XAS and XRD Structural Characterization of Lanthanum-Modified

The evolution of these parameters with lanthanum content, multiplied by their respective c axis unit cell parameter, is presented in Figure 7. For PLT...
3 downloads 0 Views 219KB Size
Download PDF
14840

J. Phys. Chem. B 2004, 108, 14840-14849

XAS and XRD Structural Characterization of Lanthanum-Modified PbTiO3 Ceramic Materials Person P. Neves,† Antonio C. Doriguetto,*,† Valmor R. Mastelaro,† Luiz P. Lopes,† Yvonne P. Mascarenhas,† Alain Michalowicz,‡ and Jose´ A. Eiras§ Instituto de Fı´sica de Sa˜ o Carlos, UniVersidade de Sa˜ o Paulo, C.P. 369-Sa˜ o Carlos-SP, Brazil, Groupe de Physique de Milieux Denses, UniVersite´ Paris XII, Val de Marne, 94010, Creteil Cedex, France, and Departamento de Fı´sica, UniVersidade Federal de Sa˜ o Carlos, Sa˜ o Carlos-SP, Brazil ReceiVed: October 20, 2003; In Final Form: June 22, 2004

For the first time combined X-ray absorption spectroscopy (XAS) and X-ray diffraction (XRD) experiments were carried out to probe the short- and long-range order in lanthanum-modified PbTiO3 ceramic materials (Pb1-xLaxTiO3) for x ranging from 0 to 30 atom % of La. XRD results show that a tetragonal structure fitted well to all samples, except to the one with 30 atom % of La for which XRD results highlighted a cubic unit cell. XRD refinements also confirmed the existence of only A-site vacancies. X-ray absorption near edge spectroscopy (XANES) measurements indicate that the local structure around Ti atoms has a different compositional dependence from the one obtained from XRD, mainly for the sample with 30 atom % of La. According to XANES data, for the sample with 30 atom % of La, a local distortion around Ti atoms persist. The analysis of extended X-ray absorption fine structure (EXAFS) spectra shows a decrease in the local disorder around Ti atoms as the content of La increases. Structural details, in terms of short and long-range order, are presented and the correlation between the XRD and XANES/EXAFS data is discussed.

1. Introduction Lanthanum-doped lead titanate ceramics (Pb1-xLaxTiO3 or PLT) have been studied in detail due to their interesting physical properties.1 The isomorphic substitution of lead by lanthanum atoms induces some interesting changes in the physical properties of PbTiO3 material. When the lanthanum content is higher than 25 atom %, a diffuse character of the ferroelectricparaelectric phase transition (DPT) is observed.1 The modification in the phase transition character is followed by a linear decrease in the Curie temperature (Tc) to room temperature for x ∼ 30 atom %.1-3 It has been suggested that the lanthanuminduced modification concerning PbTiO3 results in structural changes that can be directly related to the nature of the phase transition. 4-7 The PbTiO3 crystal structure has been exhaustively studied.8-10 Its room-temperature structure is tetragonal, P4mm (a ) 3.902(3) and c ) 4.156(3) Å), with four atoms in the asymmetric unit: Pb at (0, 0, 0); Ti at (1/2, 1/2, 0.5377(4)); O(1) at (1/2, 1/2, 0.1118(3)); and O(2) at (0, 1/2, 0.6174(3)).11 Assuming the perovskite structure notation, the sites occupied by Pb and Ti, are labeled as A and B, respectively. In La3+-modified PbTiO3 materials, La3+ replaces Pb2+ rather than Ti4+ in Pb-based perovskites.12 To keep the neutrality of the charge, since that replacement is aliovalente, site vacancies are created.12 Considering that the PbO-La2O3-TiO2 solid solution is lying within a ternary diagram formed by PbTiO3, and the two * To whom correspondence should be addressed: Departamento de Fı´sica e Informa´tica, Instituto de Fı´sica de Sa˜o Carlos, Universidade de Sa˜o Paulo, C.P. 369, CEP 13560-970, Sa˜o Carlos-SP, Brazil. Telephone: 55-162739887. Fax: 55-16-2739881. E-mail: [email protected]. † Universidade de Sa ˜ o Paulo. ‡ Universite ´ Paris XII. § Universidade Federal de Sa ˜ o Carlos.

metastable defect compounds La2/3-1/3TiO3 and La[Ti3/4-1/4]O3, these compositions have been described by the general defect equation13

3(1 - Ry) y(2R - 1.5) 3y Pb La 0 × 3 + y(1.5 - R) 3 + y(1.5 - R) 3 + y(1.5 - R) y(1.5 - R) 3 Ti O 0 3 + y(1.5 - R) 3 + y(1.5 - R) 3

[

]

where 0 is a site vacancy and R is the Pb elimination factor. The R factor can adopt values ranging from 1.5 (A-site vacancies only) to 0.75 (B-site vacancies only). There is a consensus in the literature that lanthanum exclusively occupies the A site.14,15 Raman scattering and X-ray diffraction techniques have been respectively used to probe the short and long-range order in the Pb1-xLaxTiO3 ceramic system.14-16 Tavares et al.,16 using the Raman scattering technique, probed the short-range structure for x ranging from 0.0 to 30 atom %. According to this work, in highly doped samples (x g 27 atom %) and for temperatures above the phase transition, X-ray diffraction measurements indicate a cubic structure. However, Raman scattering measurements show a residual short-range structural disorder in this cubic phase. These authors suggested that the existence of this short-range structural disorder could be correlated to the relaxor behavior in highly doped samples. Recently, Tae-Yong Kim et al.14 also observed the presence of Raman-forbidden bands that appeared in temperatures above the phase transition in lead titanate ceramics containing 30 atom % of La. They explain the observation of these Raman-forbidden bands at temperatures substantially higher than the phase transition temperature by the existence of short-ranged clusters with tetragonal 4mm (C4V) symmetry distributed over the globally cubic m3hm (Oh) symmetry matrix.

