Irreversible oxygen uptake and thermal expansion in the solid electrolyte oxygen-deficient perovskite Ca(Ti_0.8Fe_0.2)O_3−δ

Several perovskite-type oxides, including CaTiO₃, exhibit the orthorhombic Pbnm structure under ambient conditions.^4,11 This structure type exhibits a combination of in- and out-of-phase tilts of the TiO₆ octahedra (notated as a⁺b⁻b⁻ in Glazer’s notation).^12–14 The degree of orthorhombic distortion typically decreases as the temperature increases, with CaTiO₃ transitioning to tetragonal I4/mcm (a⁰a⁰b⁻) above 1230 °C and then the ideal untilted cubic Pm$3̄$m (a⁰a⁰a⁰) aristotype above 1310 °C (Fig. 1).¹⁵ Although Kennedy et al. postulated an orthorhombic Cmcm intermediate phase between the orthorhombic Pbnm and tetragonal I4/mcm phases, Ali et al. later disproved this.^4,15 There are numerous examples of ABX₃ materials, including SrZrO₃ and SrRuO_3, undergoing phase transitions from lower symmetry structures to the cubic Pm$3̄$m aristotype upon heating.^16,17 An additional method of increasing the symmetry of the perovskite system is to dope with appropriate cations to increase the Goldschmidt tolerance factor toward t ∼ 1.0. For example, SrZrO₃ exhibits the orthorhombic Pbnm structure (t ∼ 0.95) at room temperature, but substituting Zr⁴⁺ for smaller Ti⁴⁺ cations in the Sr(Zr_1−xTi_x)O₃ solid-solution series increases the symmetry to tetragonal I4/mcm (for 0.40 < x ≤ 0.95) and finally cubic Pm$3̄$m for SrTiO₃ (t ∼ 1.0).¹⁸ 

Perovskite-type structures with either oxygen vacancies or interstitial oxygen ions have found application as functional materials.^19,20 Tuning the oxygen stoichiometry is often achieved by substituting cations of different oxidation states or through modified synthetic methods.²¹ Oxygen-deficient perovskite-type structures such as (La_1−xSr_x)BO_3−δ (B³⁺ = Mn, Fe) are used as solid electrolytes in solid-oxide fuel cells (SOFCs) and for carbon monoxide oxidation,^22,23 as well as the oxygen-rich perovskite (Ba_1−xLa_x)CoO_3+δ as a battery cathode material.²⁴ The Ca(Ti_1−xFe_x)O_3−δ series of perovskites have been intensively studied due to their stability in reducing atmospheres, high ionic conductivity, and ability to act as oxygen separation membranes.^5–7,25 Substituting Ti⁴⁺ with Fe³⁺ introduces oxygen vacancies, which can lead to a complex relationship between chemical substitution and the symmetry of the perovskite structure.^26,27 This was recently shown, where despite substituting Ti⁴⁺ with larger Fe³⁺ cations (ionic radii 0.605 vs 0.645 Å as BO₆ octahedra),²⁸ competing phenomena, including the presence of oxygen vacancies, underbonding of the A-site cation, and changing cation coordination environments, can lead to a sporadic relationship between the amount of Fe-doping and the unit cell volume.²⁹ Furthermore, the general prediction from the Goldschmidt tolerance factor that symmetry increases as the tolerance factor approaches unity is less well established for nonstoichiometric structures, with CaTiO₃ (t ∼ 0.97) exhibiting the orthorhombic Pbnm structure, whereas Ca(Ti_0.6Fe_0.4)O_3−δ (t ∼ 0.96) exhibits the cubic Pm$3̄$m structure.^29,30

