Pore morphology evolution and atom distribution of doped Fe2O3 foams developed by freeze-casting after redox cycling

Original Article Pore morphology evolution and atom distribution of doped Fe 2 O 3 foams developed by freeze-casting after redox cycling P.J. Lloreda-Jurado, Jesu ´s Hern andez-Saz, E. Chicardi, A. Pau ´l, R. Sepu ´lveda * Departamento de Ingenierı´a y Ciencia de los Materiales y del Transporte, E.T.S. de Ingenieros, Universidad de Sevilla, Avda. Camino de los Descubrimientos s/n., 41092 Sevilla, Spain article info Article history: Received 5 March 2021 Accepted 3 June 2021 Available online 9 June 2021 Keywords: Chemical looping system Fe 2 O 3 Pore evolution Freeze-casting Camphene Al 2 O 3 abstract Chemical looping water splitting systems operate at relatively high temperatures (450 e800 C) to produce, purify, or store hydrogen by the cyclic reduction and oxidation (redox) of a solid oxygen carrier. Therefore, to improve long-term operation, it is necessary to develop highly stable oxygen carriers with large specific surface areas. In this work, highly interconnected doped Fe 2 O 3 foams are fabricated through the freeze-casting technique of a submicrometric camphene-based suspension to prevent Fe sintering and pore clogging during redox operation. The influence of the dopant elements (Al and Ce) over the pore morphology evolution, and redox performances are examined. The use of an Fe 2 O 3 porous structure with initial pore size above 100 microns shows a significant reduction of the sample densification, and the addition of Al 2 O 3 by the citrate method prevent the rapid formation of an Fe 3 O 4 layer at the foams struts that diminish the reoxidation rate step in the redox processing. ©2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction In recent decades, the reduction of CO 2 emissions from fossilfuel consumption has become a necessity due to the undeniable effects of global warming [1]. Switching to a clean and renewable energy source is an unavoidable route to counterbalance this harmful effect. Hydrogen seems a valid alternative, since no CO 2 is emitted at its consumption, although current large-scale production is still based on reforming methane, oil, or coal [2]. Therefore, a more realistic option for cleaner hydrogen production involves utilising renewable (or CO 2 neutral) resources in combination with carbon-capture systems [3]. The Chemical-Looping Water-Splitting process (CLWS) or Steam-Iron Process (SIP), wherein iron is employed, are promising technologies for hydrogen production and/or purification that can combine both these technological requirements. The SIP is a two-step process, where first a suitable solid oxide (i.e., an oxygen carrier) is exposed to a reducing gas stream to turn it into metal iron. The gas stream *Corresponding author. E-mail address: [email protected] (R. Sepu´ lveda). Available online at www.sciencedirect.com journal homepage: www.elsevier.com/locate/jmrt journal of materials research and technology 2021;13:1887e1898 https://doi.org/10.1016/j.jmrt.2021.06.008 2238-7854/©2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/). could be a mixture of hydrogen and CO, commonly referred to as syngas [4]. A steam flow then re-oxidizes the Fe into iron oxide and produces pure hydrogen by splitting the water molecule. Finally, the re-oxidized iron oxide is employed to restart the process. Hematite (Fe 2 O 3 ) presents an attractive oxygen carrier capability for the production, purification, or storage of hydrogen by SIP since it is abundant and inexpensive and also due to its redox properties [5e7]. Iron could act as a secondary source of hydrogen storage if it is oxidized by steam to form Fe 3 O 4 and if the hydrogen released is then collected, theoretically, at standard temperature and pressure conditions, whereby 0.534 L of H 2 can be stored per gram, which is comparable with the storage density of conventional metal hydrides [8]. Nonetheless, the performance of Fe 2 O 3 as an oxygen carrier is deeply affected by the sintering phenomena, also called deactivation, since many applications require numerous redox cycles at high temperatures. In order to prevent or delay particle sintering and to control the redox reaction kinetic, many studies have doped the hematite with diverse elements, such as Al [9], Ce [10,11], Zr [12], Mo [13], La [14], and Ca [15,16]. Other researchers have demonstrated the use of structural porous materials as oxygen carriers to replace the fluidized-bed systems. Honeycomb monoliths [10,17] and Fe 2 O 3 foams [18] proved to possess