Emergence of Dynamically‐Disordered Phases During Fast Oxygen Deintercalation Reaction of Layered Perovskite

Abstract Determination of a reaction pathway is an important issue for the optimization of reactions. However, reactions in solid‐state compounds have remained poorly understood because of their complexity and technical limitations. Here, using state‐of‐the‐art high‐speed time‐resolved synchrotron X‐ray techniques, the topochemical solid‐gas reduction mechanisms in layered perovskite Sr3Fe2O7− δ (from δ ∼ 0.4 to δ = 1.0), which is promising for an environmental catalyst material is revealed. Pristine Sr3Fe2O7− δ shows a gradual single‐phase structural evolution during reduction, indicating that the reaction continuously proceeds through thermodynamically stable phases. In contrast, a nonequilibrium dynamically‐disordered phase emerges a few seconds before a first‐order transition during the reduction of a Pd‐loaded sample. This drastic change in the reaction pathway can be explained by a change in the rate‐determining step. The synchrotron X‐ray technique can be applied to various solid‐gas reactions and provides an opportunity for gaining a better understanding and optimizing reactions in solid‐state compounds.


Introduction
Understanding the mechanisms behind chemical reactions is one of the biggest issues in materials chemistry, since this makes it possible to create compounds with the desired structures or to optimize reactions toward functional uses. However, reactions in solid-state crystalline compounds are rather poorly understood [1] compared with molecular reactions whose steps have been analyzed with nuclear magnetic resonance (NMR) and other spectroscopies. This is partly because solid-state reactions generally require high temperatures; thus, it is relatively difficult to monitor them by using in-situ methods. Inhomogeneous reactions derived from the difference in surface and bulk states as well as the existence of numerous numbers of atoms in crystals also prevent precise analyses of reactions in solid states. In order to achieve the rational design of solid-state compounds and their reactions, the development of in-situ measurement techniques is necessary.
In-situ X-ray diffraction (XRD) is a powerful tool for monitoring reactions in bulk materials, [2] and the recent development of synchrotron X-rays enables us to access high-resolution data with a short time window. Time-resolved synchrotron XRD measurements have been used to investigate many reactions in crystalline phases, such as solid-state reactions, [26] hydrothermal reactions, [3] gas absorption, [4,5] and solid-gas catalytic reactions. [6] However, the development of a time-resolved XRD measurement of a solid-gas reaction that can reveal structural evolution precisely on a subsecond scale is still a challenge. When the time scale of the measurement becomes faster, one can capture intermediate phases with a shorter lifetime. This will provide further opportunities to develop optimization of the reactions and/or synthesis of metastable structures.
Here, we focus on the Ruddlesden-Popper layered perovskite Sr 3 Fe 2 O 7− , which has recently attracted attention as a highperformance oxygen storage material. [7][8][9] This compound shows a reversible topochemical redox reaction between Sr 3 Fe 2 O 7− ( ∼ 0.4) and Sr 3 Fe 2 O 6 ( = 1.0) under O 2 and H 2 at 773 K and excellent performance as an environmental catalyst material. Our previous study revealed that Pd loading dramatically promotes the oxygen release rate and decreases the release temperature under H 2 flow on Sr 3 Fe 2 O 7− ( ∼ 0.4), [8] but the reaction pathways and structural evolution during the reduction were still unclear. In the study reported here, we monitored the reduction reaction from Sr 3 Fe 2 O 7− ( ∼ 0.4) to Sr 3 Fe 2 O 6 by using a high-speed time-resolved synchrotron X-ray diffraction (XRD) technique with the time window of a few hundred milliseconds. We found a gradual single-phase structural evolution during the reduction of pristine Sr 3 Fe 2 O 7− ( ∼ 0.4), while 1.0 wt% Pd loading drastically altered the reaction pathway, where there was a first-order transition from a dynamicallydisordered phase to a stable-ordered phase. This reaction was further investigated by using the simultaneous quick X-ray absorption fine structure (QXAFS)-XRD technique. These high-speed synchrotron X-ray techniques allow us to determine the reaction pathways of the fast solid-gas reactions and will provide an opportunity to understand and optimize reactions in solid-state compounds.

