conduction paths parallel to
the thin films may increase the apparent conductivity of ultra-
thin single layer YSZ by at least one order of magnitude,
especially when the conductivity is measured in an in-plane
configuration using electrodes spaced far apart (i.e., when the
expected resistance of the film is high). Leakage resistances
were found to be as low as 150 kΩat 600 °C. The resistance of
a film structure with a thickness of just a few nanometers and
inter-electrode spacing of a few millimeters may approach or
exceed this value, leading to incorrectly reported, anomalous-
ly low “film”resistances. Measurement of a sample blank is
recommended in order to compare the resistance with that of
samples that are identical save the inclusion of a film. Such a
comparison was reported in a few of the studies summarized
above, and Kim et al. show that taking a few precautions
indeed allows the accurate determination of the conductivity
of 40 nm thick (and presumably thinner) YSZ films.
When considering reported conductivity of YSZ, it is help-
ful to separately consider the concentration and the mobility of
the mobile oxygen vacancies. For the relatively high dopant
concentrations used in nearly all studies of doped zirconia, the
concentration of oxygen vacancies is fixed. The short Debye
length attendant with such dopant levels means that space
charge regions can only affect the defect concentration over
the span of a few lattice parameters or less. Increased conduc-
tivity is thus likely to arise not from increased vacancy con-
centration but rather from increased vacancy mobility within
the lattice structure. This situation can be compared to studies
that use charge carrier redistribution near interfaces in
undoped materials to increase the mobile defect concentration
[46]. As discussed above, decreasing the defect migration and/
or association enthalpy is key to affecting significant change
in the mobility, and doing so should lead to reduced activation
energy. Tailoring the mechanical strain or introducing a high
concentration of linear defects remains a theoretically valid
way to change the lattice structure in a way that may increase
(or, for that matter, decrease) the ion mobility.
Activation energies of the conductivity-temperature prod-
uct are typically around 1.1 eV –1.2 eV for bulk YSZ.
Sillassen, et al. [19] suggested that the observed activation
energy in their films of 1.24 eV at high temperature corre-
sponds to bulk-like conduction, while the activation energy of
0.71 eV at low temperature corresponds to an interfacial
conduction mechanism. Kosacki, et al. obtained [31]activa-
tion energies as low as 0.62 eV for the thinner films,
supporting the notion of a high mobility interfacial conduction
path. Still, for films with intermediate thickness, it was report-
ed that the activation energy was 0.62 eV at measurement
temperatures greater than 650 °C but was close to bulk values
at temperatures below 650 °C. It is not clear how parallel bulk
and interfacial conduction paths could lead to a reduced
activation energy at increased temperature, since whichever
path is more conductive should dominate the behavior. Jiang
et al. [34] also found that the activation energy decreased as
the film thickness decreased, from 0.99 eV for 100 nm thick
films to 0.79 eV for 6 nm thick films. Unlike the other two
studies, however, there was no obvious change in activation
energy as a function of temperature for any sample.
J Electroceram (2014) 32:37–46 41
2.1 The effect of thickness
Since the interfaces with the substrate and the surrounding
environment are the defining features that separate thin film
and bulk samples, the relative portion of the volume that is in
proximity to the interfaces—in other words, the film thick-
ness—may be of key importance. The decreasing thickness
may induce different microstructure, phase, stress state, stoi-
chiometry, lattice parameter, or other feature correlated to the
conductivity. Indeed, most of the reports that found increased
conductivity in YSZ films also found that the conductivity
increased as the film thickness decreased. Karthikeyan and
Kosacki found that conductivity increased when the thickness
was below 210 nm and 58 nm, respectively. Taking all
references into account, increased (or decreased) con-
ductivity is generally found when the film thickness is
below about 500 nm.
