Vous êtes sur la page 1sur 9

Materials Research Bulletin 43 (2008) 1761–1769

www.elsevier.com/locate/matresbu

Locating the normal to relaxor phase boundary in


Ba(Ti1xHfx)O3 ceramics
Shahid Anwar *, P.R. Sagdeo, N.P. Lalla
UGC-DAE Consortium for Scientific Research, University Campus, Khandwa Road, Indore 452017, India
Received 4 January 2007; received in revised form 28 June 2007; accepted 6 July 2007
Available online 13 July 2007

Abstract
In this article, we report our studies on the relaxor behavior of Ba(Ti1xHfx)O3 ceramics, made with close compositions between
0.20  x  0.30, to locate the hafnium concentration boundary for the normal to relaxor crossover. X-ray diffraction followed by
Rietveld refinement shows the occurrence of single-phase cubic structure for the synthesized Ba(Ti1xHfx)O3 ceramics.
Temperature and frequency dependence of the real (e0 ) and imaginary (e00 ) parts of the dielectric permittivity has been studied
in the temperature range of 90–350 K at frequencies of 0.1, 1, 10, and 100 kHz. A diffuse phase transition accompanying frequency
dispersion is observed in the permittivity versus temperature plots revealing the occurrence of relaxor ferroelectric behavior. The Tm
verses Hf concentration plot shows a discontinuous jump and change in the slope at x = 0.23. Quantitative characterization based on
phenomenological models has also been presented. The plausible mechanism of the relaxor behavior has been discussed.
Substitution of Hf4+ for Ti4+ in BaTiO3 reduces the long-range polar ordering yielding a diffuse ferroelectric phase transition.
# 2007 Elsevier Ltd. All rights reserved.

Keyword : C. X-ray diffraction (XRD)

1. Introduction

Eversince its discovery in 1943, the high permittivity ferroelectric ceramic BaTiO3 and related ceramics have been
widely studied for its use as passive components in electronic industries. Since BaTiO3 has characteristic anomalies in
the dielectric constant near the ferroelectric–paraelectric phase transition at 393 K, to make it usable at lower
temperatures, several homovalent and hetrovalent substitutions have been tried [1,2]. The dielectric response of
relaxor ferroelectrics are quite different from that exhibited by the normal ferroelectrics. Relaxors show certain
unusual behavior like the occurrence of strong frequency dispersion in its dielectric properties, absence of
macroscopic polarization and anisotropy at a temperature far below Tm (the temperature at which the real part of
dielectric permittivity shows a maximum). Very high response coefficient of dielectric and mechanical properties and
the enhanced width of these response regime around the maxima temperature Tm (‘‘Curie range’’), makes relaxors
popular material for its applications as piezoelectric/electrostrictive actuators, sensors, multilayered capacitors and
other ferroelectric applications [2]. Lead-based perovskite relaxor ceramics have been studied extensively for the last
couple of decades. To extend their applicability to modern electronics further detailed information on their physical

* Corresponding author. Tel.: +91 7312463913; fax: +91 7312462294.


E-mail addresses: shahidanwr@gmail.com (S. Anwar), nplalla@csr.ernet.in (N.P. Lalla).

0025-5408/$ – see front matter # 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.materresbull.2007.07.013
1762 S. Anwar et al. / Materials Research Bulletin 43 (2008) 1761–1769

properties are necessary. Much effort has been made in last few decades especially about complex lead-based
perovskite oxide [3].
Recently, the attention has moved towards lead free materials, as lead free relaxor materials present a great interest
for environment protection and fundamental studies. Impurity doping in ceramics is a common way to improve its
performance. Several dopants like Ca, Zr, Ce, etc. have been tried and many enhanced properties are obtained [4–6].
Among these Ba(Ti1xZrx)O3 has received renewed attention due to its enhanced piezoelectric properties both in
single crystals and ceramics [5]. Homovalent substitution of Hf at Ti site in BaTiO3 has recently come up as a new
possible substitute for the lead free relaxor ceramics [6,7].
In this communication we present the results of structural and dielectric studies, carried out to refine the
composition of the homovalent substitution in Ba(Ti1xHfx)O3 (BHT) for normal to relaxor crossover. The dielectric
behavior has been monitored through permittivity versus temperature measurement in the temperature range of 90–
350 K and frequency range of 0.1–100 kHz. In our earlier studies [7,8] we have reported the occurrence of relaxor
behavior for x = 0.30 in BHT. But here we present our studies on BHT samples made at relatively narrower
composition steps between x = 0.20 and x = 0.30 of the Hf concentration. We observed emergence of relaxor behavior
for x  0.23, i.e. for Ba(Ti0.77Hf0.23)O3 and above. Knowledge of the substitution level for relaxor transition is
important for the study of basic physics behind relaxor transition. In literature, for certain compounds, anomalies have
been reported at the normal to relaxor transition boundary [9].

