Inhibiting the interaction between FeO and Al2O3 during chemical looping production of hydrogen

Hydrogen of high purity can be produced by chemical looping using iron oxide as an oxygen carrier and making use of the reaction between steam and either iron or FeO. However, this process is viable only if the iron oxide can be cycled between the fully-oxidised and fully-reduced states many times. This can be achieved if the iron oxide is supported on refractory oxides such as alumina. Unfortunately, the interaction between alumina and oxides of iron to form FeAl2O4 hinders the kinetics of the reactions essential to the production of hydrogen, viz. the reduction of Fe(II) to metallic iron by a mixture of CO and CO2 prior to the oxidation by steam. Here, oxygen carriers containing Fe2O3 and Al2O3 were doped with Na2O and, or, MgO, in order to inhibit the formation of FeAl2O4 by forming NaAlO2 or MgAl2O4, respectively. The performance of the modified oxygen carriers for producing hydrogen, i.e. cyclic transitions between Fe2O3 (or Fe3O4) and metallic Fe at 1123 K were investigated. It was found that the interaction between FeO and Al2O3 was successfully mitigated in an oxygen carrier containing Mg, with an Al: Mg ratio of 2, resulting in consistently stable and high capacity for producing hydrogen by chemical looping, whether or not the material was oxidised fully in air in each cycle. However, the oxygen carrier without Mg only remained active when a step to oxidise the sample in air was included in each cycle. Otherwise it progressively deactivated with cycling, showing substantial interaction between Al2O3 and oxides of Fe.


Introduction
Hydrogen is a potential fuel for a low-carbon economy, provided it can be manufactured with minimal release of CO 2 to the atmosphere.Generally, to generate hydrogen, carbonaceous fuels need to be consumed, usually via largescale, energy-intensive processes such as steam-methane reforming.Where the hydrogen is to be used in polymericelectrolyte-membrane fuel cells; an additional process is required to purify the hydrogen to avoid poisoning of the Pt electrodes by the low concentrations of CO which contaminate the product of steam-methane reforming. 1However, CO may be less of a problem in other types of fuel cells, e.g.solidoxide fuel cells or solid molten carbonate fuel cells. 1 An alternative process, capable of producing pure hydrogen directly from carbonaceous fuels, at small to medium scales with simultaneous capture of CO 2 , is chemical looping hydrogen production. 2 Chemical looping techniques rely on the ability of transition metal oxides to transport oxygen from an oxidising environment to a reducing environment.When CO is used as a reductant and the oxygen in air is used as an oxidant, the reactions can be generalised: CO reduction: MeO x(s) + yCO (g) → MeO x−y(s) + yCO 2(g) ( 1 )   Air oxidation: MeO x−y (s) + O 2(g) → MeO x(s) ( 2 ) where MeO x and MeO x−y represent the oxidised and reduced forms of the transition metal oxide, referred to here, together with any supporting matrix, as the oxygen carrier.In this 50 example, oxygen is transported from air by MeO x to combust CO in a nitrogen-free environment, resulting in a stream of pure CO 2 , which is suitable for carbon storage.Here, the additional cost associated with the separation of CO 2 from N 2 for carbon capture in conventional coal-fired power plants is 55 avoided.In a variant of the above, if iron oxide is used as an oxygen carrier, hydrogen can be produced during the oxidation stage of the steam-iron process: 3/4Fe (s) + H 2 O (g) ⇌ 1/4Fe 3 O 4(s) + H 2(g) , ∆H o 1123 K = −26.3kJ/mol ( 3 ) originally patented by Messerschmitt 3 .Reed and Berg 4 proposed a system to produce pure hydrogen continuously 60 using the steam-iron process, by circulating iron oxides between three interconnected fluidised bed reactors.In their process, the iron oxide oxygen carrier was subjected to repeated redox reactions, i.e. chemical looping.When synthesis gas, e.g. a mixture of CO and H 2 is used as a fuel, the 65 reactions in the three-reactor system would be: 3Fe 2 O 3(s) + H 2(g) ⇌ 2Fe 3 O 4(g) + H 2 O (g) ∆H o 1123 K = −5.9kJ/mol ( 4 )  9) would be conducted, where haematite (Fe 2 O 3 ) is reduced by CO and H 2 sequentially to magnetite (Fe 3 O 4 ), wüstite (Fe 0.947 O) and iron (Fe).Both wüstite and iron can be then oxidised by steam in the second reactor, viz. the steam reactor, 5 according to the reverse of reactions ( 3 ) and ( 5 ).The solid product, magnetite, is further oxidised to haematite according to the exothermic reaction ( 10 ), in the third reactor, viz. the air reactor, which provides heat to the system.Overall, the chemical energy of the synthesis gas is used to produce 10 hydrogen by splitting water via chemical looping of the iron oxide.Of course, the hydrogen used during the reduction is mixed with CO, whereas the hydrogen produced in this manner is effectively free from carbonaceous contamination, and has high purity. 2 One of the major challenges in the production of hydrogen by chemical looping is the development of a stable oxygen carrier capable of being reduced and oxidised over many cycles without deactivation.When almost pure iron oxide (99 wt% haematite) is used as an oxygen carrier, the reduction of Fe 3 O 4   20   to Fe in each cycle led to rapid deactivation within the first few cycles. 2However, by limiting the extent of reduction to wüstite (FeO) in every cycle, a stable hydrogen yield could be obtained over more than 10 redox cycles. 2 Nevertheless, the transition between Fe and Fe 3 O 4 produces about four times as 25 much hydrogen as the transition between wüstite (FeO) and magnetite (Fe 3 O 4 ), for a given mass of iron, so that reduction to iron is desirable to maximise yield.6][7] However, it 30 was also found that hercynite, FeAl 2 O 4 , was inevitably formed in the presence of Al 2 O 3 and this slowed down the rate of reduction of the supported iron oxide when the latter was reduced beyond Fe 3 O 4 . 8Here, FeAl 2 O 4 belongs to the spinel group of crystalline phases with a general formula of AB 2 O 4 , 35 where the A-sites are typically occupied by divalent cations and B-sites by trivalent cations.On the other hand, it was hypothesised by Kidambi et al.,7 that the formation and destruction of solid phases on each redox cycle might be important in maintaining the activity of the oxygen carrier.40 However, Liu et al. 9 have demonstrated the stabilising effect of a ZrO 2 support on the performance of iron oxide oxygen carriers for chemical looping hydrogen production.The support in this system is chemically stable, and no phases involving iron oxide are formed.Therefore, albeit desirable, it 45 is not strictly necessary for there to be changes of phase to promote a carrier's stability over many cycles.Nevertheless, one drawback of using ZrO 2 as a support material for chemical looping is its high cost relative to common refractory materials, e.g.Al 2 O 3 .