Figure 2plots the conductivities reported at 500 °C of YSZ
films as a function of film thickness. At 500 °C, bulk YSZ is
consistently reported to have a conductivity of around 10
−3
S/
cm. For film thicknesses below 1 μm, and especially below
100 nm, reported conductivity values range over four orders
of magnitude, from 10
-4
S/cm to nearly 1 S/cm. A bias
towards the reporting of conductivity values that are increased
and not decreased from bulk values may be expected. In spite
of this, no clear correlation can be directly found between
conductivity values and thickness except that the variability
clearly increases as thickness decreases. Though thickness
measurements become increasingly difficult—and therefore
may be more uncertain—as a film becomes thinner, the var-
iability in reported conductivity values greatly exceeds any
expected film thickness measurement uncertainty.
As discussed above, Kim, et al. [46]showthatleakage
current parallel to the films under test can lead to an illusional
effect where the conductivity seems to increase with decreas-
ing thickness. Since the same parallel resistance is divided by
decreasing film thickness, an inverse relationship between
conductivity and film thickness might be expected in these
cases; a slope of −1 should be found in log-log plots of
conductivity vs. film thickness. Interestingly, the inset of
Fig. 2indicates slopes that are smaller (more negative) than
this value.
2.2 The effect of substrate
If the substrate-film interface determines the film’sproperties,
then the substrate composition, crystal structure, and orienta-
tion may be expected to play a role. The data presented in
Fig. 1included reports using a number of different substrates,
including MgO, Al
2
O
3
, Si, and SiO
2
. No clear effect of
substrate can be found. Several groups have directly measured
films deposited identically (including thickness) except for the
substrate, and these data are presented in Fig. 3. Karthikayan
et al. [33] deposited 17 nm YSZ on MgO and Al
2
O
3
and
Sillassen et al. [19] prepared 58 nm YSZ on MgO with (110)
and (111) orientations. In both cases, the conductivities were
quite close no matter the substrate. Based on these reports, the
effect of the substrate on the YSZ film conductivity seems
surprisingly negligible and other causes seem to be inducing
Fig. 2 The values reported for electrical conductivities at 500 °C of YSZ
thin films as a function of their thickness indicate a positive correlation
between film thickness and experimental repeatability. The inset presents
the same information in a double log scale. The symbols represent data
from the same references as indicated in Fig 1(a).UnlikeFig.1(a),plotted
here are the data from all films reported in these references, not just the
films that maximally deviate from the expected value. Note that some
values are interpolated or extrapolated from original data to 500 °C. The
black dashed line represents an average conductivity of bulk YSZ at
500 °C. Dotted lines are guides for the eye
Fig. 3 Arrhenius plots of electrical conductivities of YSZ thin films with
similar thickness grown on substrates of MgO or Al
2
O
3
. The symbols
represent the same data as indicated in Fig. 1(a)
42 J Electroceram (2014) 32:37–46
the increased conduction. Obtaining the same result despite
changing the substrate composition also suggests that leakage
current through the substrate is not significant, though other
sources of leakage current remain possible [45].
2.3 The effect of microstructure
Depending on the deposition conditions, thin films can be
polycrystalline or epitaxial and thus free of grain boundary
blocking effects. In addition, epitaxial thin films can exhibit
different crystallographic texturing. This effect is usually at-
tributed to lattice mismatch and free surface energy minimi-
zations. As an example of the different microstructures that
can be created, Guo et al. [35] obtained polycrystalline YSZ
thin films via PLD at a substrate temperature of 700 °C, while
Kosacki [31] et al. achieved epitaxial YSZ thin films in the
[100] direction by PLD on MgO (100).
The blocking effect of grain boundaries is well established in
doped zirconia. Even in the absence of impurity phases, the
grain boundary resistivity can betwoorthreeordersofmagni-
tude higher than the bulk resistivity [47]. This fact suggests that
polycrystalline thin films, which typically exhibit nanometric
grain sizes and thus a large number of grain boundaries between
laterally separated electrodes, are likely to exhibit reduced
conduction relative to conventionally fabricated samples. Ac-
cordingly, Navickas et al. [37] found that the across-plane
conductivity of YSZ thin films was approximately one order
of magnitude higher than the in-plane values. The activation
energy of across-plane conduction (≈0.8eV)wasalsolower
than that of in-plane conduction (≈1 eV). The films were
deposited by PLD and had a columnar grain structure. Thus,
the difference between the across-plane and in-plane conduc-
tion behavior was attributed to the effect of grain boundaries
partially blocking the in-plane ion transport.
