Unusual Relaxor Ferroelectric Behavior in Stairlike Aurivillius Phases

Aug 15, 2016 - Institute of Physics, Czech Academy of Sciences, Na Slovance 2, 182 21 Prague, Czechia. Inorg. ... *E-mail: [email protected]...
1 downloads 0 Views 9MB Size
Article pubs.acs.org/IC

Unusual Relaxor Ferroelectric Behavior in Stairlike Aurivillius Phases Gwladys Steciuk,† Philippe Boullay,*,† Alain Pautrat,† Nicolas Barrier,† Vincent Caignaert,† and Lukas Palatinus‡ †

Laboratoire CRISMAT, UMR CNRS 6508, ENSICAEN, 6 Bd Maréchal Juin, F-14050 Caen Cedex 4, France Institute of Physics, Czech Academy of Sciences, Na Slovance 2, 182 21 Prague, Czechia



S Supporting Information *

ABSTRACT: New ferroelectric layered materials were found in the pseudobinary system Bi5Nb3O15−ABi2Nb2O9 (A= Ba, Sr and Pb). Preliminary observations made by transmission electron microscopy indicate that these compounds exhibit a complex incommensurately modulated structure. A (3 + 1)D structural model is obtained using ab initio phasing by charge flipping based on the analysis of precession electron diffraction tomography data. The (3 + 1)D structure is further validated by a refinement against neutron powder diffraction. These materials possess a layered structure with discontinuous [Bi2O2] slabs and perovskite blocks. While these structural units are characteristics of Aurivillius phases, the existence of periodic crystallographic shear planes offers strong similarities with collapsed or stairlike structures known in high-Tc superconductors and related compounds. Using dielectric spectroscopy, we study the phase transitions of these new layered materials. For A = Ba and Sr, a Vögel−Fulcher-like behavior characteristic of the so-called relaxor ferroelectrics is observed and compared to “canonical” relaxors. For A = Sr, the absence of a Burns temperature separated from the freezing temperature appears as a rather unusual behavior.



INTRODUCTION The Aurivillius family of bismuth-layered oxides1 of general formula Bi2mAn−mBnO3(n+m) can be described as the regular stacking of [Bi2O2]2+ slabs and perovskite [Ap−1BpO3p+1]2− blocks, where p denotes the number of sheets of cornersharing BO6 octahedra (Figure 1a). Members having a unique size for all the perovskite blocks correspond to simple terms (m = 1), while intergrowth compounds correspond to m > 1 cases. The majority of them exhibit a ferroelectric behavior at room temperature and have been widely investigated as potential high-temperature lead-free piezoelectric devices, ferroelectric nonvolatile memories, oxide ion conductors, and photocatalysts.2−6 In Aurivillius phases, by tuning the chemical composition, it is possible to modify the nature of the ferroelectric−paraelectric (FE-PE) transition from a roomtemperature relaxor-like behavior to a very high-temperature sharp transition. In terms of chemistry, Aurivillius phases are characterized by a large compositional flexibility originating mainly from the presence of the perovskite blocks, since the substitution of Bi3+ by other cations in the [Bi2O2]2+ slabs is limited. In the general formula, A is ideally a 12-fold coordination cation, like Na+, K+, Ca2+, Sr2+, Ba2+, Pb2+, Bi3+, or RE3+, while B is a sixfold coordination cation. Aurivillius phases with low n values are stabilized with B cations having empty d orbitals such as W6+, Nb5+, and Ti4+. Examining the charge balance of the general © XXXX American Chemical Society

formula, it is seen that the synthesis of members with high n values, that is, with a large thickness of the perovskite block, can be promoted by introducing cations with lower valence state on the B site. This strategy has been used in the 1970s7 with the study of the Bi4Ti3O12−BiFeO3 system, where several high-n members have been stabilized. This system recently attracted a renewed interest because these compounds were reported to be ferroelectric (FE) and anti-ferromagnetic (AFM) at room temperature.8 Several groups have succeed in synthesizing compounds of the series Bin+1(Fen−3Ti3)O3n+3 up to members n = 8 using conventional solid−solid reactions,9 ball-milling,10 or sol−gel routes.11 Such Aurivillius phases with high n values are difficult to stabilize as single-phase compounds,12−15 and the formation of ordered long period simple members (m = 1) or long period intergrowths (m > 1) can hardly be considered as a potential source for new single-phase oxide ferroelectrics. Recently,16 using precession electron diffraction tomography (PEDT), we revisited the structure of Bi5Nb3O15 (referred to as type IV) confirming17,18 its incommensurate nature with the presence of aperiodic crystallographic shear planes leading to the formation of an original layered structure containing both continuous and discontinuous [Bi2O2]2+ slabs and perovskite blocks. While it possesses the structural units characteristic of Received: June 13, 2016

