Full text of DOI:10.1557/JMR.2002.0351

Dielectric and piezoelectric properties of textured
Sr0.53Ba0.47Nb O ceramics prepared by templated
2
6
grain growth
a)
Cihangir Duran, Susan Trolier-McKinstry, and Gary L. Messing
Materials Science and Engineering Department, Materials Research Institute, Pennsylvania State
University, University Park, Pennsylvania 16802
(Received 11 November 2001; accepted 24 June 2002)
Fiber textured Sr0.53Ba0.47Nb O ceramics were reactively sintered to ജ95% of the
2
6
theoretical density from a mixture of SrNb O and BaNb O powders. Texture in 〈001〉
2
6
2 6
was obtained by templated grain growth on 〈001〉-oriented acicular KSr Nb O
2
5 15
template particles. The most highly textured ceramics had a peak dielectric constant of
2
2
3,600 (at T ס 128 °C), a remanent polarization (P ) of 20.3 ␮C/cm , a saturation
c r
2
polarization (P ) of 24 ␮C/cm (69–96% of single crystal), and a piezoelectric charge
sat
coefficient (d ) of 84 pC/N (76–93% of single crystal). A model, correlating
3
3
polarization with grain orientation, predicts that P increases sharply when a
r
percolating grain network forms to transfer charge between elongated grains.
I. INTRODUCTION
the matrix were found to broaden the texture distribution
in the polar axis, particularly at low template loadings. In
this paper, we report a reactive templated grain growth
process to obtain textured SBN53 in undoped and
Nb O -rich matrices to avoid the deleterious effects of
Electroceramics are used in a variety of applications
such as sensors, actuators, capacitors, and varistors. In
these applications, the ceramic microstructure often con-
sists of equiaxed, randomly oriented grains. Ferroelectric
transducers require poling to obtain useful piezoelectric
properties. Poling is difficult for some ferroelectric ce-
ramics because the symmetry-related orientations avail-
2
5
V O on dielectric properties. The electrical properties
2
5
of textured SBN53 are compared to those of random
samples, textured samples prepared with V O , and
2
5
1
single crystals. A method is proposed to model the po-
larization behavior as a function of texture fraction and
the degree of orientation distribution.
able for the spontaneous polarization are limited.
2
Nagata et al. achieved anisotropic properties by hot
pressing SBN ceramics (x ס 0.30–0.65), but their po-
larization values were well below the single-crystal val-
ues. In an earlier paper, we reported enhanced electrical
properties in highly textured Sr0.53Ba0.47Nb O (SBN53)
II. EXPERIMENTAL
2
6
3
ceramics fabricated by templated grain growth (TGG).
Acicular KSr Nb O (KSN) particles (5 to 15.4 wt%)
SrNb O (SN) and BaNb O (BN) powders were used
2 6 2 6
to prepare SBN53 ceramics by reactive sintering. Acicu-
lar KSN templates were used to texture SBN53 by TGG.
Templates of 1–3 ␮m in diameter and 10–20 ␮m in
length were physically separated by differential settling
in water, to narrow the texture distribution. A detailed
study of template particle and matrix powder synthesis
(together with reactive versus conventional sintering) is
2
5 15
and V O (0.8 mol%) were used as template and liquid
2
5
former, respectively. It was found that texture develop-
ment started at T < 1000 °C due to melting of V O at
2
5
690 °C. Highly grain-oriented ceramics (texture fraction, f,
approximately 0.93) had 60–84% of single-crystal satura-
tion polarization and 80–98% of single-crystal strain
coefficient (d ) in the polar (c) direction. However, the
3
,4
given elsewhere. The fiber axis is [001] (or c direc-
tion), which corresponds to the spontaneous polarization
and fast crystal growth direction of SBN.
33
peak dielectric properties were found to be much lower
than the single-crystal values due to the presence of
V O -based nonferroelectric phase(s) on the grain
Table I summarizes the sample designation, KSN con-
centration, initial amounts of additives, and excess BN in
the SBN53 samples. Because the composition of tem-
plate particles is different from that of the matrix powder,
extra BN powder was added to compensate for the stron-
tium and niobium introduced by the KSN templates.
However, the potassium and some of the excess niobium
2
5
boundaries. In addition, small template particles (espe-
cially <5 ␮m in length) and anisotropic grain growth in
a)
Now with the Department of Materials Science and Engineering,
Gebze Institute of Technology, Gebze, Turkey.
J. Mater. Res., Vol. 17, No. 9, Sep 2002
© 2002 Materials Research Society
2399
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
TABLE I. Sample designation, KSN content, and additive contribu-
tion in SBN53 ceramics.
The dielectric constants of normal ferroelectric ceram-
ics can be expressed by the Curie–Weiss law above Tc
(
or Tmax). However, the broad dielectric spectra of the
Additive contribution (wt%)
Sample
designation
KSN content
(wt%)
Excess
BN (wt%)
relaxor ferroelectrics more appropriately follow a qua-
dratic law. To quantify the phase transition modes for
the intermediate type ferroelectrics (e.g., neither normal
nor relaxor), a power law combining the Curie–Weiss
and the quadratic laws was proposed by Uchino and No-
a
K O
Nb O5
2
2
9
T5
T9
T12
T15
T5-9
T5-15
5
9.1
12
15.4
5
0.26
0.47
0.62
0.79
0.47
0.79
0.72
1.32
1.74
2.23
1.32
2.23
4.05
7.41
9.71
12.45
4.05
4.05
1
0
mura and is given as
5
␥
1
ր
K = ͑1 Kmax͒ + ͓͑T − Tmax͒ ͔
ր
ր
C
,
(3)
a
T refers to textured samples.
where K is the dielectric constant at T, Kmax is the di-
electric constant at Tmax, ␥ is a diffuseness coefficient,
and C is the Curie–Weiss constant. Equation (3) de-
scribes a normal ferroelectric when ␥ ס 1 (the Curie–
Weiss law) and a relaxor ferroelectric when ␥ ס 2 (the
from the KSN could not be compensated in the final
composition. Therefore, the concentration of K O and
2
Nb O increases with increasing KSN concentration. For
2
5
the T5-9 and T5-T15 samples, only 5 wt% KSN tem-
plates were used. However, their compositions were ad-
1
1
quadratic law).
