Sie sind auf Seite 1von 18

Solid State Ionics 125 (1999) 285–302

www.elsevier.com / locate / ssi

Aspects of the formation and mobility of protonic charge carriers


and the stability of perovskite-type oxides

K.D. Kreuer
¨ Festkorperforschung
Max-Planck-Institut f ur ¨ , Heisenbergstr. 1, D-70569, Stuttgart, Germany

Abstract

Proton conducting acceptor-doped perovskite-type alkaline earth cerates, zirconates, niobates and titanates have been
investigated experimentally and by numerical simulations. For all cubic perovskites the concentration of protonic defects
almost reaches the acceptor dopant concentration under appropriate conditions, and the mobility of protonic defects fall into
a narrow range. Any symmetry reduction, however, leads to a reduction of the concentration and mobility of protonic
defects. For all oxides, dynamical hydrogen bonding is suggested to lead to a local lattice softening, which provides an
advantageous environment for high proton-mobility. This effect may explain the very high proton conductivity in covalent
acceptor-doped BaZrO 3 , which has been found experimentally for the first time. Since this oxide also shows good
thermodynamic phase stability, it is an interesting candidate as separator material in high-drain electrochemical applications
such as fuel-cells.  1999 Elsevier Science B.V. All rights reserved.

Keywords: Perovskite-type oxides; Protonic charge carriers; BaZrO 3 ; Electrochemical applications

1. Introduction but prevents them from being used in high-drain


applications such as fuel cells.
Although high proton conductivity has been re- This paper is a report on our recent progress in the
ported for many perovskite-type oxides in humid search for perovskite-type oxides with high proton
atmosphere [1] the combination of high proton conductivity and high thermodynamic phase stabili-
conductivity and stability, which is a prerequisite for ty. The research strategy is based on a few relation-
the application of such compounds as a separator ships between chemistry and structure on one side
material in electrochemical cells, is generally consid- and the above addressed properties on the other side,
ered to be a key problem [2]. Oxides with high which have been worked out for some well-known
conductivity (e.g. BaCeO 3 -based compounds) gener- proton conducting perovskite-type oxides. Since the
ally show very low phase stability (e.g. with respect quality of the data is critical to this approach, some
to carbonate or hydroxide formation [2]), while more initial comments on the acquisition of experimental
stable oxides (e.g. In-doped CaZrO 3 ) show very low and simulation data are necessary.
conductivity, which only permits their use in low-
drain applications (e.g. potentiometric sensors [3])
2. Acquisition of experimental and numerical
data
E-mail address: kreuer@chemix.mpi-stuttgart.mpg.de (K.D.
Kreuer) IR-spectroscopy in the diffuse back scattering

0167-2738 / 99 / $ – see front matter  1999 Elsevier Science B.V. All rights reserved.
PII: S0167-2738( 99 )00188-5
286 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