A

DOI: 10.1021/acs.inorgchem.6b01373 Inorg. Chem. XXXX, XXX, XXX−XXX

Article

Inorganic Chemistry

In the present paper, we show that type IV Bi5Nb3O15 can be used to produce new layered ferroelectrics having a composition ABi7Nb5O24 (A = Ba, Sr, and Pb) made of discontinuous [Bi2O2]2+ slabs and perovskite blocks resembling the type IV* phase.



EXPERIMENTAL SECTION

Polycrystalline samples of ABi7Nb5O24 (pale yellow) were synthetised by solid-state reaction. Stoichiometric amounts of Bi2O3, Nb2O5 and SrCO3, BaCO3, or PbO were mixed, pressed into pellets, and calcined in air at 1050 °C for 24 h. After they were reground and repressed, a second calcination step at 1075 °C for 24 h was applied. X-ray powder diffraction (XRPD) patterns were recorded in the 2θ range of 5−100° (scan step: 0.009°) with a Bruker D8 Advance diffractometer (Cu Kα1) equipped with a Lynx-Eye detector (see Figure S1 in Supporting Information). Neutron powder diffraction (NPD) experiments were performed on all ABi7Nb5O24 samples. SrBi7Nb5O24 and BaBi7Nb5O24 were analyzed on the 3T2 2-axis high-resolution diffractometer (Orphée reactor at the Laboratoire Léon Brillouin, Saclay, France) with an incident neutron wavelength of 1.227 Å from 4.50 to 121.20° (2θ) with a step of 0.05° (2θ). The intensities were measured at room temperature by a bank of 50 3He detectors apart 2.4° during 24 h. The PbBi7Nb5O24 NDP patterns were obtained on the D1B diffractometer (Institue Laue-Langevin, Grenoble, France) at 300 K for 1.28 Å with a range of 128° (2θ) by step of 0.2° (2θ). For all experiments, the powdered samples (∼8 g) were contained in a vanadium container (⌀ = 8 mm). For transmission electron microscopy (TEM) investigations, a small quantity ( 3σ(I).

B

DOI: 10.1021/acs.inorgchem.6b01373 Inorg. Chem. XXXX, XXX, XXX−XXX

Article

Inorganic Chemistry

Figure 2. (a) Schematic drawing of the [010] ZAP of SrBi2Nb2O9 and experimental SAED patterns of SrBi7Nb5O24 (b) and Bi5Nb3O15 (c). By analogy with (a, c), the c* direction in (b) shall be associated, in real space, with the layer stacking. From the insets, representing enlarge areas of the SAED patterns, the position of reflections in (b) is different from the ones observed for the incommensurately modulated Bi5Nb3O15 compound with a deviation from an orthorhombic lattice.

Figure 3. (a) [0100] ZAP of the incommensurately modulated SrBi7Nb5O24 phase. In red: [0010]* direction associated, in real space, to the layer stacking in Aurivillius phases. (b) Enlarged area of (a), showing how the indexation can be performed. (c) Corresponding HREM image. were painted with a conductive silver paste and annealed at 400 °C for 3 h. The XRPD were collected on each pellet without indication of a preferred orientation.

