J O U R N A L O F M AT E R I A L S S C IE N C E: M ATE R IA LS I N EL E C T RO N I C S 1 1 (2 00 0 ) 64 5± 65 1
Analysis of the dielectric characteristics for
polycrystalline Ba
0:65
Sr
0:35
TiO
3
(II) ± d.c. ®eld
dependence with a modi®ed bias equation
JIH-WEI LIOU, BI-SHIOU CHIOU
Department of Electronics Engineering & Institute of Electronics,
National Chiao Tung University, Hsinchu 300, Taiwan
E-mail: bschiou@cc.nctu.edu.tw
The effect of direct-current (d.c.) biasing on the dielectric properties of Ba
0:65Sr
0:35TiO
3(BST)
with various grain sizes (0.87 mm to 5.43 mm) have been studied in the paraelectric state.
The leakage current of BST samples is proportional to the biasing ®eld. Specimens with
larger grains exhibit higher leakage current under biasing. The bias equation used in the
literature to predict the d.c. ®eld dependence of permittivity is modi®ed by introducing a bias
exponent parameter b. This modi®ed bias equation gives a better ®t to the experimental
results both in this study and in the literature. Similar trends of a (representing the
distribution of relaxation times), b, and the variance of lattice distortion with grain sizes
suggests that b is a material parameter. It is concluded that the temperature dependence, the
frequency response and d.c. ®eld effect for the dielectric characteristics of Ba
0:65Sr
0:35TiO
3are
all suppressed by the transition broadening.
# 2000 Kluwer Academic Publishers
1. Introduction
Ferroelectric barium strontium titanate (BST) is of great interest for many practical applications. The dielectric characteristics of polycrystalline BST depend strongly both on its microstructure and the environment during usage. The effects of d.c. ®eld on the dielectric behavior of BST have been reported [1, 2] and applied in various devices such as tunable capacitors [3] and microwave phase shifters [4, 5]. However, only a few studies [3, 6, 7] have been conducted on the d.c. ®eld related properties of paraelectric BST, especially the BST system with various grain sizes. The non-linearity of dielectric permittivity with d.c. biasing ®eld in the paraelectric state is believed to originate from the anharmonic interaction of Ti ions in the perovskite cubic structure. Johnson [2] started from Devonshire's theory and proposed a bias equation to correlate the dielectric constant and d.c. ®eld. Outzourhit et al. [3] ®tted this phenomenological equation to their experimental data and argued that deviations from the theoretical prediction were attributed to the dielectric dispersion caused by non-uniform grain sizes. In our previous work [7], it was observed that the d.c. effect is suppressed for BST with ®ner grains due to the smaller dielectric constant caused by transition broadening.
In this study, the d.c. ®eld dependence of dielectric constant and dielectric loss for Ba0:65Sr0:35TiO3 with various grain sizes are investigated in the paraelectric
state. Dopants MgO and MnO2 are used to obtain
uniform grained, low-loss dielectrics [8, 9]. Samples with various grain sizes are obtained by varying sintering temperatures and times. The purpose is to analyze the
frequency response and d.c. effect for Ba0:65Sr0:35TiO3 and to ®nd the correlation between these characteristics and the grain size. The frequency response, reported previously [10], was analyzed with two power laws which are obtained from the Cole±Cole expression at low frequencies. The d.c. ®eld dependence of the dielectric permittivity is analyzed with a modi®ed bias equation which is proposed in this study.
2. Experimental procedures
The Ba0:65Sr0:35TiO3 polycrystalline samples with var-ious grain sizes were prepared by the conventional solid state reaction method with commercial powders of BaCo3, SrCo3, TiO2, MnO2, and MgO. Powder mixtures containing additives of 1.0 mol % MgO and 0.05 mol % MnO2were calcined at 1100C for 2 h in air and pressed
to form disk-shaped samples at 180 MPa. Samples with various grain sizes were prepared with a fast-®ring, two-stage, sintering pro®le described previously [7]. Phase identi®cation and average grain size were analyzed by powder X-ray diffraction and scanning electron micro-scopy (SEM) examination, respectively. Lattice distortion in polycrystalline BST is calculated by the X-ray diffraction peak broadening of a single peak pro®le. For dielectric characteristic measurement, the bulk sample was polished to 0.4 mm in thickness before making electrodes. The current±voltage curve was obtained with an HP4140 pA meter (Hewlett Packard). The capacitance and tangent loss were measured with an HP4275A LCR meter (Hewlett Packard). A 200 V d.c.
measuring frequency range in this study is from 10 kHz to 4 MHz.
