Effect of atomic substitution on the sodium manganese ferrite thermochemical cycle for hydrogen production

Effect of atomic substitution on the sodium manganese ferrite thermochemical cycle for hydrogen production Francesco Torre a , Teresa Aguilar Sanchez a , Stefania Doppiu a , * , Mikel Oregui Bengoechea b , Pedro Luis Arias Ergueta b , Elena Palomo del Barrio a , c a Centre for Cooperative Research on Alternative Energies (CIC energiGUNE), Basque Research and Technology Alliance (BRTA), Alava Technology Park, Albert Einstein 48, 01510 Vitoria-Gasteiz, Spain b Universidad del Pais Vasco - Euskal Herriko Unibertsitatea, Escuela de Ingeniería, Plaza Torres Quevedo, 1, 48013 Bilbao, Spain c Ikerbasque, Basque Foundation for Science, Bilbao 348013, Spain article info Article history: Received 8 April 2022 Received in revised form 6 July 2022 Accepted 7 July 2022 Available online 25 August 2022 Keywords: Thermochemical water splitting Sodium manganese ferrite cycle Carbonation Decarbonation Fuel production Cation substitution abstract This work presents the effect of atomic substitution on the MnFe 2 O 4 eNa 2 CO 3 thermochemical cycle for H 2 production. The non-oxidative decarbonation/carbonation reaction of the MnFe 2 O 4 eNa 2 CO 3 mixture is investigated as the starting reference. Repeated cycling results in a 30% loss of reversibility due to an overall reduction of the reactive interfaces. The substitution of Na 2 CO 3 for Li 2 CO 3 decreases the decarbonation onset temperature by about 100  C, but almost no reversibility is observed during the cycles due to the irreversible Li þ intercalation. The effect of partial Mn substitution for Ca, Ni, and Zn is presented. The 5% Zn mixture shows the best decarbonation/carbonation reversibility and is tested for H 2 production together with MnFe 2 O 4 eNa 2 CO 3 . The reference mixture produces more H 2 during the first cycle (z1.1 vs. 0.7 mmol/g), but its production drastically drops by two orders of magnitude upon cycling and becomes negligible after 5 cycles. By contrast, the Zn-doped mixture exhibits a stable H 2 production of 0.22 mmol/g with no decreasing trend observed from cycle 2 to cycle 5. As result, in the fifth cycle, the Zn-doped mixture produces 23 times more H 2 than MnFe 2 O 4 eNa 2 CO 3 . Thermogravimetry and X-ray diffraction confirm that doping with Zn significantly improves the regeneration of the reactants. ©2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Thermochemical water splitting (WS) represents a very promising and elegant solution for sustainable large-scale hydrogen production [1e3]. This technology uses water and heat as input, to produce H 2 and O 2 . Although direct water thermolysis can be achieved in principle, the extreme temperatures (T >2200  C) and the necessity to separate H 2 from O 2 make this solution impractical [4]. Improvements can be achieved by dividing the overall process into consecutive reactions (minimum two) that are cyclically repeated through a so-called thermochemical cycle [5]. This allows to lower the operating temperatures and produce H 2 and O 2 in separate steps, reducing the risk of accidents. Among more than 300 proposed so far, two-step thermochemical cycles based on metal oxides (e.g. SnO 2 , ZnO, CeO 2 , Mn 3 O 4 ,Fe x O y ) rely on relatively low-cost materials and do not involve dangerous or corrosive reaction intermediates [6e12]. However, the regeneration (reduction) step typically requires temperatures >1500  C, which negatively affects the long-term materials performances and poses significant challenges when choosing the materials for the thermochemical reactors [13]. The addition of secondary compounds has been proposed to lower the reaction temperatures below 1000  C. In this regard, a recent assessment considering process economics, environmental impact, cyclability, and simplicity of operation highlighted the MnFe 2 O 4 eNa 2 CO 3 cycle as one of the best for practical applications [14]. This cycle was initially proposed by Y. Tamamura et al. [15,16] and then further investigated by C. Alvani et al. [17e25]. Beyond lab-scale experiments, the feasibility of this thermochemical cycle has also been proven in a small solar concentration facility [25]. In first approximation, this thermochemical cycle consists of two steps [15] and can work at temperatures around 800e750  C: *Corresponding author. E-mail address: [email protected] (S. Doppiu). Contents lists available at ScienceDirect Materials Today Energy journal homepage: www.journals.elsevier.com/materials-today-energy/ https://doi.org/10.1016/j.mtener.2022.101094 2468-6069/©2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Materials Today Energy 29 (2022) 101094 2MnFe2O4ðsÞþ3Na2CO3ðsÞþH2OðgÞ/6NaFe 2 = 3 Mn 1 = 3 O2ðsÞ þH2ðgÞþ3CO2ðgÞ (1) 6NaFe 2 = 3 Mn 1 = 3 O2ðsÞþ3CO2ðgÞ/2MnFe2O4ðsÞþ3Na2CO3ðsÞ þ0:5O2ðgÞ (2) However, a more complex reaction path was experimentally observed, as it was clearly described by Varsano et al. [23]. In particular, the hydrogen production reaction (1) was shown to proceed through two separate steps. First, the non-oxidative partial decarbonation takes place, where 2/3 of Na 2 CO 3 reacts with MnFe 2 O 4 . As a result of this reaction, CO 2 is released, and sodium intercalates in the cubic spinel phase to form a fine mixture of NaFeO 2 and MnO: 2MnFe2O4ðsÞþ3Na2CO3ðsÞ/2ðMnO*2NaFeO2ÞðsÞþNa2CO3ðsÞ þ2CO2ðgÞ (3) The formation of the (MnO*NaFeO 2 ) compound paves the way to the subsequent reaction with H 2 O steam to produce H 2 (4). Mn 2þ is oxidized to Mn 3þ to form NaFe 2/3 Mn 2/3 O 2 , which has the same crystal structure as NaFeO 2 . At this point, more Na þ can be intercalated, which drives the decomposition of the remaining Na 2 CO 3 . 