Introduction

One important direction in the developments of novel fast ionic conductors is associated with nano-heterogeneous solids where the concentration and mobility of ionic defects at the interfaces are both higher than those in the bulk [15]. The cation-conducting nanocomposites may be relatively easily prepared by mixing melts of the corresponding ionic salts with nano-sized powder of a refractory oxide, such as Al2O3, chemically stable with respect to the melt [46]. Such an approach is not applicable, however, for the oxygen ion-conducting composites, which are usually characterized by rather high melting points of the components and by their chemical interaction during preparation. A simple and versatile method to form nano-heterogeneous oxide systems relates to the use of “disorder ↔ order” phase transitions. For instance, the A2B2O5 brownmillerite crystal lattice can be derived from the cubic ABO3 perovskite by removing and long-range ordering of 1/6 oxygen anions in parallel rows. Alternatively, oxygen removal from perovskite oxides may lead to the formation of a brownmillerite nanodomain structure [79]. This phenomenon was extensively studied for mixed-conducting strontium ferrites doped with pentavalent cations, e.g., SrFe1−xVxO3−δ and SrFe1−xNbxO3−δ [10, 11], where the domain size and volume fraction both depend on the dopant content. Although undoped Sr2Fe2O5 undergoes the brownmillerite to perovskite phase transition on heating above ∼800 °C [12], the oxygen vacancy-ordered domains may still persist and coexist with a disordered cubic perovskite matrix at higher temperatures. Due to large interfacial boundary and effects on the mobile vacancy concentration in both phases, the domain structure should significantly impact on ion transport.

On the contrary to nano-heterogeneous systems based on ionic salts where the interface morphology is essentially governed by alumina particles, the micro- and nano-structure of the oxide composites with vacancy-ordered and disordered domains should be temperature-dependent. Another important feature is related to the difference between the temperature range where oxygen ion transport becomes significant (typically above 500 °C) and temperatures that can be experimentally used for local structure studies. Up to now, the latter analysis can only be performed at room temperature or under moderate heating, far from the prospective application conditions. As the high-temperature defect states may not always be quenched, in many cases, the relationships between the domain formation and transport properties cannot be assessed by any direct experiment. Nevertheless, enhanced catalytic activity toward partial oxidation of methane and high oxygen ion diffusivity in SrFe1−xAlxO3−δ ceramics at elevated temperatures seem to manifest the presence of morphotropic phase boundaries due to coexisting perovskite- and brownmillerite-like domains [13, 14]. In SrFe1−xTaxO3−δ system at x ≥ 0.3, the anion conduction at 700–950 °C was found to increase following the reduction and separation of Ta-enriched double perovskite and an oxygen-deficient disordered phase with primitive cubic perovskite structure [15, 16]. One can expect that variations in domain sizes and volume ratio may further facilitate ion diffusion.

The present work was centered on the analysis of relationships between structural features and transport properties of SrFe1−xТаxO3−δ with moderate dopant concentration favorable for the formation of nano-heterogeneous states with perovskite- and brownmillerite-type domains. Though incorporation of higher-valence cations into the iron sublattice of strontium ferrite seems to be a necessary prerequisite for forming stable domain structures (e.g., [12] and references therein), the vacancy-ordering processes may be suppressed at excessively high dopant contents when the oxygen deficiency becomes exceedingly small [15, 16]. Particular emphasis in this work was focused on the effects of redox cycling, which may substantially alter ion conduction due to the vacancy-ordered domain reconstruction [17].

Experimental

The procedure used for solid state synthesis of single-phase SrFe1−xTaxO3−δ (х = 0.03 − 0.10) powders at atmospheric oxygen pressure was similar to that described elsewhere [15, 16]. The ceramic samples with ∼93 % density were sintered at 1350 °C in air with subsequent slow cooling in order to achieve equilibrium oxygen stoichiometry. For selected samples used for x-ray diffraction (XRD) and transmission electron microscopy (TEM) studies, an additional reductive treatment during 10 h was performed at 950 °C in Ar–CO–CO2 gas mixture where the oxygen partial pressure (\( {p}_{{\mathrm{O}}_2} \)) was adjusted to ∼10−12 atm.