Other studies cannot so clearly link grain boundary resistance
to noted changes in overall conductivity. Kosacki et al. [32]
prepared 8.7 mol% nanocrystalline YSZ thin films in the thick-
ness range of 0.4 μm–0.7 μmonAl
2
O
3
substrates. The con-
ductivity of the nanocrystalline YSZ thin films was about one
order of magnitude higher compared with bulk polycrystalline
YSZ. Still, the conductivity of nanocrystalline YSZ thin films
with a thickness around 0.4 μm–0.7 μm was higher than that of
29 nm thick YSZ thin film with epitaxial growth on MgO (001)
reported by the same group [31] (not shown in Fig. 1(a)).
2.4 The effect of composition
Jung et al. [8] reported that nanometric grain sizes in YSZ thin
films helped stabilize the cubic phase at dopant contents less
than normally required for bulk samples. This effect has been
noted by others [48]. Since the highest conductivity in YSZ is
normally achieved at the lowest dopant content where cubic
phase is stabilized, this enabled a means to reduce the yttria
content and achieve improved conductivity. As shown in
Fig. 4, the peak conductivity for YSZ films prepared by
reactive RF sputtering occurred at 6.5 mol% Y
2
O
3
(for T ≥
400 °C) vs 9 mol% Y
2
O
3
in bulk YSZ.
2.5 The effect of strain
Lattice strain is a suspected means to improve ion conduction
in solid electrolytes, including in YSZ. At the same time, very
high residual stresses are known to exist in thin films, espe-
cially polycrystalline films [49], and YSZ films in particular
[50]. The stresses can arise from thermal expansion mismatch
with the substrate, lattice parameter mismatch at coherent
interfaces, ion bombardment during deposition, grain coales-
cence, impurity removal during thermal treatment, recrystal-
lization, and other causes. Despite this, residual stress remains
underreported in the literature. It can be difficult to control the
residual stress in a film, and the biaxial stress state of a thin
film may in fact change enormously with differing process
parameters or during the course of a seemingly innocuous
thermal treatment. For example, a 1 μmthickYSZfilmsputter
deposited onto an unheated fused silica substrate was mea-
sured to have around 0.5 GPa of compressive residual stress.
After heating the film to 450 °C and cooling back to room
temperature, the film was measured to have around 0.3 GPa of
tensile residual stress [8]. Similar effects were noted in films
depositedonSisubstrates[50]. With such difficulty in
controlling film stress, it is difficult to test the role of
stressonionconductivityinYSZfilmsinasystematic
manner (a notable exception is to be found in Ref. [20]
and other works from this group).
Further, Rupp [51] reported on the importance of lattice
strain to ionic conduction in metal oxide thin films. It was
Fig. 4 Ionic conductivity as a function of Y
2
O
3
compositions of YSZ
thin films. Solid circles represent YSZ thin films, solid triangles and
squares represent bulk YSZ ceramics. Peak conductivities for bulk and
thin film YSZ are indicated by arrows. (Figure reprinted from Ref. [8])
J Electroceram (2014) 32:37–46 43
suggested that different thin film processing methods and
thermal histories might induce or change lattice strains, in-
cluding intra-grain microstrains. Lattice strain changes the
bond strength between cations and anions and thus can alter
ion migration barriers. The author proposed to take lattice
strain as an additional microstructural measurement to char-
acterize atomic disorder in metal oxide thin films. Such mea-
surements are made difficult by microstrains that yield no
change in average lattice parameter but significantly increased
spread in their values across regions as small as one grain.