2ðMgO*2NaFeO2ÞðsÞþNa2CO3ðsÞþH2OðgÞ/6NaMn 1 3 Fe 2 3  O2ðsÞþ2CO2ðgÞþH2ðgÞ(4) Once the WS reaction is completed, NaFe 2/3 Mn 2/3 O 2 is exposed to a CO 2 -rich gas, which promotes the sodium oxide deintercalation to form Na 2 CO 3 . As result, the Na content in the NaFe 2/3 Mn 2/3 O 2 progressively decreases and the layered structure collapses. The reduction of Mn 3þ to Mn 2þ and the corresponding release of molecular oxygen eventually leads to the regeneration of MnFe 2 O 4 and Na 2 CO 3 : 6NaFe 2 = 3 Mn 1 = 3 O2ðsÞþ3CO2ðgÞ/2MnFe2O4ðsÞþ3Na2CO3ðsÞ þ1 2O2ðgÞ (5) Unfortunately, particle sintering and coalescence of a Na 2 CO 3 layer tend to hinder the complete regeneration of the starting reactants, leading to a drastic decrease in the H 2 production upon prolonged cycling [24]. In general, the improvement of the reaction kinetics and reversibility, as well as the decrease of the operating temperatures, is highly important to increase the efficiency of thermochemical cycles, as well as to broaden their potential fields of application [26]. Speed up the reaction kinetics means increasing the H 2 production per unit of time, while long-term stability of the materials is fundamental for the overall process to be feasible and competitive. In this sense, lowering the operating temperature reduces the sintering of the used oxides and reduces the requirements for the materials needed to construct thermochemical reactors. Atomic substitution, reduction of the crystallite size, and addition of inert compounds are common strategies to improve materials performances [7,27e35]. For instance, atomic substitution significantly improved the performances of CeO 2 - and perovskitesbased thermochemical cycles for fuel production [36e41]. A straightforward relation between cycle performances and the microstructural changes induced by atomic substitution was reported in some cases. The regeneration temperature of some doped perovskites decreased when introducing doping elements with lower atomic size. Such behavior was attributed to an increase in the atomic size mismatch that, in turn, led to a higher local disorder [38]. Concerning ferrite-carbonate cycles, investigations performed on single metal oxides (Mn 3 O 4 ,Fe 3 O 4 , and Co 3 O 4 ) and carbonates (Na 2 CO 3 ,Li 2 CO 3 , and K 2 CO 3 ) showed that the H 2 evolution rate was found to vary depending on the specific cation combinations [42]. To the best of our knowledge, similar approaches have not yet been reported for the MnFe 2 O 4 eNa 2 CO 3 thermochemical cycle. Thus, the present work represents the first attempt in this direction. Particular attention is given to the non-oxidative decarbonation reaction (reaction 3) as it represents the first step of the overall thermochemical cycle. The decarbonation-carbonation reaction of MnFe 2 O 4 eNa 2 CO 3 is followed during 10 consecutive cycles, and the related structural changes are investigated. The effect of atomic substitution is then explored by using MnFe 2 O 4 eNa 2 CO 3 as the basis for comparison. Starting from the MnFe 2 O 4 eNa 2 CO 3 system, the effect of atomic substitution on its reactivity is faced through two distinct approaches. Evaluating the substitution of Na for Li and K cations and partially substituting Mn 2þ for Ca 2þ ,Ni 2þ , and Zn 2þ . In the latter case, ferrites with general formula Mn 1x A x Fe 2 O 4 , with A being Ni, Zn, Ca, and x ¼0.05, 0.10, and 0.50 are synthesized and tested through decarbonation/carbonation cycles together with Na 2 CO 3 . The most promising doped mixture is then tested for the H 2 production reaction, and its performances are compared with those of the reference MnFe 2 O 4 eNa 2 CO 3 system. 2. Material and methods 2.1. Synthesis of Mn ferrite oxides Manganese ferrite-based oxides, with spinel structure, were synthesized following the self-combustion method; glycine (C₂H₅NO₂, 98.5%, Sigma Aldrich) was used as complexant and fuel agent [43]. An aqueous solution of Fe(NO 3 ) 3 $9H 2 O (99%, Alfa Aesar) and Mn(NO 3 ) 2 $4H 2 O (99%, Alfa Aesar) was prepared; a molar ratio of 2:1 between Fe 3þ and Mn 2þ and a total cation concentration of 1 M were ensured. A 1 M solution of glycine was then added and the final solution was stirred for 2 h. The solution was transferred into a Petri-type dish and kept at 100  C until a red-brownish gel was formed. Finally, the self-combustion reaction was carried out at 350  C, where a black-brownish, foamy product was obtained. The same methodology was used to synthesize modified manganese ferrites with general formula Mn 1-x A x Fe 2 O 4 , with A being Ni, Zn, Ca, and x ¼0.05, 0.1, and 0.5. To this end, part of the Mn 2þ precursor was substituted by the nitrate of the corresponding divalent dopant cationdi.e. Zn(NO 3 ) 2 $6H 2 O (98%, Alfa Aesar), Ni(NO 3 ) 2 $6H 2 O (EssentQ®, Scharlau), and Ca(NO 3 ) 2 $4H 2 O(99%,SigmaAldrich).On the other side, the Fe 3þ concentration was not changed. In this way, the ratio between the divalent and trivalent cations was maintained constant at 1:2. The atomic distribution of divalent and trivalent cationsin the octahedral and tetrahedral sites of the spinel lattice can change depending on the specific chemical composition, the synthesis method, and the microstructural features [44,45]. The determination of the exact atomic distribution is not straightforward and willnotbeaddressedinthiswork.Infact,theMnsubstitutionwerefer to should be intended only in terms of chemical compositiondi.e. the partial replacement of Mn 2þ with Ca 2þ ,Ni 2þ ,andZn 2þ . For the sake of simplicity, the different oxides will be indicated by referring to the percentage of Mn atomic substitution; for F. Torre, T.A. Sanchez, S. Doppiu et al. Materials Today Energy 29 (2022) 101094 2 instance, Mn 0.5 Ca 0.5 Fe 2 O 4 will be referred to as 50 at.% Ca, Mn 0.95 Ni 0.05 Fe 2 O 4 as 5 at.% Ni, etc. (see Table 1). 2.2. Thermal analysis 2.2.1. Non-oxidative decarbonation reaction of the MnFe 2 O 4 eNa 2 CO 3 mixture Temperature programmed desorption (TPD) experiments were performed to investigate the decarbonation/carbonation behavior of the different mixtures of spinel ferrites and carbonates. For each experiment, the selected spinel ferrite and carbonate were mixed at a 1:1 M ratio using a mortar. The choice of this stoichiometry can be understood by looking at the non-oxidative decarbonation step described by Eq. (3). In non-oxidizing conditions (absence of H 2 O), only 2/3 of the Na 2 CO 3 reacts with MnFe 2 O 4 , while the remaining 1/ 3ofNa 2 CO 3 does not participate in the reaction. This corresponds to a MnFe 2 O 4 :Na 2 CO 3 molar ratio of 1:1. About 15e20 mg of the mixture were transferred to an alumina crucible and located in a NETZSCH STA 449 F3 thermobalance. Preliminary isothermal experiments were performed on the MnFe 2 O 4 eNa 2 CO 3 mixture at 600, 