High-Speed Time-Resolved XRD Measurements for the Topochemical Reductions of Sr 3 Fe 2 O 7− and Pd/Sr 3 Fe 2 O 7−
A pristine Sr 3 Fe 2 O 7− and 1.0 wt% Pd-loaded sample (Pd/Sr 3 Fe 2 O 7− ) were prepared as reported previously. [8,11] The XRD patterns for Sr 3 Fe 2 O 7− did not change by Pd-loading ( Figure S1, Supporting Information) and the peaks corresponding to Pd species were not detected due to the small amount of Pd. In a static XAFS measurement of Pd-loaded sample before reduction, the Pd K-edge extended X-ray absorption fine structure (EXAFS) oscillation in Pd/Sr 3 Fe 2 O 7− was different from that of PdO as a standard sample and resembled with Fe K-edge EXAFS oscillation in Pd/Sr 3 Fe 2 O 7− ( Figure S2, Supporting Information). These results imply that most Pd species were substituted in the Fe site of Sr 3 Fe 2 O 7− , rather than existing as PdO. Thermogravimetric analyses confirmed that the oxygen deficiency under O 2 at 773 K was ≈0.4, regardless of the presence of Pd ( Figure S3, Supporting Information). In the stable structure of Sr 3 Fe 2 O 7− ( ∼ 0.4), most of the vacancies are at the O1 site in the perovskite layer, which consists of the apical sites between FeO 6 octahedra (Figure 1a). [12] High-resolution time-resolved XRD measurements were carried out on the BL02B2 beamline at SPring-8 ( Figure 1b). The data collection for the reduction was started under vacuum conditions after a pretreatment at 973 K (details are in the experimental section). The interval between the diffraction measurements on Sr 3 Fe 2 O 7− (or Pd/Sr 3 Fe 2 O 7− ) was 200 (or 100) ms. H 2 gas (≈0.5 atm) was injected 3 seconds after starting the data collection. By injecting H 2 gas above 773 K, the following reaction proceeds: Regarding the state of Pd species in Pd-loaded sample, EXAFS oscillations clearly indicate that Pd 2+ species in Pd-loaded sample was reduced to Pd metal nanoparticle by the H 2 reduction at 773 K ( Figure S2, Supporting Information). We previously confirmed that the particle size of Pd metal nanoparticle was below 1 nm by comparing the theoretical average coordination number (CN) of mono-dispersed spherical particles having a cuboctahedral structure with the CN estimated by EXAFS curve fitting analysis. [8] Figures 1c,d show the time-resolved XRD patterns for the reduction of Sr 3 Fe 2 O 7− and Pd/Sr 3 Fe 2 O 7− at 773 K (the full angle data are shown in Figures S4 and S5, Supporting Information). We noticed that the sample temperature immediately increased (within 100 ms) when gas was injected. This is because the sample temperature under the vacuum conditions was unintentionally lower than the setting temperature due to cooling by adiabatic expansion. As a result, the peak positions jump to a lower angle immediately after the H 2 injection (Figure 1c,d). This phenomenon was confirmed by injecting inert gas to Sr 3 Fe 2 O 6 and Si standard (Figures S6 and S7, Supporting Information). Thermal equilibrium was achieved within 100 ms, and the increase in temperature due to gas injection was estimated to be ≈60 K at 723 K.
