Electrochimica Acta 48 (2003) 1817
/1828

www.elsevier.com/locate/electacta

Electron-hole transport in (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d electrolyte:

effects of ceramic microstructure
V.V. Kharton a,b,*, A.L. Shaula a, N.P. Vyshatko a, F.M.B. Marques a
a
Department of Ceramics and Glass Engineering, UIMC, University of Aveiro, 3810-193 Aveiro, Portugal
b
Institute of Physicochemical Problems, Belarus State University, 14 Leningradskaya Str., 220080 Minsk, Belarus

Received 10 January 2003; received in revised form 19 February 2003; accepted 25 February 2003

Abstract

The oxygen ion transference numbers of a series of (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d (LSGM) ceramics with different
microstructures, prepared by sintering at 1673 K for 0.5
/120 h, were determined at 973
/1223 K by a modified Faradaic efficiency
technique, taking electrode polarization into account. In air, the transference numbers vary in the range 0.984
/0.998, decreasing
when temperature or oxygen partial pressure increases. Longer sintering times lead to grain growth and to the dissolution of Sr-rich
secondary phases and magnesium oxide, present in trace amounts at the grain boundaries, into the major perovskite phase. This is
accompanied with a slight decrease of the total grain-interior resistivity and thermal expansion, while the boundary resistance
evaluated from impedance spectroscopy data decreases 3
/7 times. The electron-hole transport in LSGM ceramics was found to
decrease when the sintering time increases from 0.5 to 40 h, probably indicating a considerable contribution of acceptor-enriched
boundaries in the hole conduction. Due to reducing boundary area in single-phase materials, further sintering leads to higher p-type
conductivity. The results show that, as for ionic conductivity, electronic transport in solid electrolytes significantly depends on
ceramic microstructure.
# 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Lanthanum gallate; Solid electrolyte; Transference number; Ceramic microstructure; Electron-hole conductivity

1. Introduction domains, providing a significantly better stability in

reducing environments with respect to ceria-based solid
The impetus to find materials with greater oxygen-ion electrolytes, and also moderate thermal expansion
conductivity than yttria-stabilized zirconia, the current coefficients [2,4,10
/12]. In oxidizing conditions LSGM
solid oxide fuel cell (SOFC) electrolyte of choice, ceramics exhibit an almost pure oxygen ionic conduc-
resulted in recent research devoted to a class of tion; the ion transference numbers in air are higher than
perovskite phases based on lanthanum gallate, LaGaO3 0.98. In order to compensate gallium oxide volatilization
[1
/12]. High ionic conduction is achieved by doping the in the course of sintering at high temperatures, the use of
perovskite (ABO3) with lower valence cations on the A an A/B cation concentration ratio lower than 1.0 may be
and/or B sites, including the substitution of La with useful. Significant A-site cation deficiency leads, how-
alkaline-earth cations (Sr, Ca, Ba) and Ga with bivalent ever, to a minor decrease of the ionic conduction and to
metal cations such as Mg. One of the highest oxygen-ion a considerably higher n-type electronic transport in
conductivities of any material occurs for the solid reducing atmospheres [6,8].
solution series La1x Srx Ga1y Mgy O3d (LSGM), Solid-state synthesis and sintering of LSGM ceramics
where x /0.10
/0.20 and y /0.15
/0.20 [3
/6]. Other occur via complex mechanisms, which are often asso-
advantages of LSGM include relatively large electrolytic ciated with formation of intermediate phases and
strongly depend on temperature and cation composition
* Corresponding author. Tel.: /351-234-370-263; fax: /351-234-
[6,8,13]. As a rule, impurity phase segregation has a
425-300. deteriorating effect on the ionic conductivity. Although
E-mail address: [email protected] (V.V. Kharton). relationships between ceramic microstructure and elec-
0013-4686/03/$ - see front matter # 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0013-4686(03)00247-0
1818 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828

tronic transport are still unknown, reported data [6]

suggests that the segregation of secondary phases in A-
site deficient (La0.9Sr0.1)x Ga0.8Mg0.2O3d correlates
with increasing n-type conduction in reducing condi-
tions. The difference in p-type conductivity values,
reported in the literature [10,11,14], may also result
from microstructural effects. In particular, the electron-
Fig. 1. Equivalent circuit describing transport processes in a mixed
hole conduction is expected to substantially depend on
ionic-electronic conductor.
ceramic grain size, which is typical for oxide semicon-
ductors [15].
through the sample by an applied voltage:
This work continues our studies of ionic and electro-
nic transport in LaGaO3-based materials [14,16
/19] and tobs (1)
o Ioe =I
is focused on the evaluation of p-type conductivity of
(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics as a function of If the electrode polarization is negligible and electric
microstructure. The electron-hole conductivity was fields are moderate, this ratio is equal to the transference
calculated using transference numbers, determined un- number (to) expressed in terms of the partial conductiv-
der zero oxygen chemical potential gradient in air by the ities of the material, including both grain and boundary
modified Faradaic efficiency technique, taking electrode contributions:
polarization into account [20]. When combined with
electrodes having a high polarization resistance, this so s
method considerably enhances the accuracy of the to 1te   o (2)
so  se s
determination of even minor electronic contributions
to the total conductivity of solid electrolytes due to where te is the electron transference number, and s , so
increasing detected signal, namely the difference be- and se are the total, oxygen ionic and electronic
tween the total and ionic currents through an electro- conductivities, respectively. In terms of the elements of
chemical cell [20]. equivalent circuit (Fig. 1), Eq. (2) can be written as