The XRD patterns were recorded for powdered samples using a STADI-P (STOE) diffractometer (CuK α -radiation) in 2Θ angle range 10–115 ° with a ΔΘ step 0.03 ° and counting for 60 s at each step. Polycrystalline silicon was used as an external standard (a = 5.43075(5) Å). The lattice parameters and volume fractions of constituent phases were calculated employing PCW 2.4 software (http://powdercell-for-windows.software.informer.com/2.4). The TEM images and electron diffraction patterns were obtained with the help of a JEM-200CX and a Philips CM-30 electron microscope. For these studies, the powders were dispersed in isobutyl alcohol bath by ultrasonic stirring; then, the sediments were deposited onto a copper microgrid coated with a thin holey carbon film. For the electron diffraction pattern simulations, CaRIne Crystallography 4.0 software (http://carine.crystallography.pagespro-orange.fr) was employed. Mössbauer spectra were collected at 295 and 4 K in transmission mode using a conventional constant-acceleration spectrometer and a 25 m Ci 57Co source in an Rh matrix. The velocity scale was calibrated using α-Fe foil. Isomer shifts (IS) are given relative to this standard at room temperature. The absorbers were obtained by packing the powdered samples (5 mg of natural Fe/cm2) into a perspex holder. The 4 K spectra were collected in a bath cryostat with the sample immersed in liquid He. The spectra were fitted to Lorentzian lines using a non-linear least-square method.

The total electrical conductivity (σ) was measured by 4-probe d.c. technique, varying \( {p}_{{\mathrm{O}}_2} \) in the range from 0.5 atm down to 10−20 atm and backward in isothermal conditions. These measurements were carried out at temperatures from 950 down to 700 °C with 50 °C steps. The oxygen pressure was set and controlled in dry O2–CO2 and CO–CO2 atm using yttria-stabilized zirconia (YSZ) electrochemical cell, equipped with an oxygen pump and a sensor, as described elsewhere [18]. The experimental data points were collected only when equilibration was achieved of a sample with the ambient gaseous phase at a given oxygen pressure and temperature; the equilibrium criterion was selected as d(log[σ/S ⋅ cm− 1])/dt < 0.001 min− 1. Since the oxygen pressure interval 10−10–10−5 atm is characterized with very slow equilibration kinetics and, consequently, with poor reliability of the results at moderate temperatures [19], this \( {p}_{{\mathrm{O}}_2} \) range was excluded from the measurements below 950 °C. The standard protocol for the conductivity measurements included also a reproducibility check after each \( {p}_{{\mathrm{O}}_2} \) cycle and after the entire temperature cycle.