700, 750, 800, and 850  C. The samples were heated under N 2 at 10  C/min to reach the desired temperature and kept in isothermal conditions until the reaction was completed (the duration of the isothermal step varies depending on the temperature). During these experiments, the CO 2 release was analyzed by using a mass spectrometer (NETZSCH QMS 403C). The reversibility of the non-oxidative decarbonation-carbonation was then investigated through decarbonation-carbonation cycles. The temperature was increased from room temperature to 800  C at a constant heating rate of 10  C$min 1 under N 2 and kept constant for 30 min in N 2 to induce the complete release of the CO 2 . The carbonation process was carried out under CO 2 atmosphere with a flow of 50 mL$min 1 decreasing the temperature to 400  C at a constant cooling rate of 10  C$min 1 . Cycling experiments were also performed on the MnFe 2 O 4 eNa 2 CO 3 1:1 mixture. To this aim, the temperature was again increased to 800  C, and the temperature program was repeated depending on the desired number of cycles. A maximum of 10 cycles was performed. The same methodology was subsequently used to test the effect of Mn atomic substitution. The synthesized oxides with 5 and 10 at.% of Ca, Ni, and Zn (see Section 2.1) were mixed with Na 2 CO 3 at a 1:1 M ratio. The oxides with 50 at.% of Mn substitution were excluded due to the presence of secondary phases (see structural characterization in the Supporting information). The substitution of Na 2 CO 3 for Li 2 CO 3 and K 2 CO 3 on reaction 3 was also taken into consideration. The decarbonation/carbonation of equimolar MnFe 2 O 4 eLi 2 CO 3 and MnFe 2 O 4 eK 2 CO 3 mixtures was investigated by cycling experiments. Heating was performed at 10  C/min under N 2 (50 mL$min 1 ); different temperatures were tested with a maximum of 925  C. After a short isotherm of 10 min, the sample was cooled down to 400  Cat10  C/min under 50 mL/ min of CO 2 . The program was then repeated two more times to get three decarbonation/carbonation cycles. 2.2.2. Hydrogen production H 2 production experiments were performed using an STA 449 F3 Jupiter (NETZSCH) thermobalance coupled with a water vapor generator provided byaDROP GmbH. The outlet gas of the STA oven was connected to an H 2 Clark-type microsensor (detection limit z10 2 vol%) interfaced with an amplifier (Unisense, Denmark). The sensor was calibrated before each measurement by using a two-point calibration as recommended by the provider. To this aim, pure Ar and an Ar with 2 vol% H 2 standards were used. For each experiment, the desired spinel was mixed with Na 2 CO 3 at a 2:3 ratio (see reaction 1) using a mortar and then pressed at 1 ton for 30 s to obtain a 400 mg pellet (d ¼10 mm and h z2 mm). Despite the use of powders would have ensured higher gas-solid interfaces, the use of pellets was necessary to work with a sample size large enough to allow H 2 quantification. The H 2 production experiments were performed as follows. The temperature was increased to 800  C(10  C$min 1 ) and kept constant for 30 min to release 2/3 of the total CO 2 (reaction 3); only Ar flowed during this first part (30 mL/min). Water vapor was then introduced for 90 min at a rate of 1 g H 2 O/h while keeping the Ar flow at 30 mL/min. The water supply was then stopped, and the Ar flow was maintained for 10 min to remove the remaining H 2 O. CO 2 was then introduced (60 mL/min) and the sample cooled at 10  C/min. When the temperature reached 400  C, the gas was then shifted to Ar and the temperature increased again to 800  C to start a new cycle. A total of 10 cycles were performed. Additional experiments were performed in isothermal conditions at 750  C and using an H 2 microsensor with a lower detection limit (z10 3 vol%; Unisense, Denmark). The samples were heated at 10  C/min to 750  C under a 100 mL N 2 flow. Water vapor was then introduced into the gas flow at a rate of 0.5 g H 2 O/h for 5 h. When the water supply was stopped, the N 2 flow was maintained for 5 min to remove the remaining H 2 O. CO 2 was then introduced (100 mL/min) and the sample was left at 750  C overnight. The gas flow was then shifted again to start another WS step. A total of 5 cycles were performed. As the H 2 production was observed to be faster from the second cycle, the 4 remaining WS steps were shortened from 5 to 3 h. 2.3. Structural characterization The as-synthetized ferrites and the studied mixtures were characterized by X-ray diffraction (XRD) analysis using a Bruker D8 Discover equipped with a LYNXEYE XE detector and a monochromatic Cu K a 1 radiation source of l ¼1.54056 Å. For some mixtures, a sample holder equipped with a Kapton film was used to avoid oxidation and/or hydration during the XRD pattern acquisition. For instance, MnO is very prone to oxidation to Mn 3þ species and dry K 2 CO 3 rapidly forms sesquihydrate K 2 CO 3 . The use of Kapton results in a strong background at low scattering angles (10 >2 q <30). In situ XRD measurements were performed at different temperatures using a Bruker D8 Advance diffractometer equipped with a LYNXEYE detector, a Vantec-1 PSD detector, a Co tube (K a 1of l ¼1.78896; K b 1¼1.6210334), and an Anton Parr HTK2000 high-temperature furnace. Patterns were recorded under N 2 atmosphere at room temperature and 400, 450, 500, 550, 600, 700, and 800  C in a stepwise manner. The decarbonation reaction was analyzed as follows. After a stabilization time of 30 min at 800  C, CO 2 was then introduced, and the temperature was then maintained at 800  C for 1 h to guarantee a complete saturation of the chamber. Patterns were then collected at 800, 600  C, and room temperature in a stepwise manner with a cooling rate of 10  C/min after each step. Phase identification was performed using the EVA Table 1 Compositions of the different synthesized oxides together with the abbreviation used in the text. Composition Abbreviation Mn 0.95 Ni 0.05 Fe 2 O 4 5% Ni Mn 0.9 Ni 0.1 Fe 2 O 4 10% Ni Mn 0.5 Ni 0.5 Fe 2 O 4 50% Ni Mn 0.95 Zn 0.05 Fe 2 O 4 5% Zn Mn 0.9 Zn 0.1 Fe 2 O 4 10% Zn Mn 0.5 Zn 0.5 Fe 2 O 4 50% Zn Mn 0.95 Ca 0.05 Fe 2 O 4 5% Ca Mn 0.9 Ca 0.1 Fe 2 O 4 10% Ca Mn 0.5 Ca 0.5 Fe 2 O 4 50% Ca F. Torre, T.A. Sanchez, S. Doppiu et al. Materials Today Energy 29 (2022) 101094 