Re
2. Basic relations to  (3)
Re  Rbulk
o  Rboundary
o

2.1. Faradaic efficiency measurements where Re, Rbulk

o and Rboundary
o are the electronic, grain-
bulk ionic and boundary ionic resistances, respectively.
As shown by numerous experimental and theoretical Notice that the quantities se, te and Re relate to the sum
investigations [20
/24], electrochemical cells consisting of p- and n-type electronic transport.
of a solid electrolyte or mixed conductor with two When the sample is under an oxygen chemical
electrodes can be described by equivalent circuits potential difference, the oxygen flux is determined by
comprising two parallel branches, corresponding to the both electric and chemical potential gradients; the flux
ionic and electronic transport. The latter branch is components driven by electric and chemical forces may
usually represented as a simple resistor; the former be determined in consecutive measurements of oxygen
should include at least two resistance-capacitance (RC) permeation (OP) and Faradaic efficiency (FE). In this
elements, connected in series and related to the bulk and case, Eqs. (1)
/(3) give a transference number averaged
electrode processes. Though the electrode signal is in the range of oxygen chemical potential (m) between
typically more complicated than (RC) and requires electrodes:
additional elements, such as Warburg elements, for
m2
adequate description, this has no effect on the total 1
DC resistance of the cell and on the high-frequency
intercepts of impedance spectra on the real axis. When
to 
m2  m1
×
gt
m1
o dm (4)

the grain-boundary contribution is negligible, the high-

frequency intercept corresponds to the bulk resistance of However, if the electrode polarization is significant,
the material. If the boundary resistance is significant, the apparent transference numbers (tobso ) calculated by
one additional RC element should be added (Fig. 1) Eq. (1) differ from the true to values, leading to errors
according to the so-called brick-layer model, which is, when adopting the classical Faradaic efficiency techni-
again, simplified but adequate in most cases [22,25,26]. que approach [20,24]. These errors can be suppressed by
For classical Faradaic efficiency measurements, the a modified technique [20] based on a combination of
oxygen ion transference number (tobs o ) is defined as the Faradaic efficiency measurements and impedance spec-
ratio of the oxygen ionic (Ioe) to total (I) current driven troscopy (IS):
 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828 1819

Rtotal  I Rtotal case of a significant grain-boundary effect, the bulk and

to 1 (1tobs
o )1 (1tobs
o ) (5)
U R dc boundary transference numbers can be estimated using
FE and IS data. The intermediate- and high-frequency
where U is the voltage applied to the cell, and the total intercepts of impedance spectra on the real axis can be
(grain bulk/boundary) material resistance used to determine total and bulk material resistances,
Re  (Rbulk  Rboundary ) respectively (Rtotal and Rbulk). The former quantity is
o o
Rtotal  boundary (6) defined by Eq. (6); the latter is
Re  Rbulk
o  Ro
Re × Rbulk
can be determined using IS. The dc resistance, Rdc /U / Rbulk  o
(11)
I , can be calculated from the results of Faradaic Re  Rbulk
o

efficiency tests. In this manner it is possible to take Combining Eqs. (3), (6), (10) and (11), one may
electrode polarization into account [20], but the total obtain for the bulk and boundary transference numbers:
transference numbers are affected by both grain-bulk
and boundary resistivities to ionic transport. Rbulk
tbulk
o 1 × (1to ) (12)
Rtotal
2.2. Transference numbers as function of the oxygen Re 1
partial pressure tboundary
o   (13)
Re  Rboundary
o 1  [tbulk
o ]
1
 t1
o

Within the electrolytic domain, the ionic transport in where the total transference number, to, can be derived
solid electrolytes such as LSGM is p(O2)-independent; from FE data using Eq. (5). It is easily seen that Eq. (12)
the oxygen partial pressure dependencies of p- and n- is quite similar to Eq. (5), due to the similar nature of the
type electronic conduction can be described by power corrections for electrode polarization and for grain-
functions, at least in a limited range of oxygen chemical boundary resistance. In addition, the bulk transference
potential [6,10,11]. In oxidizing atmospheres, the p-type number can be directly calculated from FE and IS
conductivity (sp) in LaGaO3-based electrolytes is pre- results:
dominant if compared to the n-type conduction [10,11]: Rbulk
tbulk 1 (1tobs
o ) (14)
se sn sp :sp (7) o
Rdc
The ion transference numbers under isothermal equili- Finally, if the transport model expressed by equivalent
brium conditions (Eq. (2)) may hence be expressed as circuit in Fig. 1 is valid, the relationship between the
so total, grain-bulk and boundary transference numbers is
to  (8)
so  sp0 × [p(O2 )]1=m 1 1 1
  1 (15)
where sp0 is the electron-hole conductivity at unit to tbulk
o tboundary
o