Results and discussion

Crystal structure and iron states in oxidized Sr(Fe,Ta)O3−δ

XRD analysis showed that all as-prepared SrFe1−xТаxO3−δ (x = 0.03 − 0.10) powders and ceramics equilibrated with atmospheric oxygen are single-phase and have a cubic perovskite structure, in agreement with Mössbauer spectroscopy. As an example, Fig. 1 presents one XRD pattern and Mössbauer spectra of as-prepared SrFe0.93Ta0.07O3−δ . The 4 K Mössbauer spectrum of oxidized SrFe0.93Ta0.07O3−δ equilibrated in air can be analyzed by three magnetic hyperfine field (B hf) distributions. The IS and B hf values (Table 1) are typical of Fe3+ in the case of the main distribution, and of Fe4+ for the remaining two distributions. At room temperature, only two doublets may be adequately fitted. The estimated IS can be attributed to Fe4+ for the doublet with smaller relative area, and to an intermediate oxidation state (3+/4+) for the larger doublet. The electron system of oxidized SrFe0.93Ta0.07O3−δ is, therefore, partially delocalized with higher electron density at the Fe3+/4+ sites and lower electron density at more localized Fe4+, similar to SrFeO2.87−δ (δ < 0.1) [20]. On cooling, the delocalized Fe3+/4+ states undergo a charge-ordering transition resulting in the formation of localized Fe3+ and Fe4+ [21]. Another important observation is that at 4 K, the fraction of Fe4+ states with higher localization degree, which also exist at room temperature, is close to the fraction of iron having one or more Ta5+ cations in the second coordination sphere. In fact, assuming a random distribution of the B-site cations, the fraction of iron with at least one neighboring Ta5+ ions is equal to 35 %. The corresponding Fe4+ states have a lower B hf compared to the other Fe4+ cations, which is consistent with the presence of diamagnetic Ta5+ in the nearest neighborhood. The TaO6 octahedra stabilizing the cubic perovskite lattice are expected to increase local distortions of the iron-oxygen bonds, thus promoting electronic localization in the nearest coordination sphere. The Fe4+ states originating from Fe3+/4+ disproportionation have IS similar to that of tetravalent iron cations in SrFeO3, in accordance with the more delocalized nature of these states. Finally, whatever the microscopic mechanisms of the electronic defect formation, Mössbauer spectroscopy unambiguously shows an absence of any traces of tetrahedrally coordinated Fe3+ and, hence, brownmillerite-like domains in oxidized SrFe0.93Ta0.07O3−δ . Similar conclusions were also drawn for other as-prepared Sr(Fe,Ta)O3−δ equilibrated with atmospheric oxygen [16].

Fig. 1
figure 1

Room temperature XRD pattern of as-prepared SrFe0.93Ta0.07O3−δ ceramics equilibrated in air (a), and Mössbauer spectra of this material measured at room temperature (b) and 4 K (c)

Table 1 Parameters estimated from the Mössbauer spectra of oxidized SrFe0.93Ta0.07O3−δ at 4 and 295 K

Phase composition and domain structure in reduced SrFe1-xTaxO3−δ

In the case of SrFe1-xTaxO3−δ subjected to reducing treatment at \( {p}_{{\mathrm{O}}_2} \) ≈ 10−12 atm and T = 950 °C, two phases, brownmillerite- and perovskite-like, were found to coexist in the composition range 0 ≤ x < 0.07 while in the region 0.07 ≤ x ≤ 0.10 of higher tantalum content, the cubic perovskite-like phase (space group \( Pm\overline{3}m \)) becomes prevailing. As expected, the volume fraction of the brownmillerite-type phase (s.g. Ibm2) estimated from the XRD data tends to rapidly decrease with increasing tantalum content, from 85 % in undoped SrFeO3−δ down to 30 % in SrFe0.97Ta0.03O3−δ and to about 1 % in SrFe0.93Ta0.07O3−δ . These changes can be illustrated by characteristic fragments of XRD patterns in the vicinity of the (220)cub Bragg peak (Fig. 2). As compared to undoped SrFeO3−δ , the x-ray diffraction peaks of the brownmillerite phase in SrFe0.97Ta0.03O3−δ are substantially broadened due to pronounced grain size effect. As estimated from full-profile refinement the average size of brownmillerite-like domains is about 70 nm for SrFeO3−δ and 30–40 nm for SrFe0.97Ta0.03O3−δ . Further increase of the tantalum content up to x = 0.07 results in even smaller brownmillerite-like domains with the size of 3–4 nm as estimated from the broadening of the (220)cub reflection in SrFe0.93Ta0.07O3−δ , Fig. 2. Similar correlations between phase composition and crystallite size of the brownmillerite-type component were reported for reduced SrFe1-xVxO3−δ with nanodomain structure [10].

Fig. 2
figure 2

The experimental (solid line) and calculated (dash and dotted lines) XRD patterns for the dual-phase reduced SrFe1-xTaxO3−δ at 2Θ = 60–70 °