3 diffraction commercial software. The collected patterns were then analyzed according to the Rietveld method [46] by using the MAUD software [47]. The morphology of the powders before and after carbonation-decarbonation cycles was analyzed using scanning electron microscopy (SEM) using a Quanta 200 FEG (FEI Company, Hillsboro, OR, USA) operating in high vacuum mode. 3. Results and discussion 3.1. Non-oxidative decarbonation reaction of the MnFe 2 O 4 eNa 2 CO 3 mixture The morphology of the as-synthesized oxides was investigated using SEM. Representative images of the MnFe 2 O 4 are reported in Fig. 1aeb. The obtained powders show a hierarchical network with well-connected macropores due to the consistent amount of gases that is rapidly released during the reaction [48]. The synthesized ferrites were characterized using XRD, and the phase composition and the microstructure were investigated according to the Rietveld method (Section 1 of Supporting information). The obtained powders present a fine microstructure with the average crystallite sizes in the range of 50e100 nm, which is due to the fast cooling experienced during the combustion synthesis. In all cases, the main crystallographic phase is the cubic spinel characteristic of the MnFe 2 O 4 compound, followed by minor amounts of FeO and MnO (Fig. S1a). Minor amounts of other secondary phases were detected in the samples with 50% of atomic substitution. The atomic substitution also affected the lattice parameter of the cubic spinel phase (Fig. S1a). Different trends were observed depending on the atomic mismatch between the doping element and Mn. While Zn did not induce any significant changes, Ca and Ni led to lattice expansion and shrinkage, respectively. The non-oxidative decarbonation reaction (3) of the MnFe 2 O 4 eNa 2 CO 3 system was investigated under isothermal conditions in the 600e850  C temperature range. The TPD profiles obtained for the 1:1 mixture at increasing temperatures are reported in Fig. 1c. For all the experiments, a small mass decrease (z0.5 wt%) is detected at temperatures around 100  C because of the desorption of absorbed water. The decomposition of the mixture starts at approximately 550  C and proceeds at different rates depending on the temperature isotherm. As expected from reaction 3, mass spectrometr (MS) analysis of the outlet gases confirms that CO 2 was the only product. The intensity of the characteristic peak of the CO 2 þ ion (44 m/z) is reported in Fig.1dasa function of time. For the sample heated at 600  C, the decarbonation reaction proceeds very slowly and only a small mass change (z2 wt%) is detected after more than 300 min. No significant changes can be observed in the MS signal apart from a small peak between 25 and 30 min. As the released CO 2 is diluted in the N 2 carrier gas, it is possible that the resulting concentration was too low to be detected. The reaction proceeds faster at 700 and 750  C. A significant mass loss is observed, together with an increase of the MS signal that reaches a maximum after 33 and 36 min, respectively. Then, the CO 2 evolution continues at a lower rate and the slope of the mass change profile progressively decreases. After 180 min at 700  C, a mass loss of about 10 wt% is observed, meaning that the full desorption could not be achieved within a reasonable time at this temperature. On the other hand, a plateau is reached Fig. 1. (a), (b) SEM images of the as-synthesized MnFe 2 O 4 at different magnifications. (c) TPD profiles of the MnFe 2 O 4 eNa 2 CO 3 mixture (1:1) in the 600e850 C temperature range. (d) MS signal of the CO 2 þ ion (m/z ¼44) detected during the isothermal experiments. (e) In situ XRD patterns of the MnFe 2 O 4 eNa 2 CO 3 mixture (1:1) upon heating from RT to 800 C under N 2 . SEM, scanning electron microscopy; TPD, temperature programmed desorption; XRD, X-ray diffraction. F. Torre, T.A. Sanchez, S. Doppiu et al. Materials Today Energy 29 (2022) 101094 4 after about 60 min at 750  C. The overall mass loss is around 12 wt%, which according to reaction 3 is close to the theoretical one (13 wt %). The mixtures heated at 800 and 850  C follow the previous trend. The maximum CO 2 concentration is revealed after 38 and 40 min, while desorption ends after 85 and 60 min, respectively. In both cases, the mass stabilizes at about 87.5 wt%, thus suggesting that the reaction almost goes to completion. XRD analysis of the MnFe 2 O 4 eNa 2 CO 3 mixture after the TPD at 850  C is in line with the previous observation (Fig. S2a). About 70 wt% and 25 wt% were obtained for NaFeO 2 and MnO, respectively, together with a 5 wt% of unreacted spinel. This agrees with the theoretical values calculated based on reaction 3 (75.5 wt% and 24.5 wt% for NaFeO 2 and MnO, respectively) proposed by Varsano et al. [23]. However, the same authors did not observe the formation of MnO due to the highly dispersed and disordered nature of the manganese compound. To obtain further insight, XRD patterns were acquired in situ upon heating the MnFe 2 O 4 eNa 2 CO 3 under an N 2 atmosphere Fig. 1e. A significant increase in the lattice parameter of the spinel phase can be appreciated due to thermal expansion (Fig. S2b). In addition, the transition of g -Na 2 CO 3 to a -Na 2 CO 3 takes place at 400  C according to what was previously reported [23]. At 600  C, the NaFeO 2 starts forming as highlighted by the peaks at 2 q values of 19 and 47, while MnO is not detected probably due to its low amount. At 700  C, the amount of NaFeO 2 increases while the intensity of the Bragg peaks of MnFe 2 O 4 significantly reduces. Moreover, a small peak of the MnO can be appreciable at 2 q ¼47. Finally, the pattern collected at 800  C confirms that reaction 3 went to completion, as only NaFeO 2 and MgO are detected. The cyclability of the MnFe 2 O 4 eNa 2 CO 3 mixture was investigated through 10 consecutive decarbonation/carbonation cycles. An equilibration cycle was initially performed to reduce any difference among the various batches of synthesized spinel. The sample was initially heated up to 800  C, and this temperature was kept constant for 2 h. Cooling to 400  C was then carried out under CO 2 . This cycle will be referred to as cycle 0. Ten additional cycles were then performed in the 