10.1021/jp037166h CCC: $27.50 © 2004 American Chemical Society Published on Web 09/03/2004

Lanthanum-Modified PbTiO3 Ceramic Materials

J. Phys. Chem. B, Vol. 108, No. 39, 2004 14841

Despite the scientific-technological importance of this system, conclusive investigations are scarce considering its crystalline structure, mainly regarding phase transition induced by a cation replacement. In this paper, we present a short and long-range structural study of the Pb1-xLaxTiO3 ceramic system. The XAS technique was used to probe the local structure around Ti atoms whereas the XRD technique was used to obtain information concerning the crystallographic structure of the samples, as well as the distribution of vacancies. To the best of our knowledge, it is the first time that the short- and long-range order structure of the PLT ceramic system is simultaneously studied using XAS and XRD techniques. 2. Experimental Section Seven powdered PLT samples, of the nominal composition Pb1-xLaxTiO3 with x ) 0.0, 5, 10, 15, 20, 25, 30 atom % (abbreviated as PLTx, from PLT00 to PLT30), were prepared by the conventional mixed oxide method. The oxides, PbO, La2O3, and TiO2, weighed according to stoichiometry, were mixed by ball milling in distilled water for 10 h. The slurry was dried and calcined in a covered alumina crucible at 1100 °C for 3 h. Ceramic bodies were then formed by isostatic pressure, and fired at ∼1175 °C for 3 h. The densities of all fired ceramics were found to be around 95% of the theoretical one defined by the Archimedes method. Surface analyses, by scanning electronic microscopy (SEM), show that the grainsize of all fired samples has about 1.5 µm. The PLT05 and PLT30 sample compositions were examined point-to-point using an electron microscope equipped with an electron microprobe spectrometer (Wavelength Dispersive Spectrum (WDS) accessory). Disklike ceramic bodies were polished to ensure parallel surfaces, and silver electrodes were then deposited onto surfaces by sputtering. The complex dielectric permittivity was measured from -180 to +500 °C using an HP4192A impedance analyzer at 1 kHz. The X-ray powder diffraction patterns were measured at room temperature on a Rigaku Denki powder diffractometer with geometry θ-2θ, a rotating anode X-ray source (Cu KR radiation, λ ) 1.542 Å), and a scintillation detector. The measurements were carried out using a current of 100 mA and tension of 50 kV. Each pattern consists of 5851 steps ranging from 3 to 120° in 2θ. The data were collected with a step size of 0.02°. The count time was 5 s per step. The refinement of the structure was carried out using the Rietveld method.17 The program used in the refinements was the General Structure Analysis System (GSAS) package.18 The least-squares refinements of all samples were started considering the PbTiO3 structure published by Nelmes and Kuhs in the space group P4mm.11 Since P4mm is a quiral space group, the unit cell origin was defined by fixing the Pb atom at the 0, 0, 0 position. In this way, all other atoms in the unit cell are described by only three positional parameters, δzTi, δzO(1), and δzO(2), the shift along the c axis from the ideally cubic perovskite structure: 1/2, 1/2, 1/2 + δzTi; 1/2, 1/2, 0 + δzO(1); and 0, 1/2, 1/2 + δzO(2) for Ti, O(1), and O(2) atoms, respectively. However, in each case, the refinement was started by setting δzO(1) and δzO(2) equal to zero, whereas δzTi was given a small positive value to define the positive direction of the c axis. Scattering factors of ionized atoms were used in all refinements. The O2scattering factors were taken from Rez et al.19 Refinements of the anisotropic temperature factors were tried, but the trial results were not physically significant. Therefore, the temperature factors were kept isotropic (U) during the refinements.

Figure 1. Temperature dependence of the relative dielectric permittivity of PLTx samples.

The titanium K-edge X-ray absorption spectra were collected at the LNLS (National Synchrotron Light Laboratory) facility using the D04B-XAS1 beamline.20 The LNLS storage ring was operated at 1.36 GeV and 100-160 mA. The sample pellets, obtained after sintering, were grounded for XAS measurements. XAS data were collected at the Ti K-edge (4966 eV) in transmission mode at room temperature using a Si(111) channelcut monochromator. Ionization chambers were used to detect the incident and the transmitted flux. XANES spectra at the Ti K-edge were recorded for each sample between 4910 and 5200 eV using energy steps of 0.5 eV. Because of the presence of the lanthanum LIII edge at 5483 eV, EXAFS measurements at the Ti K-edge (using steps of 2 eV) were carried out for each sample between 4840 and 5400 eV. Four acquisitions were recorded for each sample to improve the signal-to-noise ratio. To provide good energy reproducibility during the XANES data collection, the energy calibration of the monochromator was checked during the collection of the sample data using a Ti metal foil. The EXAFS analysis was carried out by using the program set written by Michalowicz,21 according to the recommended procedures described by the International Workshop on Standards and Criteria in XAFS.22 After atomic absorption removal and normalization, the K3χ(k) weighed EXAFS signal was Fourier transformed to R distance space in the 2.0-10 Å-1 K range. Each spectrum was Fourier transformed using a Kaiser apodization window with τ ) 2.5. A qualitative interpretation of XANES spectra obtained at the Ti K-edge was performed using the software package developed by Michalowicz and Noinville.23 For comparison purposes among different samples, all spectra were background removed and normalized using as unity the first EXAFS oscillation. 3. Results and Discussion a. Relative Dielectric Permittivity. Figure 1 shows the temperature dependence of the relative dielectric permittivity of PLTx samples. As the content of La increases, we observed an increase in the dielectric diffuseness followed by a gradual decrease in the temperature of maximum dielectric permittivity (Tc). As can be seen in Figure 1, depending on the La content,