3. Results and discussion
When the electrical signal passes through dielectrics, the transmitted information is deteriorated by leakage currents, especially when a d.c. ®eld is applied. Therefore, it is crucial to understand the d.c. ®eld dependence of leakage current. The current densities for each samples at various applied d.c. ®elds are shown in Fig. 1. A larger leakage current for Ba0:65Sr0:35TiO3 is observed for samples with larger grams. Dopants MgO and MnO2used in this study are known to serve as grain growth inhibitors barium titanate because they segregate preferentially at grain surfaces. However, there is a small fraction of these dopants which may dissolve in the grains and become p-type acceptors [11, 12]. Lewis et al. [11] calculated the surface defect energies for BaTiO3
with various dopants, such as Mg and Mn. They reported some accompanied oxygen defects with the acceptors substituting on Ti sites. Recently, Langhammer et al. [13] found that the defect concentration in Mn-doped BaTiO3 increased as more Mn was added. In this study, samples with larger grains were obtained with longer sintering time or higher sintering temperature, the acceptor-type dopants were believed to dissolve more into the grains, thus more oxygen defects were produced due to valence charge compensation. The I±V characteristics for barium titanate have been explained by the space charge limited currents (SCLC) theory [14] for polycrystalline insula-tors. This states that strongly defective surfaces of the crystalline grains provide a source of deep trap states in insulators where the trapped charge carriers are excited by the applied voltage. Therefore, Ba0:65Sr0:35TiO3with larger grains has more defects and consequently yield larger leakage current, as shown in Fig. 1. The SCLC theory suggests that below an onset voltage, an ohmic current ¯ows. At higher voltages, the current transfers to a rapidly increasing behavior of the form of I ÿ Vawhere
a is in the range of 1.2 to 2.0 [14]. It is seen in Fig. 1, that the current density is linearly dependent on the applied ®eld. The d.c. resistivities are then calculated and summarized in Table I. Resistivity decreases with the increase of grain size due to the increase of defects.
Considering the frequency response of an electrically tunable capacitor, an equivalent circuit model containing a d.c. resistor and a voltage-variable capacitor in parallel combination can be applied adequately. Both the d.c.
resistance as well as the capacitance are d.c. ®eld dependent under d.c. biasing.
R V V=I V 1
C V C0e0 V ÿ je00 V 2
where C0is the geometrical capacitance of the capacitor
and the relative permittivity e V e0 V ÿ je00 V has
its real and imaginary parts which are both d.c. ®eld dependent. The current±voltage characteristics I V for Ba0:65Sr0:35TiO3 with various grain sizes exhibit ohmic behavior, as illustrated in Fig. 1. Assuming that the onset of I±V behavior-transformation (SCLC theory) is not reached by the d.c. ®eld applied when a.c. measurements are being carried out, the resistance of these samples is then constant with ®eld. The real part of the relative permittivity can be calculated by the measured capaci-tance. As for the measured dielectric loss tand, it contains two terms, one for the imaginary permittivity, the other for loss due to the leakage current. Therefore at some speci®c measuring frequency o, the voltage variable loss tangent can be expressed as:
tan d V ee000 V VoRe0 VC1
0 3
The ®rst term on the right hand side of Equation 3 is the intrinsic loss of the dielectrics, and the second term is the
Figure 1 Leakage current density as a function of d.c. ®eld for Ba0:65Sr0:35TiO3with various grain sizes.