It has been postulated that at room temperature, Ca(Ti_0.8Fe_0.2)O_3−δ (x = 0.20) is the “ideal” composition for various applications.⁶ At this composition, the long-range average and short-range local structures exhibited orthorhombic Pbnm features, such as Ca²⁺ cation off-centering and tilts of the BO₆ octahedra.^29,30 The delicate balance between introducing oxygen vacancies at the expense of cation point defects that can act as diffusion “traps” was used to explain why this material has the highest ionic conductivity in the Ca(Ti_1−xFe_x)O_3−δ solid-solution series.^6,31 However, a robust understanding of the structure of Ca(Ti_1−xFe_x)O_3−δ materials at the operating temperatures of SOFCs (850–1000 °C) is required to determine their potential use as high-temperature solid electrolytes.³² A study by Becerro et al. using neutron powder diffraction (NPD) found that Ca(Ti_1−xFe_x)O_3−δ (x = 0.10, 0.20) undergoes the same sequences of phase transitions as the undoped CaTiO₃ material studied by Kennedy et al., albeit at lower temperatures.^15,30 Studies of the Ca(Ti_1−xFe_x)O_3−δ system using X-ray spectroscopy, Mössbauer, and neutron total scattering have revealed a complex distribution of Fe³⁺ coordination environments, including four-, five-, and six-coordination, detailing the role of the oxygen vacancies on the structures of these materials.^29,33–35 Despite the need for these materials to be heated and cooled for normal solid-oxide fuel cell operations, no reported studies have examined their thermal history or the effect of temperature or sample preparation on their structure.

This paper seeks to address this by determining the effect of thermal treatment on the reported crystal structure of the SOFC material Ca(Ti_0.8Fe_0.2)O_3−δ. The material was synthesized, and synchrotron X-ray diffraction (SXRD) data were collected upon heating and cooling. The SXRD measurements reveal an unexpected expansion in the unit cell volume associated with the uptake of oxygen upon heating and cooling.

The Ca(Ti_0.8Fe_0.2)O_3−δ sample was prepared using a conventional solid-state synthesis method described in a previous publication.²⁹ Variable temperature SXRD data were collected on the powder diffractometer BL-10 at the Australian Synchrotron (16.0 keV, λ = 0.773411 Å). Rietveld refinement of a LaB₆ standard was used to determine the precise wavelength and the Thompson–Cox–Hastings instrument resolution function.³⁶ The sample was loaded into a 0.2 mm diameter quartz glass capillary that was rotated during the measurements to minimize preferred orientation effects. Variable temperature measurements were collected upon heating and cooling over a 50–1000 °C temperature range, with the temperature controlled using an FMB-Oxford hot air blower. Additional high resolution neutron powder diffraction data were measured on Echidna at the Open Pool Australian Lightwater reactor, operated by the Australian Nuclear Science and Technology Organisation (ANSTO), and are presented in the electronic supplementary material.³⁷ 

Structures were refined using the Rietveld method implemented in the program TOPAS 6.³⁸ A 12th-order Chebyshev polynomial was used to model the background, with an additional pseudo-Voigt peak used to model the contribution of the quartz glass capillary to the SXRD data. The scale factor, lattice parameters, atomic positions, and atomic displacement parameters (ADPs) were refined simultaneously with the peak profile parameters. The ADPs were taken to be isotropic and constrained to be equal across the B-site of each structure, regardless of the mixed Ti and Fe occupancy. In the final refinement cycle, the parameters were fully relaxed, and all free parameters were refined. Rietveld refinements were undertaken sequentially for the variable temperature data, with the structural model from the previous dataset used as a starting point. The crystal structures were drawn using VESTA.³⁹ 

Thermogravimetric analysis (TGA) was performed using a NETZSCH STA 449F3 instrument. Samples were predried as powder at 100 °C overnight before analysis. Approximately 20 mg of sample was weighed onto a platinum/rhodium pan and heated from room temperature to 1000 °C at a rate of 10 °C min⁻¹ in synthetic air (80% N₂, 20% O₂) and nitrogen (20 ml min⁻¹). Data were also collected upon cooling from 1000 to 50 °C at a rate of 10 °C min⁻¹. The CaTiO₃ sample used for the TGA analysis was also prepared using a conventional solid-state synthesis method described in a previous publication.²⁹ 