Polarization–electric field (P–E) hysteresis data were
measured at 10 Hz, using a modified Sawyer–Tower cir-
cuit. P–E data were recorded in two different ways; for
the first, the maximum voltage was increased sequen-
tially (e.g., 10, 20, 30 kV/cm) (type A), and for the sec-
ond one, the sample was depoled by annealing at 500 °C
for 1 h before each measurement (type B).
The fatigue behavior of the samples was also meas-
ured as a function of cycling using the same P–E circuit
at 50 kV/cm, 10 Hz, and 25 °C. Poling was performed by
justed to T9 and T15, respectively, by adding K O
2
(
J.T. Baker, Philipsburg, NJ) and Nb O (H.C. Starck,
2 5
Newton, MA). For comparison, random samples (de-
noted by R) were also prepared by tape casting, using
KSN powder rather than acicular particles. Procedures to
obtain oriented template particles in a mixture of SN and
BN and to prepare samples for microstructure analysis
3
were described earlier.
After binder burnout, samples were heated at 7 °C/min
to 1200–1450 °C and held for 1 min to 12 h. Sections
were cut parallel (denoted by “//”) and perpendicular
applying 10 kV/cm at approximately 10–15 °C above Tc
for 15 min and then cooling to room temperature while
(
denoted by “֊”) to the casting direction from the tex-
maintaining the field. Procedures to measure electrical
tured samples. In other words, the symbols // and ֊ stand
for a–b oriented (non-polar) and c-oriented (polar)
samples, respectively.
3
properties were described earlier.
Crystallographic texture was characterized using x-ray
diffractometry (XRD) on the ֊-cut samples and the
texture fraction, f, was calculated, using the Lotgering
factor,
f ס (P − P°)/(1 − P°)
,
(1)
where P and P° are [I(001) + I(002)]/∑I
in the textured
(hkl)
and random cases, respectively. Scanning electron mi-
croscopy (SEM) micrographs of //-cut samples were used
to characterize the morphological texture. Crystallo-
graphic orientation distribution was obtained from cor-
5
,6
rected rocking-curve (or ␻-scan) measurements, using
the (002) SBN53 peak. The orientation distribution was
calculated by fitting corrected rocking curves to the
7
,8
March–Dollase equation:
2
2
−1
2
−3/2
Q(f,r,␣) ס f(r cos ␣ + r sin ␣)
+ (1 − f ) , (2)
where f is the texture fraction, r is the degree of orien-
tation parameter, and ␣ is the angle from the texture
direction (e.g., [001]). The r is closely related to the
width of the orientation distribution of the anisotropic
grains and ranges from r ס 1 for random to r ס 0 for
perfectly textured materials.
FIG. 1. XRD patterns of the T9֊ samples after sintering for 4 h at (b)
1200 °C, (c) 1300 °C, and (d) 1400 °C. Curve “a” is for the random
(
R9) sample (heated at 1250 °C for 5 h) and given as an untextured
reference pattern. Major SBN53 peaks (JCPDS No. 39-265) are
marked.
2
400
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
III. RESULTS
of the samples in Fig. 3, calculated by the stereological
1
3
2
2
method, is 36 × 6 ␮m for T5-T15, 24 × 4 ␮m for T5,
A. Densification and texture development
2
2
1
7 × 3 ␮m for T9, and 14 × 3 ␮m for T15 samples.
The relative theoretical density (TD) was based on the
Crystallographic texture analysis was performed on the
1
2
density of an SBN52 single crystal (5.33 g/cc), which
is close to the composition of interest. The samples
reached ജ95% TD after sintering at temperatures
ജ1300 °C for 4 h, and then densification leveled off re-
gardless of the sintering temperature (for 4 h) and tem-
plate content. For sintering at 1400 °C, the densities were
nearly the same (approximately 96–98% TD) in samples
heat-treated between 1 min and 12 h.
XRD patterns of the T9֊ samples as a function of
sintering temperature as well as a reference pattern
of random SBN53 (R9) sintered at 1250 °C for 5 h are
given in Fig. 1. For the T9 samples, the {00l} peaks
dominate the diffraction pattern as the sintering tempera-
ture increases. In addition, all samples transformed to the
tetragonal tungsten bronze (TTB) structure by T ס
֊
-cut samples, using the Lotgering factor (Fig. 4).
The texture fraction (f) increases with increasing sinter-
ing temperature for all samples. Moreover, holding the
samples at 1400 °C for 12 h resulted in enhancement in f.
For example, the T5-T15 samples attained f ס 0.98 com-
pared to 0.61 for the T5 samples.
Orientation distribution curves for the samples sin-
tered at 1400 °C for 12 h were obtained by fitting the
corrected rocking curve data to Eq. (2), and then the r
values were determined. The r values were found to de-
crease with decreasing template content; that is, r ס 0.40
for T15, 0.34 for T12, 0.32 for T9, and 0.29 for T5 and
T5-T15. Increase in the r parameter corresponds to a
broadening of the orientation distribution in the c direc-
tion (i.e., [001]).