400e800  C temperature range. The mass profile and the mass changes measured during the decarbonation and carbonation steps are reported in Fig. 2a and b, respectively. During cycle 0, almost full desorption is achieved after about 30 min at 800  C, with an overall mass loss of around 11.5 wt %(Fig. 2b). A 10 wt% mass increase is observed during the subsequent carbonation step, thus marking a reversibility loss of about 1.5 wt%. A small but constant loss of reversibility is observed from cycle 1 to cycle 4, while the mass change stabilizes at a value around 8 wt% of the initial mass during the following cycles. SEM analysis was carried out after the third desorption step and after three full cycles, i.e., after 455 and 485 min (Fig. 2ced and Fig. 2eef, respectively). Partial sintering occurred due to the prolonged exposure to high temperatures, and the sample showed grains ranging from 0.5 to 4 m m with an average value of z1.3 m m(Fig. 2c). Despite this, the desorbed sample still presents a significant porosity. A closer look unravels the presence of a laminar structure that can be attributed to the NaFeO 2 phase (Fig. 2d) [24]. The sample after the carbonation step shows a more irregular morphology (Fig. 2eef). The backscattered electron detector highlights the formation of Na 2 CO 3 (brighter areas) that results in high contrast with the underlying MnFe 2 O 4 /NaFeO 2 matrix. XRD analysis was performed after 10 cycles without showing any parasite phases (Fig. S2f). However, the quantification of the crystallographic phases suggests that after 10 cycles almost half of the initial spinel phase do not participate in the reaction, and rather acts as an inert phase. Based on these results, the loss of reversibility can be explained as follows (see Fig. 2g). As the carbonation reaction takes place and the Na ion are deintercalated from NaFeO 2 , the regenerated Na 2 CO 3 tends to coalesce and covers the newly formed MnFe 2 O 4 . The specific interfaces between MnFe 2 O 4 and Na 2 CO 3 and CO 2 are then reduced, and the subsequent decarbonation reaction is negatively affected. In this sense, phenomena like partial sintering, loss of porosity, and grain growth may further boost this behavior, which explains the loss of reversibility highlighted by thermogravimetry. 3.2. Effect of different alkali carbonates In this section, we present the effect of the substitution of Na for K and Li on the kinetic and reversibility of the decarbonationcarbonation reaction. A total of three decarbonation-carbonation cycles were sufficient in this case to highlight the difference between the three compositions. The mass profiles of the MnFe 2 O 4 eK 2 CO 3 and MnFe 2 O 4 eLi 2 CO 3 mixtures (1:1) are compared with the MnFe 2 O 4 eNa 2 CO 3 reference system (Fig. 3aeb). The XRD patterns acquired for the three mixtures before and after the three cycles are reported in Fig. 3c. As the behavior of the MnFe 2 O 4 eNa 2 CO 3 mixture has been extensively discussed in the previous section, it will not be addressed in the present one and will be only used as the term of comparison to describe the effect of Na substitution for Li and K. The MnFe 2 O 4 eK 2 CO 3 mixture shows a similar onset temperature to the MnFe 2 O 4 eNa 2 CO 3 one, but the overall decarbonation kinetic is slower than the reference system (Fig. 3a). Duringthe first desorption step, the mixture with K 2 CO 3 shows a 10.8% mass loss, which is slightly lower than the theoretical one (11.9 wt%). As CO 2 is introduced, the carbonation reaction immediately starts and 96.5 wt% of the initial mass is achieved in less than 1 min. A decrease in performance is observed during the following two cycles, and the desorption kinetics of MnFe 2 O 4 eK 2 CO 3 further slows down compared to MnFe 2 O 4 eNa 2 CO 3 . Moreover, the system loses reversibility upon cycling and desorbs 9.38% and 8.73 wt% of its initial mass during the second and third cycle, respectively. The XRD analysis of the MnFe 2 O 4 eK 2 CO 3 mixture after three cycles confirms the partial regeneration of the starting reactants, meaning MnFe 2 O 4 and K 2 CO 3 sesquihydrate (COD 2200592). The latter was detected also in the starting mixture and is due to the highly hygroscopic nature of anhydrous K 2 CO 3 , which causes its rapid hydration during the XRD sample preparation. Moreover, the cycled MnFe 2 O 4 eK 2 CO 3 mixture shows a significant amount of potassium beta ferrite, K 2 Fe 10 O 16 (COD 1529668), which explains the loss of reversibility observed upon cycling. The formation of KFeO 2 would be expected if the MnFe 2 O 4 eK 2 CO 3 followed reaction 3. However, no KFeO 2 could be detected, which suggests a different reaction mechanism compared to the MnFe 2 O 4 eNa 2 CO 3 system. Itisworthsayingthatthepotassium beta ferrite shows a lamellar structure that allowsfora broad range of non-stoichiometric compositions corresponding to different Fe to K ratios [49]. The compensation for non-stoichiometry can be achieved by the inclusion of bivalent atoms, which in this case may be Mn 2þ [50]. For these reasons, the rationalization of these results to find a reaction mechanism explaining the loss of performances observed in the MnFe 2 O 4 eK 2 CO 3 mixture is not straightforward. The MnFe 2 O 4 eLi 2 CO 3 system demonstrates a different behavior. During the first heating step, the system shows impressive performances in terms of kinetics and shows an onset decarbonation temperature of about 100  C lower than the MnFe 2 O 4 eNa 2 CO 3 mixture (Fig. 3a). It starts releasing CO 2 at around 420  C and almost completes the decarbonation when the MnFe 2 O 4 eNa 2 CO 3 and MnFe 2 O 4 eK 2 CO 3 mixtures start decarbonizing. The observed overall mass loss is about 11.37 wt%, which is 78% of the 14.4 wt% theoretical value. However, the subsequent exposure to CO 2 does not induce any significant mass increase and the mass of the sample remains almost unvaried during the subsequent two cycles (Fig. 3b). Such poor reversibility of the MnFe 2 O 4 eLi 2 CO 3 system is F. Torre, T.A. Sanchez, S. Doppiu et al. Materials Today Energy 29 (2022) 101094 5 confirmed by the XRD analysis (Fig. 3c), which highlights the presence of MnFe 2 O 4 and cubic LiFeO 2 (COD 1541312). Thus, it seems that the reaction follows the mechanism