2. Experimental

Ba(Ti1xHfx)O3 ceramics, with x = 0.20, 0.22, 0.23 and 0.30, here after termed as BHT20, BHT22, BHT23 and
BHT30, were prepared following the conventional solid-state reaction technique using high purity (99.99%)
ingredient powders of BaCO3, TiO2 and HfO2. These ingredients were pre heated at 150 8C for 6 h to remove the
adsorbed moisture, if any. The dried powder were weighed in stoichiometric proportions and wet-mixed, taking
acetone as mixing medium. After well mixing and drying, the mixture was calcined at 1100 8C for 6 h. The calcined
powder was reground, mixed with polyvinyl alcohol as binder, and then pelletized into a 15 mm diameter pellet at
100 kN/cm2 pressure. The pellets were then sintered at 1250 8C for 12 h. The thickness of the final sintered pellet was
1 mm.
The as prepared samples were subjected to phase purity and microstructural characterizations using powder X-ray
diffraction (XRD) and scanning electron microscopy (SEM), respectively. XRD was carried out using Cu Ka radiation
employing Rigaku make u–2u diffractometer mounted on a rotating anode X-ray generator working at 10 kW output
powder. For microstructural characterization JEOL-5600 SEM was used. Before proceeding for the dielectric
measurements electroding of the pellets was carried out. This was accomplished by painting the flat surfaces with
high-temperature silver-paint and then firing them at 500 8C for 15 min. Dielectric measurements were carried out
from room temperature down to 90 K during heating and cooling cycles at temperature intervals of 0.5 K. Temperature
variation of dielectric permittivity (e0 ) and dielectric loss (tan d) were monitored in the temperature range of 90–350 K.
These measurements were performed using a computer interfaced programmable LCR meter (Hioki-3532) and a
Lakeshore make temperature controller with a Pt-100 temperature sensor. The Pt-100 temperature sensor was used in
4-probe configuration. The applied ac signal was varied in the frequency range of 102 to 105 Hz with peak-to-peak
amplitude of 1.0 V. The stability of temperature measurement was better than 20 mK in the entire studied temperature
range. Thus the measurement ensures that no anomaly is left unnoticed in the studied temperature range.

3. Results and discussions

3.1. Structural studies

The XRD data of all the Ba(Ti1xHfx)O3 compositions with x = 0.20–0.30 were subjected to Rietveld refinement
for phase purity and structural characterization. The refinement was performed using Fullprof software [10]. A Pseudo
Voigt function was chosen to create the profile shape. The XRD data was refined with tetragonal (P4mm) and cubic
(Pm3m) symmetries both. The x2 values for Pm3m were invariably lower to that of the P4mm. On an average x2 value
was found to be 3.6 for P4mm and 2.3 for Pm3m. It would be imperative to mention here that in the relaxor state
the structure may be pseudo cubic [11] but at temperatures much higher than Tm the phase is long range cubic. Since
S. Anwar et al. / Materials Research Bulletin 43 (2008) 1761–1769 1763

the structural studies in the present investigation were performed at room temperature, which is well above the
transition temperature (156 K for BHT23). Therefore the observed cubic symmetry is in accordance with the
expectations. Fig. 1 depicts a typical Rietveld refined powder XRD data of the BHT23 ceramic. The observed data is
shown as circles and the continuous line shows a fit to it. The quality of the fit ensures the Pm3m cubic phase.
The absence of any unaccounted diffraction peak confirms that the synthesized BHT ceramics are all single phase. The
refined cubic lattice parameters were plotted against composition in the lattice parameter versus concentration
phase diagram as given earlier [7]. It is found that the lattice parameters of these new compositions of BHT trace exactly
the same curve in the phase diagram. The lattice parameter–composition phase diagram including the new points is
shown as inset in Fig. 1. The percentage densification of the pellets was also estimated using the X-ray density and the
pellet density [7]. A densification of 90% was estimated for all the studied pellets. The grain size, as measured using
SEM micrographs of the pellets, was found to be 2–5 mm for every composition [8]. Fig. 2a and b represent typical SEM
micrographs of BHT20 and BHT30 ceramics, which lie across the ferroelectric to relaxor phase boundary.