mode [4,5], 1 H-pulsed-field-gradient NMR [4–6] and measuring time, the samples more closely approach
X-ray diffraction including Rietveld refinement [7] equilibrium. The equilibration rate is generally de-
have been described elsewhere. Some critical aspects termined by the chemical diffusion of water (am-
of the sample preparation, thermogravimetry, impe- bipolar diffusion of V ??O and 2 OH ?O ) which is written
dance spectroscopy and MD-simulations are dis- [8]:
cussed in what follows.
(2 2 X)DOH ?O DV ??O
D˜ H 2 O 5 ]]]]]]]
XDOH ?O 1 2(1 2 X)DV ??O
2.1. Samples where the degree of hydration X 5 [OH ?O ] /S and S is
the hydration limit.
In the present work, acceptor-doped perovskites of Fig. 1 shows this diffusion coefficient as a func-
the type ABO 3 (A 5 Ba, Sr; B 5 Ce, Zr, Ti) and the tion of the degree of hydration for two perovskites.
complex perovskite Ba(Ca 0.39 Nb 0.61 )O 32d For completely dehydrated samples, DH 2 O equals the
(Ca:Ba 3 CaNb 2 O 9 ) have been prepared as dense
self diffusion coefficient of protonic defects, which is
ceramics. The preparation procedure described in
high for the compounds under consideration. But for
Ref. [8] has been adopted for each composition.
compounds with low self diffusion coefficients of
Especially for the zirconates, high final sintering
oxygen ion vacancies, this initially decreases rapidly
temperatures ( . 17008C) were required to obtain the
upon hydration (see curve for Ba 2 YSnO 5.5 ) and
desired densities ( . 98% of theoretical) and to
eventually approaches the diffusion coefficient of
minimise grain boundary resistances. Two ‘single-
oxygen ion vacancies. The concentration of protonic
crystalline’ specimens prepared by induction melting
defects in compounds with high oxygen ion con-
[6], Y-doped BaCeO 3 and BaZrO 3 , (Y:BaCeO 3 ,
ductivity (e.g. BaCeO 3 based compositions) therefore
Y:SrCeO 3 ) are also included in this study. The
equilibrates much faster. The self diffusion coeffi-
experimental results are discussed together with
cient of oxygen ion vacancies of the compounds
those obtained for a series of BaCeO 3 ceramics
considered, obtained from the ionic conductivities
doped with 2–20 mol.% Y [4] and for solid solutions
under high vacuum ( p,10 21 Pa), are shown in Fig.
of Ba 2 YSnO 5.5 and BaSnO 3 [9].
2. The approximate maximum grain sizes for
equilibration within 1 day are indicated. While a 10
2.2. Hydration isobars mm powder of Y:BaCeO 3 may be equilibrated within
1 day even at room temperature, for Ba 2 YSnO 5.5 this
Since the dominant reaction leading to the forma- is only possible at 3708C. Accordingly, we have
tion of protonic defects is the dissociative uptake of adjusted the cooling and heating runs in the TGA
water, the concentration of protonic defects can experiments for each compound and grain size
readily be determined by thermogravimetric analysis (typically 10 mm after 2 h ballmilling). The absence
(TGA) where the temperature T and the water-partial of any visible hysteresis provides additional con-
pressure pH 2 O are the experimental parameters. Con- firmation that the thermodynamic equilibria were
ventionally, pH 2 O is varied for a given temperature, achieved.
which yields the equilibrium constant K of the The concentration of protonic defects among
hydration reaction for that temperature (e.g. Ref. titanates does not saturate at low water partial
[6]). Since K(T ) generally shows Arrhenius behav- pressures. Overnight treatments in an autoclave at
iour, i.e. temperature-independent standard hydration T54508C and pH 2 O 53.5 MPa allows one to de-
enthalpies and entropies, the thermodynamic parame- termine the saturation concentration of protonic
ters may also be obtained from the hydration isobars. defects for these compounds.
Recording the weight at a constant water partial-
pressure, pH 2 O 523 hPa, as a function of temperature 2.3. Impedance spectroscopy
has two advantages: (i) the sample may be slowly
hydrated (with decreasing T ) or dehydrated (with The protonic conductivity is commonly deter-
increasing T ) only once, and (ii), for a given mined by impedance spectroscopy which allows one
K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302 287

Fig. 1. The hydration rate is frequently controlled by the chemical diffusion coefficient of water. This is shown as a function of the degree
of hydration X for both an oxide with a high and a low diffusion coefficient of oxygen ion vacancies.

Fig. 2. Oxygen ion vacancy diffusion coefficients as obtained from conductivities under high vacuum ( ptotal ,10 23 Pa). The maximum grain
size for equilibration within 1 day is indicated on the right axis (see text).
288 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