The features of the domain structure in reduced SrFe1-xTaxO3−δ (0 < x ≤ 0.10) were analyzed by TEM. Typical examples of the selected-area electron diffraction (SAED) patterns for SrFe0.97Ta0.03O3−δ are displayed in Fig. 3. In addition to the Bragg spots of the cubic perovskite phase in Fig. 3a, superstructural reflections are observed on the [100]cub diffraction pattern at positions <hkl> *cub + 1/2 <001> cub* and <hkl> *cub + 1/2 <010> cub* that can be assigned to brownmillerite-like superstructure. Simulation of the composite diffraction pattern showed that the superstructure reflection arrangement for the [100]cub SAED pattern includes two cross-sections of the reciprocal brownmillerite lattice with [10-1]br zone axis. These are contributed by the domains with two orientation relationships between the perovskite matrix and the brownmillerite crystallite lattice, namely [010]br || [001]cub and [010]br || [010]cub. The presence of the superstructure domains is also evidenced by SAED patterns with zone axes [110]cub, [120]cub, and [115]cub, Fig. 3b–d. The zone axes of brownmillerite phase diffraction patterns calculated for three orientations of the domains are summarized in Table 2. In addition, the lattice images along [100]cub axis display two areas with the periodic fringes distanced at approximately 0.8 nm, in a good agreement with the brownmillerite period d 020br = 0.7728 nm, Fig. 4. Again, the arrangement of these lattice fringes corresponds to two orientations: [010]br || [001]cub and [010]br || [010]cub. The typical size of the brownmillerite domains in SrFe0.97Ta0.03O3−δ is estimated to be about 20 nm. The lattice image in Fig. 4 reveals also a region with the cubic perovskite structure where the lattice fringe spacing is equal to 0.4 nm.

Fig. 3
figure 3

SAED patterns of reduced SrFe0.97Ta0.03O3−δ and corresponding simulated composite diffraction patterns for the brownmillerite-type domain structure: a zone axis [100]cub; b zone axis [110]cub; c zone axis [120]cub; d near the zone axis [115]cub. The cubic perovskite reflections (encircled) are indexed and the brownmillerite reflections are marked by arrows. Brownmillerite-type domains: [010]br || [001]cub (diamond), [010]br || [010]cub (dark circle), [010]br || [100]cub (cross). The squares show diffuse spots associated with intersections of the nearest smeared brownmillerite reflections and Ewald sphere

Table 2 Zone axes for the brownmillerite phase domains in SAED patterns, Fig. 4, as calculated for three mutual orientation relationships between perovskite and brownmillerite crystal lattices
Fig. 4
figure 4

HREM image of reduced SrFe0.97Ta0.03O3−δ . Brownmillerite domains of two orientations, [010]br||[010]cub and [010]br || [001]cub, are observed in cubic perovskite matrix; d 010(cub) = 0.4 nm, d 010(br) = 0.8 nm

Although the fraction of the brownmillerite-like phase revealed by XRD in reduced SrFe0.93Ta0.07O3−δ was apparently small (∼1 vol.%), the electron diffraction data made it possible to identify a much larger content of the brownmillerite domains, up to 30–40 vol.%. The [100]cub SAED pattern of SrFe0.93Ta0.07O3−δ is shown in Fig. 5 where superstructural reflections can be clearly observed from the brownmillerite domains with different orientations. Furthermore, vestigial traces of brownmillerite-type nanodomain structure are also observed in SrFe0.90Ta0.10O3−δ . The dark-field images reveal small regions with the size of about 5–7 nm, which are most likely associated with early stages of the local inhomogeneities formation in the perovskite matrix, Fig. 6a. Notice that the dark-field image obtained from the reflection of the cubic phase matrix displays no diffraction contrast, Fig. 6b.

Fig. 5
figure 5

SAED pattern of reduced SrFe0.93Fe0.07O3−δ , displaying superstructural reflections for two orientations of the brownmillerite domains. The cubic perovskite reflections (encircled) are indexed

Fig. 6
figure 6

Dark-field TEM images of reduced SrFe0.90Ta0.10O3−δ : a in the (022)br direction of the brownmillerite superstructure, [100]br zone axis; b in the (1\( \overline{1} \)1)cub matrix direction, [110]cub zone axis. The reflections are marked on the corresponding SAED patterns (insets). The formed nanodomains of the brownmillerite-type phase, shown by the arrow, are observed in the image (b) with the superstructure reflection. The cubic perovskite reflections are indexed

Therefore, the incorporation of even a small amount of tantalum in the crystalline lattice of reduced SrFeO2.5+δ promotes the “brownmillerite → perovskite” phase transformation which leads to formation of a peculiar, structurally inhomogeneous state where oxygen vacancy-ordered brownmillerite-type domains occur interspersed in the cubic perovskite-like matrix with disordered vacancies. The domain size and content of the brownmillerite phase both decrease when the tantalum concentration increases. These results clearly correlate with the obtained data on oxygen ion and electron conductivities, presented below.