comparable redox performances. CeO 2 as an additive induces crystal modification and stabilisation of the iron oxides [19], despite the limited solubility observed among Fe x O and CeO 2 phases since Fe 3þ ions can be incorporated at the CeO 2 fluorite crystal structure at lattice (Ce 4þ ) and interstitial position by hydrothermal [20] or coprecipitation [21] routes rather than by the standard milling process. Furthermore, during high temperature operation, the formation of hematite-like solid solutions and perovskite CeFeO 3 phases (Ce 3þ ) is promoted. Therefore, the increase in vacancy concentration enhances the redox performance of Fe 2 O 3 eCeO 2 systems by higher oxygen mobility [22]. Al 2 O 3 ,as an additive, improves the cycle stability of Fe 2 O 3 during continuous operation [7]. Fe 2 O 3 eAl 2 O 3 shows limit solubility: at 1300 C, about 20 wt% Al 2 O 3 can be incorporated into a hematite-like solid solution (H ss )[23], while at room temperature, solubility is neglected. The thermal cycling induced the formation of hercynite (FeAl 2 O 4 )[24], which has a spinel crystal structure where Fe and Al occupy divalent and trivalent cation positions. This spinel phase is more favourably formed during the reduction reaction and early stage of the oxidation reaction since Fe and FeO phases react exothermically in the presence of Al 2 O 3 . In the case of the Fe 3 O 4 eAl 2 O 3 system, a miscibility gap is formed below 860 C and two phases are found: magnetite-based solid solution (M ss ) and hercynite. At 750 C, M ss shows a solubility maximum of 12 wt %Al 2 O 3 . The use of up to 3 wt% Al 2 O 3 enhances the reduction process of Fe 3 O 4 due to the formation of a network-like structure of wu ¨stite [25,26]. Freeze-casting (FC) is a promising manufacturing technique for the production of porous materials with a wide variety of porosity range, pore size, and morphology [27,28]. It is also cost-effective, eco-friendly, and scalable on an industrial level. However, pure Fe-foams manufactured by FC using water [29] and camphene-based [30,31] hematite slurries have shown significant modification of the pore structure after being subjected to several redox cycles. Authors have found a continuous reduction in the pore size due to the sintering of the Fe cell walls, which is especially critical when initial pore sizes were under 30 mm in diameter. Nevertheless, Lloreda- Jurado et al. [32] have reported doped Fe 2 O 3 foams manufactured by FC with diameter pore sizes from 30 to 130 mm, simply by controlling the processing parameters. Therefore, the aim of this research is to determine the influence of the redox cycling over the pore morphology and atom distribution on doped Fe 2 O 3 foams manufactured by FC with initial pore sizes above 100 mm. The Fe 2 O 3 powder doped with Al and Ce was synthesized by the citrate method and subsequently suspended in camphene. 2. Experimental procedure 2.1. Powder synthesis Doped Fe 2 O 3 powder is fabricated using the citrate method previously described by Lorente et al. [33]. Initially, aqueous solutions of selected metal nitrates and citric acid are stirred at 80 C until a gel is obtained. The resulting gel is dried at 110 C overnight and then calcined at 350 C for 2 h followed by a final step of 700 C for 6 h to ensure a good level of crystallinity. The resulting powder, as-synthesized, is later crushed and sieved to separate particles within 160e200 mm in diameter. The final powder composition is approximately 98 wt% Fe 2 O 3 , 1.75 wt% Al 2 O 3 , and 0.25 wt% CeO 2 . 2.2. Sample preparation Doped Fe 2 O 3 foams were fabricated using the freeze-casting method previously described by Lloreda-Jurado et al. [32]. The method starts by a further reduction of the particle size by ball milling down to 0.2 mm in diameter. A 5 vol.% powder suspension in camphene (Sigma Aldrich, Spain) is then obtained by ball-milling at 60 C for approximately 12 h. First 3 wt% of dispersant agent KD4 (oligomeric polyester provided by CRODA Ib erica) is mixed for 30 min in camphene, and the powder is subsequently incorporated and ball-milled for 8 h. Finally, 20 vol.