to extract bulk resistances from the complex impe- tion of two protonic defects, the equilibrium con-
dance as a function of frequency. In Cole–Cole dition is written [4]:
representations the bulk response shows up as a high
frequency arc corresponding to a specific capacitance K 5 [OH ?O ] 2 /([V ??O ][O Ox ] pH 2 O ).
of about 1–10 pF cm 21 (´¯10–100). Particularly in
Together with the site restriction (because of high
the presence of high grain-boundary impedances,
defect concentration):
which often respond at only slightly lower fre-
quency, the bulk arc may easily be overlooked either [O xO ] 1 [V ??O ] 1 [OH O? ] 5 3
because of the limited frequency range or because of
poor resolution between bulk and grain boundary and the electroneutrality condition:
responses. As can be seen in the spectra shown in 2[V ??O ] 1 [OH ?O ] 1 [Y Ba
?
9 ]50
] 2 [Y Ce
Fig. 3, for zirconates, titanates and stannates the bulk
arc may be resolved at lower temperatures. Of the concentration of protonic defects as a function of
course, the availability of single crystals makes this pH 2 O and K is written:
problem obsolete (Fig. 3b). The very low impedance
of the Y-doped BaZrO 3 crystal supports the assump- [OH ?O ] 5
]]]]]]]]]]
tion that the high frequency arc in the spectrum of 3KpH 2 O 2 KpH 2 O (9KpH 2 O 2 6KpH 2 O S 1 KpH 2 O S 2 1 24S 2 4S 2 )
]]]]]]]]]]]]]
œ
the polycrystalline Y-doped BaZrO 3 indeed origi- KpH 2 O 2 4
nates from bulk proton conductivity.
where kT ln K 5 TDS8 2 DH8 and S is the effective
2.4. Quantum molecular dynamics simulations dopant concentration (water solubility limit).
For cubic perovskites, which exhibit only one kind
Important tools to attack the problem of the of oxygen site, this expression fits the hydration
underlying proton conduction mechanism are quan- isobars (Fig. 5) quite well. Despite the slight ortho-
tum-MD simulations which have been described rhombic distortion of Y:BaCeO 3 [7] the fit with
elsewhere [10]. But still there is a gap between single standard chemical potential differences for the
typical diffusion rates for protonic defects in oxides different species occupying the two oxygen sites is
(1 /t ¯10 210 –10 29 s 21 at T51000 K) and the time still surprisingly good. Only for Y:SrZrO 3 and
increments accessible by such techniques (a few ps). Y:SrCeO 3 , which deviate highly from an ideal cubic
Since no long range diffusion is observed during perovskite, the hydration isobars cannot be fitted
such simulations, transition state energies may be with this simple expression. In particular, the exist-
estimated by extrapolating Helmholtz potential ence of different oxygen sites seems to reduce the
curves, obtained for example from pair correlation water uptake [7]. Additionally Y:SrCeO 3 slowly
functions, to a point in phase space which is assumed starts to decompose into hydroxides below T¯4008C
to correspond to the transition state. In order to avoid (as confirmed by X-ray diffraction and IR-spectros-
such model assumptions, we have carried out MD- copy). For the cubic cases, however, the good fit
runs at extremely high temperatures (T .2000 K) with single standard chemical potentials indicates the
which produce a number of successful proton jumps absence of discrete associates, typically between
(Fig. 4) and therefore allows the identification of the protonic defects and the negatively charged acceptor
transition state in phase space [5]. As will be pointed dopants.
out below (Section 4.1), this allows one to determine Fig. 5 also shows the thermodynamic parameters
the diffusion paths and some relevant interactions of the hydration reaction for the cubic perovskites
determining the transition state energy. except for the titanates, which do not take up
sufficient water under the experimental conditions to
allow a fit of sufficient significance. The data support
3. Formation of protonic defects the general observation that the hydration enthalpy
for perovskite-type oxides become more negative in
Assuming ideal behaviour of all species involved the order: titanates → niobates → zirconates →
in the hydration reaction, which leads to the forma- stannates → cerates, indicating that basicity is a
K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302 289

Fig. 3. Impedance spectra for representative polycrystalline (a) and single crystal samples (b) (see text).
290 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

Fig. 4. The evolution of the mean square displacement with time for three protons in a box of 33333 unit cells of BaZrO 3 as obtained
from a quantum-MD-simulation [25].

major parameter. This is also supported by the the protonic defect (see also Section 4)) is not yet
evolution of the hydration enthalpy for the solid clear.
solutions Ba(Ce 12xY x )O 32d and Ba 2 Sn 22xY x O 32d Apart from the equilibrium constant K of the
(Fig. 6) where the increasing basicity with increasing hydration reaction, the saturation limit S determines
Y-content is accompanied by a more negative hydra- the concentration of protonic defects. Determinations
tion enthalpy. of S are often found irreproducible and significantly
Since hydration enthalpy and entropy are not lower than the acceptor dopant concentration. The
found to be related, the hydration entropy has to be hydration isobars normalised with respect to the total
considered as a separate parameter determining the dopant level S0 , obtained in this work (Fig. 8),
equilibrium constant of the hydration reaction. This however, show saturation limits well above 80% for
is demonstrated in Fig. 7 which shows the normal- all cubic perovskites in accordance with recent
ised hydration isobars of a 2% Y-doped BaCeO 3 [4] observations [4,11]. Even for 5% Sc-doped SrTiO 3 a
and a 10% Y-doped BaZrO 3 compared to that of a proton defect concentration of 3.7% has been found
10% Y-doped BaCeO 3 . Despite the less negative gravimetrically after hydrothermal treatment. The
hydration enthalpy of BaCeO 3 with the lower dopant remaining reduction of the saturation limit with
level, the dehydration temperature is distinctly high- respect to the dopant concentration may then simply
er, reflecting the considerably less negative hydration be the consequence of a slight substitution of the
entropy. The low hydration entropy of Y:BaZrO 3 is dopant on the A-site, where it acts as a donor [6].
also the reason for the stabilisation of protonic This is obviously a minor effect for ceramics pre-
defects up to quite high temperature despite the pared at relatively low temperatures (1400–17208C).
relatively less negative hydration enthalpy. To what For single crystals prepared well above 20008C,
extent the hydration entropy is determined by con- however, this effect may drastically reduce the water
figurational or vibrational terms (e.g. degrees of solubility [6]. The low saturation limits of the two
freedom related to the local rotational diffusion of orthorhombic perovskites (Y:SrCeO 3 and Y:SrZrO 3 )
K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302 291