It is obvious that the reaction is accelerated by Pd loading. The reaction finishes in ≈30 s for pristine Sr 3 Fe 2 O 7− , while it finishes within a few seconds for Pd/Sr 3 Fe 2 O 7− . This result can be explained by H 2 spillover effect at the surface. [13,14] Most surprisingly, the reaction pathway dramatically changes through promotion of the surface reaction. In the pristine Sr 3 Fe 2 O 7− , the diffraction peaks smoothly move without jumping or broadening (Figure 1c), meaning a single-phase process (or a second-orderlike behavior) occurs. In contrast, the 110 peak undergoes an obvious discontinuous jump at ≈4 -5 s in Pd/Sr 3 Fe 2 O 7− (inset of Figure 1d), suggesting the existence of a first-order transition. The inset of Figure 1d and Figure   pathways can be explained by a change of the rate-determining step, and a nonequilibrium phase appears in the reduction of Pd/Sr 3 Fe 2 O 7− through fast deintercalation of oxide ions, as will be discussed later.

Simultaneous High-Speed Time-Resolved QXAFS -XRD Measurements for the Topochemical Reduction of Pd/Sr 3 Fe 2 O 7−
We also carried out simultaneous QXAFS and XRD measurements on the BL36XU beamline at SPring-8 (Figure 2e), to reveal the correlation between the dynamical changes in the electronic and crystal structures. Figure 2a,b show time-resolved Fe K-edge X-ray absorption near edge structure (XANES) spectra and XRD patterns for the reduction reaction of Pd/Sr 3 Fe 2 O 7− at 773 K. The data were recorded at a 100 ms interval (80 ms and 20 ms exposure times for XAFS and XRD, respectively). The initial condition was a vacuum; then, H 2 gas (≈98 kPa) was injected 3 s after starting the data collection. The detailed conditions are given in the experimental section.
Both the XANES spectra and the XRD patterns immediately changed after the H 2 injection (Figure 2a,b). Their reactions finished in 1 s. This is slightly faster than what the data obtained on the BL02B2 beamline show (Figure 1d), probably because of the different experimental condition such as gas pressure. The XANES spectra (Figure 2c) shift to a lower energy state because of the reduction; this finding is consistent with our previous study. [9] The jump in the 110 peak at ≈3.3 seconds in the XRD patterns (Figure 2b) can also be seen in Figure 1d. Thus, this result reproduced the reaction observed on BL02B2 (Figure 1d). Figure 2d shows the time dependence of the lattice parameters determined from the XRD data and the fraction of the final state calculated from the fitting using the initial and final states of XANES data. The lattice parameters between 3.2 and 3.5 s are not plotted in Figure 2d due to the low resolution of the data, while two phases coexisting between 3.2 and 3.5 s were clearly seen ( Figure S9, Supporting Information). Note that we analyzed the energy shift from the normalized XANES spectra at half of the intensity ( Figure S10, Supporting Information), and these results are consistent with the phase fraction in Figure 2d. Remarkably,  the phase fraction and the lattice parameter simultaneously change without a time lag, suggesting that the electronic and crystal structures change simultaneously at this time scale.
Previously, we reported nonmonotonic time-dependence of the Sr K-edge XANES spectra acquired from energydispersive XAFS (DXAFS) measurements for the reduction of Pd/Sr 3 Fe 2 O 7− . [8] On the basis of the suggested findings of the current in-situ XRD measurements, the Fourier transformed EX-AFS spectra acquired after the H 2 injection were investigated in detail. As shown in Figure 3, the peak at 2.2 Å corresponding to Sr-O bonds in pristine Sr 3 Fe 2 O 7− gradually decreases, implying that the coordination number of the Sr-O bond decreases through oxygen deintercalation. In contrast, in the case of Pd/Sr 3 Fe 2 O 7− , the peak intensity drops as soon as the H 2 gas is injected, and it is lower than that of Sr 3 Fe 2 O 6 between 0.2 and 0.6 s (Figure 3c). This also confirms that Pd loading caused the change in the reaction pathways and the emergence of an intermediate phase.