T A B L E I Dielectric constant, dielectric loss, and resistivity for MgO and MnO2doped Ba0:65Sr0:35TiO3with various average grain sizes* Specimen (grain size (mm)) Dielectric constant Dielectric loss d.c. resistivity (GO cm)
10 kHz 4 MHz 10 kHz 4 kHz L1(0.87) 3783 3251 0.0115 0.1241 2477 L2(1.72) 5269 4801 0.0132 0.1200 1376 H1(2.16) 6959 5557 0.0288 0.2063 907.2 H4(3.74) 9808 7915 0.0134 0.2992 597.5 H8(5.43) 9947 8233 0.0168 0.3733 444.7
conduction loss. With the resistivity and dielectric constant listed in Table I, the conduction losses for the Ba0:65Sr0:35TiO3with various grain sizes are estimated to be less than 10ÿ7for all frequency and ®eld ranges in this
study. Hence, the conduction loss term can be neglected so that the imaginary permittivity e00is approximately the
product of dielectric constant and tangent loss.
A bias equation on the basis of Devonshire's theory to correlate the relative permittivity e0 and applying d.c.
®eld E is proposed phenomenologically [2]: e0 E e0 0
1 AE21=3 4
where e0 0 is the initial permittivity under zero d.c.
biasing. The ®eld coef®cient A represents the strength of the d.c. ®eld effect. The value of A equals ae03 where a
is the anharmonic coef®cient. As the anharmonic interaction becomes stronger, a larger value of A yields a larger value of ®eld coef®cient and thus the d.c. ®eld is more effective to reduce the permittivity. Johnson claimed that some deviations due to poly-crystallinity can be absorbed in the ®eld coef®cient A. Therefore Equation 4 can be applied with validity to the polycrystalline barium strontium titanate system. As for the imaginary part of dielectric permittivity, Johnson proposed another bias equation based on a RC series circuit:
Figure 2 Relative permittivity e00as a function of d.c. biasing ®eld for Ba
0:65Sr0:35TiO3with various grain sizes (a) L1, (b) H1 and (c) H8. Curves are ®tted with the bias equation and the modi®ed bias equation. Note that the modi®ed equation gives better ®ts than the original one.
Recently, Outzourhit et al. [3] investigated the d.c. ®eld dependence of the dielectric constant of Ba0:1Sr0:9TiO3 ceramics prepared by the wet method
and the solid state method. They found that good ®tting of this bias equation to the experimental results can only be made separately in low and high ®eld ranges. For the BST system in this study, it is found that the suppression of the d.c. ®eld effect by the grain size is remarkable [10]. The validity of the bias equation was questioned when there is a transition broadening. In Outzourhit's work, the bias equation did not ®t the experimental data, presumably due to the diffuseness of the transition of Ba0:1Sr0:9TiO3with grain sizes of about 1 mm and 10 mm. Attempts to employ the bias equation for the dielectric characteristics of Ba0:65Sr0:35TiO3in this study (shown in
Fig. 1 of Liou and Chiou [10]), reveal some controversies such as: (a) the bias equation does not predict well the dielectric behavior under d.c. biasing though the best ®tted ®eld coef®cient A is chosen, (b) if the ®eld coef®cient is calculated by Equation 4, it is no longer a constant with ®eld, and (c) the ®eld coef®cient is frequency dependent. Since the bias equation was explained adequately for single crystal or ceramics with high crystallization [2], it is argued that the controversies were attributed to the various microstruc-tures of the polycrystalline ceramics, especially the grain size. It was reported previously that there are various distributions of Curie temperatures occurring in the Ba0:65Sr0:35TiO3 with various grain sizes [10]. This implies that each grain has its own initial permittivity e0 0 which is different from each other. Thus, the
dielectric behavior under d.c. ®eld changes profoundly due to the grain size effect. To realize the d.c effect by a distribution of e0 0 in polycrystallites, the denominator
of the bias equation is modi®ed by introducing a bias exponent b:
e0 E e0 0
1 BE21ÿb3 6
The bias exponent b ranges from 0 to 1 and represents the order of suppression of the d.c. ®eld effect. The related ®eld coef®cient B is supposed to be ®eld independent and represents the strength of the d.c. ®eld effect. The ®ttings of bias equation and modi®ed bias equation for samples L1, H1, and H4 at 10 kHz, 1 MHz and 4 MHz are shown in Fig. 2a to c for comparison. For the ®tting of the modi®ed bias equation, a two-parameter least square method was employed. The initial values of the two parameters, B and b, were calculated by solving two equations
constructed by two experimental data, e E1 and
e E2. The optimum b was obtained by the least squares
method with the calculated B. The optimum B was then obtained by the least squares method with the optimum b. As shown in Fig 2a to c, the magnitude predicted by the bias equation in all samples is too high in the low ®eld region and too low in the high ®eld region. This is why Outzourhit et al. ®tted their results with the bias equation separately in high ®eld and low ®eld regions. Meanwhile, less deviation of the modi®ed bias equation
( 5 0.6%) is observed at d.c. ®elds up to 5000 V cmÿ1.