50
1][12] Interestingly, when CuO is used as a chemical looping agent, researchers have 60 reported that the formation of the spinel phase CuAl 2 O 4 is inhibited in the presence of trace amounts of alkali metals (e.g.4][15] The source of these alkali metal "contaminants" is most likely to be the Na-and K-dawsonites, viz.NaAlCO

Preparation of oxygen carriers
The oxygen carriers containing (a) oxides of Fe, Al and Na, and (b) oxides of Fe, Mg, Al and Na are hereinafter referred to, respectively, as NAFO and NMAFO.Both oxygen carriers were synthesised using the method of co-precipitation with a 85 slowly decreasing pH.In both cases, the precipitant used was an aqueous solution containing NaOH (≥ 98 wt%, Fisher Scientific) and Na 2 CO 3 (≥ 99.5 wt%, Fisher Scientific) mixed at a molar ratio of 1:1 to give a total concentration of [Na + ] ~ 4 M.The precursors were aqueous solutions containing nitrates 90 of Mg 2+ , Al 3+ and Fe 3+ , viz.Mg(NO 3 ) 2 •9H 2 O (BioXtra, ≥98 wt%, Sigma-Aldrich), Al(NO 3 )  in the alkali solution (the precipitant) was 3 times the total number of moles of metal ions (Mg 2+ , Al 3+ and Fe 3+ ) in the nitrate solution (the precursor).The precipitation was undertaken isothermally at 373 K, at atmospheric pressure.During precipitation, the precursor was added at a constant rate of ca. 1 drop per second to a 1 L flask containing the alkali solution, which was stirred constantly at 600 rpm.The temperature of the solution in the flask was held constant at 100°C, below the point at which the slurry boiled.Loss of water by evaporation was prevented by condensing and refluxing all the vapour at the outlet of the flask.After the addition of the last drop of nitrate solution, the precipitate was kept stirred at 373 K for an additional 24 h.The final pH of the resulting slurry was found to be 10.0 ± 0.2.Then, the slurry was washed repeatedly with de-ionised water (with an ionic 20 conductivity of ~20 µS cm −1 ) until the ionic conductivity of the wash water fell below 120 µS cm -1 .Approximately 120 L of de-ionised water was used, over 4 exchanges, to wash each batch of precipitate which yielded 20 g of calcined oxygen carrier.The washed precipitates were filtered, dried at 353 K in 25 air for 24 h, and fired at 1223 K for 3 h.The resulting cake of oxygen carrier was crushed and sieved to a size fraction of 300 -425 µm for experiments in a packed bed reactor.