oxygen pressure, and (1/m ) is a constant typically

varying from 1/4 to 1/6. Substitution of Eq. (8) into
Eq. (4) and subsequent integration result in
1=m  
k  p2 1 p2 1 3. Experimental
to (p2 ; p1 )mln 1 1=m  ln (9)
k 1  p1 1 p1
Dense ceramics of (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d
where k1 /so/sp0, p1 and p2 are the oxygen partial (LSGM) were prepared from a commercially available
pressures at the electrodes. This model was used for the powder (PSC, Seattle). Disk-shaped samples pressed at
analysis of p(O2) dependencies of the ion transference 280
/300 MPa were pre-fired in air at 1373 K for 4 h and
numbers of (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d solid electro- then slowly heated (1 K min1) up to 1673 K. The
lyte. sintering time at 1673 K was 0.5, 4, 40 and 120 h. These
conditions were selected on the basis of preliminary
2.3. Grain boundary effect on the transference numbers experiments, in order to obtain maximum density and to
minimize formation of secondary phases. After sinter-
Analysis of the equivalent circuit (Fig. 1) shows that ing, the materials were annealed in air at 1273 K for 3
/4
the bulk ion transference numbers (tbulk o ) can be h, slowly cooled and then polished in order to eliminate
determined as Ga-depleted surface layers. The thickness of layers
Re removed from each surface was no less than 1 mm.
tbulk
o  (10) The density of LSGM ceramics was 95
/97% of theore-
Re  Rbulk
o
tical (Table 1). X-ray diffraction data (XRD) were
When the boundary resistance is negligible, tbulk
o /to. In collected at room temperature using a Rigaku D/Max-
1820 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828

Table 1
Properties of (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics

Sintering time at 1673 K (h) Space group Unit cell volume (Å3) Relative density (%) Activation energy for total grain-bulk and boundary
conductivity (kJ mol 1) (623
/973 K)

Boundary Bulk

0.5 I 2/a 239.7(6) 95.1 1299/7
/

4.0 239.8(7) 94.9 1159/1 999/8
40.0 239.7(3) 96.0 1129/16 959/12
120 239.8(4) 96.8 1119/4 999/5

B diffractometer (Cu
/Ka radiation, 10B/2U B/1108,

step 0.028, 5 s/step); structural parameters were refined
using FULLPROF program [27]. Data [7] on the crystal
lattice of La0.9Sr0.1Ga0.8Mg0.2O3d obtained using neu-
tron diffraction (space group I2/a) was used as structure
model. One example of the final Rietveld plot, showing
the quality of the structural refinement, is presented in
Fig. 2. The overall cation composition was verified by
inductively-coupled plasma (ICP) spectroscopic analy-
sis; the composition of selected zones in LSGM ceramics
was studied using energy dispersive spectroscopy (EDS)
coupled with transmission electron microscopy (TEM).
General characterization of the materials included also
scanning electron microscopy (SEM), dilatometry and
impedance spectroscopy; the equipment and experimen-
tal procedures were described earlier ([14
/20] and
references cited).
The oxygen ion transference numbers were deter-
mined using the modified Faradaic efficiency method,
Fig. 3. Schematic drawing of electrochemical cells used for FE
proposed in Ref. [20]. The measurements were per- measurements: (1) dense LSGM membrane with porous Pt electrodes;
formed using electrochemical cells (Fig. 3) made of 8% (2) glass-ceramic sealant; (3) YSZ solid electrolyte; (4) oxygen sensor;
yttria-stabilized zirconia (YSZ) solid electrolyte. The (5) oxygen pump.
cells consist of several hermetically-sealed ceramic parts
and comprise a LSGM membrane with porous Pt YSZ ceramics was 3
/4 mm. Platinum wires (0.2 and 0.3
electrodes, an oxygen pump and a sensor, separated by mm in diameter) were used as current collectors. Each
insulating layers of glass-ceramic sealants. Thickness of part was fabricated independently and then sealed into a
single cell. This construction was chosen in order to
prevent the voltage applied to the electrodes of the
oxygen pump or LSGM membrane affecting the sensor
emf reading; this was checked after assembling the cells
and also in the course of experiments. At temperatures
below 1300 K, electric resistance between the parts was
higher than 108 V. In addition to gas-tightness control
after the cell fabrication, the cells were also checked for
gas leakages by sealing another YSZ oxygen pump onto
the cells and simulating known oxygen-permeation
fluxes. The total oxygen-leakage currents did not exceed
1
/3 mA, which is less than 0.5% of the measured oxygen
fluxes pumped through LSGM samples. If compared to
the electrolytic oxygen permeation caused by electronic
transport in LSGM ceramics placed under an oxygen
Fig. 2. Observed, calculated and difference XRD pattern of chemical potential gradient, the oxygen-leakage currents
(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics sintered for 120 h at 1673 K. were however significant, up to 20%. Therefore, no
 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828 1821