Atomistic simulations suggest that increased oxygen ion
mobility can come from tensile strain in YSZ [52]. Kushima
and Yildiz predicted an enhancement of oxygen diffusivity in
biaxially tensile strained 9 mol% YSZ using kinetic Monte
Carlo simulations at different temperature and strain states.
Dilation is suggested to decrease the ion migration barrier
largely by weakening the oxygen-cation bond strength.
Though increases in conductivity of up to 3–4ordersof
magnitude are predicted to be possible at low temperature,
the effect of strain is also predicted to diminish as temperature
is increased.
The mismatch of thermal expansion coefficient between a
YSZ thin film and the underlying substrate may induce stress,
as a sample is typically measured in a range of temperatures.
However, thermal stresses are much less likely than other
sources of residual stress to cause appreciable changes in
conductivity. The averaged linear thermal expansion coeffi-
cient between 25 °C and 1000 °C is 10.5·10
−6
K
−1
for 8 mol%
YSZ [53]. MgO has a thermal expansion coefficient of 13.9·
10
−6
K
−1
between 25 °C and 1000 °C [54] and Al
2
O
3
is
similar. Thermal expansion mismatch strain is, to first order,
(α
s
−α
f
)∗ΔT.ForΔT= 1000 K, the mismatch strain is only ≈
0.001. According to Kushima and Yildiz [52], this amount of
strain would cause a roughly 50 % increase in oxygen diffu-
sivity, which is negligible relative to the reported orders of
magnitude changes in conductivity. Lattice parameter
mismatch between the YSZ thin film and substrate may also
cause stress in thin films. However, as shown in Fig. 3,the
effect of the substrate is negligible for at least some substrates
(MgO and Al
2
O
3
in particular). So, it is supposed that the
substrate mismatch contribution to film stress is less than that
arising from the other mechanisms by which residual stresses
are known to develop in thin films. These other mechanisms
will be highly dependent upon specifics of the film deposition
process, and thus may lead to the low repeatability demon-
strated in Fig. 1a. It should be mentioned that the amount of
strain required to accommodate lattice mismatch between
YSZ and either MgO or Al
2
O
3
is sufficiently large to suggest
incoherent interfaces.
A few of the reports of improved YSZ film conductivity
suggested significant residual stresses in the films. Sillassen
et al. found that the cubic YSZ XRD peaks were shifted
towards larger d-values in the out-of-plane direction [19].
Roughly −1 % in-plane elastic (compressive) strain was cal-
culated from the apparent peak locations. Jiang et al. [34]also
used x-ray and electron diffraction measurements to find that
films that were 50 nm thick had an effective in-plane lattice
parameter of 5.11 Å and out-of-plane lattice parameter of
5.19 Å, while for 6 nm thick film they were 5.10 Å and and
5.21 Å, respectively. As the film thickness decreased from
100 nm to 6 nm, the out-of-plane strain increased from 1 % to
2 % and the ionic conductivity increased by a factor of 5 at
650°Candafactorof15at300°C.Unfortunately,other
papers described above provide little direct data for
comparison.
Janek’s group has reported a series of theoretical and ex-
perimental studies on the influence of interface and elastic
strain on ionic conductivity in nanoscale solid electrolyte thin
films [20–22,55–57]. Nanoscale multilayer thin films,
consisting of YSZ and insulating rare earth metal oxides
(RE
2
O
3
=Y
2
O
3
,Lu
2
O
3
and Sc
2
O
3
) were fabricated with
coherent or semicoherent interfaces and the in-plane conduc-
tivity was measured. It was found that in YSZ/Y
2
O
3
system,
where the lattice misfit was 3.09 % dilative, the ionic conduc-
tivity increased, with the amount of increase in proportion to
the reciprocal thickness of the YSZ layers. In the YSZ/Sc
2
O
3
system, where lattice misfit was 4.28 % compressive, the ionic
conductivity decreased and again the effect increased in pro-
portion to the reciprocal thickness of YSZ layers. In the YSZ/
Lu
2
O
3
system, where lattice misfit was only 1.02 %, the ionic
conductivity remained unchanged and was independent of the
thickness of YSZ layers.