Characterisation of the solids
To characterise the crystalline phases in the solid samples, X-30 ray powder diffraction (XRD) were performed on a Panalytical X'Pert Pro diffractometer using Cu Kα radiation (wavelength = 1.5418Å), operated at 40 kV and 40 mA.The angle of reflection, 2θ, was varied between 5 and 60º, at a rate of 0.0167° per step where each step takes 16.5 s.The diffraction patterns were collected at ambient temperature and atmosphere.The collected diffractograms were preliminarily inspected with the aid of reference patterns found in the Inorganic Crystal Structure Database (ICSD).Rietveld refinements were further performed on the collected diffractograms using GSAS with the graphical interface EXPGUI. 16,17o investigate the elemental composition of the oxygen carriers, loose powders of the oxygen carriers were analysed by an S4 Explorer X-ray fluorescence spectrometer (XRF) system (Bruker AXS GmbH).The surface morphologies of the solid samples were inspected by a scanning electron microscope (NOVA NANOSEM, FEI), with a secondary electron detector, an accelerating voltage of 5 kV and probe current of 2 nA under high-vacuum (6 × 10 −5 mbar).The samples were sputtercoated with Pt prior to examination by SEM.Energydispersive X-ray spectroscopy (EDS) analyses were performed on selected spots on the surface of the solid samples within the same SEM, using an accelerating voltage of 15 kV.The elemental composition at these spots was quantified using the 55 collected EDS spectra and existing calibrations.The oxygen content available for reduction and oxidation in the oxygen carriers was quantified by reducing the samples isothermally in a thermogravimetric analyser (TGA/DSC1, Mettler Toledo).In a typical TGA experiment, ca.20 mg of 60 fully oxidised sample was placed in a 70 µL alumina crucible, and heated to 1223 K at a rate of 20 K min -1 under a flow of N 2 (BOC Ltd.) of 50 mL min −1 (measured at 295 K, 1 atm).After degassing for one hour at 1123 K, an additional flow of 50 mL min −1 (measured at 295 K, 1 atm) of 5 vol.%H 2 in N 2 65 (Spectrashield, BOC Ltd.) was introduced to the reaction chamber for 10 h before cooling down in N 2 at a rate of −20 K min −1 .The change in mass of the samples was recorded and the mass fraction of Fe 2 O 3 in the fully oxidised sample was calculated, as shown later in section 5. 1.   70   To investigate the performance of the oxygen carriers, the samples were cycled under conditions suitable for producing hydrogen by chemical looping in a vertical packed bed.The wall of the tubular reactor (nominal I.D. = 9 mm) was made of recrystallised Al 2 O 3 (>99 wt% purity, Multi-Lab Ltd.).The 75 results of an earlier study by Liu et al. 9 confirm that the wall of the tube does not react with Fe-based oxygen carriers under chemical looping conditions, and can be therefore considered as inert.The packing arrangement, from top to bottom, consisted of the following layers of materials: (i) 10.0 g α-80 Al 2 O 3 (1400 -1700 µm), (ii) approximately 0.3 g of oxygen carrier particles (300 -425 µm), (iii) 2.0 g of α-Al 2 O 3 (300 -425 µm), (iv) 2.0 g of α-Al 2 O 3 (1400 -1700 µm), and (v) a ceramic distributor plate.The distributor, consisting of three holes of diameter 1.5 mm each in triangular array, was cast 85 using an Al 2 O 3 -based cement (AL/CS -chemical setting adhesive cement, Multi-lab Ltd.) inside the tube.Both ends of the alumina tube were fitted with brass fittings with Viton rubber O-rings acting as seals.The control and gas sampling system for the packed bed has been described elsewhere by Liu 90 et al. 9 .During the cycling experiments in the packed bed, the temperature of the active bed was maintained at 1123 K.In each cycle, the gas supplied to the bed consisted successively of: (i) 5 min of reduction by 9.5 vol% CO in N 2 (BOC Ltd.), 95 (ii) 2 min of N 2 purge (Air Liquide UK), (iii) 3 min of oxidation by 12 vol% CO 2 (Air Liquide UK) in N 2 and (iv) 1 min of N 2 purge and (v) 2 min of oxidation by air.Switching between the feed-gases was achieved using solenoid valves.In an additional set of experiments, the freshly-prepared oxygen 100 carriers were cycled without step (v), in order to examine the presence of interaction between Al 2 O 3 and oxides of Fe.Should there be interaction between Al 2 O 3 and oxides of Fe, the absence of step (v) would result in a decay in reactivity and consequently deterioration in the conversion of solid over where p i stands for partial pressure of species i.The validity of using CO 2 to stimulate steam oxidation for chemical looping hydrogen cycles has also been experimentally confirmed by Liu et al. 9 .The flowrates of gases at the inlet were controlled and monitored using rotameters or mass flow sensors (Honeywell AWM5103VN).In all stages, the nominal gas flowrates were 1.5 L min -1 (measured at 295 K, 1 atm).For each sample, the oxygen carrier was cycled 20 times, with or without air-oxidation, viz.step (v).The effluent gas from the reactor was analysed by a non-dispersive infrared analyser (NDIR, Uras26, EL3020, ABB), which detects the mole fractions of CO and CO 2 in the range of 0 -10 vol% and 0 -12 vol%, respectively.X-ray diffraction patterns of the precipitates were collected to investigate the crystalline phases formed during coprecipitation but before calcining at 1223 K; they are shown in Fig. 1.In Fig. 1a, it can be seen that the dried precipitate consisted of a mixture of NaAlCO 3 (OH) 2 (dawsonite) and Fe 1.67 H 0.99 O 3 , a dehydrated form of Fe(OH) 3 .In Fig. 1b, the peaks at 11.7° and 23.5° are characteristic of the (0 0 2) and (0 0 4) planes, respectively, of Mg 4 Al 2 (OH) 12 CO 3 •3H 2 O (quintinite).The fact that the position of the peaks of the quintinite phase matches closely a reference pattern of pure quintinite (Mg 4 Al 2 (OH) 12 CO 3 •3H 2 O, ICSD-182293) suggests that the trivalent sites in the quintinite phase contain little or no Fe 3+ , i.e., almost all Fe 3+ precipitated in the form of Fe(OH) 3 .Peaks of dawsonite are also seen in Fig. 1b, together with weak peaks of Fe 1.67 H 0.99 O 3 .If there are no amorphous phases present, the precipitates of (a) NAFO and (b) NMAFO, should contain mixtures of (a) dawsonite and Fe(OH) 3 and (b) quintinite, dawsonite and Fe(OH) 3 , respectively.In particular, for NAFO, the composition detected by XRD is also in agreement with the results of XRF shown in Table 1 below.

40
Overall, all peaks in Fig. 1 can be indexed by Fe 1.67 H 0.99 O 3 , dawsonite and quintinite.The precipitates of both samples were also examined by SEM, and the images are shown in Fig. 2. Both images in Fig. 2 show large rod-shaped crystallites, which are characteristic of 45 the dawsonite (NaAlCO 3 (OH) 2 ) phase. 18In Fig. 2(a), it can be seen that the precipitate also contains spherical agglomerates.Assuming no amorphous phase is present, these spherical agglomerates should be the dehydrated precipitate of Fe(OH) 3 .Fig. 2(b) shows that the precipitates for NMAFO consist of a 50 mixture of rod-shaped crystals and some small grains.Based on the results of XRD, the latter may be assigned to a mixture of quintinite and Fe(OH) 3 .x , the mass fraction of Fe 2 O 3 in the calcined oxygen carrier, was estimated using:   3, suggesting that all the Fe(III) in both oxygen carriers is eventually reducible in H 2 to Fe(0).However, the proportion of 65 Fe(III) practically reducible to Fe(0) during chemical looping is discussed in section 5.2, below.

Cyclic performance in chemical looping
Particles of both oxygen carriers, sized d p = 300-425 µm, were investigated in a packed bed for the chemical looping off-gas during the first two cycles as well as the last cycle in a typical cycling experiment are shown in Fig. 5. Here, the concentration of O 2 during the air-oxidation was not measured.
It can be seen from Fig. 5 that, in the case of NMAFO, the profiles were reproducible over the cycles, showing stable performance.The profile of CO 2 produced during reduction, when CO was fed to the inlet, showed an initial sharp rise followed by decay.This is characteristic of the reduction of Fe 2 O 3 successively to Fe 3 O 4 , FeO and Fe. 19It should be noted that the concentration of CO 2 did not fall to zero at the end of the reduction, suggesting a conversion of Fe 2 O 3 to metallic Fe of less than 100% in each cycle.On the other hand, the rate of oxidation of the reduced carriers by CO 2 fell to zero after 90 s, suggesting fast apparent rates c.f. reduction.The spike of CO 2 at the very beginning of the oxidation in each cycle was a result of insufficient mixing between CO 2 and N 2 in the feed, immediately after gas-switching, which was achieved using solenoid valves.The conversion from Fe 3 O 4 to Fe, X, was calculated in two ways, when the oxygen carriers were oxidised in air to Fe 2 O 3 at the end of each cycle:

01667
. 0 Here, equations ( 15) and ( 16) calculate X during the reduction and the subsequent CO 2 -oxidation, respectively.The constants 0.00208 and 0.01667 mol are the stoichiometric amounts of CO  and X ox was taken as the average value of X in each cycle.The measured X were plotted against cycle number in Fig. 6.Fig. 6 (a) shows that, when NAFO was oxidised by air at the end of each cycle, X ~ 100% consistently over 20 cycles.In contrast, without air, X fell quickly, tending towards 50% by the 20 th cycle.This diminution is similar to those reported by Bohn et al. 5  consistent performance was seen for NMAFO, as shown in Fig. 6b, regardless of whether or not air oxidation was used.
The measured X, with and without air, equals to ca. 80% over the 20 cycles at 1123 K.These values of X were less than unity owing to the fact that reduction did not reach completion in 20 each cycle, as discussed above.Overall, the evolution of the XRD patterns in Fig. 7b shows that, the MgFe 2 O 4 phase (with some dissolved Fe 3 O 4 ) was decomposed during reduction, and could be regenerated in CO 2 .Despite the formation of MgFe 2 O 4 , the nearly 80% X seen in Fig. 6(b) demonstrates that this spinel compound did not significantly compromise the reactivity of Fe 2 O 3 .In addition, the fact that MgFe 2 O 4 may be regenerated in CO 2 means there is a potential to produce more hydrogen per mole of Fe by the steam-oxidation of reduced MgFe 2 O 4 than of pure Fe; this is discussed below in section 6.2.Lastly, similar to the case of NAFO, NaAlO 2 was deemed stable under the conditions used for the cycling tests.Another important aspect indicating the stability of an oxygen carrier is its change in surface morphology with redox cycling.This was investigated by SEM; the results in Fig. 8 show the morphologies of NAFO when (a) fresh, (b) after 20 cycles without air-oxidation and (c) after 20 cycles with air-oxidation, respectively, as well as the surface of particles of NMAFO when (d) fresh and (e) + (f) after 20 cycles without oxidation in air.In all cases, the disappearance of the rod-shaped grains seen in Fig. 2 can be correlated to the decomposition of dawsonite (NaAlCO 3 (OH) 2 ).Fig. 8(a) shows that fresh NAFO, the surface of which is relatively porous, is composed of small, irregularly-shaped grains of uniform size.After 20 cycles without air-oxidation, some large hexagonal crystals were found in NAFO, together with some agglomerated small grains 60 of no particular crystalline feature.When cycled with airoxidation, the surface of NAFO was found to be mainly composed of uniformly-sized grains similar to those seen in Fig. 8 (a).A large grain exhibiting a prismatic crystal habit is also seen in Fig. 8(c); this means that the crystallite has an 65 orthorhombic crystal structure, suggesting that it is made of NaAlO 2 .On the other hand, both fresh and cycled (without airoxidation) NMAFO were found to contain large hexagonal crystals of ca. 1 µm in size mixed with small grains of ca. 100 nm in size.Interestingly, none of the crystalline phases 70 identified in Fig. 7 (b)(iv) crystallises hexagonally.To investigate the nature of the hexagonal crystals seen in energydispersive X-ray spectra (EDS) of the corresponding samples were taken at points shown in Fig. 9; and the results of the quantitative analysis are presented in Table 2.  Here, it should be noted that, the volume of interaction 80 between the electron beam (with an accelerating voltage of 15 kV) and the solid material is several microns wide.Hence, the EDS collected for the hexagonal grains, e.g.spectra 2, 10 and 11, might also contain signals produced by the surrounding materials.Therefore, the numbers shown in Table 2 should 85 only be regarded as giving a qualitative indication, which can only be used to estimate the chemical compositions of the hexagonal grains, in combination with other experimental results.
The result of EDS analysis (Fig. 9 and Table 2) suggests that these micron-sized hexagonal grains are predominantly made of Al 2 O 3 , which means that the "host-phase" of these crystals could be corundum (α- where n = 1 -1.5), or β''-alumina ((Na 2 O)•(Al 2 O 3 ) 6 ).However, 5 the lack of detection of the corundum peaks in Fig. 7 suggests that, the content of Al 2 O 3 relative to the other phases is low in the bulk of the solid, despite its strong presence at the surface.Furthermore, the strongest peak of Al 2 O 3 at around 35° could overlap with a peak of NaAlO 2 , making detection of the former even more difficult.From Fig. 9 and Table 2, it can be also seen that, the metal-based elemental composition of the small, 100 nm-sized grains in NAFO was comparable to the analysis by XRF, although the content of Fe in the small grains in NMAFO appears to be underestimated by EDS.Thus, a plausible explanation to the behaviour of the oxygen carrier is proposed and depicted in Fig. 11.Here, the reaction schematic shown in Fig. 11 postulates that, as a result of the formation of NaAl 1−y Fe y O 2 , where y increased gradually with 60 time, an increasing amount Al 3+ was displaced from the This journal is © The Royal Society of Chemistry 2012 NaAl 1−y Fe y O 2 phase, and became free to interact with oxides of Fe.Evidence for this explanation comes from XRD measurements on fresh carrier and carrier cycled many times.Initially, the freshly-calcined oxygen carrier contained a mixture of NaAlO 2 and Fe 2 O 3 , both of which were detected by 5 XRD, as seen in Fig. 7 (a)(i).In the first cycle, the only reactive species in the fresh NAFO was Fe 2 O 3 , whilst the NaAlO 2 was more or less stable, as seen in Fig. 7 (a), (i)-(iii).
In other words, almost all the Fe 3+ in NAFO was present as Fe 2 O 3 , which could be reduced completely to Fe, resulting in 10 the observed X of ~ 100%.However, peaks of FeAl 2 O 4 can be clearly seen in the sample after 20 cycles without air-oxidation (Fig. 7 (a)(iv)), suggesting that a substantial amount of Al 3+ was made available to interact with Fe 2 O 3 , Fe 3 O 4 and FeO.In particular, the formation of FeAl 2 O 4 has been known to be 15 responsible for the decaying performance of the Fe-based oxygen carriers in the absence of an air-oxidation step. 6,7Thus far, the theory matches the experimental facts.This section considers the picture when there is air oxidation included in each cycle and how the mechanism just posited is altered; this is also depicted in Fig. 11.Firstly, despite an increasing value of y in the NaAl 1−y Fe y O 2 phase, the consistently high conversion over the 20 cycles with air-30 oxidation shown in Fig. 6(a) means that, the accumulation of Fe 3+ in NaAl

