Sie sind auf Seite 1von 9

Domain wall motion effect on the anelastic behavior in lead zirconate titanate piezoelectric ceramics

El Mostafa Bourim, Hidehiko Tanaka, Maurice Gabbay, Gilbert Fantozzi, and Bo Lin Cheng Citation: J. Appl. Phys. 91, 6662 (2002); doi: 10.1063/1.1469201 View online: http://dx.doi.org/10.1063/1.1469201 View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v91/i10 Published by the American Institute of Physics.

Related Articles
Effect of annealing under O2 and H2 on the piezoelectric parameters of the Ca12Al14O33 single crystals J. Appl. Phys. 111, 054107 (2012) Piezoelectric properties of epitaxial Pb(Zr0.525, Ti0.475)O3 films on amorphous magnetic metal substrates J. Appl. Phys. 111, 07D916 (2012) Direct piezoelectric properties of (100) and (111) BiFeO3 epitaxial thin films Appl. Phys. Lett. 100, 102901 (2012) Piezoelectric 36-shear mode for [011] poled 24%Pb(In1/2Nb1/2)O3-Pb(Mg1/3Nb2/3)O3-PbTiO3 ferroelectric crystal J. Appl. Phys. 111, 034107 (2012) Highly piezoelectric biocompatible and soft composite fibers Appl. Phys. Lett. 100, 063901 (2012)

Additional information on J. Appl. Phys.


Journal Homepage: http://jap.aip.org/ Journal Information: http://jap.aip.org/about/about_the_journal Top downloads: http://jap.aip.org/features/most_downloaded Information for Authors: http://jap.aip.org/authors

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

JOURNAL OF APPLIED PHYSICS

VOLUME 91, NUMBER 10

15 MAY 2002

Domain wall motion effect on the anelastic behavior in lead zirconate titanate piezoelectric ceramics
El Mostafa Bourim and Hidehiko Tanakaa)
National Institute for Materials Science (NIMS), Namiki 1-1, Tsukuba, Ibaraki 305-0044, Japan

Maurice Gabbay and Gilbert Fantozzi


Institut National des Sciences Appliquees (INSA) de Lyon - GEMPPM, UMR 5510 CNRS, Batiment B. Pascal, 69621 Villeurbanne Cedex, France

Bo Lin Cheng
IRC in Materials, University of Birmingham, Edgbaston, Birmingham B15 2TT, United Kingdom

Received 11 December 2000; accepted for publication 22 February 2002 Three undoped lead zirconate titanate PZT ceramics were prepared with compositions close to the morphotropic phase boundary: Pb Zr0.50Ti0.50 O3 , Pb Zr0.52Ti0.48 O3 , and Pb Zr0.54Ti0.46 O3 . Internal friction Q 1 and shear modulus G were measured versus temperature from 20 C to 500 C. Experiments were performed on an inverted torsional pendulum at low frequencies 0.1, 0.3, and 1 Hz . The ferroelectricparaelectric phase transition results in a peak P 1 of Q 1 correlated with a sharp minimum M 1 of G. Moreover the Q 1 (T) curves show two relaxation peaks called R1 and R2 respectively, correlated with two shear modulus anomalies called A1 and A2 on the G(T) curves. The main features of the transition P 1 peak are studied, they suggest that its behavior is similar to the internal friction peaks associated with martensitic transformation. The relaxation peak, R1 and R2 are both attributed to motion of domain walls DWs , and can be analyzed by thermal activated process described by Arrhenius law. The R2 peak is demonstrated to be due to the interaction of domain walls and oxygen vacancies because it depends on oxygen vacancy concentration and electrical polarization. However, the R1 peak is more complex; its height is found to be increased as stress amplitude and heating rate increase. It seems that the R1 peak is inuenced by three mechanisms: i relaxation due to DWpoint defects interaction, ii variation of domain wall density, and iii domain wall depinning from point defect clusters. 2002 American Institute of Physics. DOI: 10.1063/1.1469201