1200 °C. Figure 2 shows that morphological texture de-
velopment increases with sintering temperature and/or
time in the T9// samples. At 1200 °C, the matrix is still
very porous and there is no overgrowth on the KSN
templates. During the later stage of densification, the
porosity starts to concentrate near template particles
B. Dielectric properties
Figure 5 shows the orientation and temperature
dependence of the dielectric properties at different fre-
quencies for the T9 samples sintered at 1400 °C from
1 min to 12 h (samples are ജ96.5% dense). Figure 5(a)
shows that the peak dielectric constant (Kmax) strongly
improves in the c direction with sintering time. Dielectric
spectra indicate relaxor behavior with a gradually de-
creasing Kmax with frequency. In addition, the breadth of
the phase transformation diminishes with sintering time.
The frequency dependence of Tmax is more pronounced
for the samples sintered for 1 min. Room-temperature
[
Figs. 2(b) and (c)].
Figure 3 compares the effect of template concentration
on texture development in the samples sintered at
400 °C for 12 h. For the T5-T15 samples [Figs. 3(a)–(d)],
1
a template content of 5 wt% is insufficient for full texture
development. Porosity in the samples increases with the
template content. In addition, elongated grains are larger
in dimension with decreasing template content except for
the T5-T15// sample. The average length and thickness
dielectric constant (K ) and K
are much lower in the
RT
max
FIG. 2. Texture development in the T9// samples as a function of sintering conditions: (a) 1200 °C for 4 h; (b) 1300 °C for 4 h; (c) 1400 °C for
min; (d) 1400 °C for 4 h.
1
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2401
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
FIG. 3. Texture development as a function of template concentration after sintering at 1400 °C for 12 h: (a) T5//; (b) T9//; (c) T12//; (d) T15//;
e) T5-T15//.
(
nonpolar a–b plane, shown as an inset in Fig. 5(a), and
decrease with increasing sintering time. In addition, the
frequency dispersion of the dielectric spectra diminishes
with sintering time. The anisotropy in the dielectric prop-
erties is consistent with data on SBN single crystals. The
two orientations also show different dielectric loss be-
haviors such that the loss is lower in the a–b plane than
the c direction. Unlike the a–b plane, loss increases with
sintering time in the c direction [Fig. 5(b)].
lowest and highest Kmax values (5600 versus 23,600) and
corresponding f values (0.6 versus 0.93), respectively.
␥ and C/Kmax decrease with increasing f.
C. Switching and piezoelectric properties
When hysteresis (P–E) loops were recorded using pro-
gressively higher applied fields on highly textured T9֊
and T15֊ samples (type A), it was found that extremely
rapid fatigue was observed [see Fig. 7(a)]. As a result, the
Kmax in the c direction was found to increase with
increasing initial template content and f. A maximum
Kmax of 23,600 at f ס 0.93 was measured in the T9֊
samples. Kmax values of 1525 and 860 in the random
remanent polarization (P ) decreased and coercive field
r
(E ) increased on successive measurements. The de-
c
crease in switchable polarization with increasing meas-
urement field could be avoided by heating the sample
(
R9) and T9// were obtained, respectively. Although the
above T between measurements [type B in Fig. 7(b)]. In
c
T5-15 samples reached the highest texture fraction (f ס
this case, P increases with the applied field. Highly tex-
r
0.98), they showed relatively lower Kmax (of 17,000).
tured T5-T15֊ samples, however, did not show strong
The diffuseness coefficient (␥) and degree of diffuse-
fatigue dependence and reached the highest P values
r
2
ness (C/Kmax) were calculated using Eq. (3). Figure 6
[e.g., P ס 20.3 ␮C/cm at 50 kV/cm in Fig. 7(b)]. In the
r
␥
shows an analysis of (Kmax/K) versus (T − Tmax) at
a–b plane, P decreased with increasing texture fraction,
r
100 kHz for the T9֊ samples sintered at various sinter-
and there was no evidence for fatigue.
ing conditions. Each line has a correlation coefficient
Comparison of the P–E loops as a function of crystal-
lographic orientation and type of measurement is given
in Fig. 8 for the T5-T15 samples sintered at 1400 °C for
12 h. The random sample (R15) was sintered at 1350 °C
2
(R ) ജ0.99 to a linear fit. The inset shows f, ␥, and
C/Kmax (in °C) as a function of sintering conditions.
Samples sintered at 1400 °C for 1 min and 12 h have the
2
402
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
FIG. 4. Texture fraction (f ), calculated using the Lotgering factor, as
a function of sintering temperature.
for only 4 h to prevent abnormal grain growth that de-
creases P and increases E . The textured samples show
r
c
strong anisotropy.
The remanent polarization was modeled on the basis
of the processing parameters (i.e., f and r). SBN has a
spontaneous polarization only in the c direction; that is,
P ס P ס 0 and P ꢀ 0. Therefore, for misoriented
1
2
3
grains, the polarization distribution can be expressed as
P3 cos ␣, where ␣ is the angle from the texture axis.
The orientation distribution in the c direction can be
calculated using Eq. (2). The volume of material oriented
5
at an angle ␣ changes as sin ␣. Therefore, Eq. (2) can be
multiplied by the sine function to determine the fraction
of material as a function of ␣ (i.e., normalization of the
March–Dollase function).
The polarization at any angle, therefore, can be found
by multiplying the polarization distribution function with
the normalized March–Dollase function,
FIG. 5. Dielectric (a) constant and (b) loss as a function of sintering
time and crystallographic (polar c and nonpolar a or b) direction for
the T9 samples. Samples were sintered at 1400 °C. The inset in (a)
gives the data for measurements in the a–b plane.
␲
2
P͑r,f͒ =
P3 cos ␣ ͓Q͑ f,r,␣͒ sin ␣͔ d ␣
.