Performance of NMAFO
Here, the changes in composition during the preparation and investigation of NMAFO are depicted in Fig. 12  The results of EDS analysis, shown in Fig. 9 and Table 2, suggest that this Al-containing phase could be Al 2 O 3 (corundum), which appears to be stable in the form of large hexagonal crystals before and after the cycling experiments, with or without air-oxidation.In addition, the absence of  The phase diagram of the Mg-Fe-O system at 1123 K, calculated using MTDATA 21 and plotted in Fig. 13, can be used to explore the thermodynamics further.Fig. 13 shows that, in the Mg-Fe-O system, the reduction of NMAFO, in  support).Thus, the use of Mg 2+ as an inhibitor as well as a promoter might be optimised.Finally, Fig. 13 shows that, during the cycling experiments, the MgFe 2 O 4 -Fe 2 O 3 mixture in NMAFO would be transformed between single-phase regions and two-phase mixtures repeatedly.It is believed that cycling between a two-phase 35 mixture and a single-phase solid solution could give oxygen carriers a regenerable porous structure, which renders superior cyclic stability over conventional oxygen carriers, e.g. Dunstan et al. 22 have suggested that a similar mechanism might explain the 40 longevity of their high-temperature sorbents for the capture and release of CO 2 , where the solid sorbents undergo changes in molar volume cyclically at a temperature similar to that used in this study.Nevertheless, it must be remembered that, in practice, the oxygen carrier would have to withstand several 45 thousand of cycles of reduction and oxidation.Therefore, other properties, such as resistance against fracture due to repeated change in molar volume and resistance against attrition, must be considered for the future development of the oxygen carrier.
In conclusion, the oxygen carrier NMAFO worked well,  23 for the production of hydrogen by chemical looping.Accordingly, Aston et al. 23

Conclusion
Fe-based oxygen carriers, consisting of oxides of (i) Na, Al conditions suitable for producing hydrogen at 1123 K, both oxygen carriers showed stable performance over 20 cycles, demonstrating ~100% and ~80%, respectively, of their theoretical capacities to produce hydrogen, when they were oxidised in air at the end of each cycle.However, without airoxidation, the performance of NAFO was found to decay with cycling, attributed to the gradual, progressive formation of FeAl 2 O 4 .In contrast, the performance of NMAFO was found to be stable and consistent with cycling regardless of the inclusion of air-oxidation.The experiments suggested that, although there was some evidence that NaAlO 2 resisted the formation of FeAl 3 (OH) 2 and KAlCO 3 (OH) 2 , respectively, during 65 the preparation of the oxygen carriers by co-precipitation.The aim of the present study is to investigate the potential to use oxygen carriers containing a mixture of Fe 2 O 3 , Al 2 O 3 , Na 2 O and, or, MgO for producing hydrogen by chemical looping, because the use of Na 2 O and MgO could potentially 70 inhibit the formation of FeAl 2 O 4 by forming NaAlO 2 and MgAl 2 O 4 with Al 2 O 3 instead.Two oxygen carriers were synthesised: one containing oxides of Fe, Al and Na and another containing oxides of Fe, Mg, Al and Na.Both oxygen carriers were investigated under conditions which simulate the 75 production of hydrogen by chemical looping.The formation and interaction of different phases of mixed oxides during repeated redox cycling were studied using ex-situ characterisation techniques.
105 cycles, as shown by Kidambi et al. 7 .It should be noted that CO 2 was used in step (iii) to simulate the oxidation of reduced iron oxide by steam to produce hydrogen, because the equilibrium values of p CO 2 /p CO and p H 2 O / p H 2 for the oxidation of Fe to wüstite and wüstite to Fe 3 O 4 are similar at 1123 K, 110 This journal is © The Royal Society of Chemistry 2012

5 .
Fig. 1 X-ray diffraction patterns of the precipitates during the preparation of (a) NAFO and (b) NMAFO, before thermal treatment.All diffraction peaks between 5° and 60° are identified and marked with the corresponding crystalline phases.

Fig. 2
Fig. 2 Scanning electron micrographs of the precipitates during the preparation 55 Fe 1.67 H 0.99 O 3 Dawsonite NaAlCO 3 (OH) 2 , Quintinite Mg 4 Al 2 (OH) 12 CO 3 •3H 2 of the active iron oxide in the calcined precipitates, i.e. the freshly-prepared oxygen carriers, was determined by isothermal reduction of the calcined samples in a diluted stream of H 2 in N 2 at 1223 K in the TGA.The change in mass of the samples during reduction was rendered dimensionless: m(0) is the mass of the sample just before reduction by H 2 , at t ~ 133 min, when there was no further weight loss by dehydration and degassing of the solid sample.The resulting profiles of ) ( ˆred t m are shown in Fig. 3. Assuming complete 10 reduction of the samples at the end of the experiment, Fig. 4 XRD patterns of the freshly-prepared oxygen carriers of (a) NAFO and (b) NMAFO.

4
70production of hydrogen at 1123 K.As mentioned above, the production of hydrogen was simulated by the production of CO by oxidising the reduced oxygen carrier in CO 2 , the redox potential of which is almost identical to that of steam at 1123 K.A total of 20 cycles were performed in each 75 experiment.Profiles of mole fractions of CO and CO 2 in the 5 This journal is © The Royal Society of Chemistry 2012

Fig. 5
Fig.5Variation of the mole fractions of CO and CO2 in the off-gas of a packed bed, in which a small amount of NMAFO was cycled at 1123 K under conditions equivalent to those suitable for producing hydrogen by chemical looping.The oxygen carrier is tested for 20 cycles, where the profiles of the first two cycles and the last cycle are shown here.The packed bed was oxidised in air at the end of each cycle, although the concentration of oxygen is not measured during the experiments.