With the ®ttings to the dielectric characteristics shown in Fig. 1 of [10], the bias exponents b and ®eld coef®cients B for Ba0:65Sr0:35TiO3 with various grain
sizes at different frequencies are shown in Fig. 3a and b, respectively. No apparent frequency dependence of b and B suggest that these two parameters are material parameters. The increase of b and the decrease of B with the decrease of grain size is clearly seen. Larger b and smaller B make the denominator in the modi®ed bias equation (Equation 6) closer to one, thus the reduction of dielectric constant by the d.c. ®eld is suppressed. This suppression is presumably due to the broad distribution of Curie temperatures in polycrystalline materials.
Figure 3 (a) Bias exponent b and (b) ®eld coef®cient B of permittivity e0 as a function of frequency for Ba
0:65Sr0:35TiO3 with various grain sizes.
Equation 5 is modi®ed in a similar way as: e00 E e00 0
1 BE22 1ÿb3
7 Fig. 4 shows the ®ttings with the bias equation (Equation 5) and the modi®ed bias equation (Equation 7). The modi®ed equation gives a better ®t, as in the case of e0 E. As reported previously, dopants contribute to the
split of dipolar relaxation, and the losses at frequencies below 100 kHz are higher than those predicted by the approximated power law [10]. It is noted in Fig. 4 that the behavior of the dielectric loss e00 at 10 kHz is also in
agreement with the prediction of the modi®ed bias equation. As shown in Fig. 5, the bias exponent b
increases and the ®eld coef®cient B decreases with the increase of grain size. Similar trends of e0 E and e00 E
are observed, b ranges from 0.27 to 0.70 for e0and from
0.29 to 0.70 for e00. Although the bias ®eld coef®cients
for both e0and e00are also comparable, more ¯uctuations
of b and B for e00at different frequencies are observed in
Fig. 5. The calculation of the imaginary permittivity e00is
based on Equation 3 neglecting the second term as the leakage current is very small. There are differences between the b and B in Fig. 3 and those in Fig. 5. It is believed that ®tting data with Equation 6 (Fig. 3) gives more accurate b and B values, since the contribution of conduction loss and dopant-related relaxation to e00 are
neglected in Equation 7 (Fig. 5).
Using the modi®ed bias equation proposed in this
Figure 4 Relative permittivity e00as a function of d.c. biasing ®eld for Ba
0:65Sr0:35TiO3with various grain sizes (a) L1, (b) H1 and (c) H8. Curves are ®tted with the bias equation and the modi®ed bias equation. Note that the modi®ed equation gives better ®ts than the original one.
study, the ®ttings of Outzourhit's results are shown in Fig. 6. The bias exponent b is 0.42 for Ba0:1Sr0:9TiO3 prepared by the wet method and is 0.28 for that prepared by the solid state method (ss method). The grain sizes of Outzourhit's samples are about 1 mm (wet method) and 10 mm (ss method), respectively. Outzourhit also showed that samples prepared by the wet method had broader Curie peaks than those prepared by the solid state method. The larger value of bias exponent is explained by the broader transition distribution in the sample prepared by the wet method.
One may argue that an equation with two ®tting parameters always gives better ®ts than one with only one ®tting parameter. This is true with regard to the mathematics. The bias exponent b varies, the trend of modi®ed bias equation intrinsically, i.e., different b values determine different function forms of Equations 6 and 7. Together with the frequency independence of b, as shown in Figs 3a and 5a, these features suggest that b is a material parameter. Perhaps the bias coef®cient B plays a tuning role during the ®tting procedure. Nevertheless, once these two parameters are
Figure 6 Cure ®tting with the modi®ed bias equation for the permittivity e0. Ba
0:1Sr0:9TiO3prepared by wet and solid state methods (data obtained from Outzourhit et al. [15]).