I. INTRODUCTION

II. MATERIALS AND EXPERIMENTAL PROCEDURE A. Studied chemical compositions

Lead zirconate titanate Pb Zr1 x Tix )O3 usually called PZT ceramics are used as the active material for various sensors and actuators. These devices require high electromechanical coupling constants as well as low dielectric and mechanical losses. The energy dissipation from dielectric and mechanical losses causes a change of physical properties in piezoelectric materials. It is important, therefore, to investigate the mechanisms that control these losses in order to reduce them. The losses are mainly associated with domain wall motion,1 but also with point defects.2 4 Postnikov et al.5 have shown that the peaks of mechanical losses observed in PZT ceramics are linked to the interaction between DWs and mobile point defects. In this paper, the anelastic behavior of PZT ceramics is studied through the analysis of internal friction peaks due on the one hand to the ferroelectricparaelectric phase transition, and on the other hand to the relaxation linked to the motion of DWs. This is in order to reveal the physical mechanisms responsible for mechanical losses in the ferroelectric PZT ceramics.
a

Electronic mail: tanaka.hidehiko@nims.go.jp 6662

Undoped PZT ceramics were prepared by conventional sintering solid solution from the starting powders of PbO, ZrO2 , and TiO2 of 99.9% pure reagent grade. Mixtures of desired molar ratio were chosen near the morphotropic phase boundary with the following Zr/Ti ratios: Pb Zr0.50 Ti0.50)O3 , Pb Zr0.52 Ti0.48 O3 , and Pb Zr0.54 Ti0.46 O3 : socalled PZT 50/50, PZT 52/48, and PZT 54/46, respectively. X-ray diffraction showed that the structures of the PZT 50/50 and the PZT 52/48 compositions were tetragonal, as identied by the associated (002) T /(200) T doublet lines Fig. 1 . In the PZT 54/46 composition, the x-ray diagram indicated the presence of two phases, a tetragonal one related to the (002) T /(200) T doublet lines, and a rhombohedral one related to the (200) R line Fig. 1 . The coexistence of this double structure is specic to the morphotropic phase boundaries.6,7 The plateau shape, which corresponds to the additional diffracted intensity between the doublet lines, is a ngerprint of the ferroelectric microstructure related to the 90 DWs.8 Thus, for two associated 90 domains indicated, for example, by DI and DII respectively , the corresponding x-ray diffraction is a doublet, in which the rst line would result from DI domain diffraction, the second line would result
2002 American Institute of Physics

0021-8979/2002/91(10)/6662/8/$19.00

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

J. Appl. Phys., Vol. 91, No. 10, 15 May 2002

Bourim et al.

6663

nealed 500 C in air for 1 h in order to remove residual stresses. An inverted torsion pendulum is used to measure the internal friction Q 1 and the shear modulus G as a function of temperature from 20 to 500 C.9 A sample was xed at its lower end to a xed clamp and at its higher end to a mobile clamp interdependent to a rigid transmission stem via which the sample was excited by torsional oscillations. In an anelastic solid, the application to the i t induces sample of a sinusoidal torsional stress (t) 0 e a sinusoidal strain with the same pulsation but with a i( t ) behind the stress: (t) . The lag phase lag 0 e angle is typical of an anelastic behavior of the material and allows to compute the mechanical loss Q 1 . The dynamic elastic shear modulus is given by the ratio between the applied stress and the measured strain; it is a complex number G * expressed by the following relation: G* i i G exp i G iG .

is calculated by the Thus, the internal friction Q 1 tan ratio of G /G , and the shear modulus G is estimated to be equal to G in the case of weak mechanical losses. Three sinusoidal vibration frequencies were used: 0.1, 0.3, and 1.0 Hz with a strain amplitude lower than 10 5 . To neglect the air damping to the oscillation movement of sample, all measurements were performed under vacuum ( 10 3 Torr .
III. RESULTS AND DISCUSSION

FIG. 1. X-ray diffraction 002/200 doublet lines evolution with phase structure.

Figure 2 shows Q 1 (T) and G(T) curves recorded during the rst heating run for the three different vibration frequencies 0.1, 0.3, and 1 Hz on the PZT 52/48 and PZT 54/46 ceramics. The following anelastic events can be identied: i The P 1 peak of Q 1 correlated with a sharp minimum M 1 of shear modulus G are due to the tetragonalcubic phase transition. ii The P 2 peak of Q 1 and the M 2 minimum of G are associated with the rhombohedraltetragonal phase transition in the case of the PZT 52/48 and PZT 50/50 ceramics, the rhombohedraltetragonal phase transition characteristics are not visible because their temperatures are about 69 and 139 C, respectively .911 iii Two relaxation peaks R1 and R2 correlated with two shear modulus anomalies A1 and A2 , respectively, on the G(T) curves. Similar results were observed for the PZT 50/50 composition ceramic. The analysis of the three main peaks P 1 , R2 , and R1 is presented below.
A. P 1 peak analysis

from DII domain diffraction, and the plateau would arise from the connectiona zone frontier existing between the two domains. Consequently, the 90 domain wall corresponds to a distorted zone with continuous variation of the interplanar distance from DI domain to its associated DII domain.
B. Internal friction and shear modulus measurements