The room-temperature structure of Ca(Ti_0.8Fe_0.2)O_3−δ was refined using the orthorhombic Pbnm (a⁺b⁻b⁻) model and agreed well with previous studies.^4,30 At higher temperatures, features consistent with changes in the symmetry of perovskites, such as anomalies in the lattice parameters and cation positions, can be challenging to identify.¹⁵ Therefore, the identification of weak superstructure reflections due to tilting of the BO₆ octahedra was used to establish changes in symmetry (Fig. 2). Upon heating, superstructure reflections consistent with the M-points (indicating out-of-phase tilts) gradually decreased in intensity before becoming indistinguishable from the background [Figs. 2(c), S1, and S2]. This is consistent with the observation of an orthorhombic Pbnm to tetragonal I4/mcm phase transition between 800 and 810 °C before a final transition to the cubic Pm$3̄$m structure between 840 and 850 °C (Fig. 2 and Table I).³⁰ 

The structure of Ca(Ti_0.8Fe_0.2)O_3−δ between 50 and 800 °C was refined to the orthorhombic Pbnm space group (Fig. 3). As in the undoped CaTiO₃ counterpart, the BO₆ octahedra are not strictly regular but elongated slightly.¹⁵ As a result, the tilt angles are best estimated using the atomic coordinates of the oxygen anions. The relationships between the atomic coordinates and tilt angles are given in supplementary material. Upon heating, the magnitude of the in- and out-of-phase tilts decreases, indicating that as the unit cell thermally expands, the strain in the structure is relieved (Fig. 4). The disappearance of the weak (021)_o and (113)_o superstructure reflections (at Q ∼ 2.45 and 2.95 Å⁻¹ corresponding to an M- and X-point, respectively) at 810 °C is consistent with the loss of the in-phase rotations of the BO₆ octahedra (Fig. 1). This indicates an increase in symmetry from orthorhombic Pbnm to tetragonal I4/mcm. The BO₆ octahedra remain slightly distorted, with a slight compression along the axial direction. Above 840 °C, the (202)_t R-point superstructure reflection (at Q ∼ 2.80 Å⁻¹) indicative of out-of-phase BO₆ octahedra tilts was absent, and the cubic Pm$3̄$m structural model was deemed most appropriate.

The lattice parameters are presented in Fig. 5 as reduced unit cell parameters (see supplementary material), with the reduced unit cell volume displayed in Figs. S4 and S5. The orthorhombic Pbnm lattice parameters increase anisotropically upon heating, with the b lattice parameter expanding in a quasi-linear fashion at low temperatures (Fig. S6). This effect has been noted in the undoped CaTiO₃ perovskite and the Fe-doped Ca(Ti_1−xFe_x)O_3−δ analogs,^15,30 ruling out a potential magnetostriction contribution. Additional neutron powder diffraction measurements did not reveal a magnetic contribution to the data (Figs. S7 and S8).³⁷ Above 200 °C, the unit cell volume expands linearly, consistent with the previous literature.^27,30

The B–O bond lengths calculated from the Rietveld refinements are plotted in Fig. S9. The number of different bond lengths within the BO₆ octahedra decreases from three in the orthorhombic Pbnm structure to two in tetragonal I4/mcm, and finally to a unique B–O distance in cubic Pm$3̄$m.³⁰ We note a difference in the orthorhombic B–O bonds for the temperature regions of ∼50–500 and 500–800 °C that resembles a phase transition. Kennedy et al. described a subtle orthorhombic Pbnm to orthorhombic Cmcm high-temperature transition in CaTiO₃.¹⁵ Similarly, an intermediate orthorhombic Imma phase was noted between orthorhombic Pbnm and tetragonal I4/mcm in the SrZrO₃.⁴⁰ However, careful observations of the superstructure reflections and peak splitting showed no features consistent with these structures in the current SXRD study of Ca(Ti_0.8Fe_0.2)O_3−δ (Figs. S10 and S11). Similarly, the high-resolution neutron powder diffraction study by Becerro et al. dismissed the presence of an orthorhombic Pbnm to orthorhombic Cmcm phase transition, but no B–O bond lengths were reported in that work that would enable a direct comparison.³⁰ 