% polystyrene (Sigma Aldrich, Spain) with a M W ¼350,000 g mol 1 was added as a binder and mixed for another 3 h. Both organic additives were formulated according to the initial powder load and were incorporated sequentially to aid in the proper particle dispersion. The camphene suspension was then poured into a 30 mm in diameter by 15 mm in height PFTE mould, preheated at 60 C placed inside an incubator. Directional solidification was promoted by running water at 42.5 C through the mould base and gradually reducing the incubator temperature until reaching complete sample solidification. A slow solidification rate was employed to enhance pore enlargement during the FC process. Finally, after demoulding, and a 3-day camphene sublimation process at ambient conditions under forced airflow, the sample is sintered in air at 600 C for 2 h for organic-specimen burn-out, plus a sintering of 1100 C for 2 h for particle sintering. After the sintering, the sample as-fabricated measures approximately 9 mm in height by 24 mm in diameter. journal of materials research and technology 2021;13:1887e18981888 2.3. Microstructural and redox performance characterisation The pore morphology characterisation of the as-fabricated sample is determined by Optical Microscopy (OM) and Field Emission Scanning Electron Microscopy (FESEM) across the entire height. Initial pore size and cell wall width were determined by OM images using the non-redundant maximum-sphere-fitted algorithm [34] of the ImageJ software. To estimate the average pore diameter at different areas of the sample height, 5 OM images covering an area of approximately 4 mm 2 in total were used. In turn, the redox behaviour is determined on subsamples of 1.2 mm 1.2 mm 4.5 mm, sectioned from the centre-top part of the as-fabricated sample (to reduce the presence of pores under 50 mm in diameter), while maintaining the freezing direction parallel to the longest dimension. Some additional manual grinding was necessary to obtain a 20 mg subsample. This study is carried out in a thermogravimetric analyser (TG) Netzsch®STA 449 F3 Jupiter using an alumina plate as support for the porous subsample. Two redox conditions, dynamic and isothermal, were analysed under the following setup: 1) Dynamic: reduction up to 700 C with a gas flow of N 2 -20 vol.% H 2 , followed by an oxidation up to 800 C with a gas flow of N 2 -12 vol.% H 2 O. Data points were collected at a rate of 5 data s 1 . 2) Isothermal: 3 consecutive cycles of reduction at 750 C, and oxidation at 450 C. Data points were collected at a rate of 12 data s 1 . Heating rate and flow rates were established at 5 C/min and 100 NmL/min, respectively. In both redox conditions, the porous samples were removed from the STA after the oxidation stage is completed. Two subsamples were tested at each redox condition. After the redox test, subsamples were subjected to X-ray computed tomography (X-CT) to determine the pore morphology variation using a YXLON micro-tomograph model Y.COUGAR-SMT. Optical configuration achieved a voxel of 7.56 mm/pixel with the X-ray source operating at 52 KV and 10.09 mA. The subsamples were rotated 360over the largest axis to obtain 1440 projections during approximately 49 min of scan, the collected 2D cross-sectional images were pre-processed with ImageJ software (8-bit binarisation, threshold, and removal of 1-pixel outliers), and 3D visualisations were made using Avizo software. Dopant element distribution was determined by Transmission Electron Microscopy (TEM) in the as-synthesised and as-fabricated condition. An FEI TALOS F200S is used operating at 200 kV with a field emission gun and equipped with Super-X dispersive X-ray spectrometry systems including two silicon drift detectors (FEI ChemiSTEM), provided with their specific software for acquisition and data processing (Thermo Scientific Velox®). The Crystallography Open Database was used to identify the unknown planes of the CeO 2 and iron oxide particles. Subsequently, X-ray Photoelectron Spectroscopy (XPS) is employed to determine the oxidation state and chemical content of doped Fe 2 O 3 in the as-synthesised,as-fabricated, and Fig. 1 eHAADF-STEM image of the initial doped Fe 2 O 3 powder (as-synthesised) showing an aggregate of particles. (a) EDX analysis shows the distribution of the principal elements: Fe (b), O (c), Al (d), and Ce (e). An HRTEM image of a selected area of the particle (f), displays the atomic arrangement of the crystal plane (110) (g). journal of materials research and technology 2021;13:1887e1898 1889 after isothermal redox processing specimens. A SPECS Hemispherical Energy Analyzer model Phoibos 150 MCD is used. The deconvolution of XPS spectra is performed according to the GaussianeLorentzian Cross Product expression as a fitting function (E.q.1) using PeakFit software (Systat Software Inc.) y¼a0 1þa3xa1 a22 expð1a3Þ1 2xa1 a22(1) where a 0 is the amplitude, a 1 is the centre of the curve, a 2 is the width (a 2 >0), and a 3 is related to the shape of the curve (0 a 3 1, the pure