Structural changes in equimolar ceria – hafnia materials under solar thermochemical looping conditions: cation ordering, formation and stability of the pyrochlore structure †

Equimolar ceria – hafnia oxides form a pyrochlore structure Ce 2 Hf 2 O 7 , which exhibits an ordered arrangement of Ce 3+ and Hf 4+ cations under the reducing conditions of a solar thermochemical looping reactor for the two-step dissociation of water or carbon dioxide. The ceria – hafnia pyrochlore phase was prepared from oxidized ceria – hafnia powders by chemical reduction in a ﬂ ow of H 2 /He and by auto-reduction in a ﬂ ow of Ar at up to 1825 K. Full conversion of Ce 4+ to Ce 3+ was con ﬁ rmed by thermogravimetric analysis and Ce K edge X-ray absorption spectroscopy. X-ray di ﬀ raction and Hf K edge X-ray absorption spectroscopy identi ﬁ ed the pyrochlore phase. The dynamics of the structural changes were determined by time-resolved in situ Ce K edge X-ray absorption spectroscopy and in situ X-ray di ﬀ raction. Under the oxidizing conditions of the regeneration step of isothermal carbon dioxide splitting at 1800 K, the pyrochlore transformed to a mixture of ﬂ uorite-type or tetragonal ceria and monoclinic and orthorhombic hafnia phases. A k -Ce 2 Hf 2 O 8 phase, an oxidized form of the pyrochlore with an ordered arrangement of cations, was not detected.


Introduction
Solar energy can be converted and stored in chemical bonds by thermochemical dissociation of water and/or carbon dioxide in a two-step redox process. Eqn (1)-(3) describe the generation of solar fuels from water and/or carbon dioxide using ceria as a redox intermediate. Ceria-based solid solutions are the stateof-the-art oxygen storage material for this process. [1][2][3][4][5][6][7][8] Typical process temperatures are 1773 K for the reduction (eqn (1)) and 1073 K for the regeneration steps (eqn (2) and (3)). Isothermal operation at typically 1773 K has been successfully demonstrated.
CeO 2Àd red + DdH 2 O / CeO 2Àd ox + DdH 2 CeO 2Àd red + DdCO 2 / CeO 2Àd ox + DdCO Considerable effort has been made to tune the oxygenstorage properties of ceria by introducing heterocations into the uorite-type structure to increase the energy conversion efficiency of this process. [9][10][11][12][13][14][15][16][17] The fuel yields of ceria-zirconia and ceria-hafnia solid solutions are very similar [18][19][20] and higher than the fuel yield of pure ceria. [21][22][23][24][25][26][27][28] Zirconium and hafnium have a similar chemistry and their oxides are isomorphs. Tabulated cation radii in crystals 29 of Zr 4+ and Hf 4+ in 6-fold (0.72 and 0.71) and 8-fold coordination (0.84 and 0.83), respectively, are almost the same. The monoclinictetragonal phase transition of hafnia occurs at a temperature about 600 K higher than for zirconia. 30,31 The crystal structure of ceria-zirconia and ceria-hafnia compounds depends on the composition and the oxidation state of cerium. 32 The structure of bulk materials prepared from small particles can be controlled by the sintering conditions. Under oxidizing conditions, oxides of equimolar composition form solid solutions with cubic uorite-type and tetragonal structures and a random arrangement of the cations. In contrast, under reducing conditions, ordering of the cations occurs.
Determination of the local structure of heterocations is a key to pinpointing oxygen vacancies and to understanding the relationship between the structure and the redox performance. Cation ordering in ceria-based materials for solar thermochemical looping processes has not been reported thus far. In most studies, the preparation of ceria-based oxides to obtain the porous structures required for thermochemical cycles were prepared by sintering in air. 3,4,33,34 Cation ordering requires concentrations of the hetero-cation higher than those of ceriabased oxides which are typically tested for the thermochemical generation of solar fuel. Petkovich et al. 25 discussed the effect of inhomogeneities of ceria-zirconia. However, they did not expose the materials to the very high temperatures of solar thermochemical cycling. Cation ordering may be difficult to detect by XRD. 35,36 Fig. SI 1 † displays the uorite-type CeO 2 and the pyrochlore structure A 2 B 2 O 6 O (space group Fd3m). It can be derived from a uorite-type structure with an ordered arrangement of cations, such as Ce 3+ and Zr 4+ , by removing and displacing oxygen atoms from their original lattice positions. A 3+ ions are coordinated to eight oxygen atoms and B 4+ ions to six. The ratio of the ionic radii of the A and B cations is typically in the range of 1:46 # r A r B # 1:8. 37 Variations in the oxygen stoichiometry causes distortions as well as changes in symmetry. 38 Upon oxidation, cation ordering is maintained and the cubic k-Ce 2 Zr 2 O 8 phase is formed. Intermediate oxygen-rich nonstoichiometric phases, such as the Ce 2 Zr 2 O 7.5 phase, [39][40][41] exhibit an ordered arrangement of cations. The extent of order/ disorder in the arrangement of cations and oxygen vacancies strongly affects the capacity and dynamics of the release and uptake of oxygen. [42][43][44][45][46] Similar to ceria-zirconia, relatively wide ranges of concentrations of solid solutions with pyrochlore structure are present in ceria-hafnia. Homogeneity ranges become narrower as the stability of the pyrochlore phase decreases with increasing temperature and decreasing cation radius of the rare-earth. Unlike the uorite-type phases, the pyrochlore hafnates are more stable than the corresponding zirconates. 47 The case of ceria-containing pyrochlores such as hafnates, containing cerium in a 3+ oxidation state, has not been considered in many earlier studies, because their preparation requires vacuum or more reducing conditions and high temperature. Experimental evidence of structural changes in ceria-hafnia materials under reducing conditions at elevated temperatures is scarce. Andrievskaya et al. 48,49 determined ceria-hafnia-zirconia phase equilibria at 1773 K in air. Stanek et al. 50 had no data on ceria-hafnia on hand for their predictive study of hafnia-containing pyrochlore phase elds. Baidya et al. 51 compared the chemical reduction of uorite-type Ce 0.5 Hf 0.5 O 2 and Ce 0.5 Zr 0.5 O 2 prepared by solution combustion. Different hydrogen uptake proles were obtained by temperature-programmed reduction in dry hydrogen up to 1073 K. Hydrogen uptake of ceria-hafnia shied to higher temperature compared to that of ceria-zirconia. While ceria-zirconia was reduced to Ce 0.5 Zr 0.5 O 1.75 and even beyond that, ceriahafnia could be reduced only to a composition of Ce 0.5 Hf 0.5 O 1.77 and the pyrochlore crystal structure was not detected. Bonk et al. 20 prepared ceria-hafnia powders containing up to 20 mol% hafnia (Ce 0.8 Hf 0.2 O 2 ) by means of a Pechini-type synthesis and obtained ceramic bodies by sintering at 1973 K in air. The lattice expansion was related to the nonstoichiometry of uorite-type phases using in situ XRD during a switch from vacuum to 2 bar hydrogen at 873 K. Weak reections of other phases emerged during chemical reduction in hydrogen at 873 K. However, there were no indications that a pyrochlore phase had formed. The pyrochlore structure and cation ordering have not been reported for ceria-hafnia nanoparticles. Zhou and Gorte 52 reported single-phase Ce 0.5 Hf 0.5 O 2 nanoparticles and Raitano et al. 53 reported the phase boundary of the cubic domain of nanocrystals, which was close to x ¼ 0.5 in Ce 1Àx Hf x O 2 . Sharma et al. 54 performed XANES (at the O K and Ce M 5 edges) and EXAFS (at the Ce K, Zr K and Hf L 3 edges) of nanoparticles with the composition Ce 0.5 Zr 0.5 O 2 , Ce 0.5 Hf 0.5 O 2 and Ce 0.5 Hf 0.25 Zr 0.25 O 2 . Fujimori et al. 55 conrmed the existence of the tetragonal "t" phase in ceria-rich compounds with up to 20 mol% hafnia by means of Raman scattering and synchrotron XRD.
We report cation ordering in equimolar ceria-hafnia materials as well as the formation of cation-ordered ceria-hafnia pyrochlore and describe its stability under realistic solar thermochemical looping conditions. In addition to thermogravimetric analysis (TGA) and X-ray diffraction (XRD), we used our high-temperature in situ cells 56,57 for X-ray absorption near edge structure (XANES) and extended X-ray absorption ne structure (EXAFS) measurements at the Ce K and Hf K edges, and high-energy in situ synchrotron XRD with a focused beam from RT to 1823 K.