SXRD data were also collected upon cooling from 1000 to 100 °C in steps of 100 °C. These data have been plotted as hollow symbols in Fig. 5. Upon cooling, there is a measured hysteresis in the tetragonal I4/mcm to cubic Pm$3̄$m phase transition. Although the tetragonal I4/mcm to cubic Pm$3̄$m occurs between 830 and 840 °C upon heating, the SXRD pattern collected at 800 °C upon cooling shows no sign of the R-point reflections indicative of the tetragonal I4/mcm structure (Fig. S12). The tetragonal I4/mcm structure was not observed in the 100 °C steps upon cooling. Upon further cooling, the lattice parameters deviate between the heating and cooling temperatures, with the difference plotted in Fig. 6. The a and c lattice parameters significantly increase upon cooling for temperatures below 700 °C, while the b lattice parameter matches well between heating and cooling until ∼200 °C. The difference in structure is also obvious when comparing the cooling and heating diffraction patterns collected at 50 °C (Fig. S13).

The difference in the structure upon heating and cooling has not previously been reported in this perovskite-type system. To understand this further, the A- and B-site polyhedral volumes were calculated using the structures obtained from the Rietveld refinements, as shown in Fig. 7. For temperatures ≥500 °C, there is a negligible difference between the heating and cooling BO₆ volumes. However, when the temperature is below 500 °C, the BO₆ volume is significantly smaller during cooling compared to heating, which we will return to below. The CaO₁₂ polyhedral volume shows a similar but opposite trend. Observing the CaO₁₂ polyhedra, the difference in volume between the heating and cooling is negligible for temperatures between 900 and 400 °C. However, when the cooling temperature is below 400 °C, there is a significant increase in the CaO₁₂ volume relative to the heating. Observing the bond valence sums (BVS) of the Ca and the B-site shows trends that correlate with the changes in polyhedral volume (Fig. S14).⁴¹ The expansion of the CaO₁₂ polyhedra causes the BVS of the Ca²⁺ to reduce, while the contraction of the BO₆ octahedra causes the BVS of the B-site to increase. This demonstrates that the oxygen atoms are held much closer to the B cations upon cooling than the Ca²⁺ cations, and the average Ca–O bond increases, as would occur if the number of oxygen atoms increases, to maintain optimal bonding.

Given the large changes in polyhedra volume upon heating and cooling, TGA was used to determine a potential change in mass of the Ca(Ti_0.8Fe_0.2)O_3−δ material. This was performed upon heating and cooling in different environments to confirm the possibility of oxygen uptake within the structure at elevated temperatures (Fig. 8). In each measurement, between 0.6% and 1.0% mass was lost upon heating, presumably due to the loss of surface species. For Ca(Ti_0.8Fe_0.2)O_3−δ in synthetic air (80% N₂, 20% O₂) at temperatures >800 °C, a large increase in mass is observed, which correlates with the increase in symmetry from orthorhombic Pbnm to tetragonal I4/mcm. It is postulated that the combination of thermal expansion and removal of the local strains (tilts of the BO₆ octahedra) facilitates the uptake of oxygen, which is then “trapped” upon cooling. A similar increase in mass was not observed when Ca(Ti_0.8Fe_0.2)O_3−δ was heated and cooled in N₂, confirming that the increase in mass is due to oxygen uptake. Similar measurements were conducted for CaTiO₃ (x = 0.00), and no significant changes in mass were observed upon heating or cooling, regardless of the choice of environment.