14842 J. Phys. Chem. B, Vol. 108, No. 39, 2004

Figure 2. Evolution of the (102), (201), and (210) reflections on the La content doping increase.

the ferroelectric behavior can be normal or relaxor: the relative dielectric permittivity curve of the PLT20 sample (Tc ≈ 80 °C) presents a narrow peak representative of a normal ferroelectric material, whereas the PLT30 one (Tc ≈ -50 °C) presents a broad peak representative of a relaxor ferroelectric material. In

Neves et al. the first case, for temperatures higher than Tc, a centrosymmetric cubic structure (Pm3hm) is expected since the material undergoes a phase transition to a paraelectric material. In the second case, for temperatures higher than Tc, a residual short-range structural disorder has been found16 not to be compatible with the centrosymmetric cubic structure usually observed for the paraelectric phase. b. X-ray Diffraction. The tetragonal structural model fits well to all XRD data, except for the PLT30. Concerning the PLT30, the a and c cell parameters converged to very similar values and the shift parameters are δzTi ) δzO(1) ) δzO(2) ) 0. This result highlights a cubic unit cell. The overlap of the three tetragonal reflections (102), (201) and (210), confirms the tetragonal-to-cubic phase transition (see Figure 2). Until PLT25, these three reflections appear individually, but for PLT30 they coalesce according to the cubic multiplicity of (210). Therefore, a new Rietveld refinement was conducted for PLT30 using as initial model the cubic structure obtained by Glazer and Mabud10 for PbTiO3 at 550 °C. In this structure, the asymmetric unit is defined by three independent atoms, Pb (0, 0, 0), Ti (1/2, 1/2, 1/ ), and O (1/ , 1/ , 0), and no positional parameter needs to be 2 2 2 refined. The fitting obtained in the cubic perovskite phase Pm3hm is significantly better than in P4mm. The experimental and calculated XRD patterns of the two extremes of the compositional series studied here, PLT00 and PLT30, are shown in Figures 3 and 4, respectively. Figure 5 is an ORTEP-324 view of the PLT00 structure, which is tetragonal P4mm. In PLT00, the Ti, O(1), and O(2) are positively shifted along the c direction, whereas in the PLT30 structure (Pm3hm), these atoms occupy exactly the corners (A site), the face centers (oxygens sites),

Figure 3. Experimental (+) and calculated (-) XRD patterns of the PLT00. The difference between the two patterns is presented below the patterns.

Lanthanum-Modified PbTiO3 Ceramic Materials

J. Phys. Chem. B, Vol. 108, No. 39, 2004 14843

Figure 4. Experimental (+) and calculated (-) XRD patterns of the PLT30. The difference between the two patterns is presented below the patterns.

Figure 5. ORTEP-3 drawing showing (a) the unit cell, atom designation, and the Ti (site A) octahedral coordination environment and (b) the Pb,La (site B) 12-fold coordination environment. Displacement ellipsoids are drawn at the 50% level considering an idealized isotropic thermal parameter of 0.025 Å2 (symmetry codes: i x, y, z + 1; ii -y + 1, x, z; iii x, y, z - 1).

and the body center (B site) of the unit cell, which means the ideal perovskite cubic structure. The final structural parameters of all samples are shown in Table 1. For all La-containing PbTiO3 samples, the A-site occupancy was shared between Pb and La. Considering A-site vacancies only, the refinements of the sample containing La were started with the defect equation Pb(1-1.5y)Lay-0.5yTiO3, meaning that the R value is equal to 1.5. Several models considering B-site vacancies only (R ) 0.74) and intermediate R values (0.74 e R e 1.5) were carried out; however, these refinements present

unrealistic values for the atomic occupancies and worse statistical compatible parameters, even for the samples with high La content. The A-site occupation is also given in Table 1. The Pb and La content obtained from XRD data present values very similar to the nominal compositions. Nevertheless, it is important to emphasize that realistic occupancy results depend on the atomic number difference among the atoms involved in the site occupancy refinements. In this case, the Pb (82) and La (57) atomic numbers are significantly different, which usually permits a suitable occupancy refinement. However, false minima and

14844 J. Phys. Chem. B, Vol. 108, No. 39, 2004

Neves et al.

TABLE 1: Cell Parameters, Fractional Coordinates, Cation Distribution, and Powder and Reflection Data Statistics sample

A

c

δzTia

δzO(1)

δzO(2)

compositionb

wRp

χ2

R(F2)

PLT00 PLT05 PLT10 PLT15 PLT20 PLT25 PLT30

3.90197(4) 3.90809(4) 3.91502(4) 3.91687(4) 3.92357(2) 3.93367(2) 3.91690(1)

4.1373(1) 4.0733(1) 4.0181(1) 3.97939(4) 3.94878(2) 3.95371(3) 3.91690(1)

0.036(1) 0.029(1) 0.030(1) 0.023(1) 0.017(1) 0.021(1) 0

0.115(2) 0.101(2) 0.077(2) 0.057(2) 0.034(3) 0.022(5) 0

0.115(2) 0.110(2) 0.096(1) 0.077(1) 0.053(1) 0.042(2) 0

Pb1.00La0.00Ti1.0000.00O3 Pb0.91La0.06Ti1.0000.024O3 Pb0.88La0.08Ti1.0000.050O3 Pb0.81La0.13Ti1.0000.07O3 Pb0.75La0.17Ti1.0000.10O3 Pb0.67La0.22Ti1.0000.13O3 Pb0.62La0.25Ti1.0000.13O3

0.097 0.078 0.083 0.061 0.042 0.049 0.052

26.8 16.5 14.2 10.2 4.4 4.9 6.1

0.035 0.033 0.044 0.040 0.044 0.063 0.058

a The atomic fractional coordinates are as follows: Pb, La at 0, 0, 0; Ti at 1/ ,1/ ,1/ + δz ; O(1) at 1/ , 1/ , 0 + δz 1 1 2 2 2 Ti 2 2 O(1); and O(2) at 0, /2, /2 + δzO(2). Therefore, δzTi, δzO(1), and δzO(2) stand for the shift from the ideally cubic perovskite structure along the (001) direction. b The Pb and La contents were obtained from A site occupancy refinements, and 0 denotes A-site vacancies.