Material synthesis
The ceria-hafnia materials were synthesized by a polymerized complex method (Pechini-type synthesis). (NH 4 ) 2 Ce(NO 3 ) 6 (Alfa Aesar, 99.9%) and ZrO(NO 3 ) 2 , (Alfa Aesar, 99.9%) or HfCl 4 , (Alfa Aesar, 99.9%) were dissolved in deionized water. Anhydrous citric acid (CA, Sigma Aldrich, 99.5%) was added to 1,4-butanediol (BD) (Sigma Aldrich, 99%) at a 1 : 4 molar ratio (with [CA] : [M 4+ ] ¼ 4 : 1) and stirred at 373 K until all citric acid dissolved. The aqueous solution containing the dissolved metal salts was added to the CA/BD mixture and heated to 423 K to promote the esterication reaction. The resulting highly viscous polymer was dried at 353 K for 4 h and red for 10 h at 973 K in a constant air ow to remove organic matter. 19 Sintering of samples was performed in a Carbolite HTF 17/10 furnace at 1873 K for 5 h in air.

Thermogravimetric analysis
Thermogravimetric analysis (TGA) experiments were carried out using a Netzsch STA 449F3 thermogravimetric analyzer. Samples were placed inside an alumina crucible supported by an alumina rod which houses a thermocouple. Aer oxidation at 973 K in air, the temperature was reduced to 573 K and the gas ow in the analyzer was switched to argon 5.0. The heating and cooling rate was 30 K min À1 . A blank run was used to correct buoyancy effects. Between the TGA and XRD measurements samples were ground in a mortar and stored for ve months under ambient conditions.

X-ray diffraction
X-ray diffraction measurements of pressed pellets, aer calcination and aer sintering, respectively, were performed using a Bruker D8 diffractometer with a Cu Ka source. The step size was 0.014 2q and the scan speed 1 min À1 . For X-ray diffraction measurements aer TGA, the pellets were ground in a mortar and a powder-ethanol slurry was deposited onto a sample holder. Coupled theta-two-theta scans were performed on another Bruker D8 instrument with a Cu source on a 280 mm measurement cycle, 2.5 axial Soller slits, a Ni lter, and a Lynxeye detector. The step size was 0.0296 2q and dwell times were 96 (Hf10 and Hf20) and 384 ms (Hf50). The Ka 2 signal was stripped by means of the Bruker Diffrac.da vinci soware.