interpretation of the permeation data was attempted in current-independent within the limits of experimental
this work; the oxygen fluxes pumped through a sample error.
were corrected for the total oxygen-permeation flux The results on transference numbers were verified
under given conditions, as described below. Moreover, studying several LSGM samples prepared in the same
most results presented in this work correspond to zero conditions, and also using alternative measurement
oxygen-pressure gradient, when the role of oxygen techniques described earlier [20,24]. The alternative
leakage, either physical or electrolytic, can be neglected. methods, all taking electrode polarization into account,
FE measurements were carried out at 973
/1223 K included the modified emf technique (EMF), and the
under zero oxygen chemical potential gradient in air, combinations of emf measurements with Faradaic
and also isothermally as a function of the oxygen partial efficiency (EMF-FE) or closed-circuit (EMF-CC) tests.
pressure inside the cell (p1). In the latter case, the oxygen Selected examples showing the reproducibility of the
pressure at the LSGM sample external electrode (p2) was results presented in this paper are given in Fig. 4. In all
21 kPa (atmospheric air); the values of p1 calculated cases the reproducibility error was significantly lower
from the sensor emf by the Nernst equation, varied from than the transference number variations discussed
0.6 to 215 kPa. All results presented below are averaged below.
from 2 to 7 data points and correspond to steady-state
conditions. The bulk resistance (Rbulk) was determined
by impedance spectroscopy before and after each single
4. Results and discussion
test.
For steady-state FE measurements, the total oxygen
flux through the sample, driven by electric and chemical 4.1. Materials characterization
forces (Ioe/Iop), is equal to the flux through the pump
removing oxygen from or pumping oxygen into the cell: XRD studies showed that only LSGM ceramics
sintered at 1673 K for 120 h was free of minor impurity
Ioe Iop Ipump (16)
phases, within the sensitivity limits of X-ray diffraction
The experiments were performed in galvanostatic mode; analysis. Traces of secondary phases were observed in
the currents through sample (Itotal) and pump (Ipump) other materials, though their amount was quite small:
were adjusted in order to achieve a given p1 value. The the intensities of impurity reflections in the XRD
permeation flux, corresponding to this p1, was deter- patterns were less than 3% of the strongest perovskite
mined prior to FE measurements by adjusting Ipump at peak (Fig. 5).
I /0 to obtain the necessary sensor emf, independent of For the material sintered during 0.5 h, the secondary
time: phase was identified as LaSrGaO4 [28]. In this case
inhomogeneity, probably due to incomplete solid state
Iop Ipump jU0 (17)
reaction, is visible even in SEM micrographs (Fig. 6A).
The values of I, U and Ipump were used to calculate This leads to a significant deviation of the perovskite
transference numbers by Eqs. (1), (5), (16) and (17). For
the FE measurements under zero oxygen chemical
potential gradient, the currents were adjusted to achieve
a sensor emf equal to zero, when
Ioe Ipump jp1p2 (18)
The values of voltage applied to LSGM samples in the
course of FE measurements, varied from 0 to 300 mV;
the currents through pump and sample were in the range
0
/15 mA. The sum of cathodic and anodic over-
potentials, evaluated using impedance spectroscopy
results, was lower than 20 mV; major part of total
voltage drop corresponded to ohmic losses on current
collectors, made of relatively thin Pt wires in order to
ensure an absence of gas diffusion limitations at the
electrodes. In these conditions, the influence of applied
voltage on electronic conductivity of LSGM is expected
negligible. This was checked by the measurements of ion
transference numbers as function of the current; 2
/7 Fig. 4. Examples of the experimental results showing reproducibility
data points were measured for each to value. In the of the oxygen ion transference numbers of LSGM, measured by
studied voltage range, the transference numbers were different methods.
1822 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828

Fig. 5. Fragments of XRD spectra of LSGM, normalized to the

intensity of strongest perovskite peaks (Imax). Horizontal lines show
level of the phase impurity peaks.

unit cell parameters, b and c , from the values observed

for other LSGM ceramics (Fig. 7). Nevertheless, the
Fig. 7. Unit cell parameters (a, b and c), average thermal expansion
increase in b parameter is compensated by lower c coefficient (a) and phase impurity content, estimated from XRD data,
value, and the perovskite unit cell volume (Table 1) is as functions of sintering time.
essentially independent of sintering time.
A longer sintering time leads to grain growth (Fig. 6) unambiguous identification of the impurity peaks was
and to dissolution of minor secondary phases in the difficult due to very low intensity, the additional
perovskite lattice (Fig. 5). As a particular result, another reflections in XRD patterns could be attributed to a
dominant impurity phase was detected in LSGM solid solution based on Sr3Ga2O6 (JCPDS card 32-
sintered for 4 and 40 h. Although in these two cases 1230). The secondary phases completely disappear after