The specimens were rectangular bars dimensions: 40 5 1 mm3 . After machining, the specimens were an-

The height of the P 1 peak depends on the following parameters: vibration frequency f, temperature rate T Fig. 7 Fig. 8 . All these characteristics are and stress amplitude very similar to those of materials that undergo martensitic transformations, suggesting that its behavior is similar to the internal friction peaks associated with martensitic transformation. Among the various theoretical models reviewed by Van Humbeeck,12 there is an interesting qualitative agreement with the phenomenological models presented by Belko et al.,13 Delorme,14 and De Jonghe et al.15

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

6664

J. Appl. Phys., Vol. 91, No. 10, 15 May 2002

Bourim et al.

FIG. 2. Curves Q

(T) and G(T) curves recorded at low frequencies during the rst heating run.

Internal friction Q by the relation Q


1

calculated by Belko et al. is given

G a2 m kT

where G is the shear modulus, is the volume of critical nucleus, a is the non-elastic deformation at Curie tempera ture, m is the relative volume of the transformed phase per unit of time, is the angular frequency, T is the temperature, and k is Boltzmanns constant. However, this relation does not take into account the applied stress, which is also the case with Delormes model. De Jonghe et al. propose a model in which the amount of transformed material per unit of time takes into account not only the temperature rate T but also the applied stress , thus the amount of transformed phase per unit of time m f (T, ) is expressed according the following relation: dm dt m T T t m t .

Nevertheless, the applied stress only contributes to phase transformation over a critical stress C , and thus De Jonghe et al. propose the calculation of internal friction Q 1 using the following relation: Q
1

A . 2

m T1 T t f

4 3

m
0

. 1

C 0

With it, De Jonghe et al. have proposed various peak shapes for martensitic transformation according to the mobility of

the interfaces and the applied stress value. This model predicts a peak shape for the phase transition similar to the transition peak shape obtained by Gridnev et al.16 with niobium-doped PZT ceramics. Such a doping allows to suppress the relaxation peaks R1 and R2 and to make the low temperature side of the P 1 peak clearly observable. Thus, it is possible to improve our understanding of phase transition mechanisms. Moreover this also facilitates the decomposition of Q 1 (T) curves to elucidate the R1 peak. Our results show that the P 1 peak is more visible when the vibration frequency is lower Fig. 2 : the peak height 1 (Q max)P1 increases as the vibration frequency decreases Fig. 3 a , and the peak temperature is insensitive to the frequency and to the thermal cycle heating and subsequent cooling , which have an effect only on the peak height Fig. 3 a . The P 1 peak height increases with increasing tempera ture rate T Fig. 7 with a nonlinear evolution as shown in 1 Fig. 3 b from the plot Q max)P1 versus T / f , and the peak temperature remains constant. Furthermore the P 1 peak height is sensitive to the stress amplitude of measurement Fig. 8 : the increase is quasilinear Fig. 3 c . These results are roughly described by the preceding models, but for the recently observed nonlinear relations be tween Q 1 and the temperature rate T as well as Q 1 and 17 the vibration frequency, Zhang et al. suggested that the P 1 peak observed at the transition temperature arises from the motion of phase interfaces during the rst-order phase transition and obtained a relation which is well veried by the measurements. Similar relation was obtained by Wang

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

J. Appl. Phys., Vol. 91, No. 10, 15 May 2002

Bourim et al.