In situ X-ray absorption spectroscopy
In situ XAS experiments were carried out at BM01B, Swiss-Norwegian Beam Lines at the European Synchrotron Radiation Facility in Grenoble, France. BM01B, a multi-technique beam line provided access to very hard X-ray photons. 58 The storage ring (6 GeV) was operated in 7/8 multibunch lling mode at a maximum ring current of 200 mA. Measurements were performed in transmission mode. Fresh ceria-hafnia powder (31 mg) was pressed at 1 ton to obtain a thin pellet with a diameter of 5 mm. The pellet was mounted on an alumina sample holder and placed inside a laboratory-constructed hightemperature in situ XAS cell 56 consisting of an alumina tube heated by an infrared furnace (Ulvak VHT E44). The temperature was measured by a Pt/Rh thermocouple, and the cold ends of the tube were sealed by PTFE windows. The composition of the product gases was monitored by a quadrupole mass spectrometer (MS, Pfeiffer Omnistar GSD320). X-ray intensities were monitored by ionization chambers lled with 1 bar krypton/ argon (20/80 vol%) and 1.2 bar krypton before and aer the in situ cell, respectively. A ceria or hafnia/cellulose pellet was placed as a reference between the second and third ionization chamber. At both the Ce K and Hf K edges, the Si-111 monochromator was tuned to maximum intensity because rejection of higher order harmonics is unnecessary. Owing to their high energy (ca 125 keV at the Ce K edge) and their low intensity in the spectrum of the source, a bending magnet with critical energy of 20 keV, their interaction with the ion chambers was negligible. Energy scans were carried out at a step size of 1 eV. Ce K edge XANES and EXAFS scans were carried out with a dwell time of 50 and 200 ms, respectively. Hf K edge XANES and EXAFS scans with a dwell time of 100 and 200 ms, respectively, and 3-5 scans were merged (in m(E)) for XAS data analysis. The ow rate of argon 6.0 (BIP, Air Products) was 100 mL min À1 .
The absolute value of the Hf K edge energy (tabulated value: 65.351 keV) was not determined, because Hf metal was unavailable as a standard. About 25% of the incoming intensity was absorbed at 65 keV in the ion chambers that were lled with Kr at 1.1 bar. Hf K edge spectra were normalized by tting a linear background in the pre-edge range and a second order polynomial in the post-edge region, À230 to À150 and 70 to 350 (XANES) or 1500 eV (EXAFS) relative to the maximum of the rst derivative, respectively. Hafnia (Alfa Aesar, 99.95%, see ESI †) was the reference material. Data processing and analysis were performed with MATLAB code and the Athena and Artemis soware packages. 59 Fitting of Ce K and Hf K edge EXAFS data was performed by exchanging Zr by Hf in the feff6 input le which was generated by the 'atoms' code of Demeter 0.9.21 and crystallographic data of Ce 2 Zr 2 O 7 determined by neutron diffraction (Raison et al., ICSD collection code 168595). 60 Fitting was performed in R-space at 1.15 to 4.4Å with a Hanning window function for the Fourier transform.

In situ X-ray diffraction
During in situ XRD measurements, the storage ring was operated in 16 bunch mode with a ring current of 90 mA and a halflife of 10 h. The combined XAS/XRD measurements were carried out in the high temperature in situ cell. 57 The beam was focused at a photon energy of 41.507 keV, which corresponds to a wavelength of 0.2987Å. The wavelength was calibrated against a Si NIST standard using the beamline's high-resolution diffractometer. The intensity of diffracted photons was collected with a Dexela 2923 CMOS 2D pixel detector. In this conguration, a resolution of typically 0.005 2q is achievable. Sets of ve dark (background) and ve exposed images were recorded with an exposure time of four seconds and a fast shutter. From each set, one background-corrected diffraction pattern was obtained. The time resolution of the measurements was 0.876 min À1 . The position of the sample relative to the pixel detector was determined by means of a LaB 6 660c NIST standard in a 0.4 mm quartz capillary mounted on top of the inner ceramic tube. To ensure that the sample was exactly 3 mm inside the tube, the whole cell was moved 3 mm towards the source. Data reduction by azimuthal integration was performed with the PyFAI soware package 61 and background subtraction was performed in MATLAB and the baseline estimation and denoising using sparsity (BEADS) algorithm. 62 The auto-reduced pellet was decorated with platinum by dipping it into hydrogen hexachloroplatinate (H 2 PtCl 6 (6H 2 O), 99.9%, (38-40% Pt), Acros) to increase the reaction rate on the surface before the isothermal carbon dioxide splitting experiment. Gases for in situ XRD experiments were argon 6.0 (BIP, Air Products), 5% hydrogen 5.0 in helium 5.0 (Messer gases), and carbon dioxide 4.8, (O 2 # 2 vpm, CO # 1 vpm, H 2 O # 3 vpm, CH 4 # 2 vpm, N 2 # 8 vpm). The ow rates of argon and hydrogen/helium were 200 mL min À1 , and the ow rate of carbon dioxide was 100 mL min À1 .