Fig. 6
Fig. 6 The conversion of Fe3O4 to Fe, X versus cycle number, when approximately 0.3 g of (a) NAFO and (b) NMAFO were cycled in a packed bed reactor in the presence or absence of oxidation by air, at 1123 K. Error bars are 55

Fig. 7
Fig. 7 XRD patterns of (a) NAFO and (b) NMAFO at different stages of the cycling experiments, when they are (i) fresh, (ii) reduced in a mixture of CO and CO2 (pCO 2 :pCO = 1) at 1123 K for 10 min, (iii) reduced in 10% CO in N2 at 1123 K for 10 min, (iv) oxidised in CO2, after 20 cycles without oxidation in air, and (v) oxidised in air, after 20 cycles with oxidation in air.All peaks are assigned and labelled: NaAlO2, Fe2O3,  Fe-based spinel (e.g.Fe3O4 and, or MgFe2O4), wüstite, * MgO, Fe, and † Al-based spinel (e.g.MgAl2O4 and, or FeAl2O4).The peaks marked by ‡ are not indexed.The arrows indicate peak-shifts relative to a pure phase of MgFe2O4.

Fig. 7
Fig.7shows the XRD patterns of the oxygen carriers before and after cycling, as well as the fresh samples after various treatments, viz.(i) reduction in 10% CO in N 2 at 1123 K for 10 min and (ii) reduction in a mixture of CO and CO 2 (p CO 2 :p CO = 1) at 1123 K for 10 min.For NAFO, it can be seen from Fig.7(a) that, during the reduction of the fresh oxygen carrier, the Fe 2 O 3 is reduced subsequently to Fe 3 O 4 , FeO and finally Fe, without any detectable formation of FeAl 2 O 4 , i.e. the interaction between FeO and Al 2 O 3 was absent during the first cycle, and hence the initially high X seen in Fig.6(a).However, peaks of FeAl2O4 can be clearly seen in the sample after 20 cycles without air-oxidation Fig.7(a)(iv)).This result suggests that there was an accumulation of FeAl 2 O 4 in NAFO over cycles without air-oxidation.Consequently the reactivity of the oxygen carrier decreases gradually and progressively, as 75

Fig. 9
Fig. 9 SEM image of the surface of a particle of (a) NAFO after 20 cycles without oxidation by air, (b) fresh NMAFO and (c) NMAFO after 20 cycles without oxidation by air.Positions where EDS spectra were taken are marked by crosses and results are shown in Table2.
Fig. 9 SEM image of the surface of a particle of (a) NAFO after 20 cycles without oxidation by air, (b) fresh NMAFO and (c) NMAFO after 20 cycles without oxidation by air.Positions where EDS spectra were taken are marked by crosses and results are shown in Table2.

20 6 . Discussion 6 . 1 .
Performance of NAFOThe experimental results suggest that NAFO underwent several phase changes during its initial preparation and also during the subsequent cycling experiments.The interesting 25 features observed during the experimental investigation of NAFO are discussed in this section.6.1.1.BEHAVIOUR OF NAFO DURING REDOX CYCLING WITHOUT AN AIR-OXIDATION STEP Unfortunately, a comprehensive phase diagram for the Na-Fe-30 Al-O system is not available in the literature, so that the following discussion is based on experimental results from both the current study as well those in the existing literature.Firstly, the chemical inertness of the NaAlO 2 phase was investigated by XRD for 27° ≤ 2θ ≤ 33°, with a slow scanning speed of 0.00418° per 50 s per step.The results of the XRD, shown in Fig. 10, indicate that for both NAFO and NMAFO, the (1 2 0) peak of the NaAlO 2 phase at 30.3° has shifted slightly to a lower 2θ value after the cycling experiments.Here, the NaAlO 2 originates from the dawsonite in the original 40 precipitation.This shift could be explained by the dissolution of Fe 3+ into the NaAlO 2 phase, forming a solid solution with a general formula NaAl 1−y Fe y O 2 (0 < y < 1), which is stable even at room temperature. 20However, Fig.2(a) suggests, visually, that Fe(OH) 3 and dawsonite (NaAlCO 3 (OH) 2 ) precipitated 45 heterogeneously during the preparation of NAFO, so that for Fe 3+ -from either Fe 2 O 3 or Fe 3 O 4 -to dissolve in the NaAlO 2 phase, it would have to migrate a considerable distance to do so.Because the migration of ions in a solid is often a slow process below its melting point, the substitution of Al 3+ by 50 Fe 3+ in NaAlO 2 might therefore be, kinetically, slow, even though thermodynamically favourable.

Fig. 10 X
Fig. 10 X-ray diffraction patterns of the oxygen carriers before and after the 20 45 line in fact with the thermodynamic predictions from the Fe-Al-O phase diagram.Accordingly, it is concluded that (i) the addition of Na 2 O alone to a mixture of Fe 2 O 3 and Al 2 O 3 is insufficient to prevent the Fe-Al-O interaction, and (ii) the accumulation of the resulting FeAl 2 O 4 spinel will lead to a 50 progressive reduction in reactivity unless it is decomposed by air-oxidation in each cycle.

Fig. 11 A
Fig. 11 A schematic diagram showing the change of composition of the oxygen carrier NAFO at various stages of the experimental investigation.The availability of the corresponding XRD patterns is denoted at the bottom of the boxes.55

MgAl 2 O
4 peaks in Fig. 7 (b)(i)-(iii), and the absence of cubic crystals in Fig. 8 (d), suggest that the fresh NMAFO contains very little MgAl 2 O 4 .As with NAFO, NaAl 1−y Fe y O 2 was progressively formed with cycling and its presence verified by XRD, with evidence 20 shown in Fig. 10.Therefore, analogous to the case for NAFO, it is suggested here that the initially pure NaAlO 2 phase is gradually doped with Fe 3+ , progressively forming NaAl 1−y Fe y O 2 over the course of the cycling experiments.Then, the displaced Al 3+ ions are simultaneously absorbed by: 25 MgFe2O4 + Al2O3 = MgAl2O4 + Fe2O3.Consequently, MgAl2O4 accumulated over time, and is clearly seen in Fig. 7 (b)(iv) and (v).In the extreme scenario where all the Al 3+ ions are eventually displaced by Fe 3+ in the NaAl 1−y Fe y O 2 phase, a ratio of [Mg]: [Al] of 0.5 would ensure that FeAl 2 O 4 does not 30 form.Hence, the oxygen carrier showed consistent conversion from Fe 3 O 4 to Fe of ~80% with or without air-oxidation over 20 cycles at 1123 K.In passing, it should be also noted that the conversion is less than unity in Fig. 6(b) because the reduction was not completed in each cycle, as indicated by the non-zero 35 [CO 2 ] at the end of each reduction step in Fig. 5.