The Reaction Mechanisms of Sr 3 Fe 2 O 7− and Pd/Sr 3 Fe 2 O 7−
The reaction from Sr 3 Fe 2 O 7− ( ∼ 0.4) to Sr 3 Fe 2 O 6 ( = 1.0) should involve oxygen deintercalation at the surface and oxygen diffusion in the bulk. Thus, oxygen vacancies must be randomly distributed at both O1 and O2 sites during bulk diffusion since the favorable oxygen conduction pathway in Sr 3 Fe 2 O 7− is -O1-O2-O1-O2 in the perovskite layer (Figure 4e). [15] However, most of the vacancies are at the O1 site in the thermodynamically stable structures (Figure 4a,b). [12] Therefore, it is expected that a nonequilibrium dynamically-disordered phase can emerge before the oxygen vacancies are rearranged to make thermodynamically stable structures.
The experimental observation showed the pristine Sr 3 Fe 2 O 7− reacts slowly, with a continuous change in structure, while Pd/Sr 3 Fe 2 O 7− reacts quickly, with an obvious jump in the lattice parameters. We propose that this change in the reaction pathway is governed by the rate-determining step of each reaction. In the reduction of pristine Sr 3 Fe 2 O 7− , the surface reaction is slower than bulk diffusion because of slow H 2 dissociation, which was www.advancedsciencenews.com www.advancedscience.com previously determined by temperature-programmed reactions. [8] As a result, the structural relaxation, i.e. rearrangement of oxygen vacancies, is faster than the vacancy supply. This allows a continuous relaxation through thermodynamically-stable phases, resulting in a continuous change in the lattice parameters (Figure 1e). In contrast, in the case of Pd/Sr 3 Fe 2 O 7− , Pd metal nanoparticles on Sr 3 Fe 2 O 7− dramatically accelerated H 2 adsorption and dissociation (spillover effect), which facilitated the reaction between lattice oxygen and H 2 . [13,14] As a result, the surface reaction rate exceeds the bulk diffusion rate in the reduction of Pd/Sr 3 Fe 2 O 7− by surface modification. This change of the rate-determining step is supported by previous density functional theory (DFT) calculations: It has been shown that the H 2 adsorption and dissociation on Pd metal surface is energetically very favorable. [14] The activation energy for the path, in which the hydrogen dissociated on Pd metal surface transfers to the catalyst support, is only 0.58 eV. [14] This value is much lower than that (1.09 eV) for the diffusion of lattice oxygen in Sr 3 Fe 2 O 7− which is obtained from the same calculation method. [15] As the result of the change of the rate-determining step, the vacancy supply becomes faster than the structural relaxation so that the value reaches a maximum ( = 1.0) before the structure fully relaxes. This allows us to observe the first-order transition from dynamically-disordered Sr 3 Fe 2 O 6 to ordered Sr 3 Fe 2 O 6 ( Figure 4d). Now let us explain the reaction pathway for the pristine Sr 3 Fe 2 O 7− without the nonequilibrium phase ( Figure 4c). As seen in Figure 1e, the time dependence of the lattice parameters obtained by Le Bail analysis reveals that the a-axis length monotonically increases. In contrast, the c-axis length continuously increases until 3 s after the H 2 injection and then gradually decreases. The nonmonotonic evolution of the c-axis can be rationalized by considering two competing effects: One is expansion by decreasing the valence of Fe ions; the other is reduction of the Fe-Fe distance along the c-axis by deintercalation of oxide ions at the O1 site to form face-to-face FeO 5 pyramids. A schematic image of the structural evolution is shown in Figure 4c. In the initial stage of the reduction, the former effect is dominant because the residual oxide ions at the O1 site work as pillars to maintain the long Fe-Fe distance. As the value increases, the latter effect becomes dominant because the "pillar" oxide ions are deintercalated. Although the reaction rate decreases by decreasing the reaction temperature, the features of the lattice evolutions are similar at all temperatures between 673 K and 773 K ( Figure S11, Supporting Information).