Sample preparation
Samples of Hf10C (Ce 0.9 Hf 0.1 O 2 ), Hf20C (Ce 0.8 Hf 0.2 O 2 ) and Hf50C (Ce 0.5 Hf 0.5 O 2 ) were synthesized by the polymerized complex method and calcination (C) at 973 K in air. Fig. 1 presents XRD patterns of samples that were collected aer calcination (C) at 973 K as well as aer sintering (S) pressed pellets in air at 1873 K. The XRD patterns of the calcined materials indicate the uorite-type crystal structure of ceria. The peaks were substantially broadened due to the small crystallite size that is obtained with the polymerized-complex method. Reections of phases other than the uorite-type were not detected. The positions of the peaks were very similar in the diffraction pattern of calcined Hf10C and Hf20C, while the peak positions in the diffraction pattern of calcined Hf50C shied to higher angles, indicating a contraction of the lattice with increasing hafnia content. The diffraction peaks of calcined Hf20C were more asymmetric than those of calcined Hf10C and Hf50C. Peak broadening was most pronounced in the diffraction pattern of calcined Hf50C. In the diffraction patterns of calcined Hf10C and Hf20C, the widths of the reections were very similar.
While all reections of sintered Hf10CS and Hf20CS were assigned to the uorite-type crystal structure, a large number of weak reections below 10% normalized intensity were observed in the diffraction pattern of sintered Hf50CS, all of which were assigned to monoclinic and orthorhombic hafnia. In the diffraction pattern of sintered Hf50CS, the normalized intensity of the reections indicating the cubic ceria lattice were asymmetric, smaller and signicantly broader than those in the patterns of sintered Hf10CS and Hf20CS. Reections indicating the pyrochlore structure were not detected. Fig. 2 shows the mass loss of the three samples determined by thermogravimetric analysis during heating to 1773 K in a ow of argon. The oxygen release proles of Hf10C and Hf20C were very similar: auto-reduction started at about 1173 K and weight loss occurred in two steps. Before and at the rst decrease in weight at intermediate temperature, the auto-reduction proles of Hf10C and Hf20C overlapped. The mass change in Hf20C was higher than that in Hf10C in the second step at higher temperature. Hf50C also showed that oxygen was released in two steps; however, the onset of auto-reduction started already at about 573 K. The weight loss prole of Hf50C was similar to Hf10C and Hf20C at temperatures up to about 1173 K but at higher temperature there was a substantially higher rate of oxygen release. The mass loss of all three samples leveled out when they were kept at 1773 K for about 100 minutes and did not change during subsequent cooling to 573 K. Table 1 reports the weight changes and composition of the auto-reduced materials assuming complete oxidation of the pristine material and that weight loss was solely caused by the release of oxygen. While the composition of Hf10CT Hf20CT is typical of nonstoichiometric uorite-type ceria, that of Hf50CT suggests changes in the crystal structure and complete reduction of cerium from Ce 4+ to Ce 3+ . Fig. 3 displays XRD patterns of the ceria-hafnia materials aer thermal auto-reduction (T). Auto-reduced Hf10 showed only the reections of the uorite-type phase, which indicates that hafnium ions were well dispersed/randomly distributed and that they substitute cerium ions in the uorite-type lattice. Thermal reduction of the materials with a higher hafnium content gave rise to additional very weak reections with normalized intensity below 0.03, which indicate the ordering of cations in pyrochlore-type and pyrochlore-related crystal structures. The peak positions were located at slightly higher angles than that in literature data based on XRD 40 and neutron diffraction 60 of pyrochlore-type ceria-zirconia. The main component of Hf20CT is a uorite-type solid solution; however also some reections indicating the presence of a cation-  ordered phase were found. The uorite-type phase in Hf20CT had a slightly smaller lattice parameter than that of Hf10CT, indicated by a shi to higher diffraction angles.

Thermogravimetric analysis
Reections assigned to the cation-ordered phase of autoreduced Hf20CT were less intense and appeared at lower angles than corresponding reections in the diffraction pattern of Hf50CT. The reections arising from cation ordering in the diffraction pattern of Hf50CT were sharp and, in contrast to the diffraction pattern of Hf20CT, most of the reections of the pyrochlore crystal structure were visible. At the scattering angles of the reported 40 (220) indicated the presence of a very small amount of monoclinic or orthorhombic hafnia. Corresponding reections of cubic lattice planes of the cation-ordered structure appeared at higher angles relative to the uorite-type lattice of Hf10CT and Hf20CT, indicative of a smaller lattice parameter in auto-reduced Hf50CT. Peaks in the pattern of Hf50CT were asymmetric with a tail at the higher 2q side.

3.3
In situ XAS at the Ce K edge Fig. 4 shows time-resolved Ce K edge XANES spectra that were collected during the auto-reduction (R) of Hf50C in a ow of argon. The temperature was rst ramped to 1773 K at 33.3 K min À1 , kept constant for 30 minutes before cooling to room temperature at the same rate, similar to the TGA experiment described above (see Fig. 5 for more details). With increasing temperature, thermal damping lowered the amplitude of the absorption ne structure, which smeared spectral features related to structural changes. Except for a discontinuity at about 873 K, the XANES features changed gradually. Fig. 4a shows the drop in the intensity of the white line during heating. The white line intensity of the auto-reduced material changed only slightly upon cooling to room temperature ( Fig. 4c and d). During cooling, there was only a minute shi in the energy of the Ce K edge position to higher energy. In the absorption ne structure of Hf50CR, oscillations of higher frequency compared to those in the spectrum of pristine oxidized Hf50C emerged and the edge position shied to lower energy, indicating signicant changes in the local geometrical and electronic structure of cerium. Fig. 5 presents the temperature prole, the mass spectrometer signals of oxygen, and carbon dioxide of the temperature-programmed auto-reduction of Hf50 in the in situ XAS experiment. The oxygen trace (m/z ¼ 32) was in good agreement with the derivative of the mass change in the TGA experiment (2). In both experiments, two maxima in the rate of oxygen release were observed: a maximum at intermediate temperature and then the absolute maximum at 1773 K. The carbon dioxide signal indicated desorption from the sample and reactor walls and, to some extent also was due to carbon that was burned and hindered the detection of oxygen. In contrast to the TGA experiment, there was no pre-treatment of oxidation in air at 973 K and in argon at 573 K before heating. Thus, the release or desorption of carbon dioxide from the sample and reactor started at very low temperature. When the temperature decreased, all the oxygen released from the sample had been removed from the reactor because there was a very small drop in the oxygen signal to its baseline.   to Ce 3+ . In agreement with the MS signals, XAS indicated that auto-reduction to Ce 3+ was virtually complete when the temperature reached 1773 K. The non-stoichiometry reached a maximum of 0.29 at 1773 K, and the mean value determined from the 10 last spectra was 0.28 AE 0.04. This clearly indicated that, during auto-reduction, cerium changed from Ce 4+ to Ce 3+ , which is required for the formation of the ceria-hafnia pyrochlore structure. In contrast to the MS oxygen signal, there were not two inection points in the curve of d determined by XAS. The non-stoichiometry d corresponds to the cumulative oxygen release and thus its temporal derivative gives the oxygen release rate, which is proportional to the MS signal of oxygen.