of reaction 3, with LiFeO 2 being formed instead of NaFeO 2 . However, the subsequent deintercalation of Li ions seems to be hindered probably due to the strong LieO bond. As result, reaction 3 is irreversible at the present experimental conditions for the MnFe 2 O 4 eLi 2 CO 3 system. The XRD quantitative analysis indicates 49 wt% and 51 wt% for MnFe 2 O 4 and LiFeO 2 , respectively. Considering the molar mass of the two compounds, this means that the molar ratio between MnFe 2 O 4 and LiFeO 2 is around 1:2.5. If the stoichiometry of reaction 3 is valid also for the MnFe 2 O 4 eLi 2 CO 3 system, this means that only about 55% of the initial mixture reacted. This would imply a mass loss of only 7.9 wt%, which is lower than the experimentally observed and suggests that some cross-reaction may have taken place. Moreover, the spinel phase in the MnFe 2 O 4 eLi 2 CO 3 cycled mixture has an average lattice parameter of 8.38 Å, which is significantly lower than the one observed for the MnFe 2 O 4 eNa 2 CO 3 cycled in the same experimental conditions (8.53 Å). This has at least two possible explanations. First, Fe 3 O 4 is formed rather than MnFe 2 O 4 ; indeed, the two compounds share the same crystal structure with the only difference being the larger lattice parameter of MnFe 2 O 4 as a consequence of Fe 2þ substitutions for the bigger Mn 2þ . However, the formation of Fe 3 O 4 would imply the reduction of part of Fe 3þ to Fe 2þ at the expenses of Mn 2þ that should be oxidized to form NaMnO 2 or other secondary compounds, which were not detected. Another possibility is that the MnFe 2 O 4 lattice has shrunk due to a partial Li þ substitution or intercalation, as the atomic substitution for low electronegative cations was reported to induce a decrease in spinel lattice parameters [51e53]. Resuming, the MnFe 2 O 4 eNa 2 CO 3 mixture showed better performances than both MnFe 2 O 4 eLi 2 CO 3 and MnFe 2 O 4 eK 2 CO 3 . Indeed, the former showed almost negligible reversibility caused by the irreversible formation of LiFeO 2 , while the latter showed worse kinetics and reversibility due to the formation of potassium beta ferrite. For these reasons, the MnFe 2 O 4 eNa 2 CO 3 was kept as Fig. 2. (a) Mass profiles of the MnFe 2 O 4 eNa 2 CO 3 mixture during 10 decarbonation-carbonation cycles performed between 800 C and 400 C. (b) Mass change related to decarbonation and carbonation steps during the cycles. (cef) SEM images of the MnFe 2 O 4 eNa 2 CO 3 mixture after cycle 3. The images were acquired after the third decarbonation (ced) and (eef) after the third carbonation steps, respectively. (g) Schematic representation of the sintering and phase coalescence processes that lead to the decrease of reversibility during the first cycles. SEM, scanning electron microscopy. F. Torre, T.A. Sanchez, S. Doppiu et al. Materials Today Energy 29 (2022) 101094 6 the starting reference system to investigate the atomic-substituted spinel ferrites that are presented in the following section. 3.3. Effect of Mn substitution In this section, the effect of partial substitution of Mn for Zn, Ni, and Ca on the decarbonation-carbonation reaction is presented. Samples at 5 and 10 at.% of Zn Ni, and Ca were cycled 10 times. Data are reported in Fig. 4 by taking the MnFe 2 O 4 eNa 2 CO 3 mixtures as the reference system. In the first cycle, the three 5 at.%-doped samples show slightly better decarbonation kinetics than the undoped system, which can be appreciated by the small left shift of the mass profiles in the time scale (Fig. S3a). However, such difference is no further appreciated in the fifth and ninth cycles (Fig. S3b and c), as the four decarbonation profiles almost overlap. Same thing for the carbonation kinetics; the small differences between the four samples that are observed during the first cycle, progressively vanish as the number of cycles increases. Major differences can be appreciated concerning the amount of desorbed CO 2 , as can be appreciated in Fig. 4a. In the first cycle, the undoped MnFe 2 O 4 shows the highest mass loss (9.76 wt%), followed by the 5% Ni, 5% Zn, and 5% Ca ferrites, which exhibit a mass loss of 8.8, 8.15, and 6.8 wt%, respectively. However, the undoped MnFe 2 O 4 progressivelyloses capacity while the three doped ferrites seem to suffer less reversibility loss, with the Zn-doped sample even showing a slight increase in performance. After 10 cycles, the undoped MnFe 2 O 4 and the 5% Ni samples show a similar behavior, as they lose 8.14 wt% and 8.2 wt%, respectively. The Ca-doped sample desorbs only 6.5 wt%, while the Zn-doped leads to the highest mass loss (8.6 wt%). Fig. 3. Mass profiles of the MnFe 2 O 4 eNa 2 CO 3 , MnFe 2 O 4 eLi 2 CO 3 , and MnFe 2 O 4 eK 2 CO 3 mixtures during (a) the first decarbonation step and (b) during 3 decarbonation-carbonation cycles. (c) XRD patterns collected for the three mixtures before and after 3 cycles. XRD, X-ray diffraction. Fig. 4. The effect of 5 and 10 at.% Ca, Ni, and Zn atomic substitution on the MnFe 2 O 4 eNa 2 CO 3 mixture during 10 decarbonation-carbonation cycles. (a) The measured CO 2 release during 10 cycles for the 5 at.% Ca, Ni, and Zn-doped ferrites, and (b) for the 10 at.% Ca, Ni, and Zn-doped ferrites. (c) XRD patterns of different Mn 1-x A x Fe 2 O 4 eNa 2 CO 3 mixtures after 10 decarbonation-carbonation cycles. XRD, X-ray diffraction. F. Torre, T.A. Sanchez, S. Doppiu et al. Materials Today Energy 29 (2022) 101094 7 Data of the mixtures with 10 at.% of Zn, Ni, and Ca are reported in Fig. 4b and Fig. S3d. Overall, the increase in atomic substitution from 5% to 10% does not lead to any significant change. After 9 cycles, the three substituted ferrites exhibit a slight worsening in the mass loss rate in comparison to pure MnFe 2 O 4 (Fig. S3d), while the carbonation reaction proceeds faster. Among the doped ferrites, no significant differences can be appreciated in terms of kinetic performances. On the other hand, the dopant element affects the reversibility as the 10% Zn and 10% Ni ferrites show a similar weight loss during all the cycles, while the substitution for Ca results in a lower CO 2 release. All the samples show a loss of reversibility that is more evident during the first cycles. After 10 cycles, pure MnFe 2 O 4 shows the highest desorption, followed by 10% Zn, 10% Ni, and 10% Ca samples that lose 7.9 wt%, 7.8 wt%, and 7 wt%, respectively. According to the XRD analysis, the partial Mn substitution for Ni and Zn did not induce significant changes in the phase composition of the mixtures (Fig. 4c). No appreciable amounts of secondary phases were detected suggesting that the initial fine dispersion of the dopants was retained even upon prolonged cycling. On the other hand, the XRD patterns of both the 5 at.