The mass of the sample remains above 100% upon cooling, suggesting oxygen is incorporated into the structure at higher temperatures. This could potentially lead to the partial oxidation of Fe³⁺ to Fe⁴⁺. X-ray absorption measurements of this sample published elsewhere showed no evidence of Fe⁴⁺ in the as-prepared sample (Figs. S18 and S19).²⁹ Unfortunately, the experimental arrangement does not allow for the recovery of the thermally cycled sample, limiting further spectroscopy studies. We note that the thermal gradient across the capillary in SXRD studies can enhance oxygen mobility, as seen in our study of SrUO_4−x, which did not occur in neutron diffraction measurements of the same material conducted in a vacuum.⁴² Since the structures were refined against X-ray diffraction data where oxygen makes a relatively minor contribution to the overall diffraction due to the presence of the heavier Ca, Fe, and Ti cations, it was not possible to obtain precise estimates of the oxygen stoichiometry in the samples. Partial oxidation of the Fe³⁺ present in the as-prepared sample would lead to a reduction in the B–O bond distances and volume of the BO₆ octahedra. Other possibilities, such as the “trapping” of oxygen atoms around Fe³⁺ (changing from TiO₆ octahedra and FeO_4,5 polyhedra to mixed TiO₅ and FeO_5,6 polyhedra), would lead to a reduction in unit cell volume and do not explain the observed phenomena.²⁸ Previous studies of Fe⁴⁺-containing perovskites, including CaFeO₃,^43,44 have revealed the need for high temperature and high oxygen partial pressure during the synthesis.^45–47 The present study shows that such extreme conditions are not required for Ca(Ti_0.8Fe_0.2)O_3−δ. This may be correlated with the presence of mobile vacancies that allow for easy oxygen uptake, as this composition has the highest ionic conductivity in the Ca(Ti_1−xFe_x)O_3−δ solid solution series.^6,31 The addition of oxygen into the structure at elevated temperatures may be favored to relieve the strain in the structure, as the ionic radii of six-coordinate Fe⁴⁺ are closer in size to six-coordinate Ti⁴⁺ (0.585 vs 0.605 Å) compared to either four- or five-coordinate Fe³⁺ (0.49 and 0.58 Å, respectively).²⁸ 

This uptake of oxygen and the concurrent partial oxidation of Fe³⁺ to Fe⁴⁺ upon cooling would cause the unit cell to increase in size, which would significantly impact the A- and B-site environments for temperatures less than ∼500 °C. This is unusual, as we expect these environments to be affected at all temperatures. This could indicate that above ∼500 °C, the anions are sufficiently mobile to inhibit the formation of Fe⁴⁺. When the structure is cooled below ∼500 °C, the anions become fixed such that the coordination increases and six-coordinate Fe⁴⁺ forms, resulting in the observed contraction of the average B–site polyhedra volume due to the higher Fe⁴⁺ oxidation state and the overall expansion of the unit cell described earlier. This expansion of the unit cell, in turn, affects the octahedral tilting, with less tilting observed during cooling (Fig. S3). The contraction of the B–site polyhedral volume upon cooling could be due to charge disproportionation (CD) of Fe⁴⁺ into Fe³⁺ and Fe⁵⁺.^48–51 However, this is unlikely as CD is expected to occur at much lower temperatures and would result in lattice contraction (relative to heating) due to the smaller ionic radius of Fe⁵⁺.

Other work has reported on the thermal history of some oxygen-deficient perovskites. Early reports by Manthiram et al. noted that thermal hysteresis in Ba(Zr_1−xIn_x)O_3−x/2 was often accompanied by a weakening of the ceramic pellet, suggesting the intercalation of chemical species.⁵² However, this is not seen in all oxygen-deficient perovskites, with the oxygen-deficient perovskite (Sr_0.7Y_0.3)CoO_2.65−δ showing no deviation in its weight percent upon heating and cooling.⁵³ Concerning Fe-doped CaTiO₃, there are differing reports of the appropriate lattice parameters for some compositions depending on the synthesis conditions used. For example, Ca(Ti_0.9Fe_0.1)O_3−δ prepared using a solid-state reaction by Becerro et al. showed a larger unit cell volume than the sample prepared using a flux-mediated method by Novianti et al. or samples prepared using the Pechini method by Barbieri et al.^30,54,55 This suggests that not only is the thermal history of the sample important, but the availability of oxygen in the chosen synthesis method may play a role in determining the precise oxygen content for members of the Ca(Ti_1−xFe_x)O_3−δ series. As a result, it is appropriate to refer to these materials with ambiguous oxygen stoichiometry, with additional titration experiments used to determine precise oxygen content.⁵⁶ This may explain some of the discrepancies in the reported structural models. Furthermore, X-ray diffraction patterns of Ca(Ti_0.8Fe_0.2)O_3−δ collected more than six days after cooling from 1000 °C suggest these structural changes are irreversible (Fig. S16).