Fig. 6. SEM micrographs of (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics sintered at 1673 K: (A) 0.5 h, (B) 4.0 h, (C) 40 h, (D) 120 h. White rectangles in
(A) show segregated impurity phase.
 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828 1823

the heat treatment of 120 h (Fig. 5). One should also tendency for oxide solid-electrolyte ceramics [15,29],
note that, when sintering time increases from 0.5 to 120 being distributed in the boundary layers without crystal-
h, the average grain size increases from 0.5
/3 to 6
/10 lization as a separate phase. In the case of materials
mm. Therefore, the variations of transport properties sintered for long periods, Ga-depletion from the bound-
discussed in this paper, may result from two factors, aries results, obviously, from gallium oxide volatiliza-
namely the presence of secondary phases and decreasing tion. At the same time, the changes in total cation
grain-boundary area. Another necessary comment is composition due to prolonged heat treatments were
that the formation of impurity phases, such as LaSr- rather minor, comparable to the error of ICP spectro-
GaO4 and Sr3Ga2O6, in LSGM ceramics is well known scopic analysis (9/0.5 at.%). For comparison, while
in the literature [6,8,13]; their type and amount are both nominal LSGM composition is La0.88Sr0.10-
dependent on the sintering conditions. For example, Ga0.80Mg0.20O3d, the analytical compositions normal-
increasing sintering temperature was reported to de- ized to B-site cation concentration were La0.89Sr0.10-
crease the LaSrGaO4 fraction in favor of the LaSr- Ga0.79Mg0.21O3d and La0.90Sr0.11Ga0.79Mg0.21O3d for
Ga3O7 formation [8]. the ceramics sintered for 0.5 and 120 h, respectively.
TEM/EDS inspection showed that the grain bound- This suggests that the initial LSGM powder produced
aries in (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d sintered for a by PSC was depleted with gallium in the course of
shorter time are enriched with Sr and Mg, whilst combustion synthesis, and that sintering leads to further
increasing sintering time leads to Ga depletion from Ga-depletion. However, since these changes are close to
the boundaries. As an example, Fig. 8 compares EDS the level of experimental uncertainty, one may expect
spectra of grains and boundaries of LSGM ceramics that the observed variations of transport properties
sintered for 0.5 and 120 h. The boundary enrichment result from the microstructural effects, including the
with strontium is in excellent agreement with XRD data changes in grain-interior and boundary composition,
showing traces of secondary phases, where the Sr rather than from the total cation composition varia-
concentration is higher than in the perovskite. The tions.
segregation of magnesium oxide reflects a common Notice that the impurity phase dissolution and the
corresponding changes in the composition of the
perovskite phase influence the thermal expansion of
LSGM ceramics. The average thermal expansion coeffi-
cients (TECs), calculated from dilatometric data (Fig.
9), are given in Fig. 7A. Increasing sintering time from
0.5 to 120 h leads to decreasing TEC values from 11.4 /
106 to 10.4 /106 K 1, which is considerably higher
than TEC error (9/0.1 /106 K 1). This indicates, in
particular, that the scatter in literature data on thermal
expansion of LaGaO3-based materials [2,4,6,12,14] may
also result from microstructural effects.

Fig. 8. EDS spectra of grain interior and boundaries of LSGM Fig. 9. Dilatometric curves of (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics
sintered for 0.5 and 120 h. in air.
1824 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828

4.2. Total conductivity

The values of total conductivity of

(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics are quite close
to those of La0.9Sr0.1Ga0.8Mg0.2O3d electrolyte
[5,6,10,11] and increase with sintering time (Fig. 10).
Such a behavior suggests a significant role of the grain-
boundary resistivity, as a result of the secondary phase
segregation, magnesium enrichment of the boundaries,
and a large boundary area in the case of short sintering
times. Indeed, at temperatures below 1073 K the
impedance spectra consist of two or three arcs, proving
the presence of a grain-boundary effect at intermediate
frequencies (Fig. 11). The capacitances associated with
these arcs make it possible to relate the high- and
intermediate-frequency intercepts with the bulk and
total resistivities, respectively.
The values of total bulk and boundary conductivities
estimated assuming a brick-layer model behavior are
presented in Fig. 12. Increasing sintering time leads to a
slightly higher bulk conductivity due to progressive
dissolution of Sr-rich secondary phases and magnesium, Fig. 11. Examples of the impedance spectra of
(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics.
segregated at the grain boundaries, in the perovskite
phase. A similar but more pronounced behavior is
observed for the boundary conductivity. When compar-
ing LSGM ceramics sintered for 0.5 and 120 h, the
difference in sboundary values is 4
/7 times larger than
that for sbulk. Therefore, a decreasing boundary area
and disappearance of impurity phases from the grain
boundary has a greater effect on sboundary than the
influence of a slight acceptor-dopant enrichment of the
grains on the bulk conductivity. This can be easily
understood as minor amounts of secondary phases
present along the grain boundaries might have a strong

Fig. 12. Temperature dependence of the total bulk (closed symbols)

and boundary (open symbols) conductivities of
(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics, calculated from the impe-
dance spectra. Inset shows grain-boundary conductivity in the low-
temperature range.

local effect but become highly diluted when dissolved in

the grains.
As for other solid electrolytes [15,29,30], the activa-
tion energy for the grain interior conduction is lower
with respect to the boundary. The role of boundary
Fig. 10. Temperature dependence of the total conductivity of resistivity hence decreases when temperature increases.
(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics in air. At temperatures above 1000
/1050 K, the grain-bound-
 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828 1825

ary conductivity apparently deviates from Arrhenius

plots due to loss of accuracy in the determination of the
small values of the boundary resistance compared to the
grain bulk and electrode. Another factor contributing to
the error of sboundary determination at high temperature
is the increase in electron transference numbers, dis-
cussed below. Note also that the difference between
activation energies (Ea) for the bulk and boundary
conductivities becomes smaller with increasing sintering
time (Table 1).