Hf K edge XANES
Immediately aer measurements at the Ce K edge during autoreduction of Hf50CR ( Fig. 4 and 5), spectra were recorded at the Hf K edge at room temperature. Fig. 6 presents Hf K edge spectra of Hf50C and Hf50CR, recorded before and aer the in situ XAS auto-reduction experiment. Furthermore, it shows spectra of the reference material, hafnia powder that consisted of a mixture of monoclinic and orthorhombic phases (see ESI †). Spectral features were less pronounced in the spectrum of pristine Hf50C compared to the auto-reduced material, mainly due to lower crystallinity and homogeneity of the powders compared to the dense ceramic body of auto-reduced Hf50CR. A spectral feature close to the absorption edge at a normalized absorption of 0.92 was most intense in Hf50CR, less pronounced in the hafnia reference and weak in pristine Hf50C. The energy at the rst maximum (white line) was highest in the spectrum of Hf50CR. At energy higher than that of the white line, the largest amplitude of the ne structure was found in the spectrum of auto-reduced Hf50CR. The differences in the energy and intensity of the Hf K edge XANES features of Hf50C and Hf50CR suggest changes in the oxygen coordination geometry. Fig. 7 shows the Hf K edge EXAFS signals of Hf50C and Hf20C recorded before and aer thermal reduction ( Fig. 4 and 5). There were signicant differences in the amplitude and the signal/noise ratio of the data. The EXAFS signal of auto-reduced Hf50C had the largest amplitude and the best signal/noise ratio. In the Fourier-transform of all the spectra except that of calcined Hf50C, two main peaks were identied, the rst and most intense originating from the rst coordination shell of oxygen atoms and the second from the second coordination shell consisting of cerium and hafnium atoms at higher radial  Fig. 5). Black arrows indicate changes in spectral features. Spectra in (a and b) were recorded during heating from room temperature (blue line) to 1773 K and at 1773 K (red lines); spectra in (c and d) were recorded during cooling from 1773 K (red line) to room temperature (blue line). Spectra in (e and f) were recorded at room temperature before (blue solid line) and after (blue dashed line) auto-reduction and at 1773 K before cooling (red line). Correction by aligning the simultaneously recorded reference spectra was not carried out.

Hf K edge EXAFS
distance. The crystallinity of the samples is much higher in the dense ceramic body of Hf50CR compared to Hf50C, which consisted of calcined powder. In the pyrochlore ceria-zirconia structure, 60 the rst coordination shell of zirconium consists of six oxygen atoms at a distance of 2.097Å. Fig. 8 presents a t of the Hf K edge EXAFS of Hf50CR. Table  2 presents the numerical results of coordination numbers, Debye-Waller factors, and interatomic distances. The real part and the magnitude of the t are in excellent agreement with the data, which is reected in the excellent goodness-of-t parameters (reduced c 2 ¼ 462.6 and R-factor R f ¼ 0.0063). Deviations at low radial distance are small and due to low frequency components in the background. Results of the t are in very good agreement with the pyrochlore structure. The hafniumoxygen distance is slightly shorter than that of the zirconiumoxygen bond of the initial model, which is in agreement with the slightly smaller ion radius of Hf 4+ compared to Zr 4+ (0.72 and 0.71Å) in a six-fold coordination. S 0 2 was manually adjusted such that a Hf-O coordination number of 6.0 was obtained in the t. The Hf-Ce and Hf-Hf coordination numbers are in agreement with the expected theoretical value of 6.0. While the Hf-Ce coordination number is lower and the bond length is shorter compared to the starting value, the Hf-Hf coordination number is higher than that of Hf-Ce with longer bonds. Consistent with and very similar to the results with a slightly smaller statistical uncertainty were obtained by the tting of the data in a larger k-range.