Lorentzian occurs with a 3 ¼1 and the pure Gaussian occurs with a 3 ¼0). Finally, X-Ray Diffraction (XRD) was performed to bear out phase evolution in the as-fabricated and after isothermal redox processing samples. Previous to the XRD study, the specimens were mechanically milled to ensure the sample representativeness. XRD patterns were recorded using a PANalytical X'Pert Pro instrument (Malvern Panalytical Ltd, Malvern, UK) equipped with a BraggeBrentano q/qgeometry, Cu-Karadia- tion source (45 kV, 40 mA), secondary Kbfilter and using a X'Celerator detector. The XRD patterns were obtained in 2 theta degree configuration by scanning between 20and 120 with 0.03steps and a counting time of 800 s$step 1 . The openaccess Crystallography Open Database (COD) was employed to elucidate the crystalline phases developed together with their corresponding structure and space-group symmetries (SGSs). A lanthanum hexaboride, LaB 6 (Standard Reference Material 660b, NIST) pattern is utilised to calibrate the positions of the diffraction lines. 3. Results and discussion 3.1. Doped Fe 2 O 3 powder characterisation: assynthesised Figure 1a shows a High-Angle Annular Dark-Field (HAADF) image in Scanning TEM mode obtained from an aggregate of doped Fe 2 O 3 particles that exhibit an average size of 0.2 mmin diameter. The element distribution analysis performed by EDX shows a uniform distribution of Fe, O, and Al (Fig. 1bec), while Ce is segregated to a different location at the particle surface (Fig. 1e). These results suggest that Ce atoms were depleted from the Fe 2 O 3 matrix during particle synthesis. This is reasonable since oxide phases can be separated during annealing at temperatures above 600 C[35,36]. The doped Fe 2 O 3 particle (Fig. 1f) shows an interplanar distance of 0.251 nm (Fig. 1g), in agreement with the plane 110 of pure bulk a-Fe 2 O 3 . Some minor lattice distortion is detected in other interplanar distances. This is due to Al ions (with smaller ionic radii) randomly occupying cation positions in the rhombohedral crystal structure of a-Fe 2 O 3 . This result corroborates the formation of a (Fe,Al) 2 O 3 solid solution described in the literature [23,37]. 3.2. Doped Fe 2 O 3 foam characterisation: as-fabricated During FC, the directional solidification imposed on the slurry created a unique pore structure. The growing camphene dendrites push the particles into the interdendritic spaces and form particle aggregates that replicate the dendrite morphology. Since the sample heat is extracted through the bottom by a constant-temperature flow of water during solidification, the incubator temperature was continuously reduced as well. The thermal gradient decreased and the solidification front velocity increased across the sample height. This situation promotes the formation of an equiaxed and interconnected porous structure, where the average pore size increases significantly with the sample height as the solidification front moves a few millimetres away from the sample base, as shown in Fig. 2aeb. Average pore diameters were calculated in areas described by Fig. 2aed, as 29, 100, 117, and 127 mm, respectively. ANOVA and Tukey test of the pore size variation along the sample height showed a calculated p-value, q 0.05 , and critical-value of 0.0022, 4.05, and 66 mm respectively. Therefore, there is significant difference in pore size at the bottom (Fig. 2a), but non along the rest of the sample height Fig. 2 eSEM micrographs of as-fabricated doped Fe 2 O 3 foams at different sample heights: 0 mm (a), 3 mm (b), 6 mm (c), and 9 mm (d). Details of the porous cell-wall surface (e, f) and cross-section (g) are shown. Micrographs a-d were taken using the same magnification. Doped Fe 2 O 3 foams before sintering (h) shown PS aggregated into the particle walls and forming struts across them. journal of materials research and technology 2021;13:1887e18981890 (Fig. 2bed). The average thickness of the cell wall showed a much wider size distribution along the sample, from minimum values of 3e4mm to a maximum of 60 mm. Regardless of the position, the cell wall (Fig. 2e) exhibited a highly porous structure on its own. An interconnected pore structure with diameters less than 1 mm is observed at the cellwall surface (Fig. 2eef) and its cross-section (Fig. 2g). This secondary porosity is attributable to the pyrolization of C- based compound incorporated through the foam fabrication. A gas adsorption analysis performed in the as-fabricated material using a Micromeritics ASAP 2420 (N 2 adsorption at 77 K) have shown a BET surface area of 