% and 10 at.% Cadoped ferrites show the formation of Ca 2 Fe 2 O 5 as a parasite phase, which explains the decrease in performances observed during the cycles. The XRD patterns of the different Mn 1-x A x Fe 2 O 4 eNa 2 CO 3 mixtures were refined to get further insight. The average lattice parameter of the two main phases, i.e. the cubic MnFe 2 O 4 and the trigonal NaFeO 2 , are reported in Fig. S3e and f, respectively. Overall, a general decrease of the two lattice parameters is observed as the amount of dopant increases. While this behavior is less evident for Zn ferrites, it becomes way more significant for the samples doped with Ni and Ca. Data obtained for Zn and Ni samples are in line with what was expected, as a similar trend was observed for the lattice parameter of the as-synthesized Ni and Zn ferrites (Section S1 Fig. S1b). On the contrary, data are more difficult to interpret in the case of Ca-doped ferrites. Indeed, the as-synthesized CaeMnFe 2 O 4 samples showed a significant lattice expansion due to the larger Ca 2þ ions in comparison to Mn 2þ . For this reason, the lattice shrinkage detected after 10 cycles would not be expected. A possible interpretation can be found in the precipitation of the Ca 2 Fe 2 O 5 phase, which may lead to the formation of a high concentration of vacancies in the NaFeO 2 phase. In general, a straightforward connection between the lattice volume and the decarbonation-carbonation performances can be excluded for the three systems investigated. 3.4. Hydrogen production Among the composition investigated in the previous section, the spinel with 5% of Zn showed the best reversibility during the decarbonation-carbonation cycles. This composition was then selected for WS experiments. In the following, the H 2 production of the Zn 0.05 Mn 0.95 Fe 2 O 4 eNa 2 CO 3 system is presented, while using the MnFe 2 O 4 eNa 2 CO 3 mixture as a reference (Fig. 5). The samples were initially tested in dynamic conditions, by modifying the temperature program used in the decarbonationcarbonation cycles (see Section 2.2). The two mixtures were cycled ten times to investigate the reversibility of the H 2 production reaction. The first WS step is reported in detail in Fig. 5a, where the mass profiles are shown together with the H 2 evolution. The gas flow composition corresponding to each step of the experiment is also reported to facilitate data interpretation. During the first 100 min of the experiment, no water vapor was introduced. As previously carried out for the 1:1 mixtures, the samples were heated to 800  C and kept under isothermal conditions for 30 min. As can be appreciated in Fig. 5a, an initial small mass loss of 0.45 and 0.75 wt% is observed for undoped and Zn-doped mixtures, respectively; this is likely due to the desorption of some residual moisture from Na 2 CO 3 powders, which are highly hygroscopic. As expected, both mixtures start releasing CO 2 at around 550  C; the mass loss rate first increases and then approaches a plateau after around 30 min at 800  C. In this regard, the sample with 5% of Zn shows a better kinetic than the undoped MnFe 2 O 4 ,aswellasa higher CO 2 desorption (12. wt.% vs 10 wt%). Such values are in line with the theoretical mass loss expected for the release of 2/3 of the total CO 2 , which according to reaction 1, corresponds to a weight loss of 11.3 wt%. This confirms the mechanism proposed by Varsano et al. and resumed in the introduction (reactions 3 and 4) [23]. As soon as water vapor is introduced, the masses of the two samples start to decrease again and H 2 is suddenly detected. The amount of H 2 increases and reaches a maximum of 22.3 and 12.6 m mol g 1 s 1 for pure and 5% Zn MnFe 2 O 4 , respectively. Even if at a lower rate, H 2 production continues for both samples as long as water vapor is supplied. In the meantime, the masses of the two samples keep decreasing and stabilize at around 85% of the initial value. An overall mass loss of 14.4 wt% and 15.2 wt% is detected for the undoped and Zn-doped mixtures, respectively. According to the stoichiometry of reaction 1, theoretical mass losses of 14.9 and 15 wt% are expected for the undoped and Zn-doped mixtures, respectively. These values are given by the sum of two different contributions. First, the CO 2 release due to the decomposition of Na 2 CO 3 , which corresponds to a 16.9 wt% mass loss. Second, the O 2 uptake from the reaction with water steam to form the NaMn 1/3 Fe 2/3 O 2 , which corresponds to a 2 wt% mass gain. The second contribution is linked to the oxidation of Mn 2þ to Mn 3þ and, in turn, to the H 2 production. As Zn does not contribute to the redox process, a 5%-lower mass change is expected for the 5% Zn spineldi.e. þ1.9 wt%. Considering the additional mass change that is due to the initial moisture desorption, a maximum mass loss of 15.35 and 15.78 wt% are reasonably expected for the undoped and the Zn-doped mixtures; comparing the experimental data with these values, conversions of 93.8% and 96.3%, respectively, were obtained. The data obtained from the thermobalance were compared with those obtained from the H 2 microsensor. In particular, the integration of the H 2 production over time provided H 2 yields of 1.14 mmol/g for pure MnFe 2 O 4 and 0.66 mmol/g for 5% Zn MnFe 2 O 4 . As the theoretical yield for the MnFe 2 O 4 eNa 2 CO 3 mixture is 1.28 mmol/g, a relative H 2 yield of z89% is obtained, which agrees with the value obtained from thermogravimetric data. On the other hand, the Zn-doped mixture marks an H 2 yield of 54% only, with its theoretical H 2 yield being 1.21 mmol/g. Such value is significantly lower than the one calculated based on thermogravimetry and it is counterintuitive at a first glance. In fact, as the Zn-doped spinel showed a higher