3.6
In situ X-ray diffraction 3.6.1 Chemical reduction and auto-reduction. Fig. 9 displays selected reections of the diffraction patterns of Zr50 and Hf50 that were recorded during heating from room temperature to high temperatures in a ow of hydrogen/helium   (chemical reduction), in a ow of argon (auto-reduction), and in air. Fig. SI 3 † shows normalized diffraction patterns. The temperature was increased at a rate of about 50 K min À1 , and 1823 AE 50 K was reached in argon and air and 1623 AE 50 K in hydrogen/helium. In hydrogen/helium the temperature was about 200 K lower due to the high heat conductivity of helium. The 2q-range, which is limited by the X-ray energy and the geometry of the in situ cell, was about 1 to 18 2q. The scattering contribution of the quartz tube appeared as a smooth background with a very broad peak at 3.8 2q. The normalized intensity of the background was signicantly higher in the pattern of Zr50C compared to that of Hf50C.
In the diffraction patterns of materials that were exposed to reducing conditions, the reections of the cubic uorite-type structure appeared at higher scattering angles than expected for pure ceria and contained additional very weak reections, some of which indicated the formation of the cation-ordered pyrochlore structure. At the end of the heating ramp the peaks in the pattern of Hf50C that had been heated in a ow of hydrogen/helium were wider than the peaks in the pattern of Zr50C. During chemical reduction in a ow of hydrogen/ helium, more peaks appeared in the diffraction pattern of Hf50C than in that of Zr50C. Many of these reections were tentatively assigned to the monoclinic or the orthorhombic hafnia phases. The patterns of Hf50C heated in air and the pattern recorded aer sintering in air at 1873 K (Fig. 1) were very similar and differed only in the intensity of the reections. In both diffraction patterns, reections of hafnia were present and reections indicating cation ordering were absent. Most reections of ceria-zirconia pyrochlore Ce 2 Zr 2 O 7 emerged during reduction of Zr50C in a ow of hydrogen/helium at a maximum temperature of 1623 AE 50 K: the (111), (331) and (333), (511) and more reections were identiable. The peak of the very weak (442) reection of the pyrochlore could not be assigned due to potential overlap with a reection of the monoclinic zirconia phase that emerged aer the temperature ramp.
In the patterns of Hf50C recorded during reduction in hydrogen/helium, there were a larger number of small peaks, many of which were assigned to monoclinic and/or orthorhombic hafnia phases. The (111) reection of the pyrochlore phases of chemically reduced Zr50C and Hf50C in a ow of hydrogen/helium emerged at a similar temperature. The (333), (511) diffraction peaks of auto-reduced and chemically reduced Hf50C were clearly detected close to the position of the peaks of reduced Zr50C. Compared to reduced Zr50C, the intensity of the (333), (511) peaks relative to the (111) reection were slightly weaker. Close to the position of the (311) reection of the Zr50C pyrochlore, three very small peaks were detected in the pattern of chemically reduced and auto-reduced Hf50C, which could not be clearly assigned to the pyrochlore phase.
During thermal reduction of Hf50C in a ow of argon, the (1 11) reection at 5.9 2q indicated the formation of monoclinic hafnia. The (222) peak of the pyrochlore corresponds to the (111) reection of the uorite-type structure. At higher temperature, the normalized intensities of the reections of the monoclinic hafnia phase decreased, while the intensity of the peaks assigned to the pyrochlore increased further. In agreement with the pattern of chemically reduced Hf50C, the (333), (511) reection of auto-reduced Hf50C was clearly detected. There were fewer reections of other phases compared to the reduction of Zr50C and Hf50C in hydrogen/helium at a lower temperature. At the position of the notable (331) reection of ceria-zirconia pyrochlore, there was no peak with increasing intensity related to the pyrochlore phase.
The reections at low angles, indicating cation ordering, were not observed during heating in air. In the heating ramp both the monoclinic and a small amount of the orthorhombic hafnia phase formed. The peak assigned to the (121) reection, the most intense reection of orthorhombic hafnia, was observed at 5.7 2q and decreased to below 0.2% normalized intensity aer 30 min at 1823 AE 50 K (see ESI †). A peak at a very similar position as that of the (331) reection of the pyrochlore also formed by heating in air. While the intensity of the (111), (331), and (333), (511) reections associated with cation ordering increased in the pattern of reduced Zr50C, the corresponding intensity of the reections of Hf50C decreased in hydrogen/helium and argon. The intensity of the (311) reections in the patterns of reduced Zr50C and reduced Hf50C changed only slightly. In reducing atmosphere, there were only Fig. 8 Fit of the Hf K edge EXAFS of auto-reduced Hf50CR to the pyrochlore structure. Magnitude (blue line) and real part (red line) of the data (solid lines) and the fit (dashed lines). The k-range for the Fourier transform was 4.0 to 11.6Å À1 . Table 2 gives numerical details. Table 2 Hf K edge EXAFS fits of auto-reduced Hf50CR to the pyrochlore structure 60 (see Fig. 8). CN is the coordination number, s 2 the Debye-Waller factor, and DR is the change of the inter-atomic distance relative to the literature value R. The R range of the fits was 1.1 to 4.4Å. For all scattering paths, one common parameter for DE 0 was used which gave a value of À5.64 AE 1.37 eV. All parameters except for S 0 2 were "guess" parameters. S 0 2 was manually adjusted so that a Hf-O coordination number of 6.0 was obtained at S 0 small changes in the intensity of the hafnia phases. In contrast, in air, the intensity of the reections of monoclinic hafnia increased and the most intense peak split into two peaks, indicating the presence of substantial amounts of monoclinic hafnia and a uorite-type phase. Upon cooling, there was no signicant change in the intensity of the (121) reection, indicating the presence of a small amount of orthorhombic hafnia. 