and Pb < 3% for Pb compound. Despite the tendency of the sample being damaged by a focused electron beam, high-resolution electron microscopy (HREM) image was performed with an FEI Tecnai G2 30 microscope (LaB6 cathode, 300 kV). PEDT data sets of nonoriented patterns were recorded at room temperature on several different thin crystals of ABi 7 Nb 5 O24 compounds. To investigate possible change in the crystal structure and symmetry above the FE-PE transition, PEDT data sets were recorded at 400 °C for A = Ba and Sr samples using a Gatan heating holder. For all data collections (see Table 1), the precession angle was set to 1.2° with a goniometer tilt step below 1°. PEDT data were analyzed using the computer programs PETS21 and JANA200622 following a procedure described elsewhere.16,23 For each data set the result is a list of hklm indices with associated intensities and estimated standard deviations based on counting statistics. The dielectric properties of the ABi7Nb5O24 phases were measured as a function of the temperature in the range of 100−600 °C with a systematic cooling−heating procedure. Within resolution, no thermal hysteresis was observed. Frequencies from 8 to 800 kHz were applied, and the alternating current excitation was low enough to be in the linear regime. The experimental setup consists of a lock-in amplifier (EG&G model FRD1025) connected to a potentiostat (EG&G model 283) in a potentiostatic mode. The data presented here are those that are not affected by resistive losses that occur at low frequencies/high temperatures and are responsible for a characteristics increase of the loss tangent. For dielectric measurements, powders were pressed as pellets (diameters = 8 mm, thicknesses < 0.8 mm) after addition of a vinyl polymer and sintering at 1000 °C during 24 h to give ceramics (densities: Sr = 75.7% and Ba = 75.2%). The two faces of the pellets



STRUCTURAL ANALYSIS Finding suitable cell parameters to index the XRPD patterns (Figure S1 in Supporting Information) of the new ABi7Nb5O24 phases appeared as a difficult task. The reason is obvious when one investigates the samples by TEM where the incommensurate modulated nature of the structure is clearly evidenced by selected area electron diffraction (SAED). The reciprocal space of the new compound possesses similarities with the ones observed for ABi2Nb2O9-layered Aurivillius phases (Figure 2a) but with extra satellite reflections (Figure 2b) whose positions differ from the ones observed in type IV Bi5Nb3O15 (Figure 2c) and cannot be indexed on the basis of the orthorhombic lattice used for this latter compound.16 Accordingly, the data were indexed using a monoclinic lattice with the cell parameters a ≈ b ≈ c ≈ ap√2 (ap stands for the lattice parameter of cubic perovskite), β > 90°, α ≈ γ ≈ 90° and a modulation vector in the form q = αa* + γc* (Table 1). The average unit cell shows some similarity with Aurivillius phases with a ≈ b ≈ 5.4 Å and c corresponding to the stacking direction. As illustrated in Figure 3 for A = Sr, this unit cell and satellite reflections up to order 8 allow to index the reflections present in the [0100] zone axis patterns (ZAP). C

DOI: 10.1021/acs.inorgchem.6b01373 Inorg. Chem. XXXX, XXX, XXX−XXX

Article

Inorganic Chemistry

Figure 4. (a) Projection (010) in a supercell of the electrostatic potential map for BaBi7Nb5O24. (b) Model interpreted from the raw solution after defining A/Bi (red) and Nb (green) discontinuous atomic domains using crenel functions. (c) View (010) of the NPD refined structure extended along a and c directions to emphasize the stairlike nature of the new compound. Potential map and structure visualization are done using VESTA.25

To get a better view of the reciprocal space and obtain data suitable for subsequent ab initio structure determination, we performed PEDT experiments. After the lattice parameters and modulation vector (Table 1) were refined, the symmetry was further examined using reciprocal space sections reconstructed from the PEDT data. Looking at the enlarged part of one [0100] ZAP in Figure 3b, the indexing of the h = 2n rows of strong reflections follow the condition h + l + m = 2. In the enlarged part of an (h1lm)* section in Figure S2a (Supporting Information), it appears clearly that the condition h + l + m = 2 is not valid on hklm. By examining (0klm)* and (1klm)*

sections in Figure S2b,c (Supporting Information), the condition hklm: k + l + m = 2n does exist with no additional general condition observed for the (hk0m)* section leading to a lattice-centering X = (0, 1/2, 1/2, 1/2). For PEDT data recorded above the FE-PE transition (Table 1), the most striking difference is seen on the (h0lm)* section, where the h = 2n + 1 rows are systematically absent at high temperature, while they are weak but present at room temperature as illustrated in Figure S4 (Supporting Information). This leads us to assume that a a-glide plane perpendicular to b is present above the FE-PE transition. At this point the FE-PE transition D