A question remains as to why this happens in the SXRD experiment and not during the initial synthesis of the compound. These samples are sintered as bulk ceramic pellets, whereas the SXRD and TGA are conducted on powder samples, significantly increasing the surface area for oxygen uptake. Previous studies of bulk ceramics have identified surface structures that deviate from the bulk due to the volatility of certain elements at elevated temperatures.⁵⁷ To confirm this, we performed TGA on Ca(Ti_0.8Fe_0.2)O_3−δ that was heated as a powder in air at 1000 °C (Fig. S17). The higher surface area of the powder allowed for greater oxygen uptake during heating, which resulted in lower oxygen uptake during the subsequent TGA. When considering the applications of oxygen-deficient perovskites in bulk ceramic form, the thermal history will no doubt affect the surface chemistry and, hence, the physical properties in terms of ionic conductivity and mechanical toughness.

In this work, we have shown that the oxygen-deficient Ca(Ti_0.8Fe_0.2)O_3−δ perovskite can undergo irreversible thermal expansion using variable temperature SXRD, in addition to the previously described reversible structural phase transitions associated with changes in the cooperative tilting of the corner sharing BO₆ octahedra. Rietveld refinements against the SXRD data revealed lattice expansion upon cooling from 1000 °C, which was accompanied by contraction of the B-site polyhedra and expansion of the Ca²⁺ polyhedra. TGA experiments confirmed that Ca(Ti_0.8Fe_0.2)O_3−δ absorbs oxygen at high temperatures. Coupling this with the SXRD results demonstrates that partial oxidation of Fe³⁺ to Fe⁴⁺ occurs with increased oxygen content, resulting in significantly different structural trends between heating and cooling. This study provides an explanation for discrepancies between previous structural studies of this material and demonstrates how the thermal history of this and related oxygen-deficient materials will strongly influence their physical properties, which have implications for their use in SOFCs. Changes in the oxygen content of the material during thermal cycling are likely to result in strain gradients in the material that may be detrimental to its long-term practical use.

The supplementary material encompasses equations for the determination of BO6 octahedra tilts from atomic coordinates, reduced lattice parameters, and bond valence sums. Plotted integrated intensities of the R- and M-points, BO6 octahedral tilts upon heating and cooling, differences in the unit cell volume upon heating and cooling, neutron powder diffraction data, additional fits using the orthorhombic Cmcm and orthorhombic Imma structural models, and differences in the synchrotron x-ray diffraction patterns upon heating and cooling.

We acknowledge the support of the Australian Research Council for this work that was facilitated by access to Sydney Analytical, a core research facility at the University of Sydney. B.G.M. and M.K.N. acknowledge the Australian Institute for Nuclear Science and Engineering for a PGRA and RSS scholarship. This project has received funding from the European Union’s Horizon Europe Research and Innovation Program under Project No. 101109595 (MAGWIRE). This work incorporates data from the Powder Diffraction (Grant Nos. PDR17762 and M18313) beamline at the Australian Synchrotron. We acknowledge Professor Maxim Avdeev for collecting additional neutron diffraction data at Echidna at ACNS. We acknowledge Associate Professor Allyson Fry-Petit for her helpful discussions involving this work.

The authors have no conflicts to disclose.

Bryce G. Mullens: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Frederick P. Marlton: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Supervision (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Caleb J. Bennett: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Matilde Saura-Muzquiz: Conceptualization (equal); Data curation (equal); Investigation (equal); Supervision (equal); Writing – review & editing (equal). Maria K. Nicholas: Data curation (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Helen E. A. Brand: Data curation (supporting); Methodology (supporting); Writing – review & editing (supporting). Brendan J. Kennedy: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Software (equal); Supervision (equal); Validation (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.