(4)
͐0
slowed due to SBN53 phase formation from the SN and
BN mixture and compositional homogenization of
This equation calculates the maximum expected po-
larization as a function of experimentally determined r
cations in the TTB structure. The effect of K O and
2
and f values. Measured P data were fit to this model,
r
Nb O on densification and phase formation of SBN53
2
5
using f and r values for each composition. Calculated
ceramics during reactive sintering is discussed in detail
data from this modeling and measured P are given in
r
14
elsewhere.
Fig. 9. Calculated data yield the maximum polarization
that can ideally be obtained for a given set of f and r (like
single crystals cut at certain angles and assembled to-
The final solid solution (e.g., SBN53) between
the KSN and SBN starts at the template–matrix inter-
face. This process is accelerated by a liquid phase. Lee
gether). It is clear that P remains constant until f ≅ 0.5
r
15,16
et al.
found that Nb O -excess grain boundaries
2 5
and then increases sharply.
cause liquid formation at T ജ 1260 °C, and the tempera-
ture at which the liquid phase forms decreases with more
IV. DISCUSSION
1
7
excess Nb O . Therefore, liquid may form due to the
2
5
A. Densification and texture development
presence of excess Nb O contributed from the KSN
2
5
Densification increases with sintering temperature and
levels off after sintering at temperatures ജ1300 °C for
h for all samples. Densification below 1300 °C was
templates. The T5-T9 and T5-T15 samples, however,
also have a liquid phase in the matrix due to the presence
of Nb O , which makes the cation distribution more
4
2
5
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2403
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
uniform in the matrix, compared to T5-T15 samples. Upon
template growth, local homogenization (e.g., diffusion of
K , Sr , Nb , and Ba ions) can start inside the grown
region with increasing sintering temperature and/or time.
Figure 4 indicates that f increases sharply for each set
after 1300 °C, which indicates that densification of the
matrix is required for template particles to grow. How-
ever the observed crystallographic orientation in [001]
+
2+
5+
2+
(“curve b” in Fig. 1) without significant template growth
[Fig. 2(a)] suggests that texturing is mainly due to ini-
tially oriented KSN template particles at lower densifi-
cation. Increase in f with sintering conditions (Fig. 4)
results from the growth of templates in the matrix. An
initial template concentration around 9 wt% is adequate
to produce a highly textured microstructures (f ס 0.93)
with a narrow distribution of anisotropic grains (r ס
0
.32) in SBN53 ceramics when there is no liquid phase
present in the matrix (for the T5-T15 samples). However,
␥
FIG. 6. Plot of Kmax/K versus (T − Tmax) , with calculated values of
degree of diffuseness (C/Kmax in °C), as a function of sintering con-
ditions for the T9֊ samples. f and ␥ stand for the texture fraction and
diffuseness coefficient, respectively.
FIG. 8. P–E hysteresis loops of the T5-T15 samples sintered at
1400 °C for 12 h and random sample (R15) sintered at 1350 °C for 4 h.
FIG. 7. Change of average remanent polarization (P ) as a function of
r
initial template content in the c direction: (a) type A; (b) type B.
Samples were sintered at 1400 °C for 12 h.
FIG. 9. Change of average P (type B) as a function of initial template
content and f. Samples are ജ95% dense.
r
2
404
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
if template growth is assisted by a liquid in the matrix
i.e., for T5-T15), a template concentration as low as
wt% is sufficient to attain an f ס 0.98 while keeping
r ס 0.29. A similar result was also observed in alumina
textured by TGG. Comparison of the T5 and T5-T15
samples in Fig. 3 reveals that templates grow more in the
T5-T15 due to a liquid in the matrix, which results in
faster consumption of the matrix grains.
Grain boundary phases considerably suppress the ob-
served dielectric properties, especially at the Tmax. There-
fore, the T5-T15 samples had a lower Kmax (17,000)
although they reached the highest f (0.98) and lowest r
(0.29). SBN53 ceramics textured with a V O liquid
(
5
1
8
2
5
3
former had much lower Kmax (approximately 7,900).
Another reason for the lower dielectric properties may
+
be due to the presence of K ions. Analysis of dielectric
properties in ജ97% dense random samples, with no ab-
normal grain growth, indicates that Kmax decreases from
3520 for undoped samples to 2250 for 5 wt% and 1725
for 15.4 wt% KSN powder containing samples. Kmax
drops off very quickly when abnormal grain growth
B. Dielectric properties
The phase transition becomes narrower with increas-
ing sintering conditions for all compositions [see
Fig. 5(a) for the T9֊ samples]. In other words, ␥ and
C/Kmax decrease with increasing f (Fig. 6). These results
suggest that SBN53 compositions behave more like nor-
mal ferroelectrics (e.g., ␥ is closer to 1), with relatively
little dispersion with frequency. The higher C/Kmax for
samples with a lower f can be attributed to the inhomo-
+
starts for each set. Similar results were observed in K -,
+
+
23
Li -, and Na -doped SBN ceramics.
The increase in dielectric loss (tan ␦) in the c direction
and decrease in the a–b direction with increasing sinter-
ing time [Fig. 5(b)] can be attributed in part to the do-
main wall loss because the volume of the grains with
uniform polarization increases with increasing f, which
results in easier domain wall movement in the polar c
direction. The evidence for this is that the dielectric loss
of poled samples, measured as the temperature increased
from 20 to 250 °C, was almost half of the dielectric loss
of unpoled samples (e.g., 6.1% versus 12.3% at room
temperature). The difference is probably attributable to a
smaller number of domains and domain walls in the
poled samples than the unpoled samples. Matrix grains
that give isotropic properties are consumed by the tem-
plate particles, and therefore, dielectric loss decreases in
the nonpolar a–b direction with increasing f.