DOI: 10.1021/acs.inorgchem.6b01373 Inorg. Chem. XXXX, XXX, XXX−XXX

Article

Inorganic Chemistry Table 2. BaBi7Nb5O24 Crystallographic Parametersa a, Å

lattice parameters:

5.5156(9) density 8.162(3) positional parameters: atom X4/△ Bi(1) 0/0.615

x-harm s,1 s,2 s,3 s,4

Nb(1)

0.5/0.385 s,1

O(1−1)

0/0.615 s,1 c,1 s,2 c,2 s,3 c,3

O(1−2)

0/0.615 s,1 c,1 s,2 c,2 s,3 c,3

O(2)

0/0.615 s,1 c,1 s,2 c,2 s,3 c,3

profile: structure: Nobs Robs wRobs a

Rp = 2.33 No. ref param = 49 all main 5172 483 1.72 1.72 2.17 2.40

b, Å 5.5001(9) q

c, Å

β, deg

V, Å3

5.2033(12) α −0.0327(8)

90.909(15) γ 0.2304(6)

157.71(7)

Uiso, Å3 0.0189(6)

x/a

y/b

z/c

0 0 −0.0290(14) −0.0420(16) −0.0285(17) 0 0.0050(10) 0.2348(10) 0 −0.0180(17) −0.0100(17) 0 0 0 0.7275(9) 0.0155(14) −0.0264(15) 0 0 0.0055(18) 0) 0.0100(14) 0 −0.0417(18) −0.0075(14) 0 0.0215(18) −0.0013(2) wRp = 3.07

0 0 0 0 0 0 0 0.7654(7) 0 0 0 0 0 0 0.7264(7) 0 −0.0369(16) 0.0053(15) 0 0 0 0.4909(12) 0 0 −0.0511(16) 0 −0.028(2) 0

0 −0.0020(6) 0.0248(12) −0.0110(16) −0.0264(16) 0 −0.1002(8) 0.5026 (16) −0.239(2) 0.013(2) −0.039(2) 0.024(2) −0.020(2) 0 0.5026(15) −0.2129(18) 0.0201(19) −0.009(2) 0.017(2) 0.007(2) 0 0.0030(11) −0.2452(9) 0 −0.0145(11) 0 0 0 GOF = 3.91

order 1 922 2.13 2.75

order 2 944 1.77 2.17

order 3 948 1.71 2.16

0.0041(4) 0.0092(5)

0.0092(5)

0.0158(11)

order 4 930 1.48 1.78

order 5 945 1.29 1.81

Obtained from NPD Rietveld refinement with SSG X21(α0γ)0 X = (0, 1/2, 1/2, 1/2) and origin shift of 1/4, 0, 1/4, 1/4.

interpretation of the raw solution by adding crenel functions to describe the A-site (Bi/Ba), B-site (Nb), and oxygen atomic domains (AD). These discontinuities in the AD are clearly visible in the sections through the electrostatic potential reconstructed by Superflip (Figure S5). They were modeled using x-harmonics in the crenel interval assigning the highest density areas to Bi. At this stage, compositions are reduced to a B-site deficient perovskite, that is, (A,Bi)8/8Nb5/8O24/8 approximated to BiNb0.625O3 disregarding the presence of A cations. In the solution, Bi and Nb share the same atomic position in the average three-dimensional cell but have a different center for the crenel function (X4 coordinate). The widths of crenel (Δ) were then associated with their cationic ratios: Δ(Bi) = (Bi)/ (Bi + 0.625Nb) ≈ 0.615 and Δ(Nb) = 1 − Δ(Bi) ≈ 0.385. The widths of oxygen crenels were placed assuming a full occupancy in compliance with the existence of [Bi2O2] slabs and perovskite blocks; that is, Δ(O1) = 0.615, and Δ(O2) = 0.615. The obtained structural model can be described with