Fig. 5. Hydration isobars for different perovskite-type oxides. The decomposition of Sr(Ce 0.9 Y 0.1 )O 32d below |5008C (see text) is indicated
by the dashed line. The experimental data shown as rough curves are fit with the smooth curves from the model detailed in the text.

have already been explained by the decay of the rotational diffusion of the protonic defect, supporting
oxygen site into two sites with different affinities the assumption that proton transfer is the rate
towards hydration [7]. limiting step. In such Grotthuss-type mechanisms the
proton is the only mobile species while the oxygen is
localised in the vicinity of its crystallographic posi-
4. Mobility of protonic defects tion. Nevertheless, proton transfer in oxides is fre-
quently believed to be coupled to the local oxygen
It has been frequently suggested that the mobility dynamics, because of the large separations between
of protonic defects involves proton transfer between nearest neighbour oxygen ions (e.g. 320 pm in
neighbouring oxygen ions and reorientation of the BaCeO 3 [12]) and the strong localisation of the
hydroxide ion on the oxygen site [12]. For a few proton within the valence electron density of the
perovskites (BaCeO 3 - and Ba 3 CaNb 2 O 9 -based) sev- oxygen. (It should, however, be noted, that from
eral independent experiments [13–15] and numerical INS-studies especially Colomban and Filleaux claim
simulations [5,16,17] provide evidence for the the existence of ‘free protons’ [18–20].) In a first
reorientation step being brought about by rapid attempt to model proton transfer [21] thermally
292 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

available experimental results (see Ref. [22]). Then


our recent proton mobility data for simple cubic
perovskites will be discussed followed by the effects
of symmetry breaking.

4.1. Proton conduction mechanism

As opposed to earlier simulations [16,17], our


recent quantum MD-simulations of protons in
BaCeO 3 produced a few successful proton transfer
events for a very high temperature (2500 K) [5].
Although the poor statistics do not allow for a
quantitative determination of the proton diffusion
coefficient (see Fig. 4), details of the diffusion
mechanism are identified by analysing the evolution
of the system with time instead of interpreting pair
correlation functions with respect to a given model
of the diffusion mechanism [5]. This leads to two
surprising observations.

1. Even for configurations with OH–O separations


well below 240 pm, proton transfer occurs with a
very low probability because of a wrong orienta-
tion of the OH. This rather points with its proton
to one of the other seven O-neighbours.
2. With this the OH forms a transient but very
strong bent hydrogen bond reducing the OH–O
Fig. 6. The hydration enthalpy of two perovskite-type oxides as separation from 312 pm (cubic case) to an average
functions of the basic Y-dopant concentration [4,9].
of 281 pm, where the transient OH–O pair
correlation function shows a very broad maxi-
activated compression of the OH–O separation has mum [22].
been assumed to assist the proton transfer event.
Based on this concept, results of quantum molecular From this transient pair correlation function the
dynamics simulations have been interpreted. While Helmholtz energy of the system as a function of the
the obtained activation enthalpy for proton diffusion OH–O separation coordinate is obtained (Fig. 9).
was in good agreement with the experimental results This shows a very shallow minimum reflecting the
for BaCeO 3 (0.41 and 0.46 eV, respectively [5]), the extended fluctuations of the OH–O separation. The
model could not explain the decreasing proton energy the system gains by strong hydrogen bonding,
mobility under pressure and the high activation which is indicated in the OH-stretching regime of the
enthalpy of proton diffusion in SrCeO 3 (0.63 eV). IR-spectra (Fig. 10), tentatively compensates for the
The model suggests extremely low activation en- energy required to deform the CeO 6 -octahedron over
thalpies for perovskites with low lattice constants, a wide range of OH–O separations. This ‘softening’
that are not confirmed experimentally. This concept of the proton environment does not only allow for
and the experimental results are also controversial the extended OH–O fluctuations but also for the easy
with respect to the identification of the rate-limiting breaking of such strong hydrogen bonds. This may
step and the role of hydrogen bonding [5]. explain the high rates of both proton transfer and
In the following a closer view of the MD-simula- rotational diffusion.
tion data will be presented that is consistent with the To give some background to these novel observa-
K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302 293