0.613 m 2 /g, much higher than those previously reported for pure hematite scaffolds fabricated under similar conditions [18]. During slurry preparation, dispersant agent and PS were added at different moments of the ball-milling procedure. First, the dispersant agent was added at the beginning of the dispersion process, and after 8 h of ball-milling the polystyrene (PS) is incorporated. KD4 is an oligomeric polyester with a C66 long chain, and displays good adsorption properties due to its carboxylic acid group. This carboxylic group is anchored to the doped Fe 2 O 3 particle surfaces leading to a high steric barrier approximately above 10 nm [38]. Since PS has phenyl functional groups every two C atoms and camphene is a bicyclic skeleton monoterpene, both dispersant and binder show good chemical affinity. The melted camphene forms a non-polar solvent where the PS polymer chains were disentangled (dissolved). Doped Fe 2 O 3 particles could be trapped by the PS long CeC chain due to the hydrogen bond driven by the KD4 previously anchored. Moreover, during directional solidification of camphene, 20 vol.% PS was incorporated into the particles walls and formed struts across them, as showed in Fig. 2h. The sintering process volatilises the C-based species and some carbonaceous reductant gases are produced at even relatively low temperatures [39]. KD4 and PS have a chemical formulation of C 90 H 172 O 10 and (C 8 H 8 ) n-1 respectively, and therefore a certain level of carbothermic reduction could be expected according to Bell's Diagram for the FeeCeO[40], where the magnetite phase remains stable at temperatures lower than 570 C and CO/(CO þCO 2 ) gas ratio below 0.5, but no traces of Fe 3 O 4 is detected by TEM analysis after sintering (i.e., a nearly 0% degree of reduction). In this work, the mole ratio of C/O described as the carbon contend at the KD4 and PS by the oxygen at the hematite is equal to 0.2, which is lower than the value referred to in other studies (0.63 [39]or1[41]) with a degree of reduction above 55%. This result suggests that reductant gases produce during the sintering process, even the close integration between solid particle and C-based species achieved by the slurry preparation, were insufficient to promote a significant reduction reaction, despite the close integration between solid particle and C-based species during slurry preparation and the FC procedure. Fig. 3 eHAADF-STEM image of the doped Fe 2 O 3 foam (as-fabricated) showing an aggregate of particles (a). EDX analysis shows the principal element distribution: Fe (b), O (c), Al (d) and Ce (e). The HRTEM image of the selected area of a CeO 2 particle on (f) displays the atomic arrangement of the crystal plane (111) (g). journal of materials research and technology 2021;13:1887e1898 1891 Figure 3a shows an HAADF-STEM image of an aggregate of particles where the sintering exerts no significant effects on the doped element distribution. The EDX analysis performed (Fig. 3bee) on several particles shows a persistent homogeneous distribution of the Al atoms (Fig. 3d), and Ce atoms forming individual clusters (Fig. 3e). The HRTEM image of Fig. 3f shows the crystalline structure of a CeO 2 particle. A detailed magnification of the CeO 2 particles is shown in Fig. 3g. The distant measures correspond to planes (111) of stoichiometric CeO 2 . 3.3. Doped Fe 2 O 3 foam characterisation: after the redox process During dynamic redox processing (Fig. 4a), doped Fe 2 O 3 foam initiates its reduction at 448 C, which is a higher temperature than that of the powder version (300 C), and the step reaction of hematite to magnetite to iron is inhibited. However, both were completed around 600 C. While the temperature kept rising from 600 C to 700 C, a sintering process on the formed Fe walls is definitely encouraged. Since most redox processing under SIP conditions are performed at temperatures from 700 C to 900 C, there is a clear evidence that regardless of the morphology of the oxygen carrier, a sintering process (i.e., efficiency loss) is inevitably encountered. Through the oxidation reaction, both foam and powder reach a maximum weight gain corresponding to Fe 3 O 4 : this oxide phase is already predicted by the Bell Diagram [40], and has previously been reported by other authors [40,42] in the case of steam oxidation. The use of a foam morphology shows a decrease on the kinetic reaction, while the subsequent formation of iron oxides layers on top of the Fe cell walls limit the access of steam to fresh Fe. Moreover, at 634 C, the oxidation reaction to Fe 3 O 