High-temperature transport properties

The oxygen partial pressure dependencies of the total conductivity of SrFe1-xTaxO3−δ are shown in Fig. 7. One can see a series of specific features typical for perovskite-related ferrites and their derivatives ([13–17, 20, 23] and references therein). In oxidizing atmospheres, the conductivity is predominantly p-type electronic; the temperature increase and/or oxygen pressure decrease both lead to progressive oxygen losses from the crystalline lattice and consequently, to a decrease of the hole concentration and conductivity values. On further reduction, the concentrations of p- and n-type electronic charge carriers become comparable. At oxygen partial pressures below 10−9 − 10−12 atm, the total conductivity isotherms exhibit minima associated with pn transitions. The minimum at a given temperature is characterized by nearly equal concentrations of Fe2+ and Fe4+ cations. In the vicinity of these minima, the partial ion conductivity (σ i) can be assumed \( {p}_{{\mathrm{O}}_2} \)-independent since the oxygen nonstoichiometry variations are very small [18, 19]. Notice that oxygen ion transfer in perovskites occurs via the oxygen vacancy migration mechanism [21]. The total conductivity in these conditions can be approximated by the classical relationship

$$ \sigma \left(T,{p}_{{\mathrm{O}}_2}\right)={\sigma}_{\mathrm{i}}(T)+{\sigma}_{\mathrm{n}}^0(T){p}_{{\mathrm{O}}_2}^{-1/4}+{\sigma}_{\mathrm{p}}^0(T){p}_{{\mathrm{O}}_2}^{1/4} $$
(1)

where σ 0n and σ 0p are the partial electron and hole conductivities at unit oxygen pressure, respectively. The fitting results obtained using Eq. (1) are shown by thick solid lines in Figs. 7 and 8. These results are in excellent agreement with the experimental data, validating the use of Eq. (1). On the other hand, the conductivity behavior of SrFe0.97Ta0.03O3−δ differs substantially from those of SrFe0.93Ta0.07O3−δ and SrFe0.90Ta0.10O3−δ , although the dopant concentrations in these materials are rather close to one another. First, the conductivity of the former material at the minima is somewhat smaller, Fig. 7. Second, the conductivity of SrFe0.97Ta0.03O3−δ gradually increases on cycling the oxygen partial pressure, Fig. 8a, and temperature, Fig. 8b. Notice that measurements illustrated by Fig. 8a were performed in the isothermal regime until the data points collected during two consecutive runs became equal to one another within the limits of experimental uncertainty.

Fig. 7
figure 7

Oxygen partial pressure dependencies of the total conductivity of as-prepared SrFe1−xTaxO3−δ . Thick red lines show the fitting results according to Eq. (1)

Fig. 8
figure 8

Total conductivity variations in SrFe0.97Ta0.03O3−δ on redox cycling around the conductivity minima at 950 °C (a), and the second cycle of conductivity measurements for SrFe0.97Ta0.03O3−δ at different temperatures (b). Thick red lines show the fitting results according to Eq. (1). The numbers in the legend for (a) correspond to the numbering of the isothermal redox cycles

The partial contributions to conductivity were extracted from experimental data with the help of Eq. (1), and are listed in Table 3. Both σ 0n and σ 0p remain essentially invariable on redox cycling while ion conductivity values increase and tend to stabilization only after the fifth run. For instance, the oxygen pressure cycling at 950 °C results in about threefold increase of the ion conductivity compared to its initial values. This feature is quite unique and cannot be explained by any increase in the oxygen vacancy concentration, which directly correlates with the concentration of electronic charge carriers [1316, 18, 19] via the equilibrium reactions