3.6.2 Isothermal carbon dioxide splitting. Fig. SI 6 † gives the furnace prole and MS signals recorded during the in situ XRD experiment under isothermal carbon dioxide splitting using a platinum-decorated Hf50C pellet that had been autoreduced in a ow of argon. The temperature increased at about 50 K min À1 to 1800 AE 50 K. Subsequently, the gas feed alternated between 30 min in a ow of argon and 30 min in a ow of carbon dioxide. The release of oxygen (m/z ¼ 32) was not observed when the sample was heated during the rst part of the experiment. The MS signals of 28 and 16 differed in the three oxidation steps, whereas we could not infer formation of CO from carbon dioxide due to re-oxidation of the auto-reduced Hf50C. Fig. 10 presents diffraction patterns of auto-reduced ( Fig. 9), platinum-decorated Hf50 under conditions of isothermal carbon dioxide splitting. During heating in argon to 1800 K (not shown) and for 30 minutes at 1800 AE 50 K in a ow of argon, there were no signicant changes in the diffraction pattern.
Switching from a ow of argon to a ow of carbon dioxide caused a gradual decrease in the intensity of the (111) and (311) reections of the pyrochlore phase to less than a third of their initial intensity, while the intensity of the (333), (511) reection increased. The normalized intensity of the peaks of monoclinic Fig. 9 Details of XRD patterns, highlighting reflections related to the pyrochlore structure recorded during heating of (a) Zr50C in hydrogen/ helium, (b) Hf50C in hydrogen/helium, (c) Hf50C in argon, and (d) Hf50C in air during heating at about 50 K min À1 from room temperature to 1623 AE 50 K in (hydrogen/helium) and to 1823 AE 50 K in argon and air. Blue lines indicate room temperature and red lines indicate high temperature. Arrows indicate directions of changes. Reflections are assigned to the pyrochlore ceria-hafnia phase (*), monoclinic hafnia ( ‡), and orthorhombic hafnia (#). Some peaks could not be definitely assigned ($).
hafnia increased signicantly during the oxidation step. Subsequent exposure to a ow of argon led to a slight increase in the intensity of the pyrochlore (111) and (311) reections and a decrease in the intensity of the (333), (511) reection. The intensity of the (333), (511) reection increased dramatically when the gas feed was switched. Aer the second switch from a ow of argon to carbon dioxide, the (111) and (311) peaks decreased gradually, and aer about 15 min exposure to carbon dioxide they were no longer distinguishable from noise. The intensity of the peak assigned to the (333), (511) reection increased gradually and then decreased again. Its intensity was signicantly higher in the rst pattern recorded in carbon dioxide compared to the last pattern in argon before the second switch of the feed gas.
There was a weak peak at a slightly higher Bragg angle than the (440) reection of the pyrochlore structure. Its intensity decreased slightly during heating and at 1800 K in a ow of argon. In a ow of carbon dioxide, its intensity increased and formed a shoulder in the (440) reection. Its intensity went up during the rst exposure to carbon dioxide, which led to the formation of a broad shoulder in the (440) peak. During subsequent exposure to a ow of argon, the shoulder changed to an individual peak with a higher intensity compared to that in the pattern recorded in argon aer reaching isothermal conditions. During the second exposure to carbon dioxide, the intensity of the peak increased further, forming a broad shoulder of the peak assigned to the (400) reection of the pyrochlore structure. The (400) peak of the Ce 2 Zr 2 O 7.5 phase was Fig. 10 Diffraction patterns of auto-reduced Hf50C, highlighting changes in the reflections related to the pyrochlore structure under isothermal carbon dioxide splitting conditions. Normalized intensities in %. A pellet Hf50C was auto-reduced at 1823 K in a flow of argon and impregnated with platinum. Subsequently, it was heated to 1800 AE 50 K in a flow of argon and the feed gas was switched at a half period of 30 min from a flow of (a) argon to (b) carbon dioxide, (c) argon, and (d) carbon dioxide. Blue indicates the first and red the last diffraction pattern in a series. Arrows indicate changes in the intensities of the reflections assigned to the pyrochlore structure. Reflections were assigned to the pyrochlore ceriahafnia phase (*), monoclinic hafnia ( ‡), orthorhombic hafnia (#). Some peaks could not be definitely assigned ($).
found at a higher angle than that of the (400) reection of the Ce 2 Zr 2 O 7 pyrochlore. However, the peak may also be due to the (002) reection of monoclinic hafnia.