4 could be considered completed. The isothermal redox processing (Fig. 4b) shows weight loss and gain with the alternate reduction and oxidation reaction in the doped foam and powder, respectively. As discussed previously, the reduction reaction at 750 C converts Fe 2 O 3 into metallic Fe, while the steam re-oxidation at 450 Callowsthe conversion to Fe 3 O 4 . Thus, a maximum theoretical value of 97% weight change could be achievable during oxide conversion. In the first redox cycle, Fe 2 O 3 turns suddenly into Fe in just 5 min within the initial 30 min of reduction. However, the oxidation step takes almost 1 h to recover 88% of the weight lost. With the second and third redox cycle, the reduction reaction is again abrupt and nearly completed to metallic Fe in minutes, and the re-oxidation reactions takes place in 1 h, whereby 86% of the weight lost is recovered. In contrast with the powder version, where reduction and oxidation reaction follows the same trend as the foams, the redox performance dropped significantly with the cycling, from 92% in the first cycle to 85% in the third cycle. The doped Fe 2 O 3 has demonstrated a more regular and steadier redox performance, by mitigating the decrease in the oxygen weight gain in the redox cycles. Additionally, the foam is entirely converse to metallic Fe after every reduction reaction, as is the re-oxidation reaction which is primarily limited by the access to fresh Fe, among other thermodynamic considerations. Indeed, the decrease in redox performance is usually attributed to a variety of factors, such as: 1) the reduction temperature, especially above 600 C, promotes the sintering of the Fe walls thereby reducing the porosity, especially those observed in Fig. 2eeg; 2) The Fe 2 O 3 creates a coreeshell structure [43] during redox cycling, which segregates Ce and Al oxides species into the grain boundaries, and ultimately to the wall interiors, due to the lack of solubility. As discussed earlier, Ce atoms were segregated from the (Fe,Al) 2 O 3 solid solution during the powder synthesis and foam consolidation heat treatments. Therefore, the improvement in the oxygen mobility and storage capacity expected from the addition of CeO 2 [44,45] could decrease due to the reduction in the number of interfaces Fe 3 O 4 /CeO 2 . Likewise, aluminium atoms were highly scattered in the doped Fe 2 O 3 powder and foam due to the citrate method employed during their synthesis. Al 2 O 3 prevents the formation of a dense iron shell by promoting a reduction step in the sample surface, from magnetite to a network-like structure of wu ¨stite, and, finally, to iron [26,46], thereby delaying the sintering of the fresh Fe particles [47]. These aspects could counterbalance the noteworthy metal sintering during the reduction step, and decrease the loss of redox performance. The pore morphology evolution of the samples subjected to dynamic and isothermal conditions were characterised by X- Fig. 4 eWeight changes of Doped Fe 2 O 3 foam under dynamic (a) and isothermal (b) redox condition. The doped powder behaviour has been included as a comparison. Fe 2 O 3 ,Fe 3 O 4 , and Fe theoretical weight lines were incorporated as reaction guidelines. journal of materials research and technology 2021;13:1887e18981892 CT. The subsamples extracted from the doped Fe 2 O 3 foams correspond to the top half of the entire sample, where the initial pore size varies from 110 mmatthebaseto143mm at the top. Figure 5a shows the 3D reconstruction of the complete doped Fe 2 O 3 foams after the dynamic redox cycles. No significant shrinkage or crack formation is observed. During constant heating, the dynamic redox condition induced a complete reduction of the doped Fe 2 O 3 foam to Fe, and a re-oxidation to Fe 3 O 4 (magnetite) phase due to thermodynamic limitation. As the slice reconstructions (Fig. 5bed) show, the pore morphology was kept constant across the sample height. The initial equiaxed and interconnected porous structure observed in Fig. 2 was retained. Moreover, the pore size slightly varies as compared to the initial pore size: the 2D projections show an average pore size of 96, 137, 150, 149, 165, and 157 mmfromthe bottom to the top specimen, respectively. The average thickness of the cell wall is determined as 48 mmatthesample bottom, and decreases continuously until it reaches 30 mmat the top. This value is overestimated since no cell wall below 20 mm in thickness was detected by this measurement technique, due to the threshold of detection being a voxel size of 7.56 mm. The reduction of the thinner cell