Figure 6. Dependence of the cell parameters and the c/a ratio with the La content.

other problems remain such as the correlation between occupancy and the thermal displacement of the atoms. Therefore, to confirm the nominal values, the chemical composition of two samples, PLT05 and PLT30, were quantified by WDS measurements. The obtained chemical formulas for PLT05, Pb0.884La0.049Ti1.02100.046O3, and PLT30, Pb0.614La0.285Ti0.97900.122O3, where 0 denotes A-site vacancies, are compatible with the nominal and XRD compositions. Therefore, it is expected that other samples also present compositions very similar to the nominal one. Figure 6 shows the La content dependence of the a and c cell parameters and the c/a ratio. The a dimension remains almost constant from PLT00 to PLT30, whereas c dimension decreases gradually in the increase of La content. Consequently, the c/a ratio decreases very fast reaching values close to 1 starting from the PLT20 sample. For the PLT30, c/a is exactly equal to 1 due to cubic symmetry constraint. These results highlight a structural phase transition from tetragonal-to-cubic occurring between PLT25 and PLT30 samples. In Table 1 the shift from the ideally cubic perovskite structure is also shown: δzTi, δzO(1), and δzO(2). The evolution of these parameters with lanthanum content, multiplied by their respective c axis unit cell parameter, is presented in Figure 7. For PLT00, a large difference between the oxygen and titanium shifts is observed: δzO(2) ) δzO(1) ) 0.48 Å and δzTi ) 0.15 Å. Apart from the increase of La content, all of them decrease almost linearly until reaching zero for PLT30. These changes directly affect the geometric structural parameters, which are also shown in Table 2. In Figure 8, the lanthanum dependence of the Ti-O and Pb-O distances is presented. It can be observed that Ti-O(2) and Pb-O(1) separations (see also Figure 5) remain constant throughout the whole series. This behavior was expected since Ti-O(2) and Pb-O(1) lie on the perpendicular plane to the ferro-distortive axis, which is unaffected by the c-axis distortion. On the other hand, it is

Figure 7. Dependence of the shift from the ideally cubic perovskite structure (δzTi, δzO(1) and δzO(2)), multiplied by their respective c unit cell parameter, with the La content.

observed that increasing La content, Ti-O(1) also increases, whereas, as expected, Ti-O(1)i decreases due to translational symmetry. In fact, Ti-O(1), Ti-O(1)i, and Ti-O(2) distances are statistically different from PLT00 to PLT20. For PLT25, the three distances present very similar values, and for PLT30 they are obviously equal due to the cubic symmetry constraint. In contrast to what has been observed with Ti-O, the PbO(1), Pb-O(2), and Pbi-O(2) distances remain statistically different along the whole curve, except for PLT30. When the La content is increased, the Pb-O(2) distance decreases, whereas, as expected, Pbi-O(2) increases due to symmetry dependence until it reaches the constant curve of Pb-O(1) taking place just in the PLT30 sample. c. X-ray Absorption Near Edge Structure (XANES). Figure 9 shows the XANES spectra obtained from the Ti K-edge for PLT samples. The preedge region of the K-edge XANES spectra of some transition metal oxides are characterized by a pronounced feature, several volts before the main rising edge.25-27 In transition metal oxides that crystallize in centrosymmetric structures, this preedge feature is very small or absent; in noncentrosymmetric structures it can be quite large.25-27 According to the literature, the physical origin of the preedge feature, labeled A in Figure 9, is due to the transition of the metallic 1s electron to an unfilled d orbital.25 This dipole forbidden electronic transition is normally allowed by the mixture of p character from surrounding oxygen atoms into the unfilled d states. On the other hand, the preedge feature labeled B in Figure 9 was shown to be caused by Ti 1s electron transition to the unoccupied 3d-originated eg-type molecular orbital of TiO6 polyhedra neighboring the absorbing Ti atoms, which are weakly affected by the core hole potential.27 The area of the B peak does not depend strongly on small displacements of the atoms from their sites in cubic crystal lattice, but it

Lanthanum-Modified PbTiO3 Ceramic Materials

J. Phys. Chem. B, Vol. 108, No. 39, 2004 14845

TABLE 2: Selected Geometric Parameters (Å)a Pb-Ti Pb-Tiiii Ti-O(1) Ti-O(1)i Ti-O(2) Pb-O(2) Pb-O(2)iii Pb-O(1) O(1)-O(2) O(1)i-O(2) O(2)-O(2)ii a

PLT00

PLT05

PLT10

PLT15

PLT20

PLT25

PLT30

3.541(3) 3.360(3) 2.39(1) 1.74(1) 1.978(1) 3.203(6) 2.519(4) 2.800(2) 2.844 2.843 2.759

3.503(3) 3.365(3) 2.329(9) 1.744(9) 1.982(1) 3.162(4) 2.517(3) 2.794(1) 2.852 2.793 2.763

3.493(2) 3.351(2) 2.199(8) 1.819(8) 1.9755(8) 3.094(4) 2.542(3) 2.7858(7) 2.860 2.751 2.768

3.464(3) 3.358(3) 2.12(1) 1.86(1) 1.9704(8) 3.019(4) 2.581(4) 2.7788(7) 2.851 2.733 2.770

3.444 (3) 3.367(2) 2.04(1) 1.91(1) 1.9669(5) 2.935(4) 2.640(4) 2.7775(5) 2.838 2.730 2.774

3.465(3) 3.365(3) 1.98(2) 1.97(2) 1.9685(4) 2.908(7) 2.674(6) 2.7829(6) 2.845 2.734 2.782

3.39213(1) 3.39213(1) 1.95845(1) 1.95845(1) 1.95845(1) 2.76967(1) 2.76967(1) 2.76967(1) 2.770 2.770 2.770

Symmetry codes: ix, y, z + 1;

ii

-y + 1, x, z;

iii

x, y, z - 1). The symmetry codes are as in Figure 5.