6665

FIG. 4. Arrhenius plots for the three different ceramic compositions PZT 50/50, PZT 52/48, and PZT 54/46.

equation and the relaxation time can be written as: 0 exp(H/kT), where H is the activation energy, 0 the preexponential factor, T is the absolute temperature, and k is Boltzmanns constant. For a Debye peak, the peak maximum occurs when 1, that is to say: ln
0

H kT P

0,

1 FIG. 3. Variation of P 1 peak height Q max)P1 as function of vibration fre quency f and 10/f a , T / f b , and stress amplitude oscillation c.

et al.18 by using models developed for rst-order transition. The effect of stress can be introduced by considering an equivalent depinning process.
B. R2 peak analysis

The R2 peak is controlled by a relaxation mechanism because it is frequency dependent. The temperature of the peak depends on the frequency according to an Arrhenius

2 f , with f the where T P is the peak temperature and vibration frequency. The Arrhenius plots of ln( ) versus 1/T P Fig. 4 lead to the following activation parameters: activation energy H (R2) 1 0.1 eV and pre-exponential factor 0(R2) 10 (13 1) s. The magnitude of the pre-exponential factor 0(R2) is coherent with a point defect relaxation, and the activation energy H (R2) of about 1 eV is the typical value for the migration enthalpy of oxygen vacancies in perovskite oxides.219 So, what is the physical mechanism controlling the relaxation process linked to the R2 peak? As observed in BaTiO3 ,20 the R2 peak could be linked to the domain wall. As proposed by Postnikov et al.,5 the R2 peak can be related to a relaxation mechanism which involves an interaction between the DWs and mobile point defects. Postnikov et al. have proposed a model of the relaxation in ferroelectric materials based on an interaction between mobile point defects and immobile DWs. In this model, under external mechanical stress, there is an increase in the number of bound electric charges on the DWs via the piezoeffect. Electrical compensation of these charges by the migration of mobile charged point defects present in the lattice leads to a change in the electric eld within the domain with time. Thus, by the inverse piezoeffect, a mechanical relaxation takes place. The Postnikov model predicts, for small concentration of 1 mobile point defects, that the relaxation peak height Q max and the relaxation time are given by the following relations:

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

6666

J. Appl. Phys., Vol. 91, No. 10, 15 May 2002

Bourim et al.

FIG. 5. Inuence of thermal treatment on the Q

(T) curve.

1 Q max

2 d 33 d 31 2 c 0 L 2 q 2
4 2 r 0 kTs

3a

L2
2

L2
2

D0

exp

H , kT

3b

where, c 0 is the equilibrium concentration of point defects, L is the 90-domain width, d i j is the piezoelectric constant, q is the point defect charge 1.6 10 19 C , r is the relative permittivity, 0 is the vacuum permittivity, s is the unrelaxed elastic compliance, k is Boltzmanns constant, H is the activation energy, and D is the diffusion coefcient of the point defects. In these Eqs. 3a and 3b the height of internal friction peak depends on point defect concentration c 0 and domain width L, and the relaxation time varies as L 2 . Figures 5 and 6, respectively, show the inuences of thermal treatment and electrical polarization on the R2 peak. High temperature vacuum annealing can introduce excess of oxygen vacancies, the R2 peak height is clearly increased by such thermal treatment, but an annealing in air at the same high temperature reduces it. And regarding the polarization,

which has a direct effect on the variation of domain size or domain wall density, the peak height is decreased by such poling under an electrical eld of 3 kV/mm at 130 C during 30 min. These results suggest that the R2 peak is related to a relaxation mechanism involving both, point defects oxygen vacancies and DWs. A good agreement was observed between the R2 peak features and Postnikovs model. The inuence of oxygen vacancy concentration and domain size variations on the R2 peak height veries the Eq. 3a . Furthermore, the measured activation energy corresponds to the activation energy for diffusion of oxygen vacancies. This hypothesis is conrmed by the results of Postnikov et al. and Gridnev et al., who observed that the R2 peak is suppressed by niobium doping which decreases the oxygen vacancy content as foreseen by Eq. 3a . In addition to the sensitivity of the R2 peak height to thermal treatment and effect of poling, we notice also a shift in its temperature position T P . Analysis of the peak shift from the thermal treatment results by the Arrhenius equation has showed that the activation energy H (R2) remains approximately about 1 eV, independent on the changes in oxygen vacancy concentration. This means it is subject to the same energy barrier for the relaxation process of the R2 peak, although the pre-exponential factor 0(R2) increases from 10 15 s after annealing in air up to 10 12 s after annealing under vacuum. So, according to the Eq. 3b relating the relaxation time to the domain size, we can say that the oxygen vacancy concentration has also an effect on the domain dimension, which inuences the peak temperature position. In the Eq. 3b the pre-exponential factor 0 is expressed by L 2 / 2 D 0 . Therefore, at peak temperature condition 1, with constant vibration frequency and constant activation energy as was checked on the two kinds of annealing, we nd that the peak temperature T P is proportional to preexponential factor 0 , and consequently to the domain width L, as shown in the following relationship: L2
2

D0

exp

H kTp

1.