Fig. 12 A
Fig. 12 A schematic diagram showing the change of composition of the oxygen carrier NMAFO at various stages of the experimental investigation.The availability of the corresponding XRD patterns is denoted at the bottom of the boxes.
55 which Fe/(Fe+Mg) > 0.68, involves three phase transitions: (i) haematite to MgFe 2 O 4 -Fe 3 O 4 solid solution, (ii) MgFe 2 O 4 -Fe 3 O 4 solid solution to MgO-FeO solid solution, and (iii) MgO-FeO solid solution to metallic Fe.These three phase transitions are analogous to the three phases in the reduction of 60 This journal is © The Royal Society of Chemistry 2012

Fig. 13
Fig. 13 Calculated phase diagram of the Mg-Fe-O showing log(pO 2 /bar) versus cationic mole fraction of Fe at 1123 K, at the Fe-rich side.The dashed line marks 30 50probably because (i) the interaction involving Fe-Al-O was prevented by the formation of MgAl 2 O 4 and (ii) the thermodynamic properties of the Mg-Fe-O system were utilised, in which the formation of spinel phase and various solid solutions might have improved the "reducibility" of the 55 Fe-based oxygen carriers.This contrasts with the decaying reactivity seen in the Fe-Al-O system.In the light of the above results , it would also be interesting to investigate the potential use of other elements in group II, e.g.Ca and Sr, as well as other divalent metal oxides, to stabilise a 60 mixture of Fe-Al-O and to promote the reactivity of oxides of Fe in the chemical looping production of hydrogen.In fact, analogous systems, viz.Ni-Fe-O and Co-Fe-O have been studied byAston et al.
75and Fe (NAFO) and (ii) Na, Mg, Al and Fe (NMAFO) were synthesised by means of co-precipitation using a technique involving a gradual reduction in the pH of the solution containing the precursor salts.The major components in the precipitates were Fe(OH) 3 , NaAlCO 3 (OH) 2 (dawsonite), and in 80 the case of NMAFO, Mg 4 Al 2 (OH) 12 CO 3 •3H 2 O (quintinite); the corresponding oxygen carriers were obtained by calcining these precipitates.Such formulations could in theory prevent the formation of FeAl 2 O 4 , by forming instead the thermodynamically more stable MgAl 2 O 4 and NaAlO 2 , which 85 could also act as support materials to stabilise the reactivity of the active Fe 2 O 3 .The spinel FeAl 2 O 4 leads to a progressive diminution in the yield of hydrogen unless the carrier is fully oxidised on each cycle using air.It was found that, the sodium in NAFO was unable to prevent the interaction between Al 2 O 3 90 and iron oxide, whereas in NMAFO, the excess MgO used assisted in inhibiting the interaction.When cycled under 3 These ratios correspond to mass fractions of Fe 2 O 3 of 71.4% and 54.9%, respectively, in NAFO and NMAFO, if neither oxygen carriers contained Na + or any other metal elements.The ratio of [Mg 2+ ]: [Al 3+ ] of 1: 2 for NMAFO corresponds to the stoichiometric formation of MgAl 2 O 4 , ignoring Fe.The total number of moles of NaOH and Na 2 CO 3 95NAFO and NMAFO were 0: 1: 1.49 and 1: 2: 2.16, respectively.

Table 1 .
Results of the elemental analysis of the freshly-prepared oxygen carriers by XRF, on oxide basis and metal basis.Contaminants include oxides of Ca, Cr, and Si. of Mg 2+ , Al 3+ and Fe 3+ .Therefore, the 25 results of the thermogravimetric analysis indicate that both oxygen carriers contain substantial amount of Na + .The compositions of the crystalline phases of the calcined oxygen carriers were investigated by XRD, the patterns of which are plotted in Fig.4.In Fig.4, NAFO was found to 30 contain Fe 2 O 3 and NaAlO 2 .For NMAFO, phases of Fe 2 O 3 (haematite), NaAlO 2 and MgFe 2 O 4 were found.Interestingly, the product of the possible interaction between MgO and Al 2 O 3 , viz.MgAl 2 O 4 , which would generate strong diffraction peaks at e.g.36.9° and 44.9°, was not detected.
x for NAFO and NMAFO were found to be 64.7 and 49.3 wt%, respectively.Both values are lower than those calculated in section 2.1, assuming the oxygen carriers contain only oxides