Figure 8. Dependence of the Ti-O and Pb-O bond lengths with the La content.

Figure 9. Ti K-edge normalized XANES spectra of PLTx powder samples.

changes significantly when 4d atoms appear in the vicinity of the absorbing Ti atom, for instance, Zr atoms in the PZT solid solution.27 As can be seen in Figure 9, the intensity of the A feature for the PLT05 sample is quite similar to the PLT00 one. The increase of lanthanum content to 10 atom % (PLT10) leads to a significant decrease in the intensity of the A feature. A further increase of lanthanum to 30 atom % (PLT30) promotes a slight variation in the A feature intensity. This behavior indicates a small dependence of the A feature intensity in lanthanum concentration, at least for samples containing more than 10 atom % of La. Figure 10 shows the XANES spectrum of the SrTiO3 sample, in which Ti atoms are practically located at the TiO6 octahedron center. As was previously observed,25 it results in a very small A feature. The comparison of the A feature in SrTiO3 sample with that one observed in PLT samples indicates that the Ti atoms are moved off-center in TiO6 octahedron.

Figure 10. Ti K-edge normalized XANES spectrum of SrTiO3 powder samples.

Figure 11. Variation of the intensity of peak A in XANES spectra compared with Ti site distortion from the center of the oxygen octahedron obtained from XRD data. (Ti atom distortion ) (Ti-O(1)Ti-O(1)i)/2, where the Ti-O bond lengths were compiled from Table 2).

Figure 11 compares the intensity of the peak A in the XANES spectra with the magnitude of the Ti off-center position obtained from XRD data as a function of lanthanum content. The intensity of peak A in the XANES spectra was obtained using a single Gaussian function in the interval from 4965 to 4980 eV. According to the literature, the intensity of peak A is proportional to the amplitude of the Ti atom distortion.26-28 From XRD data, the Ti off-center magnitude is defined as (Ti-O(1)-TiO(1)i)/2 (Ti-O distances compiled from Table 2). As shown in Figure 11, there is a qualitative relationship between the intensity of the XANES preedge peak A and the TiO6 octahedral site symmetry: when the site is known to be asymmetric, as in pure PbTiO3, peak A is quite intense, on the contrary, when the site is symmetric, as in the SrTiO3 compound, which

14846 J. Phys. Chem. B, Vol. 108, No. 39, 2004

Neves et al.

TABLE 3: Structural Parameters of PLTx Samples Obtained from EXAFS Analysis samples

single-shell fitting

PbTiO3

N ) 6.0 R ) 1.97(1) Å σ ) 0.116(5) Å quality factor ) 5.2 N ) 6.0 R ) 1.92(1) Å σ ) 0.125(4) Å quality factor ) 1.1

PLT10

three-shell fitting N1 ) 4.0 R1 ) 1. 98(1) Å σ1 ) 0.075(6) Å quality factor ) 1.0 N1 ) 4.0 R1 ) 1.94(1) Å σ1 ) 0.10(1) Å quality factor ) 1.3

XANES spectrum is presented in Figure 10, preedge peak A intensity is very low. Concerning the La containing samples, from 0 to 10 atom % of La, the off-center magnitude obtained from the XANES spectra decreases as the content of La increases, compatible with XRD. For the samples containing more than 10 atom % of La, XANES results show that the magnitude of Ti off-center position remains almost constant. On the other hand, XRD data suggest that the magnitude of Ti off-center position decreases as the La content increases reaching an idealized cubic perovskite structure for the PLT30. For the PLT00 sample, the magnitude of the off-center position was also obtained from EXAFS data. According to the data presented in Table 3, it was found that the distortion is equal to 0.31 Å, very close to the values obtained from XRD and XANES data (Figure 11) and by others authors using XAS and XRD techniques.16-19 With regard to the analysis of the B preedge feature, since we found no significant variation in its shape and amplitude when the content of La was increased, it is reasonable to assume that La atom replaces preferably the Pb atom rather than the Ti one. This assumption is in accordance with our XRD results. d. Extended X-ray Absorption Fine Structure (EXAFS). The EXAFS signals and the corresponding Fourier transforms (FTs) of the PLT samples are compared in parts a and b of Figure 12, respectively. The presence of the lanthanum LIII edge at 5483 eV just after the Ti K-edge (4920 eV) limits the analysis of the Ti K-edge EXAFS spectra, and due to this fact the ∆k range that is available in the simulation is limited to approximately 8 Å-1. As can be seen in Figure 12, some differences appear in the FTs when lanthanum is added to the PbTiO3 perovskite structure. The first difference is a shift toward smaller distances of the Fourier transform first peak, which is correlated to the first coordination shell around titanium atoms (Ti-O bonds). Another significant difference is an increase in the amplitude of this Fourier transform first peak, mainly for the 30 atom % of the La sample (PLT30). At longer distances, the appearing of a peak around 3.5 Å for all lanthanum-containing PbTiO3 samples was observed. However, its shape and amplitude did not vary significantly as the La increases from 10 to 30 atom %. The theoretical EXAFS spectra for the PLT00 and PLT10 samples were calculated using the FEFF8.2 software package29 in order to visualize the effect of the presence of lanthanum in the short-range order structure around the Ti atom. The FEFF program applying a curved-wave multiple-scattering theory is a powerful computational technique to calculate the theoretical EXAFS spectra.30 The calculations were based on the crystallographic structure of the PLT00 and PLT10 samples obtained from our Rietveld refinements. As the effect of thermal and static disorder represented by the Debye-Waller factor was not introduced in the theoretical calculation, a large difference in the amplitude of experimental and theoretical Fourier transform curves is expected.