is thus assumed to be accompanied by a change from the centrosymmetric (CS) superspace group (SSG) X21/a(α0γ)00 to the noncentrosymmetric (NCS) SSG X21(α0γ)0. While the structure solution was conducted on all compounds (see Table 1), it will be illustrated based on the analysis of the BaBi7Nb5O24 PEDT data recorded at room temperature. At this stage we decided to use the CS SSG X21/ a(α0γ)00 for performing the structure solution with the program Superflip.24 Merging the data sets obtained from three different crystallites of BaBi7Nb5O24 (Table 1) provides an almost complete data set (at 0.714 Å resolution shell) that allows to obtain a density map where all the atomic positions can be identified. The completeness is here especially important to localize the oxygens in the structure (at least 90%), while the cationic positions can come out even with a rather low completeness (∼65%). In the density map presented in Figure 4a, the oxygens belonging to the [Bi2O2] slabs and those of the AO and BO2 layers of the perovskite blocks are clearly revealed. The structural model presented in Figure 4b corresponds to the E

DOI: 10.1021/acs.inorgchem.6b01373 Inorg. Chem. XXXX, XXX, XXX−XXX

Article

Inorganic Chemistry only four atomic positions in the X21/a(α0γ)00 symmetry (see Table S1 in Supporting Information). Before refinement, the structural model plotted in Figure 4b exhibits an alternation of discontinuous [Bi2O2] layers and p = 1/p = 2 perovskite blocks having a stacking sequence |p = 1|p = 2|p = 2| along c. Octahedral tiltings are present around the a and c directions in this solution, where we neglected the weak reflections corresponding to h = 2n + 1 on (h0lm)*. Importantly, this structural model allows to fit the XRPD and NPD patterns (see Figure S3 in Supporting Information), giving indication that, first, the structure solution from PEDT data was successful, and second, the deviation from X21/ a(α0γ)00 is indeed small and shall affect mostly the oxygens atoms. These structures were then compared to those obtained after lowering the symmetry to X21(α0γ)0 in agreement with the symmetry analysis based on PEDT data. Because of the weak X-ray scattering strength of oxygen, the refinement against XRPD data is not sensitive (no significant changes in Rvalues) to this lowering of symmetry, where only one oxygen position is doubled. Unlike XRPD, NPD refinements allowed us to confirm the SSG X2 1 (α0γ)0 for all ABi 7 Nb 5 O 24 compounds at room temperature (see Table S2 in Supporting Information). For this reason a combined XRPD/NPD refinement appears not relevant, and for the final structure refinement, only NPD refinements were considered. The final structure arising from the NPD Rietveld refinement is given in Figure 4c using BaBi7Nb5O24 as an illustration. Refinement and structural parameters are summarized in Table 2, and profile fitting is given in Figure S6 (Supporting Information; see Tables S3 and S4 in Supporting Information for SrBi7Nb5O24 and PbBi7Nb5O24, respectively).



DIELECTRIC PROPERTIES

Investigating the signature of the expected FE-PE transition, the dielectric spectroscopy measurements performed on ABi7Nb5O24 (A = Ba and Sr) samples reveal the presence of a maximum in the dielectric permittivity εr at a temperature Tm simultaneously with the appearance of nonzero losses (Figure 5a). As shown in Figure 5b,c, Tm increases with the applied frequency f = 1/τ. The frequency dispersion and its temperature dependence are similar to those of relaxor ferroelectrics with a broad maximum characterized by ΔTm (1 × 103 to 1 × 106 Hz) = 97 and 98 °C for Sr and Ba samples, respectively. For the Pb sample, no transition is detected up to 600 °C, where the measurements start to be dominated by resistive processes. In relaxors, the relaxation process is non-Arrhenius, and the characteristic relaxation time is described by a phenomenoEa logical Vögel−Fulcher relationship τ = τ0exp[ ]. Here,

Figure 5. (a) Relative dielectric permittivity and loss tangent as a function of the temperature for the Sr sample ( f = 800 kHz; see text for Tm). (b) Relative dielectric permittivity as a function of the temperature for the Sr sample. The applied frequencies are, respectively, f = 8, 25.2, 80, 252, 800 kHz. (inset) 1/εr as a function of T for f = 800 kHz. Note the linear variation consistent with a regular Curie−Weiss behavior down to T ≈ Tm. (c) Relative dielectric permittivity as a function of the temperature for the Ba sample. The applied frequencies are, respectively, f = 8, 23.3, 62, 287, 480, 800 kHz. (inset) 1/εr as a function of T for f = 800 kHz. The star indicates the temperature TB, where a departure from simple Curie−Weiss behavior can be noted (Burns temperature).