Fig. 7. Hydration isobars (fitted curves) for three perovskite-type oxides with similar dehydration temperatures but distinctly different
hydration enthalpies (see text).

Fig. 8. Hydration isobars (smoothened data from Fig. 5) normalised with respect to the acceptor dopant concentration. The curves for the
cubic perovskites are bold.

tions, some features of the thermodynamics of proton may be small (65 meV in the case of BaCeO 3
diffusion are identified (Fig. 11). [7]).
2. The proton prefers an OH orientation off of the
1. Even for strong hydrogen bonding the activation octahedral diagonal, probably because of the
energy for rotational diffusion of protonic defects strong H 1 –Ce 41 repulsion. Thus the transient
294 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

Fig. 9. Helmholtz energy of BaCeO 3 as a function of the O–O separation (obtained from the O–O pair correlation function) and the OH–O
separation (obtained from the transient OH–O pair correlation function [5,22]) indicating the ‘softening’ of the proton environment.

hydrogen bonds are not linear and additional should indeed have major contributions from com-
energy is required to bring the hydrogen bonded pressing the separation of proton donor and acceptor
complex close to linear, which favours proton (Fig. 11). The instability of intra-octahedron hydro-
transfer. This is brought about by an elongation of gen bonds is also indicated in the OH-stretching
the B–O bonds [5], which explains the increase of absorption region for the perovskites with the highest
the activation enthalpy under pressure. lattice constants, i.e. the highest structural oxygen
3. Since the protonic defect may form transient separations (Fig. 10). For these a small but distinct
hydrogen bonds with eight oxygen neighbours absorption at an unusually high frequency of ¯4250
and OH orientation and OH–O separation have to cm 21 is observed which may be assigned to non-
match in the transition state, there may be signifi- hydrogen bonded OH in this type of oxides.
cant negative contributions to the proton diffusion
entropy, which becomes apparent in reduced pre- 4.2. Mobility of protonic defects in cubic
exponential factors. perovskites

The second feature actually also explains why the Fig. 12a shows the proton conductivities in the
diffusion path changes for perovskites with reduced temperature ranges where impedance spectroscopy
lattice constants, such as SrTiO 3 or CaTiO 3 [22]. For allows a reliable separation of the bulk response (see
these, proton transfer between the vertices of neigh- Section 2.3). With the concentration of protonic
bouring octahedra is favoured over intra octahedron defects taken from the hydration isobars (Fig. 5)
transfer. The latter is probably suppressed because of proton mobilities have been obtained via the Nernst–
the strong H 1 –Ti 41 repulsion preventing the forma- Einstein relationship (Fig. 12b). Except for Sc-doped
tion of a linear hydrogen bond, which may rather be BaTiO 3 all cubic perovskites show similar proton
formed between oxygens of neighbouring octahedra. mobilities, although, with some variation in the
In this case the activation enthalpy of proton transfer activation enthalpy. A plot of the pre-exponential
K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302 295

Fig. 10. Infrared spectra in the OH-stretching regime for hydrated perovskite-type oxides in the order of decreasing unit cell volume. The
spectra have been obtained in the diffuse back scattering mode which has been found to produce similar spectra as in transmission [4]. The
spectra of the cubic oxides are shown bold and the lattice constants are indicated. Note the absorption around 4250 cm 21 for the perovskites
with large unit cell volume, which may originate from non-hydrogen bonded OH (see text).
296 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

This has important implications for the development


of stable oxides with high proton conductivity (see
Section 5).