+
2+
geneous distribution of cations (e.g., more K and Sr in
2
+
the KSN templates and more Ba in the matrix due to
3
SN compensation ) between the elongated (e.g., tex-
tured) and matrix grains in the T5-T15 samples. At this
stage, the compositional fluctuation from one grain to
another can give rise to a range of Curie points, which
1
9,20
broadens the phase transition.
The compositional
fluctuations can be reduced with increasing sintering
temperature or time. Therefore, the sharpening of the
dielectric peak with increasing f may be attributed to
the gradual homogenization of cations on the A sites.
Note that the T5 samples still have weak relaxor behavior
(Tmax changes about 2 °C between 1 kHz and 100 kHz)
even after sintering 1400 °C for 12 h presumably because
of incomplete K homogenization between textured and
Tc mainly depends on the amount of KSN and
+
the distribution of cations. T increases steadily after the
c
matrix grains [Fig. 3(a)]. The T5-T15 samples, however,
have ␥ ס 1.07 and C/Kmax ס 25.9 °C with no frequency
dispersion, probably due to more uniform distribution of
cations caused by a liquid-phase sintering.
Increase in f resulted in enhanced dielectric proper-
ties. The highest Kmax is obtained in the T9֊ samples
completion of densification at 1300 °C, which may be
due to (i) increase in f (and, thus, grain size) and (ii)
diffusion of cations (primarily K ) into the SBN lattice,
+
that is, compositional homogenization. A total of five out
2
+
2+
of six A sites are occupied by Sr and Ba ions; there-
2
4
fore, SBN is an empty bronze. Alkali-metal elements,
+
+
(
Kmax ס 23,600) because the T9 samples attained
such as Na and K partially or fully occupy the inter-
2
3,25,26
a high f (0.93) and a low r value (0.32). However,
Kmax ס 23,600 is substantially less than the single-
stitial vacant sites without vacancy formation,
which stabilizes the ferroelectric structure. Therefore, Tc
increases due to the increased degree of filling of inter-
stitial A sites with KSN content. Samples with ജ9.1 wt%
2
1
crystal values. For example, K
ס 81,000 and
max
21,22
40,000–63,000
were measured for SBN50 and SBN60
single crystals, respectively. The lower dielectric prop-
erties in textured samples can be attributed to the exist-
ence of nonferroelectric grain boundary phase(s). Excess
Nb O contributed by the KSN templates forms a liquid
templates (f ס 0.89 to 0.93) have a T of 124–133 °C
compared to 102–108 °C in the T5 samples (f ס 0.61)
after sintering at 1400 °C for 12 h. The T5-T15 samples
c
had a T of 150–154 °C, possibly due to more uniform
2
5
c
during sintering and possibly forms a grain boundary
phase(s) after cooling. This is more prevalent for the
T5-T15 samples because the excess Nb O is present
cation distribution and, thus, more A site filling by the
+
K ions. These results indicate that T in TGG samples
c
is strongly correlated with f due to changes in KSN
2
5
both in the elongated grains and in the matrix, while
excess Nb O is contained only in the elongated grains in
the T5-15 samples. Therefore, the T5-T15 samples
probably have dielectrically thicker grain boundaries.
content. Pure SBN50 has a T between 115 and 128 °C,
c
2
1,27–31
depending on the crystal quality.
This difference
2
5
3
can be attributed to the different compositional variations
and doping levels in the textured samples.
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2405
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
C. Switching and piezoelectric properties
transfer to orient the domains and hence to increase the
macroscopic polarization. Because the a–b plane is non-
polar, higher r also degrades the percolation of the
switchable grains.
The type of measurement, namely, type A and type B,
also affects the properties in the c direction. Figure 7(a)
As is the case for the dielectric properties, the polari-
zation is strongly enhanced in the c direction with in-
creasing f (Fig. 9). The T5-T15֊ samples reached a
2
maximum P ס 20.3 ␮C/cm . The estimated saturation
r
polarization (P ) from the P–E loop is approximately
sat
2
shows several possible behaviors for P as a function of
r
2
4 ␮C/cm (“c ” in Fig. 8), which is 69–96% of reported
2
the amplitude of the field. P for T5֊ (f ס 0.61) in-
r
SBN50 single-crystal values. The average P for single
sat
2
creases with the field. P for T15֊ (f ס 0.89) decreases
r
crystals is 27.4 ␮C/cm , and it ranges from 25 to 35 ␮C/
cm depending on the quality of the grown crystal.
2
31–33
monotonically. P for T9֊ (f ס 0.93) shows an initial
r
decrease followed by a shallow rise that was coupled
with increased asymmetry in the P–E loops. In type B
measurements, however, no asymmetry was observed in
While P improves with f in the c direction, it decreases
r
in the a–b direction. For example, the T5// samples have
2
2
a P ס 1 ␮C/cm compared to 0.5 ␮C/cm for the T5-
r
any of the samples, and P was found to be higher
r
T15// samples after sintering at 1400 °C for 12 h. In ad-
[Fig. 7(b)]. The T5-T15 samples reached the highest Pr
dition, P (//-cuts) continuously decreases with increasing
r
in both measurements and did not experience fast polari-
zation fatigue or asymmetric loops. The reason for higher
Pr may be higher f and lower r, fewer grain boundaries,
and more uniform cation distribution.
The above-mentioned results indicate that polarization
fatigue becomes dominant with increasing f in the T5-
T15 compositions. Polarization fatigue is defined as the
gradual reduction of switchable charge with polarization
reversal under bipolar drive. To further investigate the
fatigue behavior in textured ceramics, the T15֊ sample
sintered at 1400 °C for 12 h was subjected to a cycling
f due to the consumption of unoriented matrix grains
because random samples have higher P than the //-cut
r
samples due to averaging of properties (Fig. 8).
Figure 9 shows that the calculated maximum polari-
zation from the model [Eq. (4)] is substantially higher
than the measured data, particularly at lower f values.