It was proposed in these papers that the strained interfaces
would change the total conductivity of the multilayers by an
activation volume mechanism. The ionic conductivity of
strained interface (σ
int
) was different with the bulk conduc-
tivity (σ
vol
) and quantified as below [20]:
lnσint
σvol
≈1
3
ΔVM
Vo::
RT
YYSZ
1−vYSZ
⋅fZrO2=RE2O3ð2Þ
where ΔVM
Vo⋅⋅ is the volume of migration for the oxygen
vacancies, Y
YSZ
is the elasticity modulus and v
YSZ
the Poisson
ratio of YSZ, and fZrO2=RE 2O3is the elastic strain. ΔVM
Vo⋅⋅ is
positive for a vacancy mechanism, so when fZrO2=RE2O3is
positive, as in the YSZ/Y
2
O
3
system, σ
int
is larger than σ
vol
.
When fZrO2=RE2O3is negative, as in the YSZ/Sc
2
O
3
system,
σ
int
is smaller than σ
vol
. Thus, dilative strain increases the
ionic conductivity of the YSZ near the interface, and com-
pressive strain has the opposite effect. Upon decreasing the
thickness of YSZ layers, the volume ratio of YSZ near the
interface increased, and so these effects are proportionally
more significant to the total conductivity. Ultimately,
diffusivity/conductivity increases by about a factor of 2 at
520 °C –560 °C were found for the thinnest YSZ films in
44 J Electroceram (2014) 32:37–46
proximity to a Y
2
O
3
lattice, corresponding well to the activa-
tion volume of 2.08 cm
3
·mol
−1
that was reported previously
for a YSZ bulk single crystal [58]. Since this mechanism
effectively reduced the activation energy of migration, larger
effects are be expected at lower temperatures.
Basedonthesimulationworkof Kushima and Yildiz, the
roughly 1 % strain found in the two works mentioned above
would increase the diffusivity by a factor of 5 around 500 °C.
Such enhancement is insufficient to fully explain the orders of
magnitude conductivity increases suggested in References [19],
but are very consistent with the findings reported in [34]. More
careful characterization of residual stresses and more direct
experimental verification of the simulations are in order.
Kushima and Yildiz did not directly explore the effects of in-
plane compressive strain, however Schichtel et al., found exper-
imentally that tensile strain indeed improved conductivity while
compressive strains decreased it [20]. Thus, the fact of increased
variability in reported conductivity (if not yet its quantitative
magnitude) can be explained by largely uncontrolled residual
stresses. We recommend that, at a minimum, lattice parameters
be quantified and reported in future work in this area, with
expectations of biaxial stress states in typical thin films.
3 Conclusions
Recent reports on thin YSZ films have suggested that a very
wide variation in conductivity can be observed, especially as
the thickness decreases below 500 nm. Here, we have shown
that the variations in conductivity do not seem to correlate
with any particular substrate, measurement geometry, micro-
structure, or composition. We suggest instead that the effects
are due to lattice strains in the films, which have been sug-
gested by both theory and experimentation to yield large
changes in conductivity in YSZ. Residual stresses in YSZ
films have been measured to be 100’s of MPa with strong
dependence on processing parameters and thermal history.
Still, residual stress often goes unmeasured and unreported
in the literature. Control of this parameter is thus suggested as
crucial to understanding nanoscale YSZ thin film behavior.
Acknowledgement This work was supported by the U.S. Department
of Energy, Office of Basic Energy Sciences, Division of Materials
Sciences and Engineering under Award DE-SC0005403.