4.3. Transference numbers

Fig. 13 presents the temperature dependence of

electron transference numbers determined by FE mea-
surements in air under zero oxygen chemical potential
gradient. The transference numbers vary in the range
from 2.2 /103 to 1.5 /10 2 and increase with tem-
perature, in agreement with literature data [10,11]
showing that the activation energy for p-type electronic Fig. 14. Dependence of the average ion transference numbers of
transport is higher than that for ionic. The maximum te (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d, determined by FE measurements, on
the oxygen partial pressure inside measuring cells. Solid lines
values are observed for the material sintered for 0.5 h.
correspond to best fit to Eq. (9).
When the sintering time increases up to 40 h, the
electron-hole contribution to the total conductivity of
the range 0.19
/0.22. This is quite close to the ideal
LSGM ceramics decreases 2
/4 times; further sintering
theoretical value of the exponent for power sp versus
leads to a moderate increase in the electron transference
p (O2) dependence, equal to (1/4).
numbers.
As expected, a lower oxygen pressure leads to
decreasing electron-hole transport and, thus, to higher 4.4. Electron-hole conductivity
ion transference numbers (Fig. 14); the isothermal to
versus p1 dependencies can be adequately described by The p-type conductivity behavior as a function of
Eq. (9). The values of (1/m ) coefficient obtained by sintering time follows the trend observed for transfer-
fitting experimental data to this model behavior are in

Fig. 13. Temperature dependence of the electron transference numbers Fig. 15. Temperature dependence of p-type electronic conductivity of
of LSGM ceramics, determined by Faradaic efficiency measurements (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics in air. Data on
in air. Solid lines are for visual guidance only. La0.9Sr0.1Ga0.8Mg0.2O3d [10,11] are shown for comparison.
1826 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828

ence numbers (Fig. 15). The minimum electron-hole of grain boundaries. These values can hence be used for
conduction and maximum activation energy are ob- trend analysis; understanding of rboundary
o as ‘pure’
served for LSGM ceramics sintered for 40 h. The Ea and property of phases present at the boundary would be
sp values of this material are similar to those reported however incorrect. Analogously, the values of tboundary
o
by Kim and Yoo [10], while the results by Yamaji et al. show that the ratio of the ionic and electronic transport
[11] are close to the properties of at the boundary increases with sintering time, but
(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics sintered for cannot be interpreted as an indication that the grains
120 h. Therefore, the differences reported in literature are more conductive with respect to ions than to
on the electronic transport in LaGaO3-based electrolytes electron holes. A simple correction of the boundary
[6,10,11,14] might be due to different processing routes ionic resistivity for the capacitance in order to take the
and, thus, different microstructures. grain-boundary thickness into account [30] demon-
The activation energy for p-type conduction in strated that the ionic contribution to the boundary
LSGM ceramics sintered for different periods of time conductivity is considerably lower than tbulko values, in
varies in the narrow range 97
/115 kJ mol1 (Table 2). agreement with theoretical predictions [31]. However,
This is significantly higher than the activation energies the use of (Cbulksbulk
o /Cboundary) ratio as a quantitative
for total conductivity, 68
/73 kJ mol 1 at 973
/1273 K measure of ‘true’ ionic conductivity of the boundaries
(Table 3), leading to a greater role of electron-hole [30] seems too ambiguous for LSGM ceramics, particu-
conduction when temperature increases. One particular larly due to unknown dielectric constants of grains and
result is that the error in grain-boundary resistivity, boundaries comprising different phases.
estimated simply as the difference between Rbulk and As expected, the ionic resistivity of LSGM grains
Rtotal values, increases with temperature. Due to short- weakly depends on the microstructure, slightly decreas-
circuiting by the electronic pathway (Fig. 1), the ing with increasing sintering time due to dissolution of
difference (Rtotal
/Rbulk) is lower than the true boundary the impurity phases and, thus, increasing acceptor
/
resistance to ionic conduction; the relative error is dopant concentration in the perovskite (Fig. 16B). The
boundary resistance to ionic transport also decreases
DRboundary
o Rboundary
o  (Rtotal  Rbulk ) with sintering time and is minimum for the material
boundary 
Ro Rboundary
o sintered for 120 h, where the content of secondary
phases and magnesium segregated at the boundaries is
R2e
 1 negligible. On the contrary, the electron-hole conductiv-
(Re  Rbulk
o  Rboundary
o ) × (Re  Rbulk
o ) ity of LSGM ceramics decreases when the sintering time
 1tbulk × ttotal (19) increases from 0.5 to 40 h, and increases with further
o o
sintering (Figs. 15 and 16A). Thus, the total and bulk
This error partly explains the deviation of sboundary ion transference numbers follow the behavior of p-type
estimates from Arrhenius behavior at temperatures conductivity.
above 1000 K (Fig. 12). In the low-temperature range The results on p-type conductivity may be explained
when the electron transference numbers are lower than assuming that the hole transport is significantly en-
0.01, the effect of electron-hole conduction in LSGM on hanced by the segregation of Sr-rich secondary phases
the boundary resistance estimates is negligible. and magnesium oxide at the grain boundaries. This
The specific electronic and ionic resistivities (r /1/s) hypothesis seems consistent with data [10,14] suggesting
of the grain interior and boundary of LSGM ceramics, that the electronic conduction in La(Sr)Ga(Mg)O3d
estimated using Eqs. (12) and (13), are presented in Fig. ceramics increases with increasing concentration of Sr or
16. For comparison, the scale of all three resistivity axes Mg cations. Although the exact reasons are still
is the same. Notice that the boundary ionic resistance unknown, the strong dependence of p-type conductivity
values, shown in Fig. 16, are related to the overall on the microstructure (Fig. 15) suggests the possible
volume of the ceramic materials, but not to the volume relevance of microstructural variations resulting from