mass loss than the undoped, one would expect a higher H 2 yield for the former. Such results can be interpreted by recalling the different contributions to the mass change, meaning CO 2 release and O 2 uptake. The overall difference between the experimentally observed mass loss of the Zn-doped and that of the undoped mixture is 0.8 wt%. This value reduces to 0.5 wt% when considering the contributions of the initial moisture desorption. XRD analysis was performed after the first WS step (Fig. S4a). For both mixtures, NaMn 1/3 Fe 2/3 O 2 was identified as the main phase, followed by minor amounts of unreacted Na 2 CO 3 and MnFe 2 O 4 . It is then reasonable to assume that the CO 2 release of the 5% Zn-doped mixture took place with no significant changes compared to the undoped. The difference in terms of mass loss between the two mixtures is then related only to the O 2 uptake, which is also supported by the high difference in terms of produced H 2 . After a few F. Torre, T.A. Sanchez, S. Doppiu et al. Materials Today Energy 29 (2022) 101094 8 simple proportions, this leads to a relative H 2 yield of 63%, which is in better agreement with the value obtained by the direct integration of the H 2 signal. During the subsequent cycling in dynamic conditions, the Zndoped mixture showed a better reversibility in terms of mass change (Fig. S4b) that also corresponded to a significantly lower amount of undesired secondary phases such as the nonstoichiometric Na x Mn 3 O 7 (Fig. S4c). However, H 2 was not detected during the following 9 cycles for both samples. Despite this, the presence of significant amounts of MnFe 2 O 4 and Na 2 CO 3 in the cycled mixtures suggests that even if with a lower yield, H 2 was produced also during the remaining cycles. A possible explanation is that the incomplete regeneration of the reactants lowered the H 2 production so that the H 2 concentration was lower than the detection limit of the microsensor used. A more detailed discussion is provided in Section 4.1 of the supporting Information file. Despite the effect of Zn-doping on the H 2 production could not be directly evaluated in the above-mentioned experimental conditions, both XRD and thermogravimetry suggested a positive effect of Zn on the reversibility. To shed light on this, both the undoped MnFe 2 O 4 eNa 2 CO 3 and the 5% Zn-doped MnFe 2 O 4 eNa 2 CO 3 mixtures were subsequently tested in isothermal conditions at 750  C for a total of 5 cycles (Section 2.2.2 for details). The H 2 production of the two mixtures is reported in Table 2, and the data are plotted in Fig. 5b. As seen before for the experiments at 800  C, the undoped MnFe 2 O 4 eNa 2 CO 3 mixture produces more H 2 during the first cycle (86% yield). However, its H 2 production drastically decreases by two orders of magnitude in the second cycle and keeps decreasing during the following three. Eventually, almost no H 2 is detected in the fifth cycle. On the other hand, the Zn-doped mixture produces less H 2 during the first cycle (54% yield) but its production stays one order of magnitude higher than the undoped mixture during the following 4 cycles. Moreover, exempt from a first drop after the first cycle, no decreasing tendency can be observed in the following cycles, and the Zn-doped mixture shows a stable H 2 production of z0.2 mmol/g from cycle 2 to cycle 5. High reversibility is observed also in terms of kinetic, as the H 2 production profiles from cycles 2 to 5 almost perfectly overlap (Fig. S5). This is also in line with the high reversibility observed in terms of desorbed/captured CO 2 (Table S1). As a result of the two very different behaviors, the difference in terms of H 2 production performances between the two mixtures becomes more evident during the cycles. In the second cycle, the Zn-doped mixture produces 4.6 times more H 2 than the undoped one, while in the fifth cycle, it produces 23 times more. This completely different trend also affects the cumulative H 2 production, which is higher for the Zn-doped mixture (1.60 vs. 1.24 mmol/g). 4. Conclusions In the present work, we studied the effect of atomic substitution on the sodium manganese ferrite thermochemical cycle for H 2 production. The decarbonation/carbonation of the MnFe 2 O 4 eNa 2 CO 3 mixture was investigated as a starting reference, while both Na and Mn substitution were subsequently considered. Na substitution for K led to a decrease in performance that was attributed to the formation of potassium beta ferrite as an undesired phase. On the other hand, the use of Li decreased the onset decarbonation temperature by about 100  C, but negligible reversibility was observed under the explored experimental conditions. Despite this, the substitution of Na for Li is promising for decreasing the operating temperature of the thermochemical cycle and deserves further investigation. Mn partial substitution for Ca, Ni, and Zn improved the rate of the carbonation reaction. The Mn 0.95 Zn 0.05 Fe 2 O 4 eNa 2 CO 3 mixture showed the best reversibility and was tested for H 2 production, while the undoped mixture was used as the reference. During the first cycle, maximum instantaneous H 2 production rates of 22.3 and 12.6 m mol g 1 s 1 were observed for the undoped and the Zndoped samples. After 1.5 h at 800  C, the MnFe 2 O 4 eNa 2 CO 3 mixture reached 89% of the theoretical yield, with an overall H 2 production of 1.14 mmol/g. On the other hand, only 0.66 mmol H 2 /g (54% conversion) were produced during the same time by the Zndoped mixture. In both cases, cycling in dynamic conditions (800e400  C) did not ensure a complete regeneration of the Fig. 5. (a) Mass profiles and H 2 evolution of the MnFe 2 O 4 eNa 2 CO 3 and 5% Zn MnFe 2 O 4 eNa 2 CO 3 mixtures during the first WS step at 800 C. (b) H 2 production (mmol H 2 /g) of the undoped MnFe 2 O 4 eNa 2 CO 3 and the 5% Zn-doped mixtures during 5 cycles at 750 C. Table 2 H 2 production (mmol H 2 /g) of the undoped MnFe 2 O 4 eNa 2 CO 3 and the 5% Zn-doped mixtures during 5 cycles at 750  C. Mixture Hydrogen production (mmol H 2 /g) Cycle 1 Cycle 2 Cycle 3 Cycle 4 Cycle 5 Total MnFe 2 O 4 1.10 0.05 0.04 0.03 0.01 1.24 5% Zn 0.69 0.23 0.22 0.22 0.23 1.60 F. Torre, T.A. Sanchez, S. Doppiu et al. Materials Today Energy 29 (2022) 101094 9