In Pd/Sr 3 Fe 2 O 7− , both the a-and c-axis lengths continuously increase before the discrete transition (Figure 1g), corresponding to a volume expansion (Figure 1h) due to the decrease in Fe valence. During the transition, the a-axis jumps and the c-axis drops (Figure 1g). The drop of the c-axis corresponds to the vacancy ordering since oxide ions at the O1 site in the disordered phase must work as pillars to maintain the long Fe-Fe distance as discussed above. The small difference in volume before and after the transition (Figure 1g) indicates that the phase just before the transition has the same chemical formula of Sr 3 Fe 2 O 6 . Namely, the rearrangement of oxygen vacancies occurs without a distinct change to the chemical formula at the transition. Thus, the nonequilibrium phase can be assigned as dynamicallydisordered Sr 3 Fe 2 O 6 (or Sr 3 Fe 2 O 7− ( ∼ 1)), and the first-order transition can be rationalized as being an order-disorder transition in Sr 3 Fe 2 O 6 ( Figure 4d).
To capture the characteristics of the dynamically-disordered phase at the atomic scale, we carried out a series of DFT calculations. As reported in Ref. 15, the possible oxygen conduction pathway is -O1-O2-O1-O2 (Figure 4e). Taking this scenario into account, the oxygen vacancies in the disordered structure would be preferentially located at the O1 and O2 sites. Another possibility is that the oxygen vacancies exist at all the oxygen sites. We call these structures partially disordered (PD) and fully disordered (FD), respectively. We constructed ten models of Sr 3 Fe 2 O 6 , with a 4×4×1 supercell, for each, and calculated their lattice parameters.
As can be seen in Figure 4f, all the calculated disordered models of Sr 3 Fe 2 O 6 have a longer c-axis than that of the ordered model built from the experimental data (Figure 1b). The reason for this is that the partially occupied O1 atoms work as pillars to maintain the long Fe-Fe distance along the c-axis, as discussed above. Remarkably, the a-axis lengths in the FD models tend to elongate compared with the ordered Sr 3 Fe 2 O 6 but to decrease in the PD models. Since the trend in the PD models are qualitatively consistent with the experiments (Figure 1g), our DFT calculations imply that O1-O2 site disorder (PD) is more realistic than full disorder (FD). We should, however, note that the elongation of the c-axis in the PD model is larger (+2.9%) than experimentally observed (<0.8%, Figure 1g). The reason is not clear yet, but it is probably related to the time resolution of the experiments or the too-small unit cell of the models, where the oxygen vacancies are periodically located at short intervals even in the disordered models.

Conclusion
We showed the emergence of nonequilibrium intermediate phases during fast oxygen deintercalation reaction of Pd-loaded Sr 3 Fe 2 O 7− . The characterization of such a short-lifetime intermediate phase, which requires advanced techniques, is an essential step to understand the mechanism of the fast reactions. The current compound is promising for a high-performance oxygen storage material in catalytic systems for the purification of automotive exhaust gas, and Fe-site-substitution further improves the oxygen storage performance. [7,16] Therefore, application of the time-resolved synchrotron X-ray techniques to the substituted system to compare the difference in the reaction pathways will provide important information to optimize the oxygen storage performance.
Furthermore, we have demonstrated that the reaction process can indeed be manipulated by using a surface treatment to control the rate-determining step in crystalline material (Figure 4c,d). Regarding the current reactions, the final products are the same despite the different reaction pathways. However, the selectivity of the reaction pathways will provide an opportunity for rational design of compounds. For example, in order to design organic molecules, organic chemists have developed protecting groups that are used to control the reaction pathway by limiting the diffusion of reactant molecules to the blocked sites. As demonstrated in this study, a similar reaction control through the surface treatment may be possible in crystalline phases. In www.advancedsciencenews.com www.advancedscience.com Keywords oxygen storage material, perovskite oxide, synchrotron X-ray technique, time-resolved measurement, topochemical reaction