3.2. Dielectric studies

Fig. 3 depicts the temperature dependence of real (e0 ) (a–d) and imaginary (e00 ) (e and f) parts of the dielectric
permittivity of BHT ceramics with x = 0.20, 0.22, 0.23 and 0.30 at various frequencies. The paraelectric to
ferroelectric phase transition for these doped ceramics has considerably diffused and the transition temperature (Tc)
has also decreased as that of the BaTiO3. It should be noticed that the permittivity versus temperature curves for
x  0.22 (BHT20 and BHT22), i.e. Fig. 3a and b shows only one broad diffuse transition instead of three phase
transitions as observed in pure or low (x = 0.05) Hf doped BaTiO3 [7]. The three phase transitions observed in BaTiO3
have got pinched and merged into one broad peak for higher values of x, but we could still observe the occurrence of
hysteresis during heating and cooling cycles, see Fig. 4a and b. No frequency dispersion of the maxima temperature of
real and imaginary parts of the permittivity is observed. The details of the parameters are given in Table 1.
For x  0.23 ceramics, the dielectric behavior changes quite drastically, as shown in Fig. 3c–f. There is a broad
diffuse peak in e0 –T curve, which exhibits a large frequency dispersion of Tm corresponding to e0 –T maximum. Tm
shifts to higher temperatures but e0m (the maximum of e0 ) decreases, with increasing frequency. The temperature and
frequency variation of the imaginary part e00 is also well behaved: the value of T 0m (temperature corresponding to the
maxima in dielectric absorption), is much less than Tm, and shifts to higher values with increasing frequency. The
frequency dispersion of e00 was found to be just opposite of the e0 . The above described features of (e0 –T) and (e00 –T) are
typical to a relaxor ferroelectric state [3,5,7,12,13]. It should be noted that for x = 0.23 and 0.30 we do not see any
appreciable hysteresis, see Fig. 4c and d. Thus the occurrence of frequency dispersion is accompanied by the absence

Fig. 1. Rietveld refined XRD pattern of Ba(Ti0.77Hf0.23)O3. It shows the data with the fit profile, the difference plot and the Braggs peak positions.
Inset shows the lattice parameter variation with hafnium concentration revealing tetragonal to cubic transition at x = 0.10.
1764 S. Anwar et al. / Materials Research Bulletin 43 (2008) 1761–1769

Fig. 2. Typical secondary electron (SEM) micrographs representing the (a) ferroelectric (BHT20) and (b) relaxor (BHT30) compositions. The
microstructures for both the samples are identical showing densely packed faceted grains of 2–5 mm sizes in both the cases.

of hysteresis. It appears that the dominance of relaxor behavior appears for x  0.23 only and that the normal to relaxor
crossover boundary lies somewhere in between 0.22 < x < 0.23. Fig. 5 shows the Tm versus concentration (Hf) phase-
diagram [7]. It should be noted that Tm for x = 0.22 lies exactly on the curve for the normal ferroelectric phase and that
of for x = 0.23 on the corresponding relaxor phase line.

Table 1
Table showing parameters related to ferroelectric and relaxor behavior
BHT20 BHT22 BHT23 BHT30
Tm (K) 245 227 166 149
DTm (K) 71 85 123 132
DTdif (K) 30 30 30 33
DTrelax (K) 2 1 18 23
Here Tm (or TC), DTm, and DTrelax represent the transition temperature at 1 kHz, degree of Curie–Weiss deviation, and degree of relaxation,
respectively for the studied Ba(Ti1xHfx)O3.
S. Anwar et al. / Materials Research Bulletin 43 (2008) 1761–1769 1765

Fig. 3. Temperature dependence of real e0 (a–d) and imaginary e00 (e and f) parts of BHT20, BHT22, BHT23 and BHT30 at various frequencies
(0.1, 1, 10 and 100 kHz.). The occurrence of strong frequency dispersion for x = 0.23 and 0.30 can be seen.