Table 1 .
From Table 1, it can be seen that the ratio of [Na] and [Al] is very close to unity in NAFO, indicating that almost all Al 3+ precipitated in the form of NaAlCO 3 (OH) 2 .The ratio of Fe: Al Fe 3 O 4 and MgFe 2 O 4 .The contents of Fe 2 O 3 deduced by XRF are very close to those estimated from the TGA in Fig.
45the ratio of [Mg]: [Al ]:[Fe]in the freshly-calcined NMAFO is 1: 1.9: 2.5, which suggests some loss of Mg 2+ and Al 3+ during synthesis.However, the ratio of [Mg]:[Al]in NMAFO remains roughly the same as its nitrate precursor, in which [Mg]: [Al]: [Fe] = 1: 2: 2.16.Apart from the major metal 50 elements viz.Na + , Fe 3+ , Al 3+ and Mg 2+ in the case for NMAFO, XRF also detected some contamination by trace elements such as Si, Cr and Ca.The oxides of these contaminants make up less than 2.8 % of the total mass of the oxygen carriers.These contaminants probably come from 55 impurities in the chemical precursors.Based on the results of XRF, NAFO and NMAFO contain, respectively, 64.3 and 51.7 wt% of Fe 2 O 3 , assuming that the metal oxides are fully oxidised and do not form mixed oxides or solid solutions.Here, a solid solution is considered as a single crystal phase, 60 formed from phases with almost identical crystal structure, e.g.
2 produced, per g of Fe 2 O 3 , for the complete reduction from Fe 2 O 3 to Fe 3 O 4 and from Fe 3 O 4 to Fe, respectively; i N & is the molar flowrate of species i through the 35 packed bed, y i is the measured mole fractions of species i andis the mass fraction of Fe 2 O 3 (assuming all Fe is in the form of Fe 2 O 3 ) in the fully-oxidised sample.Finally, m 0 is the mass in g of the oxygen carrier in its fully-oxidised state and t is time in s, counting from the first instance when CO was fed 40 to the bed in each cycle.In cases where the bed was not oxidised in air, an alternative expression: red .It should be emphasised that, in this study, X is a measure of the oxygen carrier's ability to generate hydrogen, i.e. the yield of hydrogen in each cycle, if steam is 45 used as an oxidant.A value of X equal to unity means that all Fe-containing species can be reduced fully to Fe and then oxidised to Fe 3 O 4 by steam, giving 100% yield of hydrogen.Equation ( 15 ) also assumes that Fe 2 O 3 must be reduced fully to Fe 3 O 4 prior to further reduction to FeO and Fe.
and Kidambi et al.7, who studied oxygen carriers made of Fe 2 O 3 and Al 2 O 3 for chemical looping production of hydrogen and attributed the decaying conversion to the formation of FeAl 2 O 4 .Furthermore, Kidambi et al.7hypothesised two possible reasons for this reduced yield: (i) the presence of the alumina alters the equilibrium for the reduction of iron oxide, making it effectively harder to reduce and (ii) the Fe-Al-O mixture could not prevent sintering over repeated cycles owing to the lack of cyclic phase segregation.In contrast, more 15

Table 2 :
Results of quantitative analysis of EDS spectra taken at points shown in Fig. 9.The compositions of the metal elements are presented in normalised mol%, excluding signals due to C, O and the Pt coating.
5−y Fe y O 2 by itself was not deleterious to the reactivity of the oxygen carrier, and that the Fe 3+ in NaAl 1−y Fe y O 2 could be fully reduced and oxidised in each cycle.Here, it should be remembered that the Fe 3+ in the 35 NaAl 1−y Fe y O 2 phase is stable in CO 2 or H 2 O and cannot be oxidised further.Secondly, Bohn et al.5and Kidambi et al.7showed that the decay in performance of the Fe 2 O 3 -Al 2 O 3 based oxygen carriers can be minimised by oxidation in air at the end of every cycle, an observation reinforced by the results 40 in Fig.6(a).Kidambi et al. 7showed further that the unreactive spinel, FeAl 2 O 4 , formed by the interaction within the Fe-Al-O system, could be eliminated by oxidation with air because, with such treatment, any FeAl 2 O 4 formed decomposed to alumina and haematite, and simple solid solutions thereof, in and discussed below.In NMAFO, MgO has been added to the oxygen carrier such that [Mg]:[Al] = 0.5, a ratio theoretically sufficient to bind all the Al 2 O 3 in the form of MgAl 2 O 4 , even if no sodium aluminate phase forms.The results of XRF, shown in Table 1, indicate that in NMAFO the ratio [Al]:[Na] is >> 1.Therefore, besides NaAlO 2 , there should be at least one more Alcontaining phase present in the freshly-calcined oxygen carrier.
Fe 2 O 3 sequentially to Fe 3 O 4 , FeO and Fe.However, in the presence of Mg 2+ , the phase transitions from haematite to spinel (Fe 3 O 4 and MgFe 2 O 4 ) and then to halite (MgO and FeO) occur at higher values of equilibrium p O 2 than for the corresponding transitions with pure iron oxides.Thus, 5 thermodynamically, the MgO-Fe 2 O 3 mixture could in fact be more readily reducible than pure Fe 2 O 3 .On the other hand, in a commercial chemical looping process, where the reduced oxygen carrier is oxidised by steam, the hydrogen produced by the reduced NMAFO would be more dilute than that by pure O 3 .However, taking into account the dilution by MgO, the theoretical yield of hydrogen per unit mass of fullyoxidised oxygen carrier, for any MgFe 2 O 4 -Fe 2 O 3 mixture, is still lower than that for unmodified Fe 2 O 3 .To investigate the assertions mentioned above, the behaviour of a mixed oxide 20 containing only Mg and Fe, e.g.MgFe 2 O 4 , should be studied in future work.Interestingly, a molar ratio of Mg:Fe:Al = 3.5: 6: 1 could be fruitful in future investigations, because such a composition corresponds to an oxygen carrier consisting of MgFe 2 O 4 (as an active component) and MgAl 2 O 4 (as a found that both 65 NiFe 2 O 4 and CoFe 2 O 4 could be reduced to FeO at higher p CO 2 /p CO than pure Fe 2 O 3 , but at the cost of a higher equilibrium value of p H 2 O /p H 2 during the generation of H 2 .All of these mixed oxide systems have demonstrated the potential of altering the thermodynamics of conventional oxygen 70 carriers by using mixed oxides for chemical looping, a technique which can be engineered to suit a wide range of chemical processes involving redox reactions.
2 O 4 , this resistance was slowly compromised by the gradual replacement of Al 3+ by Fe 3+ in NaAlO 2 , leading to the formation of "free" Al 2 O 3 .This alumina reacted with FeO to form the unreactive FeAl 2 O 4 in NAFO.However, in NMAFO, free alumina was postulated to react preferentially with Mg 2+ to form MgAl 2 O 4 instead of FeAl 2 O 4 spinel.The unique thermodynamic properties of the Mg-Fe-O system might also be responsible for the outstanding performance of the NMAFO oxygen carrier.