Possible reasons for lower P may be porosity, insulating
r
Nb O -based grain boundary phase(s), pinning sites for
2
5
domains within the elongated grains, poor continuity in
the polarization, and randomly oriented matrix grains.
Normalized P [i.e., P (calculated)/P (measured)] at f ס
r
r
r
experiment at 50 kV/cm, 10 Hz, and 25 °C. P decreased
r
0.98 indicates that a value as high as 86% of single
2
very fast from 12.5 to 6.8 ␮C/cm after 10,000 cycles
crystal can be obtained by texturing.
2
and gradually to 5.4 ␮C/cm at 130,000 cycles. E , how-
c
Figure 9 further indicates that the P remains steady
r
ever, decreased slightly from 7 to 6.2 kV/cm after
(
and low) until nearly f ≅ 0.5 but increases considerably
1
30,000 cycles. No asymmetry was observed in the P–E
for higher f. The maximum polarization, obtained using
loops during cycling. Note that the T15֊ sample with
Eq. (4) for a random sample, is half of the single crystal
V O liquid phase former did not experience polarization
2
5
P , for 180° domain reversal. This was also theoreti-
sat
3
4
fatigue even after 130,000 cycles. Also, fatigue did not
always increase with increasing Pr.
Fatigue is commonly observed in SBN single crys-
tals. It was found in SBN50 that, after repeated switch-
ing under high fields, a major fraction of the total
polarization rapidly locks into unswitching domain con-
cally calculated elsewhere. Nevertheless, it is common
for low-symmetry samples to display much lower rema-
nent polarization values, as is the case here. This is be-
lieved to be because the random grains effectively
decrease the measured polarization at lower f values, by
preventing the charge transfer (to elongated grains) nec-
essary for switching. That is, there are so few possible
polarization directions that misoriented grains seriously
degrade the ability to maintain continuity in the polari-
zation vector during switching. As a result, compar-
atively little of the material is actually able to switch. It
is intriguing that the remanent polarization is virtually
unchanged for f <0.5. This suggests that for the coopera-
tive switching process to occur effectively, it is necessary
for well-oriented grains to have good connectivity. This
would also account for the dramatic rise in the switchable
polarization for f >0.5. This requirement for percolation
of well-oriented material should be less important in sys-
tems with more possible polarization directions (e.g.,
perovskites), since continuity of the polarization vector is
much easier to achieve there. These results suggest that
higher f and lower r are required to maximize the charge
figurations after a few initial reversals. For example, P
drops from approximately 28 ␮C/cm (1st cycle) to
sat
2
2
2
1
6
0 ␮C/cm (2nd cycle) and to 6 ␮C/cm (3rd cycle) at
1
9
0 Hz and 25 °C. A similar drastic decrease was also
3
5
observed in SBN75 single crystals. Single crystals usu-
ally have imperfections such as striation bands (perpen-
dicular to the c direction) that are mainly introduced
during crystal growth, which results in a change in the
1
9,35
stoichiometry.
Therefore, the switchable polarization
could be clamped due to charges being stored at defects
such as domain walls, dislocations, imperfections, and
grain boundaries (for ceramics), which tends to break up
the ferroelectric coupling. It was also observed that there
was no asymmetry with respect to the electric field, in-
dicating no net internal electric field, and therefore,
3
5
Ec remained almost constant.
2
406
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
The changes in polarization which occur on applica-
tion of a cyclic field are principally due to the intrinsic
(e.g., elongated grains) with more defect sites. For the
T5-T15 samples, however, lower polarization fatigue
may be due to more uniform distribution of cations be-
(
i.e., volume polarizability of the individual domains)
+
5+
and extrinsic (i.e., domain wall motion and reorientation
cause these samples have K and Nb ions both in the
matrix and elongated grains, which results in less avail-
able pinning sites inside the elongated grains. A second
possibility is that the differences in the fatigue behavior
are due to differences in the electrical conductivity of the
specimens due to changes in the grain boundary conduc-
tivity. In general, both T5-T15 and samples containing
3
6
effects) contributions. There are at least two possible
reasons for the effect of texture and processing on the
observed fatigue behavior: changes in the sample homo-
geneity; changes in the character of the grain boundaries.
As mentioned previously, SBN is an empty tungsten
2
+
2+
bronze in that 5 Sr or Ba ions occupy the 6 avail-
able interstitial sites in the TTB structure. Therefore,
high intrinsic channel defect concentrations due to vacant
sites already exist in SBN. These channel defects are
electrically neutral; however, short-range strain interac-
3
V O specimens had comparatively low dielectric loss
2
5
values. These same samples had more marked suppres-
sion of the peak permittivities due to grain boundary
phases. It is possible that lowered dielectric loss was
partially associated with a change in the electrical con-
ductivity and that resulting changes in charge injection
also affected the fatigue rates.
2
+
tions may exist due to nonuniform distributions of Ba
2+
24
and Sr in the A and A sites, respectively. For the
1
2
+
T5-T15 samples, K ions may also contribute to this
nonuniform cation and vacancy distribution. In addition,
similar to the striation bands in the single crystals, the
stoichiometry inside the elongated grains varies because
Increased P after type B measurements in Fig. 7(b)
r
suggest that fatigue is completely recoverable by anneal-
ing the samples above T (e.g., 500 °C for 1 h). These
c
2
+
+
there are more Sr and K ions in the center region (e.g.,
results indicate that domain wall pinning either by defect
sites or by injected charge during polarization reversal
is the main reason for the observed polarization fatigue in
the highly textured SBN53 ceramics. Similar results were
2
+
at the original KSN template position) and more Ba
ions in the grown region (due to excess BN in the ma-
3
2+
+
trix) . Because Ba and K ions diffuse slowly within
the grains (e.g., by solid state diffusion), there may be an
uneven distribution of cations and vacant sites, especially
with increasing f or the size of elongated grains. Similar
to charge trapping in the single crystals, compositional
gradients within each grain could behave as defect sites
for pinning the domains, which results in a decrease in Pr
during polarization reversal. The fatigue rate intensifies
with increasing f due to increased single-crystal regions
37
observed in Pb(Zn Nb )O –4.5PbTiO single crystals.