activation of polar cluster fluctuations is strongly constrained. For the rest, the characteristics of the phase transition in these phases are clearly different form those of the parent phases, that is, BaBi2Nb2O9 (relaxor with Tm ≈ 180 °C at 100 kHz), SrBi2Nb2O9 (ferroelectric with Tc ≈ 460 °C),31 and Bi5Nb3O15 (ferroelectric with Tc ≈ 360 °C).32 From dielectric spectroscopy measurements, it is possible to estimate the more or less “diffuse” nature of the phase transition in relaxors.30 Hence, in analogy with the spin glass case,33 it is expected that preformed polar clusters are present at high temperature TB > Tm,34 where TB is the Burns temperature. For T < TB, the Curie−Weiss law of normal ferroelectrics in their paraelectric state is not followed anymore.34 Within experimental accuracy, we do not find evidence for a departure from the Curie−Weiss form of the

k(Tm − Tf )

Tf stands for the freezing temperature in the quasi-static limit, and Ea is the activation energy corresponding to the fluctuations of local polar clusters. Below Tf, the time scale is so slow that the system is expected to behave as a regular ferroelectric, even if it does not present long-range polar order. We find that Tm is well-described by Vögel−Fulcher law (Figure 6a) with Tf = −6 °C, Ea = 298 meV and Tf = −46 °C, Ea = 516 meV, respectively, for Sr and Ba samples. Compared to relaxors such as PbMg1/3Nb2/3O3 and PbZn1/3Nb2/3O3, which present Ea ≈ 20−30 meV,26,27 these are relatively large values but characteristic of what is reported in m = 2 Ba-based Aurivillius phases28−30 such as BaBi2Nb2O9 (Ea ≈ 600 meV). This large activation energy indicates that, in these phases, F

DOI: 10.1021/acs.inorgchem.6b01373 Inorg. Chem. XXXX, XXX, XXX−XXX

Article

Inorganic Chemistry

[Bi2O2] and octahedral layers. Our crystallographic model (Table 2) does not directly translate such an intuitive description. Concretely the γ component of the modulation vector controls the layer stacking sequence along c (Aurivillius nature) and is, within uncertainties, independent of the nature of A cations. Fixing α to 0, our model generates long-period Aurivillius intergrowths with continuous [Bi2O2] and octahedral layers with a stacking sequence of |p = 1|p = 2|p = 2|. The α-component of the modulation vector controls the length of the discontinuous layers along a (stairlike nature) and increases significantly with the ionic radius of A2+. As inferred from temperature PEDT experiments performed on the A = Ba and Sr samples, these compounds undergo a structural phase transition above room temperature accompanied by an apparent modification of the SSG from X21/ a(α0γ)00 to X21(α0γ)0. This has some impact on the roomtemperature NPD Rietveld refinements, where a better fit could be obtained using the NCS SSG, the difference being more clear for the A = Pb compound. Structurally the NCS to CS transition is related to the apparition of an extra degree of freedom in the tilting system of the NbO6 octahedra. In the CS SSG X21/a(α0γ)00, octahedral tiltings are possible along a and c directions. In the NCS SSG X21(α0γ)0, a supplementary rotation is possible around the b direction. Like in Aurivillius phases36,37 and other layered systems,38 such a transition can induce the apparition of a polarization following a so-called “hybrid improper” mechanism, where the FE-PE transition is driven by the coupling of nonpolar lattice distortions rather than the instability of a single polar mode. From the NPD structure refinements, the A = Pb compound shows stronger amplitudes in the rotation of the NbO6 octahedra (Figure 7) compared to both A = Ba and Sr compounds. By analogy with the layered Aurivillius phases,39 with stronger amplitudes in the rotation of the NbO6 octahedra one would expect a higher FEPE transition temperature for the A = Pb compound than for the two others. For the ABi7Nb5O24 (A = Ba and Sr) phases, the FE-PE transition is accompanied by a relaxor-like behavior that can be tentatively examined in light of structure−properties relationships already observed in Aurivillius phases.40−42 First, the fact that only subtle changes are involved when describing the structures using the CS SSG X21/a(α0γ)00 or the NCS SSG X21(α0γ)0 is something quite common in relaxor Aurivillius phases, where the establishment of a long-range order polar structure is unclear and might vary from one crystallite to the other. Second, the presence of A cations in the Bi2O2 layers must be addressed, since the existence of positional static disorder in this layer is one of the “structural features” often found in relaxor Aurivillius phases. In our SSG crystallographic description, all A/Bi sites are described with one single Bi atomic position, while 12.5% of the occupancy should be Ba or Sr cations. One can use the larger difference in scattering power between Ba and Bi to explore possible A/Bi intermixing by looking at the evolution of the Bi(1) site anisotropic displacement parameters (ADPs) with the internal parameter t. Larger ADPs can be caused by other reasons than just a local difference in scattering power and cations intermixing, but this shall still give us some indication. Adding and refining a modulation wave for the Bi(1) ADP of BaBi7Nb5O24 led actually to a broad maximum at t = 0 corresponding to A/Bi sites located in the middle of the p = 2 perovskite block of the structure. The bond-valence-sum analysis (Figure S7 in Supporting Information) does corroborate this with an