4.3. Effects of symmetry reduction

This ‘universal’ behaviour is restricted to cubic


perovskites. As can be seen already from the trans-
port data presented in Fig. 12, the mobilities of
protonic defects in perovskites with structures
strongly deviating from cubic (mostly orthorhombic)
are significantly lower. This effect has recently been
investigated in detail by comparing structural and
dynamical features of protonic defects in Y:BaCeO 3
and Y:SrCeO 3 [7]. The large orthorhombic distortion
of Y:SrCeO 3 has tremendous effects on the arrange-
ment of the lattice oxygen. The cubic oxygen site
degenerates into two sites with probabilities of 1 / 3
and 2 / 3. Due to different chemical interactions with
the cations, especially the strontium on the A-site,
the oxygens on these sites show distinctly different
electron densities (basicities) and therefore different
binding energies for the proton. The different site
binding energies and the biased rotational diffusion
Fig. 11. Schematic illustration of intra- and inter-octahedra proton
transfer. While the activation enthalpy of the first, which is
in this environment is thought to increase the activa-
suggested to occur in perovskites with large lattice constants (e.g. tion energy of proton mobility. Additionally, the
BaZrO 3 ), has major contributions from the elongation of the B–O diffusion paths are restricted to one dimension with
bonds, the rate of the latter, which is proposed to occur in high energy barriers separating the different paths.
perovskites with small lattice constants (e.g. SrTiO 3 , CaTiO 3 ), is But the mobilities of protonic defects may also be
strongly dependent on fluctuations of the separation of oxygens of
adjacent octahedra. Transient hydrogen bonds are indicated by
reduced by a local symmetry reduction, due to the
dashed lines. presence of the acceptor dopant. From the results of
quasi elastic neutron scattering (QNS) experiments it
has been suggested that acceptor dopants may act as
factors versus the activation enthalpies (Fig. 13) discrete trap sites for protons. The QNS spectra have
exhibits a slope of ¯1 / 630 K. Since the order of this been fitted by the rates for proton liberation from
temperature is reminiscent of typical dehydration these trap sites and the mobility of the ‘free’ proton
temperatures of these oxides, the proton mobilities at [23]. Conceptionally, such a two state model leads to
the dehydration temperatures (K5l) for two water the general behaviour shown in Fig. 15a. At low
partial pressures are plotted in Fig. 14. The sort of temperature (strong association limit) the activation
‘universal’ curves with activation enthalpies of about enthalpy is expected to be the sum of association and
0.6 eV may be considered to correspond to the migration enthalpy and the absolute value of the
maximum conductivity achievable for a given water proton mobility is expected to be inversely propor-
partial pressure. Since the activation enthalpy of tional to the concentration of trap sites. A recent
these curves only slightly differs from the activation study of the transport properties of BaCeO 3 with
enthalpy of the proton mobilities in these oxides different levels of Y-dopant indeed showed a drastic
(0.42–0.58 eV), there is obviously no significant reduction of the proton mobility with the acceptor
correlation between the chemical potential (activity) dopant concentration [4], but this was not the result
of water in the perovskite and the proton mobility. of a reduction of the pre-exponential factor as
K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302 297

Fig. 12. Proton conductivities (a) and diffusivities of protonic defects (b) of different perovskite-type oxides. The curves of the cubic
perovskites are shown bold.
298 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

Fig. 13. Pre-exponential factor versus activation enthalpy for the diffusivities of protonic defects among various perovskite-type oxides.

Fig. 14. The diffusivities of protonic defects at the dehydration temperature (K51) of different cubic perovskite-type oxides for two
different water partial pressures.
K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302 299

Fig. 15. (a) The diffusivity of protonic defects for different acceptor dopant (trap) concentration, as anticipated by the ‘two-state-model’
[23]. (b) Experimentally determined proton diffusivities for BaCeO 3 doped with different acceptor dopant concentrations [4].
300 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