FIG. 6. Inuence of electrical polarization on the Q

(T) curve.

From the above analysis, the hypothesis that the domain dimension is dependent to the oxygen vacancy concentration is feasible when we consider oxygen vacancies as an obstacle to the movement of DWs during their apparitions to form the 90 domain frontiers in the ferroelectric phase. So, at the paraelectric-ferroelectric phase transition during the subsequent cooling after air annealing at 600 C for 6 h, the weak oxygen vacancy concentration will not be high enough to block domain wall motion, and a ne domain microstructure will be established. And thus the small L size implies that the weak 0(R2) results in low T R2 . On the other hand, during subsequent cooling under vacuum annealing 600 C for 6 h, the high concentration of oxygen vacancies slows down domain wall motion, excess oxygen vacancies are segregated preferably at the DWs as observed directly in high resolution transmission electron microscopy TEM by Tan et al..21 This kind of pinning to the motion of DWs induces the formation of a large domain microstructure, and thus the large L

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

J. Appl. Phys., Vol. 91, No. 10, 15 May 2002

Bourim et al.

6667

size implies that high 0(R2) leads to high T R2 . This is precisely what we observed in the analysis of the R2 peak behavior with the thermal treatment effect. The scenario of an oxygen vacancy concentration effect on domain size is further supported by the electrical poling which has a direct effect on changes in domain size. It is known that polarization increases domain size. According to Postnikovs model, an increase in domain size will cause the peak to shift toward high temperatures with an increase in its level, but the poling effect has resulted in the opposite behavior for the R2 peak Fig. 6 . In fact, during the poling operation of a ferroelectric ceramic, the domains rst undergo a reduction in their size22 and, when the electric eld reaches a sufcient value, an increase in the domain size starts to take place. Considering the fact that the applied electric eld to our ceramic was weak compared to a coercitive eld, it is likely that only a decrease in domain size took place. Consequently, the decrease in domain width L involves a decrease in the relaxation time as foreseen by Eq. 3b , which makes the peak R2 move toward low temperatures, and also leads to a reduction in its height according to Eq. 3a . This is again precisely what we observed in the analysis of the R2 peak behavior for the polarization effect. In conclusion, our results suggest that the R2 peak relaxation could be related to the interaction between 90 DWs and mobile oxygen vacancies. Furthermore, oxygen vacancies have a direct inuence on the conguration of ferroelectric domains at time of their formation.
C. R1 peak analysis

FIG. 7. Inuence of temperature rate on Q

(T) curve.

2. Temperature rate T dependence

The R1 peak behavior is very complex because it is inuenced on both sides by the R2 and P 1 peaks as for its temperature is near the Curie temperature T C ). Nevertheless the following three main features were observed.
1. Frequency dependence

The R1 peak is controlled by a relaxation mechanism as the R2 peak. The activation parameters deduced from Arrhenius plots are H (R1) 1.8 0.2 eV and 0(R1) 10 (18 1) s for all the PZT 50/50, PZT 52/48, and PZT 54/46 ceramics. The activation energy H (R1) is much higher than one for the diffusion of oxygen vacancies. Its value lets us suppose that the relaxation process is due to an interaction of the DWs with the point defects with signicant diffusion activation energy. Thus, the R1 peak could be attributed to the interaction of 90 DWs with Ti or Zr vacancies. This hypothesis is conrmed by the decrease of the R1 peak due to a reduction of the Ti or Zr vacancies by introducing niobium oxide Nb2 O5 in PZT Nb replaces Ti or Zr atoms as observed by Postnikov et al. However, the pre-exponential factor 0(R1) of relaxation time is very short and it is difcult to give it a physical meaning. The shortness of 0(R1) could be due to the reduction in the domain wall inertia generated by the thinning of the wall thickness with the temperature. In general, the determined activation parameter values are only the apparent ones, and suggest that the R1 peak is related to a combination process. As shown below, the R1 peak is dependent on other parameters.