Discussion
XRD aer auto-reduction in a thermogravimetric analyzer showed the formation of ceria-hafnia pyrochlore in Hf50CT. The pyrochlore was stable for several months under ambient conditions. The release of oxygen from equimolar ceria-hafnia in the thermogravimeter indicates an average oxidation state of ceria in Hf50CT higher than +3.0, which is unusual. The reduction is probably due to a systematic error, i.e. a dri of the thermobalance in the experiment of Hf50CT, which was corrected with a linear function. On the other hand, Baidya reported 51 a ceria-zirconia pyrochlore-type structure with a composition of Ce 2 Zr 2 O 6.2 prepared by reduction in hydrogen/ argon at 1023 K. In contrast, under the same reaction conditions, Baidya did not report the formation of ceria-hafnia pyrochlore, most likely because the mobility of hafnium cations (M ¼ 178.5 amu) is substantially lower than that of zirconium cations (M ¼ 91.22 amu). The Ce K edge XANES of pristine Hf50C and reduced Hf50CR indicates structural changes, but the spectral features are diffuse at 1773 K. Only aer cooling to room temperature did an oscillation indicate a different cerium-oxygen coordination geometry and cerium-oxygen interatomic distances were observed. The edge shi and the drop in the intensity of the white line during thermal reduction are clear indications of changes in the electronic structure of cerium.
In the acquisition of Hf K edge spectra at 65 keV, the accuracy of the movement of the monochromator is even more critical than at the Ce K edge. 56 The differences in the energy and intensity of the XANES features of Hf50C and Hf50CR suggest changes in the oxygen coordination geometry. XRD of Hf50CT gave compelling evidence of cation ordering.
The ordered arrangement of cations collapsed upon exposure to a ow of carbon dioxide at 1800 K. Aer 30 minutes in a ow of carbon dioxide, the normalized intensity of the (111) reection of the pyrochlore decreased by about 60% but increased by less than 10% of the initial intensity during exposure to a ow of argon. In the second cycle, the reections of the pyrochlore structure disappeared. The reversibility of the order-disorder transformation is probably possible under more intense reducing conditions. However, preparation of pyrochlore from small particles that are not stable at high temperature lead to higher mobility of the cations compared to the oxidized bulk material that consists of at least two phases. It may be that the conditions during the reduction step of isothermal carbon dioxide splitting were not reducing enough to reverse the decomposition of the pyrochlore phase. Ceriahafnia pyrochlore is unstable under oxidizing conditions at 1773 K, which is in agreement with the work of Andrievskaya et al. 48,49 who determined the phase diagram of ceria-hafnia at 1773 K in air. It is unclear if as to whether or not a temperature range enables the complete oxidation of the cation-ordered material analogous to Ce 2 Zr 2 O 8 , the k-phase of ceria-zirconia.
Otobe et al. prepared Ce 2 Zr 2 O 8 by oxidation of pyrochlore in a stream of oxygen at 873 K.
The formation of the cation-ordered phase demonstrates the inuence of the preparation method on material performance in thermochemical looping. Le Gal et al. 27,64 found substantial differences in the carbon monoxide yield obtained from commercial materials (Rhodia, Solvay group) and materials prepared by co-precipitation and synthesis by the polymerized complex (Pechini-type) method. Ceria-hafnia pyrochlore might also form in thermochemical cycles in inhomogeneous samples with a lower than equimolar hafnia content. Therefore, it is important to prepare materials with a homogeneous distribution of the heterocation, for example by means of the polymerized complex. 18,19,22,64 Inhomogeneities and small domains with high concentrations of heterocations with cation-ordered structures might explain why the synthesis route and the thermochemical history of a sample matter. Bonk et al. 19 compared materials prepared by the ceramic method and a polymerized complex method and eliminated the potential inuence of the preparation method: sintering pressed pellets at 1873 K led to qualitatively equal materials in thermochemical carbon dioxide splitting.
Kuhn et al. 65 attributed the difficulty in describing the thermodynamic properties of ceria-zirconia by a simple model aer exposure to reducing conditions at elevated temperature to ordering phenomena. Oxidation at high temperature reversed these effects. This is in agreement with the collapse of the ceriahafnia pyrochlore structure under oxidizing conditions at 1800 K (Fig. 10). The pyrochlore Ce 2 Hf 2 O 7 is an important phase in the ceria-hafnia system, because it represents an extreme case of the density of oxygen vacancies. Cation-ordering facilitates the complete reduction of cerium from Ce 4+ to Ce 3+ . The local structure of oxygen vacancies in uorite-type ceria-hafnia solid solutions is closely related to the unoccupied oxygen sites in the pyrochlore. Filling these oxygen vacancies may lead to the formation of a k Ce 2 Hf 2 O 8 phase, the analogue of the ceriazirconia k phase. Thus, the determination of short-range and long-range ordering in ceria-based oxides containing tetravalent heterocations such as Zr 4+ , Hf 4+ or Sn 4+ under relevant reaction conditions is very important. A signicant improvement in the measurement and description of ordering phenomena by quantitative determination of structural properties during oxygen uptake of pyrochlores might enable us to pinpoint the elusive oxygen vacancies and lead to a better understanding of defect association. This would lead to a signicant improvement in the accuracy of thermodynamic models such as that used by Kuhn et al. 65 Pyrochlore-type structural features can lead to a higher fuel yield and must be considered in the engineering of oxygen vacancies. A high fuel yield, however, does not necessarily imply an improvement in the efficiency of solar-to fuel energy conversion of a solar thermochemical reactor system. [66][67][68][69][70] Marxer et al. 71 hold the current world-record in the energy-tofuel conversion efficiency by using pure ceria. The thermodynamic properties of cation-ordered ceria-based oxides for solarto-fuel energy conversion efficiency of two-step thermochemical water or carbon dioxide splitting have not yet been determined.
Ruan et al. 72 recently proposed a two-step solar thermochemical water splitting based on CeO 2 /SnO 2 and the pyrochlore Ce 2 Sn 2 O 7 . In seven consecutive cycles, CeO 2 /0.15SnO 2 was thermally reduced at 1673 K in a ow of argon and subsequently reacted with steam at 1073 K. Reduction led to the formation of a pyrochlore phase and a hydrogen yield 3.8 times higher than that of pure ceria.

Conclusions
The ceria-hafnia pyrochlore Ce 2 Hf 2 O 7 with high purity can be prepared via auto-reduction of ceria-hafnia powders synthesized via a polymerized complex method. The combination of in situ XAS and XRD measurements provides insight into the dynamics of structural changes under relevant solar reactor conditions. The pyrochlore structure Ce 2 Hf 2 O 7 may form during the reduction step of solar thermochemical looping, but it is converted to a mixture of uorite-type or tetragonal ceria, monoclinic hafnia, and orthorhombic hafnia in the oxidation step at 1800 K. The ordered arrangement of cations might persist during oxidation at lower temperature. Very little is known about the thermodynamics and the structural stability of the pyrochlore and k-phases of the ceria-hafnia and ceriazirconia systems. The properties of pyrochlore-type materials may be further tuned for applications that require the storage, transport, and activation of oxygen by adding heterocations.

Conflicts of interest
There are no conicts to declare.