N2 ) 1.0 R2 ) 1.78(1) Å σ2 ) 0.075(6) Å

N3 ) 1.0 R3 ) 2.39(3) Å σ3 ) 0.075(6) Å

N2 ) 1.0 R2 ) 1.81(1) Å σ2 ) 0.10(1) Å

N3 ) 1.0 R3 ) 2.41(5) Å σ3 ) 0.10(1) Å

Figure 12. (a) Experimental Kχ(k) vs k EXAFS spectra of PLTx samples and (b) Fourier Transform of K3χ(k) EXAFS spectra presented in Figure 3. For the sake of clarity, each curve was shifted downward by a constant in relation to the preceding one.

Figure 13. Fourier transform modulus of Ti K EXAFS spectra of the PLT00 and PLT10 calculated by FEFF8.2 program.

Figure 13 shows the modulus of the Fourier transform curves obtained from the theoretical EXAFS spectra. As can be seen, the first peak of the Fourier transform, which is correlated to

Lanthanum-Modified PbTiO3 Ceramic Materials the first Ti-O neighbors, shifts toward lower distances for the PLT10 sample and presents a higher intensity. As the number of first Ti-O neighbors is the same in both samples, this increase of intensity indicates a decrease of the Ti off-center distortion for the PLT10 sample. Similar to the experimental Fourier transform curve shown in Figure 12, a more intense peak located between 3 and 4 Å is also observed in the theoretical Fourier transform modulus of the Ti K EXAFS spectrum of the PLT10 sample (Figure 13). According to the calculation, this peak is mainly formed by contributions from single (Ti-Pb, Ti-La, and Ti-Ti) and from multiple-scattering paths (Ti-O-Ti, Ti-O-O, O-Ti-O-Ti, Ti-O-Pb, and TiO-La). The increase in the intensity of this peak observed in the PLT10 Fourier transform curve could be attributed to the decrease of the structural distortion at the medium range-order structure around the Ti atoms. To obtain quantitative structural data concerning the Ti-O first coordination shell, the EXAFS spectra of the Fourier transform first peak (between 0.8 and 2.2 Å) presented in Figure 12 were extracted using a back Fourier transform and then fitted using theoretical phase and amplitude functions. To model the Ti-O pair, we used theoretical amplitude and phase functions calculated using a FEFF 8.2 program.29 In all fits, the number of free parameters was kept smaller than the number of independent points defined as Nind ) 2∆R∆K/π, where ∆R is the width of the R-space filter windows and ∆K is the actual interval of the fit in the K space.22 The reliability of the fit was determined by a quality factor.22 According to the XRD results (see Figure 5 and Table 2), for PLT samples from 0 to 25 atom % of La, the first coordination shell around the Ti atom forms a distorted octahedron with three different Ti-O bond lengths (threeshell): two of them along the axial c axis and the third one referring to the four equatorial bonds, which present the same length due to the structure symmetry. On the other hand, for the PLT30 sample, the first coordination shell around the Ti atom forms a perfect octahedron, for which the six Ti bonds are equivalent (single-shell) due to the Ti m3hm (Oh) point symmetry. Therefore, during the fitting procedures of the EXAFS spectra using the FEFF 8.2 program,29 initially applied to the PLT00 sample, two models were tested: a single-shell and a three-shell. In both models, the coordination number (NTi-O) was fixed while the interatomic distance (RTi-O) and the Debye-Waller factor (σTi-O) were free to vary. In the threeshell model, a single value of σTi-O was applied to the three types of Ti-O pairs reducing the number of free parameters used in the fitting. Table 3 presents the fitting results. As observed, the three-shell model is significantly better than the single-shell one for the PLT00. Moreover, the three bond lengths obtained from the EXAFS three-shell fitting are in accordance with those ones obtained from our XRD refinement (Table 2) and also from previous XAS work, where a three shell fitting was also used.28,31 Figure 14 shows the comparison between the filtered experimental EXAFS spectrum and the one-shell and the three-shell theoretical models for the PLT00. For La-containing samples, we followed the same procedure adopted to fit the EXAFS spectrum of the PLT00 sample. First, we carried out the fitting of the first Ti-O coordination shell of the PLT10 sample. As can be seen in Table 3, the fitting quality factors of both models (single- and three-shell) are comparable. As indistinguishable fitting quality factors were also observed for the other PLTx ceramic samples, we did not include the fitting results in Table 3.

J. Phys. Chem. B, Vol. 108, No. 39, 2004 14847

Figure 14. Fitting and back-Fourier-filtered experimental signal of the Ti-O first shell for the PLT00 sample using a three-shell model.