walls (>20 mm) might counterbalance the overall shrinkage and expansion effect during this redox cycle. The Fe 2 O 3 thinner cell walls could be rapidly integrated into thicker walls at the reduction stage, via the sintering of Fe particles. These results show the minor effect of the dynamic redox processing on the doped Fe 2 O 3 foam microstructure, and the beneficial effects of the interconnected pore structure developed by the freeze-casting technique employed in this research, where no significant sample densification or pore clogging were produced. A distinguishing feature is observed in the sample, more evident in the half-top zone: the formation of a cell wall aperture that creates a new close porosity. These effects have previously been reported [29] on pure Fe samples, but are described here for the first time for doped Fe 2 O 3 foams. During reduction, the porosity of the cell walls improves the removal of oxygen, by improving the access to more oxidised material. At peak reduction temperature (700 C), a complete reduction to metal Fe is feasible and expected, the sintering of Fe particles would be undoubtedly promoted, and the cell walls become denser. In the oxidation stage, the constant formation of Fe x O phases produces a local volumetric expansion due to the crystal phase conversion and the diffusion of Fe atoms outwards, especially when the re-oxidation leads to the formation of an Fe 2 O 3 phase. To accommodate these new phases, they are separated and, eventually, exfoliated from the remaining Fe base. This morphology feature is more evident in the top part of the dynamic redox samples (Fig. 5ced). As pore size increases, Fe walls can act as flat surfaces since their curvature ratio also increases. Therefore, the local expansion of the new phases could be released normal to the surface and then be exfoliated from the Fe base. When pore size or curvature ratio decreases, the Fe walls might resemble a cylinder, and the same volume expansion could no longer be released normal to the surface, and the Fe x O grains tend to sinter into a more compact layer, thereby diminishing any opportunity for exfoliation. After 3 complete redox cycles under isothermal conditions, the doped Fe 2 O 3 bulk sample retains the overall interconnected pore structure, as shown in Fig. 6a. The 2D projections (Fig. 6bed) show average pore sizes of: 56 ±16, 76 ±21, 132 ±34, 138 ±36, 117 ±29, and 120 ±37 mm from the bottom to the top of the sample, respectively. The average thickness of the cell wall shows a higher value at the bottom (55 mm) and top (50 mm) parts, as compared with the centre (40 mm); as discussed previously, this value is overestimated due to the Fig. 5 e3D reconstructions from X-CT of the doped Fe 2 O 3 foams after dynamic redox processing: bulk sample (a) and slices of 250 mm in thickness at different sample heights: (b) 0.2 mm, (c) 2.3 mm and (d) 4.3 mm. The 2D projections are the bottom and top part of the slice reconstructions showing the pore-size evolution. journal of materials research and technology 2021;13:1887e1898 1893 measurement technique. In this case, a significant reduction in pore size was observed at the bottom part of the sample, where the initial average pore size passes from nearly 100 mm to 56 mm. As the redox cycles occurred, during the reduction step, the Fe walls thickened, and the pore size and overall porosity decreased due to the rapid sintering at 750 C. This situation is enhanced at the bottom of the sample, as can be seen in Fig. 6b. In contrast, the doped foam sample under the dynamic redox condition shows no cell wall aperture across the sample height, which suggests a decrease in the Fe outward mobility to form Fe x O due to the lower re-oxidation temperature (450 C), which reduces the possibility of vacancy formation in the cell-wall interiors. In isothermal conditions, the reduction step leads to the formation of Fe 3 O 4 preferentially, which decreases the volume expansion during the oxidation reaction and a possible exfoliation. Furthermore, the addition of Al 2 O 3 in pure Fe has shown a reduction in the pore formation due to the Kirkendall effect with the conversion of Fe to magnetite [46]. The vacancy diffusion in magnetite is four times lower than in hematite [48], and therefore the Fe outward mobility is considerably reduced. In each re-oxidation step, some metallic Fe remains unconverted, as shown in Fig. 4b. Therefore, as steam was usedfor re-oxidation at 450 C, Fe could be mainly convert to Fe 3 O 4 [40] and the remaining metal could be estimated in 9.3 wt% after the first cycle and 11.3 wt% for the remaining second and third cycles, which yield a 6.35 vol.