Figure 6. (a) Naperian logarithm of the frequency as a function of the temperature of maximum permittivity for the two samples. The solid lines are best fits using a Vogel−Fülcher law. Parameters are discussed in the text. (b) 1 − ε/εm as a function of T/Tm − 1 in a log−log scale (εm is the value of permittivity at Tm). In this scale, slopes of 1 and 2 correspond, respectively, to Curie−Weiss and diffuse behaviors.

dielectric susceptibility down to T = Tm; that is, 1/εr is a linear function of temperature for Tm < T < 600 °C (see the insets of Figure 5b,c) with no degree of diffuseness in the phase transition for the Sr sample (Figure 6b). At least from the dielectric susceptibility, we conclude on the absence of a Burns temperature separated from the freezing temperature; that is, Tm ≈ TB. This is a rather unusual behavior, since almost all the relaxors exhibit TB > Tm, with a very large separation between the two temperatures. For the Ba sample, a departure from a simple Curie−Weiss susceptibility can be evidenced as shown in Figure 6b, with a temperature variation εr ∝ (T − Tm)−γ with γ ≈ 2 for Tm < T < Tm + 80 °C. This is consistent with a Burns temperature of TB ≈ 378 °C.



DISCUSSION From the analysis of PEDT data, we solved the complex structure of the ABi7Nb5O24 compounds and subsequently refined their structure using NPD data. Looking at the SAED patterns, HREM image (Figure 3), and structure projection (Figure 4c), some similarities with the collapsed or stairlike phases derivative from the layered HTc superconductors can be seen (see L. Elcoro et al.35 and references therein). Being constituted by structural units related to Aurivillius phases, we proposed to name these phases stairlike Aurivillius to distinguish them from classical layered Aurivillius phases. Considering a hypothetic stacking sequence |p = 1|p = 2|p = 2| along c with continuous [Bi2O2] and octahedral layers, the structure of ABi7Nb5O24 compounds can be generated by the introduction of {100} shear planes, where the stacking sequence is translated by 5c/2 (Figure 4c) giving discontinuous G

DOI: 10.1021/acs.inorgchem.6b01373 Inorg. Chem. XXXX, XXX, XXX−XXX

Article

Inorganic Chemistry

Figure 7. (010) and (001) representations of a part of the structures of (a) BaBi7Nb5O24, (b) PbBi7Nb5O24, and (c) SrBi7Nb5O24 as refined from NPD data. The A/Bi, Nb, and O atomic positions are represented, respectively, in red, green, and blue. NbO6 octahedra are represented in green.

Supporting Information). The evolution of the A/Bi environment in the Bi2O2 layers is better visualized in the structure drawing Figure 8. As in other Aurivillius phases, four short Bi− O distances (