anticipated by the two-state model. The proton tion temperature is, however, somewhat unexpected
mobility was instead reduced by an increase of the considering the relatively less negative hydration
activation enthalpy with the pre-exponential being enthalpy of 276 kJ mol 21 , which is in accordance
almost unchanged (Fig. 15b). This behaviour sug- with the amphoteric character of the zirconate.
gests a more delocalised effect of the acceptor Correspondingly, the hydration entropy is also found
dopant on the proton mobility. In this particular case, to be significantly less negative than that for other
the introduction of Y apparently leads to an increase perovskites. This behaviour may qualitatively be
of the basicity of all oxygens in the lattice and explained by the directional character of the Zr–O
therefore to a stronger binding of the protons to the bond and the strength of the hydrogen bond inter-
oxygens in general. Nevertheless, not only a good action, which are both estimated by a quantum
size matching but also acid / base matching is re- chemical calculation [5]. Compared to BaCeO 3 , both
quired in the choice of the acceptor dopant with interactions are stronger but compensate over a wider
respect to an optimisation of proton mobility. range of oxygen separations, which leads to a more
The relatively low mobility of protonic defects in effective local ‘softening’ of the lattice in the proton
Sc:BaTiO 3 (Fig. 12) may be due to the symmetry environment. (Note, that the average contraction of
reduction of the proton environment as a conse- the oxygen sublattice around the protonic defect is,
quence of the high dielectric polarisation leading to however, smaller than in BaCeO 3 [5].) This may not
an increase of the proton self-localisation [12]. But only explain the surprisingly low activation enthalpy
since the data have been obtained with metastable of proton mobility but also the high entropy of the
cubic BaTiO 3 ceramics of very low grain size, protonic defect that leads to the relatively less
transport measurements on more defined materials negative formation entropy. The covalency of the
with high dielectric constants are planned. Zr–O bond also corresponds to a less positive charge
on the Zr compared to Ce in more ionic BaCeO 3 ,
which reduces the repulsion between Zr and the
5. Stable oxides with high proton conductivity proton, and therefore also the activation enthalpy of
proton mobility (see also Fig. 11).
Empirically high thermodynamic stability and high Fig. 16 shows the H 2 O pressure of dehydration
proton conductivity still seem to exclude one (K51 for the formation reaction) and the CO 2 partial
another. Cerates, for example, show high proton pressure at which decomposition into ZrO 2 and
conductivity for a given water partial pressure up to BaCO 3 occurs. For applications in air ( pCO 2 538 Pa)
quite high temperatures, but below |7008C they and a maximum humidification of pH 2 O 5100 hPa,
react even with air to form carbonates. On the other these data indicate phase stability and stability of
hand, rather stable zirconates are reported to either protonic defects in the temperature range 330 to
dissolve very little water or show high activation 4708C. At the dehydration temperature the bulk
enthalpies for proton conductivity. A first attempt to proton conductivity clearly exceeds a value of 10 22
understand the available data, which are summarised S cm 21 , which makes the material an interesting
in Ref. [1], is based on a simple ionic model for the candidate for high drain applications. But this re-
estimation of the perovskite stability and acid / base quires a reduction of the grain boundary resistance,
considerations to estimate the energetic stabilisation which is comparable to the bulk resistance at 4708C
of protonic defects [2]. But especially the novel data for our samples. On the other hand optimisation of
on Y:BaZrO 3 presented in this paper, may be better the acid / base matching and concentration of the
understood by including effects of the oxide lattice acceptor dopant still leaves some room for a further
covalency and entropic stabilisation effects. increase of the proton conductivity.
Since Y:BaZrO 3 shows an ideal cubic perovskite On the background of these surprising findings, it
structure (as opposed to Y:SrZrO 3 ), it is not surpris- also appears attractive to look for other covalent
ing that the concentration of protonic defects ap- perovskites with strong hydrogen bonding of
proaches that of the acceptor dopant under appro- protonic defects. Apart from further optimising more
priate conditions (Fig. 8). The rather high dehydra- ionic complex perovskites, as described in Ref.
K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302 301

Fig. 16. The water partial pressure of dehydration (K51) and the partial pressure of carbon dioxide, where decomposition into carbonates
occurs [26], as a function of temperature for BaZrO 3 . The potential application range is indicated (see text).

[2,24] this should be another fruitful strategy in the 2. Apart from S0 , the standard hydration enthalpy
development of stable oxides with high proton DH8 and entropy DS8 determine the concentration
conductivity. of protonic defects. In particular, the entropic
stabilisation of protonic defects is not found to be
related to the enthalpic stabilisation, i.e. perov-
6. Conclusions skite-type oxides with less negative hydration
enthalpy may still exhibit a high concentration of
Experimental and numerical investigations re- protonic defects.
vealed a few new aspects of the formation and 3. The mobilities of protonic defects at the dehydra-
mobility of protonic defects in perovskite-type ox- tion temperatures of cubic perovskites are empiri-
ides. cally found to fall close to universal curves for
given water partial pressures.
1. For cubic perovskites, the concentration of 4. Strong, dynamical hydrogen bonding of the
protonic defects approaches the acceptor dopant protonic defect with its eight nearest or four next
concentration under appropriate conditions. Sym- nearest neighbour oxygen atoms are found. This
metry reduction leads to a reduction of the results in transient reductions of the separation
saturation limit of protonic defects S0 . between corresponding oxygens and a local sof-
302 K.D. Kreuer / Solid State Ionics 125 (1999) 285 – 302