Figure 7 shows an increase of the R1 peak height as a function of T as for the P 1 peak. Such T effect on a relaxation peak associated with a very low pre-exponential factor 0 have been attributed to microstructure changes due to an evolution of the material.23 In fact, the appearance of the R1 peak near the ferroelectric-paraelectric phase transition ( P 1 peak suggests an instability in domain arrangement. This state is identied in TEM observations:9 the density of DWs increases with temperature, then domains disappear at Curie transition T C ; inasmuch as the domain interface energy evolves to approaching zero when the temperature tends toward T C . According to the model of Wang et al.24 based on the viscous motion of the DWs,25 the number of domains N is proportional to (T C -T) 1 , hence, N increases with temperature, which results in an increase of internal friction, and when the domain density reaches a critical value such that the interaction between walls tends to reduce their mobility, the internal friction decreases, thus the formation of the peak takes place. Thus, the R1 peak could be linked to domain wall motion controlled by the nucleation of new domains over a large temperature range before the Curie temperature.
3. Stress amplitude dependence

Figure 8 shows an increase in R1 peak height with the oscillating stress amplitude. This increase could be due to dragging-depinning process similar to the dislocation depinning from point defects. According to this mechanism, mechanical loss in the R1 peak could be activated by a dragging force produced by point defects opposing the domain wall motion. Thus, for a signicant stress or deformation amplitude, the displacement of the domain wall would also be signicant, and consequently a dependence of peak height on the applied stress amplitude would be obtained. Indeed, the stress amplitude dependence can be given by the classical Granato and Lucke model26 which is relative to the dislocation depinning and can be expressed by the following relationship:

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

6668

J. Appl. Phys., Vol. 91, No. 10, 15 May 2002

Bourim et al.

FIG. 8. Inuence of stress amplitude vibration on Q

(T) curve.

C1

exp

C2

where C 1 and C 2 are constants without dimensions, C 2 is proportional to the point defect concentration along the dislocation line, and is the deformation. The Granato and Lucke plot is expected to be a straight line and predicts that its slope C 2 is inversely proportional to the dislocation loop length between soft pinning points, and its intercept C 1 at the axis 1/ 0 is proportional to the dislocation density N . Figure 9 shows the Granato and Lucke plot log Q 1 versus 1/ in the temperature range between the R1 and P 1 peaks. Such a plot exhibits a positive curvature characteristic of an interaction of dislocations with immobile point defects distributed in a glide plane, and accompanied at same time with an increase of the dislocation density which appears by

the monotonous raising of the curve slope with deformation amplitude. By analogy, this simply means an interaction of a domain wall with point defects with a parallel increase in the domain wall density under the stress amplitude effect. This inuence of mechanical stress on the nucleation of new DWs has been highlighted also by Sarrazin et al.27 by optical observations in a BaTiO3 crystal subjected to local pressure stresses. They showed that the 90 domain density increases during the progressive application of the stress. It can be noticed that the R1 peak temperature does not depend on stress amplitude, contrary to a depinning process. The peak temperature T P should decrease as the applied stress increases. The lack of temperature shift could be due to simultaneous variation in the domain wall density, which evolves with temperature and stress amplitude too. In the Wang et al. model,24 which predicts that the peak temperature is controlled by a self-locking effect when the DWs reach a critical density, this mechanism should shift the R1 peak to high temperatures as the stress amplitude increases. However, an increase in stress also helps to increase the domain wall density; hence, the additional density increasing under stress plays a counterbalancing role in retaining the peak at a similar temperature level. Thus, the R1 peak could be related to a viscous motion of DWs, and the peak height increasing could be due to a dragging-depinning process involving the interaction of point defects with DWs whose microstructure evolves with temperature and stress amplitude. Moreover, the draggingdepinning process of DWs from oxygen vacancies can be further supported, since the R1 peak shape underwent some variation when the oxygen vacancy concentration was decreased by annealing in air Fig. 5 . These different hypotheses must be veried by further experimental and theoretical studies.
IV. CONCLUSION

FIG. 9. Granato and Lucke plot at different temperatures around the R1 and P 1 peaks.