This fact was expected because the distance resolution (∆R ) π/2∆k) achieved during the fitting is determined by the useful ∆k data range, which in this case, was limited by the presence of the La LIII-edge located just after the Ti K-edge. Since the useful ∆k range was about 8.0 Å-1 for all samples, ∆R was found to be equal to 0.19 Å. This means that Ti-O bond lengths that differ less than 0.19 Å cannot be resolved. This explains why the three-shell or single-shell models are statistically indistinguishable and provide unrealistic structural results for rich lanthanum samples (10-30 atom %). However, we should note that the Debye-Waller factors obtained in the one-shell fits are always greater than 0.1 Å. Such values are incompatible with a regular TiO6 octahedra with six equal Ti-O distances (for a single shell at room temperature the DW factor should be less than 0.08 Å). Thus, even if the poor EXAFS resolution prevents any accurate distance distortion evaluation, the EXAFS fitting and the XANES results are qualitatively in agreement. Thus, EXAFS and XANES results are in accordance with previous studies using Raman spectroscopy in the PLT samples.14-16 According to the work of Tavares et al.,16 the presence of extra Raman active modes in a 27 atom % La sample can be explained by the existence of a short-range structural disorder. On the other hand, Kim et al.15,16 argue that the extra Raman active modes observed to a 30 atom % La sample measured at 300 K arise from short-ranged clusters with tetragonal 4 mm (C4V) symmetry, which are distributed over the globally cubic m3hm (Oh) symmetry matrix. Although our XANES and EXAFS results show a local disorder concerning the PLT30 sample, any conclusion could be drawn regarding the existence or not of tetragonal clusters, as suggested by Kim and co-workers.14,15 Since there is a certain discrepancy between XANES/EXAFS and XRD structural results concerning the PLT30 sample, it is important to understand the XRD results in the context of our XAS results. All XRD and XAS measurements presented in this paper have been taken at room temperature, ∼25 °C. Therefore, from PLT00 to PLT25 samples, for which the respective ferroelectric-paraelectric transition temperature (see Figure 1) is above the experimental one, and which present a ferroelectric-normal behavior below Tc, a tetragonal structure is expected. This expectation was confirmed for both techniques, XRD and XAS. Concerning the PLT30 sample for which T is around -50 °C, XRD and XAS measurements have been taken above the ferroelectric-paraelectric transition temperature (Tc). According to previous Raman spectroscopy measurements at room tem-

14848 J. Phys. Chem. B, Vol. 108, No. 39, 2004 perature, the structure of this sample should present a residual short-range structural disorder in a global cubic matrix.14-16. Here, the XAS measurements show that the Ti atom remains moved off-center in its octahedron. On the other hand, the XRD analysis shows that the Ti atom is in its ideal position (1/2, 1/2, 1/ ). 2 It is important to emphasize that XRD is a long-range probe technique. However, using high-precision X-ray diffraction single-crystal data, Abramov et al.32 have performed a charge density study, which allowed them to conclude that the cubic SrTiO3 structure is subject to two types of distortion with shortrange order. The Ti atom lies dynamically off-center in the octahedron of its ligand O atoms, and the octahedra themselves vibrate. Abramov et al.32 argued that depending on the atomic site symmetry in the high-symmetry phase, several subsidiary minima will occur equidistant from the ideal position, either lying on the circumference of a circle (for the O atoms in perovskite) or the surface of a sphere (for the A and B atoms). The mechanism of reorientation between these minima will involve motion around the circle or sphere rather than through the ideal position, since the potential barriers on this surface will be lower than the potential barrier in the center. Thus, for a high enough degree of anharmonicity, the atom may move around the circle or sphere without ever crossing the ideal position. Even at this extreme, the crystal will retain its higher symmetry, Pm3hm. A similar phenomenon quite probably takes place for the PLT30 sample as pointed out by our XANES/EXAFS results, which are techniques more sensible to probe short-range order than the XRD one. For the other samples, the atoms should remain in one minimum long enough for the surrounding structure to relax leading to lower crystal symmetry, P4mm. In this case, XANES/EXAFS and XRD techniques are compatible with each other. A similar discrepancy between XAS31 and XRD10 results was commented by Sicron et al.31 when they studied the nature of the ferroelectric phase transition in the PbTiO3 compound. They observed a discrepancy between XAS and XRD results concerning the temperature dependence of Ti distortion below and above the structural ferroelectric transition. Whereas XAS results showed an off-center position of the Ti atoms above the transition temperature, XRD data suggested that Ti atoms go back to the ideal position. Teslic and Egami also found a discrepancy between structural results obtained by XRD and neutron scattering data when they studied the local atomic structure of PbZr1-xTixO3 (PZT), Pb(Mg1/3Nb2/3)O3 (PMN) and Pb1-xLaxZryTi1-yO3 (PLZT) systems.33,34 Their structural studies yielded evidence of slight local deviations from crystallographic lattice periodicity and local chemical ordering. Summarizing, the discrepancy observed in the XAS and XRD results concerning the PLT30 sample is basically due to the fact that X-ray absorption spectroscopy probed the short-range order around the Ti atoms, whereas X-ray diffraction provides structural information about the average structure.26-28,32-34 4. Conclusions The structure of the Pb1-xLaxTiO3 ceramic samples was examined by XRD-Rietveld refinements and the XAS technique. According to XRD-Rietveld refinements, the tetragonality of the structure gradually diminishes with the increase of La content until a structural phase transition from tetragonal-to-cubic structure, which takes place in the range between 25 and 30 atom % of La samples. In fact, the samples with x ranging from