Table 2
Activation energy for the p-type electronic conductivity of LaGaO3-based solid electrolytes in air

Composition Sintering time (h) T (K) Ea (kJ mol 1) Ref.

(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d 0.5 973
/1223 979/6 This work

4.0 973
/1223 979/5
40 973
/1223 1159/9
120 973
/1223 989/4
La0.9Sr0.1Ga0.8Mg0.2O3d 1073
/1273 1179/22 [10]
873
/1273 1059/11 [11]
 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828 1827

Table 3
Apparent activation energy for total conductivity of (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d ceramics in air

Sintering time (h) Low-temperature range (623
/973 K) High-temperature range (973
/1273 K)

Ea (kJ mol 1) ln(A0/S K cm1) Ea (kJ mol 1) ln(A0/S K cm 1)

0.5 969/5 15.49/0.7 719/4 12.39/0.4

4.0 1079/8 179/1 739/6 12.69/0.7
40.0 1099/10 189/2 719/8 12.49/0.8
120 1019/4 16.49/0.6 689/2 12.39/0.2

Ea and A0 are the regression parameters of the standard Arrhenius model, s/(A0/T ) /exp(/Ea/RT ).

In summary, the results obtained in this work clearly

suggest a significant effect of the ceramic microstructure
on minor contributions to the total conductivity of
oxygen ion-conducting solid electrolytes, a subject
deserving little attention in the past. Such influence is
similar to the microstructural effects on the major
conductivity components in oxide semiconducting and
solid-electrolyte ceramics. Therefore, the electronic con-
duction in solid electrolytes should be analyzed not only
as property of a given phase, determined by defect
equilibria, temperature and oxygen partial pressure, but
also as function of the microstructure. Obviously, the
overall relevance of this effect is strongly influenced by
the homogeneity of samples on a scale hardly studied in
most cases.

5. Conclusions

A series of (La0.9Sr0.1)0.98Ga0.8Mg0.2O3d (LSGM)

ceramics, with different microstructures and density
higher than 94%, were prepared from a commercial
powder by sintering at 1673 K for 0.5
/120 h. Increasing
Fig. 16. Estimates of the ion transference numbers and partial sintering time leads to grain growth and to the dissolu-
electronic and ionic resistivities of grain interior and boundary, as tion of Sr-rich secondary phases and magnesium oxide,
functions of sintering time of LSGM ceramics, at 973 K in air.
present in trace amounts at the boundaries, into the
major perovskite phase. This is accompanied with a
compositional changes, for the behavior discussed in slight decrease in the thermal expansion and grain
interior resistivity, while the grain-boundary resistance
Refs [10,14]. When the sintering time of
decreases 3
/7 times. The electron-hole transference
(La0.9Sr0.1)0.98Ga0.8Mg0.2O3d increases from 0.5 to 40
numbers of LSGM ceramics, determined by the Far-
h, a decrease in sp values could hence be attributed to
adaic efficiency measurements in air under zero oxygen
the dissolution of Sr-enriched phases and magnesium
chemical potential, vary in the range from 2.2 /103 to
present at the grain boundaries, in the perovskite phase. 1.5 /102 at 973
/1223 K, increasing with temperature
Further annealing provides single-phase ceramics, where and oxygen partial pressure. When the sintering time at
the major transport-limiting role belongs to the grain- 1673 K increases from 0.5 to 40 h, the p-type electronic
boundary area, decreasing when the sintering time transport decreases. This effect may be explained
increases. Due to this, the grain-boundary contribution assuming that the formation of Sr-enriched phases
to the total electronic resistance starts to decrease. The and/or magnesium present at the boundaries signifi-
latter effect is well known for oxide semiconductors cantly contribute to the hole conduction. Due to
(Refs. [31,32] and references cited). As a result, the decreasing boundary area in single-phase ceramics,
material sintered for 40 h shows the minimum p-type further sintering results in higher p-type conductivity.
conductivity compared to other LSGM ceramics (Fig. The activation energies for electron-hole transport in air
16A). are 97
/115 kJ mol 1. The results demonstrate that, as
1828 V.V. Kharton et al. / Electrochimica Acta 48 (2003) 1817
/1828