Table 3 Regression parameters of the total conductivity vs. oxygen partial pressure dependencies in the course of redox cycling of SrFe0.97Ta0.03O3−δ at 950 °C
$$ \begin{array}{cc}\hfill 2{\mathrm{Fe}}^{3+}+{\mathrm{O}}^{2-}\rightleftarrows 2{\mathrm{Fe}}^{2+}+{\mathrm{V}}_{\mathrm{O}}+\frac{1}{2}{\mathrm{O}}_2,\hfill & \hfill 2{\mathrm{Fe}}^{3+}\rightleftarrows 2{\mathrm{Fe}}^{2+}+{\mathrm{Fe}}^{4+}\hfill \end{array} $$
(2)

Therefore, the effect of redox cycling-induced conductivity enhancement may only be associated with increasing mobility of ionic charge carriers, originating from alterations of the domain structure. One may suggest, in particular, that the cycling of oxygen deficiency in SrFe0.97Ta0.03O3−δ dual-phase composite is accompanied by a slow rearrangement of the domain structure due to resegregation and fragmentation of brownmillerite-like phase inclusions. The rate of the supposed structural changes is to be small enough so that the equilibrium criterion for conductivity can be fulfilled at every measurement. On the whole, this process is expected to increase the population of mobile oxygen vacancies at the domain interfaces. It should be stressed again that such a behavior was observed exclusively in SrFe0.97Ta0.03O3−δ , which contains about 30 % brownmillerite-like phase after the first reduction cycle at 950 °C followed by quenching down to room temperature. For the materials with higher tantalum concentration and, hence, with lower brownmillerite fraction, all the conductivity data arrays are characterized by an excellent reproducibility on redox cycling.

After \( {p}_{{\mathrm{O}}_2} \) cycling and domain structure stabilization, the shape of the conductivity isotherms of SrFe0.97Ta0.03O3−δ becomes similar to that of the materials with x = 0.07 and 0.10, cf. Figs. 7 and 8b. The ion conductivity values in these materials also become very similar, Fig. 9. Table 4 lists the calculated activation energies for partial oxygen ion and n-type electron conductivities at \( {p}_{{\mathrm{O}}_2} \) = 10−16 atm (E i and E n, respectively). In accordance with the qualitative hypothesis described above, the energetic barrier for ion diffusion reflected by E i tends to moderately decrease after the domain structure rearrangement while E n remains unchanged. Analysis of the reconstruction mechanism and interfacial morphology changes requires further TEM studies, which are now in progress. Nonetheless, the observed phenomenon makes it possible to develop methods for tuning oxygen ion conduction independently of electron transport in perovskite-based composites.

Fig. 9
figure 9

Arrhenius plots for n-type electron conductivity at \( {p}_{{\mathrm{O}}_2} \) = 10−16 atm (a) and oxygen ion conductivity under reducing conditions (b)

Table 4 Activation energies for the partial oxygen ion and electron conductivities of SrFe1-xTaxO3−δ

Conclusions

The x-ray diffraction and Mössbauer spectroscopy studies demonstrate that single cubic phases with disordered perovskite-like structure are formed in the SrFe1-xTaxO3−δ (x = 0.03 − 0.10) system under oxidizing conditions. As TEM data reveal, moderate substitution of tantalum for iron favors partial suppression of the reduction-promoted transition into the vacancy-ordered brownmillerite polymorph, thus leading to the formation of coexisting perovskite- and brownmillerite-type nanodomains. It is evidenced from the in situ high-temperature conductivity measurements that the domain structure remains at high temperatures and may have strong impact upon oxygen ion conductivity. This effect is particularly expressed in SrFe0.97Ta0.03O3−δ where a threefold increase of the ion contribution to total conductivity is observed as the result of isothermal cycling at 950 °C. These changes were suggested to originate from rearrangement and disintegration of brownmillerite-like domains in the perovskite matrix with ensuing formation and disappearance of oxygen vacancies. When the domain size becomes small enough, the ion conductivity attain values essentially independent of the dopant content.