In this study of anelastic behavior in PZT ceramics by internal friction, three important peaks were observed: P 1 , R2 , and R1 . The P 1 peak is due to ferroelectricparaelectric phase transition. The main features of this peak are similar to internal friction peaks associated with a rst-order phase transition. The R2 peak is controlled by a relaxation mechanism with activation energy H (R2) 1 eV and preexponential relaxation time 0(R2) 10 13 s. The R2 peak relaxation process involves interaction between DWs and point defects, such as oxygen vacancies. The R1 is a more complex peak with activation energy H (R1) 1.8 eV and 18 s. In addition to its relaxational behavior, the 0(R1) 10 height of R1 peak depends on heating rate and stress amplitude. The R1 peak is probably controlled by at least three mechanisms: i a relaxation mechanism involving interaction of DWs with point defects as Ti or Zr or oxygen vacancies , ii a domain density variation mechanism in a temperature range approaching the Curie transition, and iii a hysteresis mechanism of domain wall depinning from aggregates of point defects with a stress amplitude dependence. Finally, it is of interest to note that the existence of structural defects in the ferroelectric materials plays an important

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

J. Appl. Phys., Vol. 91, No. 10, 15 May 2002

Bourim et al.
13

6669

role on the conguration of domain size as well as on the motion of DWs.


P. Gerthsen, K. H. Hardtl, and N. A. Schmidt, J. Appl. Phys. 51, 1131 1980 . 2 K. Carl and K. Hardlt, Ferroelectrics 17, 473 1978 . 3 G. H. Jonker, J. Am. Ceram. Soc. 55, 57 1972 . 4 U. Roblels and G. Arlt, J. Appl. Phys. 73, 3454 1993 . 5 V. S. Postnikov, V. S. Pavlov, S. A. Gridnev, and S. K. Turkov, Sov. Phys. Solid State 10, 1267 1968 . 6 P. Ari-Gur and L. Benguigui, J. Phys. D 8, 1856 1975 . 7 F. Vasiliu, P. Gr. Lucuta, and F. Constantinescu, Phys. Status Solidi A 80, 637 1983 . 8 C. M. Valot, N. Floquet, M. Mesnier, and J. C. Niepce, Fourth European Powder Diffraction Conference EPDICIV, Mater. Sci. Forum 228-231, 59 1996 . 9 E. M. Bourim, Ph.D. Thesis, INSA-Lyon, France 1998 . 10 E. M. Bourim, H. Idrissi, B. L. Cheng, M. Gabbay, and G. Fantozzi, J. Phys. IV 6, C8, 633 1996 . 11 E. M. Bourim, H. Tanaka, M. Gabbay, and G. Fantozzi, Jpn. J. Appl. Phys. 39, 5542 2000 . 12 J. V. Humbeeck, in Proceedings of the Summer School on IFS, Cracow, Poland, 14 17 June 1984, p. 131.
1

V. N. Belko, B. M. Darinskii, V. S. Postnikov, and I. M. Sharshakov, Phys. Met. Metallogr. 27, 140 1969 . 14 J. F. Delorme, Ph.D. Thesis, Universite Claude Bernard-Lyon I, France 1971 . 15 W. DeJonghe, R. DeBatist, and L. Delaey, Scr. Metall. 10, 1125 1976 . 16 S. A. Gridnev, B. M. Darinskii, and V. S. Postnikov, Bull. Acad. Sci. USSR, Phys. Ser. Engl. Transl. 33, 1106 1969 . 17 J. X. Zhang, W. Zheng, P. C. W. Fung, and K. F. Liang, J. Alloys Compd. 211212, 378 1994 . 18 Y. Wang, X. Chen, and H. Shen, Chin. J. Met. Sci. Technol. 7, 157 1991 . 19 T. Ishigaki, S. Yamauchi, K. Kishio, J. Mizusaki, and K. Fueki, J. Solid State Chem. 73, 179 1988 . 20 B. L. Cheng, M. Gabbay, and G. Fantozzi, J. Mater. Sci. 31, 4141 1996 . 21 Q. Tan, Z. Xu, J. F. Li, and D. Viehland, Appl. Phys. Lett. 71, 1062 1997 . 22 A. R. Von Hippel, Molecular Science and Molecular Engineering Technology Press of M.I.T, New York, 1959 , pp. 237276. 23 J. J. Amman and R. Schaller, J. Alloys Compd. 211212, 397 1994 . 24 Y. Wang, W. Sun, X. Chen, H. Shen, and B. Lu, Phys. Status Solidi A 102, 279 1987 . 25 Y. N. Huang, Y. N. Wang, and H. M. Shen, Phys. Rev. B 46, 3290 1992 . 26 A. V. Granato and K. Lucke, J. Appl. Phys. 27, 583 1956 . 27 P. Sarrazin, B. Thierry, and J. C. Niepce, J. Eur. Ceram. Soc. 15, 623 1995 .

Downloaded 15 Mar 2012 to 203.255.190.41. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

Das könnte Ihnen auch gefallen