Neves et al. 0 to 25 atom % crystallize in the P4mm space group, whereas the sample with 30 atom % of La crystallizes in the Pm3hm space group. The refinement of XRD data also confirmed the existence of only A-site vacancies, which are introduced to the perovskite structure by the aliovalente replacement of Pb2+ by La3+ cations. According to the XANES data, the Ti atom is in an off-center position in the TiO6 octahedra (Ti point symmetry ) 4mm) for all samples, including the PLT30 samples for which XRD highlighted a cubic structure (Ti point symmetry ) m3hm). However, the incompatibility between the XAS and XRD techniques concerning PLT30 is only apparent. Recently, it has been shown32 that, even in an ideal perovskite cubic structure, where the crystal retains its higher symmetry (Pm3hm), atoms may move around its ideal position. This structural short-range order phenomenon could be probed by the XAS technique, but not by a conventional XRD one. The presence of the lanthanum LIII-edge after the titanium K-edge did not allow us to obtain quantitative structural information from the EXAFS spectra of the La-rich PbTiO3 samples. In this way, except for the PLT00 (sample without La), any attempt to fit the first shell using single- or three-shell models leads to unrealistic structural models. However, it was possible to infer that the disorder in the first shell around titanium atoms persists, in accordance with the XANES and XRD results. Finally, our XANES and EXAFS results show that the Ti atoms are in an off-center position for any lanthanum content, as mentioned in previous studies using the Raman spectroscopy in PLTx samples.14-16 Acknowledgment. Research partially carried out at LNLSNational Laboratory of Synchrotron Light, Brazil. We are also grateful to FAPESP (00/09722-9), CNPq (Brazil), and USPCOFECUB for their financial support. The authors wish to thank Dra. Ducinei Garcia (DF-UFSCar) for her help during the sample preparation. We also thank Jose´ A. L. da Rocha and J. G. Catarino for the important support during the XRD measurements. A.C.D. thanks FAPESP (01/06993-4) for a postdoctoral fellowship. Y.P.M., V.R.M., and J.A.E. also thank CNPq for a researcher fellowship. References and Notes (1) Keizer, K.; Lansink, G. J.; Burggraaf, A. J. J. Phys. Chem. Solids 1978, 39, 59. (2) Garcia, D.; Eiras, J. A. Ferroelectrics 1991, 123, 51. (3) Bhaskar, S.; Majumder, S. B.; Katiyar, R. S. Appl. Phys. Lett. 2002, 80(21), 3997. (4) Dai, X.; Xu, Z.; Viehland, D. J. Appl. Phys. 1996, 51, 4356. (5) Hennings, D. Mater. Res. Bull. 1971, 6, 329. (6) Keizer, K.; Bouwama, J.; Burgraaf, A. J. Phys. Status Solid 1976, A35, 281. (7) Kuwabara, M.; Goda, K.; Oshima, K. Phys. ReV. B 1990, 42, 10012. (8) Shirane, G.; Pepinsky, R.; Frazer, B. C. Phys. ReV. 1955, 97, 1179. (9) Shirane, G.; Pepinsky, R.; Frazer, B. C. Acta Crystallogr. 1956, 9, 131. (10) Glazer, A. M.; Mabud, S. A. Acta Crystallogr. 1978, B34, 1065. (11) Nelmes, R. J.; Kuhs, W. F. Solid State Commun. 1985, 54, 721. (12) Shannon, R. D. Acta Crystallogr. 1976, A32, 751. (13) Hennings, D.; Rosenste, G. Mater. Res. Bull. 1972, 7, 1505. (14) Tae-Yong, K.; Hyun, M. J. Appl. Phys. Lett. 2000, 77, 3824. (15) Kim, T. Y.; Jang, H. M.; Cho, S. M. J. Appl. Phys. 2002, 91, 336. (16) Tavares, E. C. S.; Pizani, P. S.; Eiras, J. A. Appl. Phys. Lett. 1998, 72, 897. (17) Rietveld, H. M. J. Appl. Crystallogr. 1969, 2, 65. (18) Larson, A. C.; Von Dreele, R. B. GSAS. Generalized Structure Analysis System; Los Alamos National Laboratory: Los Alamos, NM, 1996.. (19) Rez, D.; Rez, P.; Grant, I. Acta Crystallogr. 1994, A50, 481. (20) Tolentino, H. C. N.; Ramos, A. Y.; Alves, M. C. M.; Barrea, R. A.; Tamura, E.; Cezar, J. C.; Watanabe, N. J. Synchrotron Rad. 2001, 8, 1040.

Lanthanum-Modified PbTiO3 Ceramic Materials (21) Michalowicz, A. J. Phys. IV 1997, 7, C2-235. (22) Report on the International Workshops on Standards and Criteria in XAFS. In X-ray Absorption Fine Structure: Proceedings of the VI International Conference on X-ray Absorption Fine Structures; Hasnain, S. S., Ed.; Ellis Horwood: New York, 1991; p 752. (23) Michalowicz, A.; Noinville, V. In Gallad 2.0 code, LURE, Orsay, France, 1992. (24) Farrugia, L. J. J. Appl. Crystallogr. 1997, 30, 565. (25) Ravel, B.; Stern, E. A. Physica B 1995, 208-209, 316. (26) Ravel, B.; Stern, E. A.; Vedrinskii, R. I.; Kraizman, V. Ferroelectrics 1998, 206-207, 407. (27) Veddrinskii, R. V.; Kraizman, V. L.; Novakovich, A. A.; Demekhin, Ph. V.; Urazhdin, S. V. J. Phys.: Condens. Matter 1998, 10, 9561.

J. Phys. Chem. B, Vol. 108, No. 39, 2004 14849 (28) Miyanaga, T.; Diop, D.; Ikeda, S. I.; Kon, H. Indian Geotechnical Conference 2002, in press. (29) Ankudinov, A. L.; Ravel, B.; Conradson, S. D.; Rehr, J. J. Phys. ReV. B 1998, 58, 7565. (30) Rehr, J. J.; Albers, R. C. ReV. Mod. Phys. 2000, 72, 621. (31) Sicron, N.; Ravel, B.; Yacoby, Y.; Stern, E. A.; Dogan, F.; Rehr, J. J. Phys. ReV. B 1994, 50, 13168. (32) Abramov, Y. A.; Tsirelson, V. G.; Zavodnic, V. E.; Ivanov, S. A.; Brown, I. D. Acta Crystallogr. 1995, B51, 942. (33) Teslic, S.; Egami, T.; Viehland, D. J. Phys. Chem. Solids 1996, 57, 1537. (34) Egami, T.; Teslic, S.; Dmowski, W.; Viehland, D.; Vakhrushev, S. Ferroelectrics 1997, 199, 103.