1/3
2/3
3
3
The best piezoelectric charge coefficient (d ) of
3
3
84 pC/N was obtained in the highly textured (f ס 0.98)
T5-T15֊ samples after sintering at 1400 °C for 12 h.
This value is 76 to 93% of the SBN50 single crystals
2
8,31,38–40
(d33 ס 90 to 110 pC/N for single crystals
).
This sample also showed the highest P . For the T5-T15
r
samples, the maximum d33 ס 63 pC/N was measured in
a
TABLE II. The average room-temperature dielectric and piezoelectric properties for random and textured samples sintered at 1400 °C for 12 h.
c
d
d
e
f
2 g
g
Sample
Tc (°C)
Kmax
K at 25 °C
tan ␦ (%)
d33 (pC/N)
Pr (␮C/cm )
Ec (kV/cm)
b
R0
R5
129
138
108
102
150
124
126
162
133
128
144
144
150
154
2965
2150
1065
9525
1525
860
21950
1055
865
18915
570
16070
645
445
485
610
1485
290
520
1325
300
480
1325
325
900
430
935
6.6
5.5
1.9
7.7
3.8
1.1
10.1
2.1
1.2
10.3
0.6
8.6
0.7
7.4
37
27
1.4
27
26
0.9
61
30
1.1
52
1
3.8
3.4
1
6.7
2.7
0.8
12.9
1.5
0.9
11
12.2
12.8
8.5
5.9
15.8
11
b
T5//
T5֊
b
R9
T9//
T9֊
R15
T15//
6.9
b
13.9
13.2
6.9
5.9
9.3
T15֊
T5-T9//
T5-T9֊
T5-T15//
T5-T15֊
0.3
16.8
0.5
20.3
78
1.5
83
10.8
10.6
17000
a
At least three measurements were averaged.
Dielectric properties were measured at 100 kHz due to space charge effect.
b
c
Tc: temperature at the Kmax measured at 1 kHz during cooling.
d
Kmax and K at 25 °C: Measured at 1 kHz during cooling.
e
f
tan ␦ (%): Measured at room temperature and 1 kHz during cooling.
d33: at 100 Hz and room temperature, after poling at 10 kV/cm at temperatures approximately 10–15 °C above Tc.
g
P and E : measured at room temperature and 10 Hz, for an applied field of 50 kV/cm (type B).
r
c
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2407
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
the T9֊ samples. This lower value can be attributed to
ACKNOWLEDGMENTS
inefficient poling. This is supported by the observation
This work was supported by Office of Naval Research
Grant N00014-98-1-0527. The authors also thank Hu-
seyin Yilmaz and Edward Sabolsky for their helpful dis-
cussions at various stages of this work. C.D. gratefully
acknowledges the support of the Gebze Institute of
Technology (Gebze, Turkey) for his Ph.D. study in the
United States.
that P measured from a pyroelectric measurement is
r
smaller than Psat measured in P–E loops. For example,
the T15֊ sample sintered at 1400 °C for 4 h has a P of
sat
2
1
2.6 ␮C/cm at room temperature obtained from a pyro-
electric measurement (as the poled sample was heated
from room temperature to 250 °C). However, P
ס
sat
2
1
7 ␮C/cm was estimated from the P–E loop (type B)
recorded at 50 kV/cm. This indicates inefficient domain
orientation and, therefore, the d33 values are lower.
The pertinent average room-temperature dielectric,
switching, and piezoelectric properties for random and
textured samples sintered at 1400 °C for 12 h are sum-
marized in Table II as a function of crystallographic di-
rection and composition. Random samples have higher
Curie temperatures compared to textured samples. The
REFERENCES
1. K. Shoji and Y. Uehara, Jpn. J. Appl. Phys. 30, 2315 (1991).
2. K. Nagata, Y. Yamamoto, H. Igarashi, and K. Okazaki, Ferroelec-
trics 38, 853 (1981).
3. C. Duran, S. Trolier-McKinstry, and G.L. Messing, J. Am. Ceram.
Soc. 83, 2203 (2000).
4
. C. Duran, G.L. Messing, and S. Trolier-McKinstry, J. Cryst.
Growth (2002, submitted for publication).
+
reasons could be more effective K homogenization in
the random samples (i.e., fine KSN powder versus acicu-
5. M.M. Seabaugh, M.D. Vaudin, J.P. Cline, and G.L. Messing,
1
4
lar template particles) and stress relief due to cracking.
J. Am. Ceram. Soc. 83, 2049 (2000).
6
. M.D. Vaudin, TexturePlus (National Institute of Standards and
Technology, Ceramics Division, Gaithersburg, MD, 1999).
. A. March, Z. Kristallogr. 81, 285 (1932).
These results indicate that electrical properties are
considerably improved, compared to randomly oriented
samples, and highly anisotropic properties similar to
single crystal values can be achieved in the samples tex-
tured by TGG.
7
8
. W.A. Dollase, J. Appl. Crystallogr. 19, 267 (1986).
9. V.V. Kirillov and V.A. Isupov, Ferroelectrics, 5, 3 (1973).
10. K. Uchino and S. Nomura, Ferroelectr. Lett. Sect. 44, 55 (1982).
11. S.J. Butcher and N.W. Thomas, J. Phys. Chem. Solids, 52, 595
(
1991).