References
1. N.Q. Minh, J. Am. Ceram. Soc. 76, 563 (1993)
2. O. Yamamoto, Electrochim. Acta 45, 2423 (2000)
3. E.D. Wachsman, C.A. Marlowe, K.T. Lee, Energy Environ. Sci. 5,
5498 (2012)
4. J.B. Goodenough, Annu. Rev. Mater. Res. 33, 91 (2003)
5. J.M. Ralph, A.C. Schoeler, M. Krumpelt, J. Mater. Sci. 36, 1161
(2001)
6. E.D. Wachsman, K.T. Lee, Science 334, 935 (2011)
7. J. Jiang, W.D. Shen, J.L. Hertz, Thin Solid Films 522,66(2012)
8. W. Jung, J.L. Hertz, H.L. Tuller, Acta Mater. 57, 1399 (2009)
9. M. Gerstl, T. Fromling, A. Schintlmeister, H. Hutter, J. Fleig, Solid
State Ion. 184,23(2011)
10. A.K. Opitz, J. Fleig, Solid State Ion. 181, 684 (2010)
11. R. Radhakrishnan, A.V. Virkar, S.C. Singhal, J. Electrochem. Soc.
152, A210 (2005)
12. A. Evans, A. Bieberle-Hutter, J.L.M. Rupp, L.J. Gauckler, J. Power.
Sources 194, 119 (2009)
13. C.D. Baertsch, K.F. Jensen, J.L. Hertz, H.L. Tuller, S.T. Vengallatore,
S.M. Spearing, M.A. Schmidt, J. Mater. Res. 19, 2604 (2004)
14. H. Huang, M. Nakamura, P.C. Su, R. Fasching, Y. Saito, F.B. Prinz, J.
Electrochem. Soc. 154, B20 (2007)
15. A.C. Johnson, A. Baclig, D.V. Harburg, B.K. Lai, S. Ramanathan, J.
Power. Sources 195, 1149 (2010)
16. J. Luo, D.P. Almond, R. Stevens, J. Am. Ceram. Soc. 83,1703
(2000)
17. X. Guo, W. Sigle, J. Fleig, J. Maier, Solid State Ion. 154, 555 (2002)
18. J. Garcia-Barriocanal, A. Rivera-Calzada, M. Varela, Z. Sefrioui, E.
Iborra, C. Leon, S.J. Pennycook, J. Santamaria, Science 321,676
(2008)
19. M. Sillassen, P. Eklund, N. Pryds, E. Johnson, U. Helmersson, J.
Bottiger, Adv. Funct. Mater. 20, 2071 (2010)
20. N. Schichtel, C. Korte, D. Hesse, J. Janek, Phys. Chem. Chem. Phys.
11 , 3043 (2009)
21. C. Korte, N. Schichtel, D. Hesse, J. Janek, Monatsh. Chem. 140,
1069 (2009)
22. N. Schichtel, C. Korte, D. Hesse, N. Zakharov, B. Butz, D. Gerthsen,
J. Janek, Phys. Chem. Chem. Phys. 12, 14596 (2010)
23. S. Sanna, V. Esposito, A. Tebano, S. Licoccia, E. Traversa, G.
Balestrino, Small 6, 1863 (2010)
24. S.G. Kim, S.P. Yoon, S.W. Nam, S.H. Hyun, S.A. Hong, J. Power.
Sources 110 , 222 (2002)
25. Y.W. Zhang, S. Jin, Y. Yang, G.B. Li, S.J. Tian, J.T. Jia, C.S. Liao,
C.H. Yan, Appl. Phys. Lett. 77, 3409 (2000)
26. M. Matsuda, T. Hosomi, K. Murata, T. Fukui, M. Miyake, J. Power.
Sources 165, 102 (2007)
27. M. Boulouz, L. Martin, A. Boulouz, A. Boyer, Mater. Sci. Eng. B 67,
122 (1999)
28. T. Hata, K. Sasaki, Y. Ichikawa, K. Sasaki, Vacuum 59, 381 (2000)
29. S. Sanna, V. Esposito, D. Pergolesi, A. Orsini, A. Tebano, S.
Licoccia, G. Balestrino, E. Traversa, Adv. Funct. Mater. 19,1713
(2009)