found for the ionic conductivity, the electronic conduc- (Eds.), SOFC-V, vols. PV 97-40, The Electrochemical Society,
Pennington, NJ, 1997, pp. 1041
/1050.
tion in solid electrolytes strongly depends on the ceramic
[12] H. Hayashi, M. Suzuki, H. Inaba, Solid State Ionics 128 (2000)
microstructure, the latter becoming an increasingly 131.
important tool in designing materials performance. [13] E. Djurado, M. Labeau, J. Eur. Ceram. Soc. 18 (1998) 1397.
[14] V.V. Kharton, A.A. Yaremchenko, A.P. Viskup, F.M. Figueir-
edo, A.L. Shaulo, A.V. Kovalevsky, E.N. Naumovich, F.M.B.
Acknowledgements Marques, Ionics 8 (2002) 215.
[15] V.V. Kharton, F.M.B. Marques, Curr. Opin. Solid State Mater.
Sci. 6 (2002) 261.
This work was supported by FCT, Portugal (POCTI [16] V.V. Kharton, A.A. Yaremchenko, A.P. Viskup, G.C. Mather,
program and project BD/6595/2001), NATO Science for E.N. Naumovich, F.M.B. Marques, Solid State Ionics 128 (2000)
Peace program (project 978002), and INTAS (project 79.
00276). The authors are sincerely grateful to Ekaterina [17] V.V. Kharton, A.P. Viskup, A.A. Yaremchenko, R.T. Baker, B.
Tsipis for experimental assistance and helpful discus- Gharbage, G.C. Mather, F.M. Figueiredo, E.N. Naumovich,
F.M.B. Marques, Solid State Ionics 132 (2000) 119.
sions. [18] A.A. Yaremchenko, V.V. Kharton, A.P. Viskup, E.N. Naumo-
vich, N.M. Lapchuk, V.N. Tikhonovich, J. Solid State Chem. 142
(1999) 325.
References [19] V.V. Kharton, A.L. Shaulo, A.P. Viskup, M.Y.u. Avdeev, A.A.
Yaremchenko, M.V. Patrakeev, A.I. Kurbakov, E.N. Naumo-
[1] T. Ishihara, H. Matsuda, Y. Takita, J. Am. Chem. Soc. 116 (1994) vich, F.M.B. Marques, Solid State Ionics 150 (2002) 229.
3801. [20] V.V. Kharton, A.P. Viskup, F.M. Figueiredo, E.N. Naumovich,
[2] M. Feng, J.B. Goodenough, Eur. J. Solid State Inorg. Chem. 31 A.A. Yaremchenko, F.M.B. Marques, Electrochim. Acta 46
(1994) 663. (2001) 2879.
[3] P. Huang, A. Petric, J. Electrochem. Soc. 143 (1996) 1644. [21] V.P. Gorelov, Elektrokhimiya 24 (1988) 1380 (in Russian).
[4] J.W. Stevenson, T.R. Armstrong, L.R. Pederson, J. Li, C.A. [22] N. Matsui, Solid State Ionics 57 (1992) 121.
Lewinsohn, S. Baskaran, J. Electrochem. Soc. 144 (1997) 3613. [23] M. Liu, H. Hu, J. Electrochem. Soc. 143 (1996) L109.
[5] K. Huang, R.S. Tichy, J.B. Goodenough, J. Am. Ceram. Soc. 81 [24] V.V. Kharton, F.M.B. Marques, Solid State Ionics 140 (2001)
(1998) 2565. 381.
[6] J.W. Stevenson, T.R. Armstrong, L.R. Pederson, J. Li, C.A. [25] A.R. West, D.C. Sinclair, N. Hirose, J. Electroceramics 1 (1997)
Lewinsohn, S. Baskaran, Solid State Ionics 113
/115 (1998) 571. 65.
[7] P.R. Slater, J.T.S. Irvine, T. Ishihara, Y. Takita, J. Solid State [26] J. Fleig, J. Maier, J. Am. Ceram. Soc. 82 (1999) 3485.
Chem. 139 (1998) 135. [27] J. Rodriguez-Carvajal, Physica B 192 (1993) 55.
[8] A. Ahmad-Khanlou, F. Tietz, D. Stover, Solid State Ionics 135 [28] I. Rueter, H. Mueller-Buschbaum, Z. Anogr. Allg. Chem. 584
(2000) 543. (1990) 119.
[9] S.M. Choi, K.T. Lee, S. Kim, M.C. Chun, H.L. Lee, Solid State [29] C. Tian, S.-W. Chan, Solid State Ionics 134 (2000) 89.
Ionics 131 (2000) 221. [30] G.M. Cristie, F.P.F. van Berkel, Solid State Ionics 83 (1996) 17.
[10] J.-H. Kim, H.-I. Yoo, Solid State Ionics 140 (2001) 105. [31] X. Guo, J. Fleig, J. Maier, J. Electrochem. Soc. 148 (2001) J50.
[11] K. Yamaji, T. Horita, M. Ishikawa, N. Sakai, H. Yokokawa, M. [32] J.C.C. Abrantes, J.A. Labrincha, J.R. Frade, Mater. Res. Bull. 35
Dokiya, in: U. Stimming, S.C. Singhal, H. Tagawa, W. Lehnert (2000) 965.