V. CONCLUSIONS
12. C.L. Choy, W.P. Leung, T.G. Xi, Y. Fei, and C.F. Shao, J. Appl.
Phys. 71, 170 (1992).
13. NIH Image, V.1.56, by W. Rasband, National Institutes of Health,
Highly textured (98%) Sr0.53Ba0.47Nb O ceramics
2
6
were reactively sintered in a matrix of SrNb O and
Washington, DC.
4. C. Duran, G.L. Messing, and S. Trolier-McKinstry, J. Mater. Sci.
2
6
1
BaNb O powders. Acicular KSr Nb O particles were
2
6
2
5 15
(
2002).
used to texture the samples in [001]. Samples had a
phase-pure tetragonal tungsten bronze structure by
1
1
1
5. H-Y. Lee and R. Freer, J. Appl. Phys. 81, 376 (1997).
6. H-Y. Lee and R. Freer, J. Mater. Sci. 33, 1703 (1998).
7. J.R. Carruthers and M. Grasso, J. Electrochem. Soc.: Solid State
Sci. 117, 1426 (1970).
8. E. Suvaci and G.L. Messing, J. Am. Ceram. Soc. 83, 2041 (2000).
9. T.W. Cline, Ph.D. Thesis, Pennsylvania State University (1977).
0. G.A. Smolenski and V.A. Isupov, Zh. Tekh. Fiz. 24, 1375 (1954).
1. A.M. Glass, J. Appl. Phys. 40, 4699 (1969).
1200 °C and reached ജ95% of the theoretical density
after sintering for 4 h at 1300 °C. The dielectric and pi-
ezoelectric properties were affected particularly by tex-
ture fraction ( f ), degree of orientation parameter (r), and
grain boundary phases. Properties were enhanced in the
samples with higher f and lower r values. In addition,
the presence of nonferroelectric phases (mainly Nb O -
1
1
2
2
2
2. R.R. Neurgaonkar, W.W. Ho, W.K. Cory, W.F. Hall, and L.E. Cross,
Ferroelectrics 51, 185 (1984).
3. A. Bhanumathi, S.N. Murty, K. Umakantham, K.C. Mouli,
G. Padmavathi, K.T. Rao, and V. Syamalamba, Ferroelectrics
2
5
2
2
based) at the grain boundaries suppressed the observed
dielectric properties, especially at the T . For the samples
c
1
02, 173 (1990).
4. D. Viehland, Z. Xu, and W-H. Huang, Philos. Mag. A 71, 205
1995).
with uniform cation distribution, dielectric spectra
showed normal ferroelectric behavior and fast polari-
zation fatigue was not observed. The best electrical prop-
erties obtained are the peak dielectric constant of 23,600
(
25. S.N. Murty, K.V.R. Murty, K.C. Mouli, A. Bhanumathi,
S.B. Raju, G. Padmavathi, and K.L. Murty, Ferroelectrics 158,
325 (1994).
(
2
at T ס 128 °C), the remanent polarization (P ) of
c r
2
26. K. Umakantham, S.N. Murty, K.S. Rao, and A. Bhanumathi,
J. Mater. Sci. Lett. 6, 565 (1987).
7. R.R. Neurgaonkar, W.K. Cory, J.R. Oliver, E.J. Sharp, G.L. Wood,
and G.J. Salamo, Ferroelectrics 142, 167 (1993).
8. R.R. Neurgaonkar, W.K. Cory, and J.R. Oliver, Ferroelectrics 51,
0.3 ␮C/cm , the estimated saturation polarization
2
of 24 ␮C/cm (69–96% of SBN50 single crystal), and the
piezoelectric charge coefficient of 84 pC/N (76–93% of
SBN50 single crystal). A model was developed to cor-
relate processing parameters (f and r) to the polarization.
2
2
2
3
3
(1983).
9. R. Guo, A.S. Bhalla, G. Burns, and F.H. Dacol, Ferroelectrics 93,
97 (1989).
P increased sharply (i.e., after f ≅ 0.5) when the charge
r
3
transfer (via percolation) between elongated grains was
effectively established.
0. M. DiDomenico, Jr. and S.H. Wemple, J. App. Phys. 40, 720
(1969).
2
408
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45

C. Duran et al.: Dielectric and piezoelectric properties of textured Sr0.53Ba0.47Nb O ceramics prepared by templated grain growth
2
6
3
3
1. R.R. Neurgaonkar, W.F. Hall, J.R. Oliver, W.W. Ho, and
W.K. Cory, Ferroelectrics 87, 167 (1988).
2. S.B. Deshpande, H.S. Potdar, P.D. Godbole, and S.K. Date,
36. D.A. Payne, Ph.D. Thesis, Pennsylvania State University (1973).
37. K. Takemura, M. Ozgul, V. Bornand, S. Trolier-McKinstry, and
C.A. Randall, J. Appl. Phys. 88, 7272 (2000).
J. Am. Ceram. Soc. 75, 2581 (1992).
38. R.R. Neurgaonkar, J.R. Oliver, W.K. Cory, L.E. Cross, and
3
3
3
3. I. Camlibel, J. Appl. Phys. 40, 1690 (1969).
D. Viehland, Ferroelectrics 160, 265 (1994).
39. S.T. Liu, Ferroelectrics 22, 709 (1978).
40. T.W. Cline, L.E. Cross, and S.T. Liu, J. Appl. Phys. 49, 4298
(1978).
4. T. Takenaka and K. Sakata, Jpn. J. Appl. Phys. 19, 31 (1980).
5. B. Jimenez, C. Alemany, J. Mendiola, and E. Maurer, Ferroelec-
trics 38, 841 (1981).
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2409
http://journals.cambridge.org
Downloaded: 09 Apr 2015
IP address: 203.64.11.45