Figure 7 Grain size dependence of variance of lattice distortion and the suppressing parameters a, b for Ba0:65Sr0:35TiO3 with various grain sizes.
properly chosen, the dielectric behavior can be predicted well with the initial value e0 0 or e00 0
known. The modi®ed bias equation is then useful for applications. Furthermore, from the same trends of the tilt parameter a [10], critical exponent g [10] and bias exponent b for Ba0:65Sr0:35TiO3 with various grain sizes, the variance of b is also believed to be due to the ferroelectric to paraelectric transition broadening. These parameters are called ``suppression parameters'' to represent the suppression order of temperature depen-dence, frequency response and d.c. ®eld effect. Fig. 7 exhibits the grain size dependence of these parameters. In this ®gure the lattice distortion is obtained from the X-ray analysis at room temperature, g is obtained at 10 kHz, b is obtained at 10C and 10 kHz, and a is
obtained at 10C. All these four parameters decrease as
grain size increases.
4. Conclusions
1. A decrease of the dielectric constant for Ba0:65Sr0:35TiO3 with the increase of d.c. biasing ®eld is observed. A modi®ed bias equation with a bias exponent b, is proposed to predict the dielectric behavior under d.c. biasing. It is found that the ®ts are better than the original bias equation for Ba0:65Sr0:35TiO3 with
various grain sizes. Once the parameters in the modi®ed bias equation are properly chosen, the dielectric behavior under the d.c. ®eld can be predicted accurately with the initial value e0 0 or e00 0 known.
2. Frequency independence of the bias exponent b suggests that it is a material parameter. The increase of b with the decrease of grain size indicates that the suppression of the d.c. ®eld effect occurs in polycrystal-line Ba0:65Sr0:35TiO3.
3. The grain size dependence of suppression para-meters a, b and g obtained from the dielectric characteristics of frequency response, d.c. ®eld effect
and temperature dependence, show the same trend with the variance of lattice distortion for polycrystalline Ba0:65Sr0:35TiO3.
Acknowledgments
This work is supported by the Chung-Shan Institute of Science and Technology and the National Science Council of Taiwan (contract numbers: NSC 86-2623-E-009-002 and NSC 86-2216-E-009-019).
References
1. H . D I A M A N D, J. Appl. Phys. 32(5) (1961) 909. 2. K . M . J O H N S O N, ibid. 33(9) (1962) 2826. 3. A . O U T Z O U R H I T,J . U . T R E F N Y,T. K I TO andB . YA R A R, J. Mater. Res. 10(6) (1995) 1411. 4. R . W. B A B B I T T,T. E . KO S C I CAandW. C . D R AC H, Microwave Journal June (1992) 63. 5. F. S E L M I,R . H U G H E S,V. K . VA R A D A NandV. V. VA R A D A N, Proc. SPIE 1916 (1993) 180.6. J . W. L I O UandB . S . C H I O U, J. Am. Ceram. Soc. 80(12) (1997) 3093.
7. J . W. L I O UandB . S . C H I O U, Jpn. J. Appl. Phys. 36(7A) (1997) 4359.
8. F. B AT T LO, E . D U V E R G E R, J . C . J U L E S, J . C . N I E P C E,
B . J A N N OTandM . M A G L I O N E, Ferroelectrics 109 (1990) 113.
9. B . S . C H I O U,C . M . KO HandJ . G . D U H, J. Mater. Sci. 22 (1987)
3893.
10. J . W. L I O UandB . S . C H I O U, submitted to J. Mater. Sci.: Mater.
Electron.
11. G . V. L E W I S,C . R . A . CAT LOWandR . E . W. CA S S E LTO N,
J. Am. Ceram. Soc. 68(10) (1985) 555.
12. Y. C . C H E N,G . M . LO,C . R . S H I H,L . W U,M . H . C H E Nand
K . C . H U A N G, Jpn. J. Appl. Phys. 33 (1994) 1412.
13. H . T. L A N G H A M M E R, T. M UÈ L L E R, A . PO L I T Y, K . H . F E L G N E RandH . P. A B I C H T, Mater. Lett. 26 (1996) 205. 14. W. O S A K andK . T K AC Z. J. Phys. D: Appl. Phys. 22 (1989)
1746.
Received 23 August 1999 and accepted 17 May 2000