tening of the proton environment. High proton ¨


[8] K.D. Kreuer, E. Schonherr, J. Maier, Solid State Ionics
70–71 (1994) 278.
mobility as a typical phenomenon between the
[9] K. Eberman, P. Murugaraj, K.D. Kreuer, J. Maier, in
solid and the liquid state [27] is favoured in this preparation.
environment. ¨
[10] W. Munch, G. Seifert, K.D. Kreuer, J. Maier, Solid State
5. The formation of fast proton conduction paths by Ionics 97 (1997) 39.
this local lattice softening and the entropic stabili- [11] F. Krug, T. Schober, Solid State Ionics 92 (1996) 297.
[12] K.D. Kreuer, Chem. Mater. 8 (1996) 610.
sation of protonic defects allows even hard,
[13] M. Pionke, T. Mono, W. Schweika, T. Springer, H. Schober,
covalent, stable, less basic perovskite-type oxides, Solid State Ionics 97 (1997) 497.
such as BaZrO 3 -based materials, to show high [14] Th. Matzke, U. Stimming, Ch. Karmonik, M. Soetramo, R.
proton conductivity. ¨
Hempelmann, F. Guthoff, Solid State Ionics 86–88 (1996)
621.
¨
[15] R. Hempelmann, M. Soetratmo, O. Hartmann, R. Wappling,
Solid State Ionics 107 (1998) 269.
Acknowledgements ¨
[16] W. Munch, G. Seifert, K.D. Kreuer, J. Maier, Solid State
Ionics 86–88 (1996) 647.
The author wishes to thank Yu. Baikov (St. [17] F. Shimojo, K. Hashino, H. Okazaki, J. Phys. Soc. Jpn. 66
Petersburg) for supplying a Y:BaZrO 3 single crystal, (1997) 8.
[18] Ph. Colomban, J. Tomkinson, Solid State Ionics 97 (1997)
A. Fuchs and P. Senk for technical assistance, W.
123.
¨
Munch and J. Maier for discussion and K. Eberman [19] G.J. Kearley, F. Fillaux, M.H. Baron, J. Tomkinson, S.M.
for reading the proofs. Bennington, Science 264 (1994) 1285.
[20] F. Fillaux, N. Leygue, R. Baddour-Hadjean, S. Parker, Ph.
Colomban, A. Gruger, Chem. Phys. 216 (1997) 281.
[21] K.D. Kreuer, A. Fuchs, J. Maier, Solid State Ionics 77
References
(1995) 157.
¨
[22] W. Munch, K.D. Kreuer, J. Maier, this issue.
[1] T. Norby, Y. Larring, Curr. Opin. Solid State Mater. Sci. 2 [23] R. Hempelmann, Ch. Karmoniak, Th. Matzke, M. Cap-
(1997) 593. padonia, U. Stimming, T. Springer, M.A. Adams, Solid State
[2] K.D. Kreuer, Solid State Ionics 97 (1997) 1. Ionics 77 (1995) 152.
[3] T. Yashiina, K. Koide, N. Fukatsu, T. Ohashi, H. Iwahara, [24] K.C. Liang, Y. Du, A.S. Nowick, Solid State Ionics 69
Sensors Actuators B 13–14 (1993) 697. (1994) 117.
¨
[4] K.D. Kreuer, W. Munch, M. Ise, T. He, A. Fuchs, U. Traub, ¨
[25] W. Munch, K.D. Kreuer, J. Maier, in preparation.
J. Maier, Ber. Bunsenges. Phys. Chem. 101 (1997) 1344. [26] I. Barn, Thermochemical Data of Pure Substances, VCH,
¨
[5] K.D. Kreuer, W. Munch, U. Traub, J. Maier, Ber. Bunsenges. 1989.
Phys. Chem. 102 (1998) 552. [27] K.D. Kreuer, Solid State Ionics 94 (1977) 55.
[6] K.D. Kreuer, Th. Dippel, Yu.M. Baikov, J. Maier, Solid State
Ionics 86–88 (1996) 613.
¨
[7] W. Munch, K.D. Kreuer, St. Adams, G. Seifert, J. Maier,
Phase Transitions, in press.

Das könnte Ihnen auch gefallen