% and 7.80 vol.%, respectively. Figure 7a shows a SEM micrograph of the doped Fe 2 O 3 foam surface after the isothermal redox condition. The faceted surface of the magnetite grains is interrupted by a spider-like Fe network; these Fe phases correspond to the unconverted metal. An XRD analysis (Fig. 7b) under both the as-fabricated and after redox process conditions is carried out to corroborate this assertion. This analysis detects only the hematite phase (Fe 2 O 3 , R-3c, ref. no. 1546383 in the COD) in the as-sintered specimen. In turn, for the after redox process two major phases were elucidated, magnetite (Fe 3 O 4 , Fd-3m, ref. no. 7228110 in COD) and ferrite (a-Fe, Im-3m, ref. no. 9016601 in the COD). Moreover, slight peaks of unreacted hematite were also detected in this specimen. All these observations therefore corroborate that the initial Fe 2 O 3 evolves into Fe 3 O 4 with some elemental Fe after the redox cycles. Furthermore, the position of the Fe 2 O 3 and Fe 3 O 4 peaks were not appreciably displaced, showing no influence of Al or Ce atoms. However, the spider-like Fe network created around the growing magnetite grains does indeed corroborate the influence of the Al atom clusters well-dispersed within the as-sintered sample. 3.4. XPS characterisation The XPS spectra shows the valence state and compositional studies of Fe, Al, and Ce elements in the different stages of the doped Fe 2 O 3 as-synthesised (before slurry preparation), asfabricated (before FC and sintering), and after redox processing (dynamic). Figure 8 shows the high-resolution photoelectron spectrum for Fe2p (a-c) and Al2p (d). The deconvoluted XPS spectra of the as-synthesised (Fig. 8a) and as-fabricated (Fig. 8b) doped Fe 2 O 3 powder shows similar convoluted XPS spectra, where peaks at 709.4, 723.1, and 717.3 eV were identified with the Fe 3þ ions in a-Fe 2 O 3 assigned to Fe 2p 3/2 ,Fe2p 1/ 2 , and its satellite peak, respectively. The two major peaks for Fe 2p 3/2 and Fe 2p 1/2 are separated by spin energy of 13.7 eV, and the associated satellite peak (717.3 eV) is located at þ7.9 eV of the Fe 2p 3/2 peak (709.4 eV), which are in good agreement with the reported value for a-Fe 2 O 3 in the literature Fig. 6 e3D reconstructions from X-CT of the doped Fe 2 O 3 foams after isothermal redox processing: bulk sample (a) and slices of 250 mm in thickness at different sample heights: (b) 0.2 mm, (c) 2.3 mm, and (d) 4.3 mm. The 2D projections are the bottom and top part of the slice reconstructions showing the pore-size evolution. journal of materials research and technology 2021;13:1887e18981894 [49,50]. The shake-up satellite (712.8 eV) is also confirmed in the literature for doped Fe 2 O 3 prepared by the co-precipitation method from metal nitrate precursors [51] and with the influence of Al 3þ [52]. In this work, the broadening of the peaks associated with Fe 3þ to lower binding energies could confirm the change in the oxidation state to Fe 2þ due to the influence of the Al 3þ ions being incorporated into the hematite-like (Fe,Al) 2 O 3 solid matrix. With the sintering process, the satellite peaks of Fe 3þ in the as-fabricated powder moves to a lower binding energy of 712.5 eV (Fig. 8b) due to the local formation of a near-hercynite (FeAl 2 O 4 ) phase on the surface of the particles [24]. After the dynamic redox process (Fig. 8c), the deconvoluted XPS spectrum shows three peaks at 709.5, 716.6, and 723.1 eV corresponding to a non-stoichiometric FeO phase and a peak at 712.1 eV that could be attributed to the aforementioned shake-up satellite. The difference in binding energy between Fe2p 3/2 (709.5 eV) and the satellite peak (716.6 eV) is 7.1 eV, which is 1.1 eV higher than those values reported for oxidation state Fe 2þ [50,53]. Nevertheless, the broadening of the Fe2p 3/2 peak might hinder Fe 3þ ions corresponding to the Fe 3 O 4 , since the binding energies are similar. In this regard, the XPS spectrum of the after redox processing samples shows a significant broadening of the major peak down to lower binding Fig. 7 eSEM micrograph (a) of the doped Fe 2 O 3 foams after isothermal redox processing, and XRD patterns (b) for the assintered and after redox processing specimens. Fig. 8 eHigh resolution XPS spectra of Fe2p (aec) for the doped Fe 2 O 3 samples as-synthesised (a), as-fabricated (b), and after redox processing (c) with the deconvoluted peaks, and Al2p (d) for all samples. journal of materials research and technology 2021;13:1887e1898 1895