In the literature related to the relaxor studies the Tm versus dopant concentration plot has been invariably used
[9,14,15]. Identical to our observation, the discontinuous jump in Tm and change in the slope of Tm variation at well-
defined composition has been clearly shown for different types of compounds [9,14]. The ferroelectric to relaxor
crossover boundary has invariably been attributed to the composition showing discontinuous jump. Thus in our case
also the ferroelectric to relaxor crossover transition can be attributed to the observed discontinuous jump at x = 0.23,
see Fig. 5. This is further supported by the occurrence of frequency dispersion accompanied by the absence of
hysteresis at the same composition, x = 0.23. The compositions below x = 0.23 are predominantly ferroelectric,
although diffuse. This is realized on the basis of the occurrence of clear hysteresis, although narrow, in the e0 –T plot,
Fig. 4a and b. The compositions at and above x = 0.23 do not show hysteresis but only frequency dispersion, see
Fig. 3c–f and Fig. 4c and d. It should be noted that out temperature measurement is in the intervals of 0.5 K with
temperature stability better than 0.02 K. Hence from our data it is possible to detect the frequency dispersion of even
<1 K, which we could not observe for the compositions with x < 0.23, Fig. 3a and b.
Keeping in view the characteristic change in the e0 –T variation across x = 0.23 accompanied by discontinuous jump
in Tm and change in the slope of the Tm variation across this composition, it is natural to assign this as the composition
1766 S. Anwar et al. / Materials Research Bulletin 43 (2008) 1761–1769

Fig. 4. Enlarged views of cooling and heating cycles of e0 –T variation at 1 kHz of (a and b) BHT20 and BHT22, and (c and d) BHT23 and BHT30
ceramics. The clear presence of hysteresis for BHT20 and BHT22 and no hysteresis for BHT23 and BHT30 can be seen.

boundary separating the normal and relaxor ferroelectric states [9]. In order to further confirm the relaxor behavior,
quantitative characterizations as described in the following have also been carried out.

3.3. Characterization of e0 –T behavior on high temperature side

Normal ferroelectrics follow Curie–Wiess behavior in the paraelectric state, i.e. above transition temperature Tc.
The Curie–Weiss law is described as
ðT  T c Þ
1=e0 ¼ ðT > T c Þ (1)
C

Fig. 5. Plot showing permittivity (e0 ) maxima temperature (Tc or Tm) dependence on Hf concentration, as derived from e0 –T measurement done at
1 kHz. A discontinuous jump in Tm at x = 0.23 followed by a change in the slope can be seen.
S. Anwar et al. / Materials Research Bulletin 43 (2008) 1761–1769 1767

where Tc is the Curie temperature and C is the Curie–Weiss constant. A deviation from Curie–Weiss law starts at a
temperature Tdev. The occurrence of the Curie–Weiss behavior at temperature much higher than Tc, suggests that the
behavior is not like a normal ferroelectrics. Such a behavior has generally been attributed to a modified ferroelectrics,
i.e. relaxors. A parameter DTm, which is often used to show the degree of deviation from the Curie–Weiss law [8], is
defined as
DT m ¼ T dev  T m (2)
The Tdev and DTm as determined according to Eq. (2) using 1 kHz data are presented in Table 1. For relaxors, a
modified Curie–Weiss law has been proposed by Uchino and Nomura [16], to describe the diffuseness of the phase
transition. This is defined as in the following:

1 1 ðT  Te0m Þg
 ¼ (3)
e0 e0m C1
where g and C1 are the modified constants, with 1  g  2. The exponent g defines the character of the phase
transition. For the limiting cases, i.e. for g = 1 and g = 2, the expression (3) reduces to Curie–Weiss law, for normal
ferroelectrics and quadratic dependent, for ideal relaxors respectively [7,12,17]. Thus the value of g can also
characterize the relaxor behavior. The value of the exponent g also quantifies the degree of the diffuseness of the phase
transition. It can be derived from least square fit of the logð1=e0  1=e0m Þversus logðT  Te0m Þ plot, as shown in Fig. 6.
We obtained the value of the parameter g to be 1.86 (BHT23) and 1.88 (BHT30), which are close to 2, suggesting the
occurrence of relaxor behavior [15]. The diffuseness of the phase transition can also be described empirically by a
parameter DTdif, defined [14,15] as in the following:
DT dif ¼ T 0:9e0 mð100 HzÞ  T e0 mð100 HzÞ (4)

here T 0:9e0 mð100 HzÞ is the temperature corresponding to 90% of the maximum dielectric constant (e0m ) at 100 Hz towards
the high temperature side and T e0 mð100HzÞ is the temperature corresponding to the maxima in the e0 –T curve at 100 Hz.
The diffuseness DTdif obtained from the available data is also presented in Table 1. Yet another parameter, which is
used to characterize the degree of relaxation behavior in the frequency range of 100 Hz to 100 kHz, is described [17] as
DT relax ¼ T e0 mð100 kHzÞ  T e0 mð100 HzÞ (5)

The value of DTrelax, as determined from the permittivity data, is presented in Table 1. The above characterization done
on the basis of Curie–Weiss law and the value of empirical parameters like DTm, g, DTdif and DTrelax suggest that the
permittivities of BHT20 and BHT22 corresponds to the normal ferroelectrics and that of BHT23 and BHT30 to
ferroelectric relaxors. For composition x  0.23, Curie–Weiss law follows at temperatures much higher than Tm. Thus

Fig. 6. logð1=e0  1=e0m Þversus logðT  Te0m Þ plots for Ba(Ti0.77Hf0.23)O3 and Ba(Ti0.70Hf0.30)O3 ceramics at 10 kHz. Symbols represent experi-
mental data and the solid lines show fits to the data.
1768 S. Anwar et al. / Materials Research Bulletin 43 (2008) 1761–1769

Fig. 7. Frequency dependence of Tm for Ba(Ti0.77Hf0.23)O3 and Ba(Ti0.70Hf0.30)O3 ceramics. The symbols and solid line show respectively the data
points and its fits to the Vogel–Fulcher relation.

large deviation from the Curie–Weiss behavior, large relaxation temperature DTrelax, and g ffi 2, suggest that
compositions above x  0.23 are relaxors.
The frequency dependence of Tm is shown in Fig. 7 as ln f versus 1000/Tm plot. This frequency dependence of Tm is
empirically evaluated using Vogel–Fulcher relation as
 
Ea
f ¼ f o exp  (6)
kB ðT m  T f Þ

here Ea is the activation energy, Tf the freezing temperature of polarization-fluctuation, and f o is the pre-exponential
factor. The values of Ea, Tf and f o for BHT23 and BHT30 are given in Table 2. Similar values have been reported for
BaTi0.75Zr0.25O3 also [17].
As already proposed by various authors [3–5,13], the dipoles giving rise to ferroelectric ordering in the perovskites,
are constituted of Ti4+ displacement in the equatorial plane of the O2 cage. In pure or weakly substituted BaTiO3 the
strongly correlated and highly cooperative interactions among the Ti–O dipoles lead to long-range ferroelectric order,
but in relaxors the ordering gets limited to the formation of polar nano-clusters. The relaxor ferroelectric behavior in
BHT23 and other higher Hf compositions of BHT appears due to the existence of ‘‘nano polar-clusters’’ [7,17,18]. This
may be due to substitution of Ti4+ ions by Hf4+, which has a larger ionic radius of 0.71 Å than that of Ti4+ (0.605 Å)
[19]. When Hf replaces Ti, the crystal structure and the ferroelectricity of BaTiO3 gets modified. The replacement of
Ti4+ by Zr4+ ions in ferroelectric BaTiO3 is well studied and understood [5,17]. The increasing substitution decreases
the Curie temperature, which is related to the ionic radius [7,18]. Similar behavior is expected for BHT ceramics also.
It has been suggested that polar nanodomains are responsible for the observed relaxation behavior in PMN, PLZT and
other systems also [3,13,20].
The substitution of Hf at Ti site tends to elongate the distance between off center Ti–O dipoles and thus weakening
the correlation between them. The mismatch in the size of Ti and Hf ions causes substitutional distortion of the oxygen
octahedra, giving rise to the local electric and strain fields. These local fields are usually random in nature and tend to
break the long-range ferroelectric domains into nano polar-clusters, which stabilize at lower temperatures [5,12,18].
Thus the competition between long-range ordering, owing to strong correlation between the off centered Ti dipoles,
and the random fields induced by Hf doping, induces relaxor behavior.