30. L.C. De Jonghe, C.P. Jacobson, S.J. Visco, Annu. Rev. Mater. Res.
33, 169 (2003)
31. I. Kosacki, C.M. Rouleau, P.F. Becher, J. Bentley, D.H. Lowndes,
Electrochem. Solid-State Lett. 7,A459(2004)
32. I. Kosacki, T. Suzuki, V. Petrovsky, H.U. Anderson, Solid State Ion.
136, 1225 (2000)
33. A. Karthikeyan, C.L. Chang, S. Ramanathan, Appl. Phys. Lett. 89,
183116 (2006)
34. J. Jiang, X.C. Hu, W.D. Shen, C.Y. Ni, J.L. Hertz, Appl. Phys. Lett.
102, 143901 (2013)
35. X. Guo, E. Vasco, S.B. Mi, K. Szot, E. Wachsman, R. Waser, Acta
Mater. 53, 5161 (2005)
36. E. Navickas, M. Gerstl, G. Friedbacher,F. Kubel, J. Fleig,Solid State
Ion. 211,58(2012)
37. E. Navickas, M. Gerstl, F. Kubel, J. Fleig, J. Electrochem. Soc. 159,
B411 (2012)
38. J.H. Joo, G.M. Choi, Solid State Ion. 177, 1053 (2006)
39. A. Rivera, J. Santamaria, C. Leon, Appl. Phys. Lett. 78 ,610
(2001)
J Electroceram (2014) 32:37–46 45
40. M. Gerstl, G. Friedbacher, F. Kubel, H. Hutter, J. Fleig, Phys. Chem.
Chem. Phys. 15, 1097 (2013)
41. I. Kosacki, C.M. Rouleau, P.F. Becher, J. Bentley, D.H. Lowndes,
Solid State Ion. 176, 1319 (2005)
42. M. Sillassen, P. Eklund, N. Pryds, E. Johnson, U. Helmersson, J.
Bottiger, Adv. Funct. Mater. 20, 3194 (2010)
43. A. Karthikeyan, M. Tsuchiya, S. Ramanathan, Solid State Ion. 179,
1234 (2008)
44. M. Aoki, Y.M. Chiang, I. Kosacki, I.J.R. Lee, H. Tuller, Y.P. Liu, J.
Am. Ceram. Soc. 79, 1169 (1996)
45. H.R. Kim, J.C. Kim, K.R. Lee, H.I. Ji, H.W. Lee, J.W. Son, Phys.
Chem. Chem. Phys. 13,6133(2011)
46. N. Sata, K. Eberman, K. Eberl, J. Maier, Nature 408, 946 (2000)
47. X.Guo,J.Maier,J.Electrochem.Soc.148, E121 (2001)
48. A.S. Kao, J. Appl. Phys. 69, 3309 (1991)
49. C.V. Thompson, J. Mater. Res. 14, 3164 (1999)
50. D.J. Quinn, B. Wardle, S.M. Spearing, J. Mater. Res. 23 , 609 (2008)
51. J.L.M. Rupp, Solid State Ion. 207, 1 (2012)
52. A. Kushima, B. Yildiz, J. Mater. Chem. 20, 4809 (2010)
53. I. Yasuda, M. Hishinuma, Electrochem. 68, 526 (2000)
54. Y.S. Touloukian, R.K. Kirby, R.E. Taylor, T.Y.R. Lee, in
Thermophysical Properties of Matter. Thermal Expansion-
Nonmetallic Solids, vol. 13 (IFI/Plenum, New York, 1977), p. 288
55. C. Korte, A. Peters, J. Janek, D. Hesse, N. Zakharov, Phys. Chem.
Chem. Phys. 10, 4623 (2008)
56. H. Aydin, C. Korte, M. Rohnke, J. Janek, Phys. Chem. Chem. Phys.
15, 1944 (2013)
57. H. Aydin, C. Korte, J. Janek, Sci. Tech. Adv. Mater. 14, 035007
(2013)
58. E.T. Park, J.H. Park, (1998). doi:10.2172/656718
46 J Electroceram (2014) 32:37–46