Table 2
Table showing parameters like Ea (activation energy), Tf (freezing temperature) and fo (pre-exponential factor) obtained from the Vogel Fulcher
fitting corresponding to the BHT23 and BHT30 ceramics
Compositions Ea (eV) Tf (K) fo (Hz)
BHT23 0.087 119 1.55  1012
BHT30 0.14 78.5 1.53  1013
S. Anwar et al. / Materials Research Bulletin 43 (2008) 1761–1769 1769

4. Conclusions

Based on the above described results and discussions on the structural and dielectric studies on Ba(Ti1xHfx)O3
ceramics with close Hf compositions between 0.20  x  0.30, it can be concluded that although the tetragonal to
cubic phase transformation occurs at x  0.10 as shown in our previous work, the normal to relaxor ferroelectric
crossover occurs somewhere in between 0.22 < x < 0.23 of Hf concentration. The temperature dependent dielectric
studies at various frequencies show that the permittivity maxima are diffusive and dispersive with respect to the
frequency, which are the characteristics of relaxors. The quantitative evaluation of other empirical parameters like
DTm, g, DTdif and DTrelax are also in accordance with the dispersive phase transition, characterizing the relaxor
behavior.

References

[1] B. Jaffe, W.R. Cook, H. Jaffe, Piezoelectric Ceramics, Academic, London, 1971.
[2] Y. Sakabe, T. Takagi, K. Wakino, D.M. Smith, in: J.B. Blum, W.R. Cannon (Eds.), Advances in ceramics, vol. 19, Am. Ceram. Soc., Inc., Ohio,
1986, p. 103.
[3] L.E. Cross, Ferroelectric 76 (1987) 241; 151 (1994) 305.
[4] J.N. Lin, T.B. Wu, J Appl. Phys. 68 (1990) 985.
[5] J. Ravez, A. Simon, Eur. J. Solid State Inorg. Chem. 37 (1997) 1199.
[6] V. Tura, L. Mitoseriu, Euro. phys. Lett. 50 (6) (2000) 810.
[7] S. Anwar, P.R. Sagdeo, N.P. Lalla, J. Phys. Condens. Mat. 18 (2006) 3455.
[8] S. Anwar, P.R. Sagdeo, N.P. Lalla, Solid State Commun. 138 (2006) 331.
[9] J. Ravez, A. Simon, J. Solid State chem. 162 (2001) 260.
[10] Rodriguez-Carvajal, J. Physica B 192 (1993) 55–69.
[11] G.A. Samara, J. Phys. Condens. Mat. 15 (2003) R367.
[12] P. Victor, R. Ranjith, S.B. Krupanidhi, J. Appl. Phy. 94 (12) (2003) 7702.
[13] Z.Y. Cheng, R.S. Katiyar, X. Yao, A. Guo, Phys. Rev. B 55 (1997) 8165.
[14] A. Simon, J. Ravez, M. Maglione, Solid State Sci. 7 (2005) 925.
[15] N. Abdelmoula, H. Chaabane, H. Khemakhem, R. Vonder Mühll, A. Simon, Solid State Sci. 8 (2006) 880.
[16] K. Uchino, S. Nomura, Integr. Ferroelectr. 44 (1982) 55.
[17] Y. Zhi, A. Chen, G. Ruyan, A.S. Bhalla, J. Appl. Phys. 92 (2002) 2655.
[18] A. Chen, J. Zhi, Y. Zhi, J. Phys. Condens. Mat. 14 (2002) 8901.
[19] R.D. Shannon, Acta Crystallogr. A 32 (1976) 751–767.
[20] D. Viehland, S.J. Jang, L.E. Cross, M. Wuttig, Phys. Rev. B 46 (1992) 8003.

Vous aimerez peut-être aussi