The interaction between CuO and Al2O3 and the reactivity of copper aluminates below 1000 °C and their implication on the use of the Cu–Al–O system for oxygen storage and production

Department of Engineering, University of Ca CB2 1PZ, UK. E-mail: wenting.hu@newcast Department of Chemical Engineering and Pembroke Street, Cambridge, CB2 3RA, UK Laboratory of Energy Science and Engineerin Engineering, ETH Zürich, 8092 Zürich, Swit † Electronic supplementary informat 10.1039/c6ra22712k. All data accompanyin within the publication or within the accom at https://www.repository.cam.ac.uk; https: ‡ Present address: Merz Court, School o Materials, Newcastle University, Newcastl Cite this: RSC Adv., 2016, 6, 113016


Introduction
CuO is an attractive candidate for the production of oxygen at temperatures around 900 C owing to its suitable thermodynamic equilibrium as well as the high capacity for the release of gaseous oxygen in The stoichiometric yield of oxygen is about 10 wt% of the CuO. The forward reaction can take place in an oxygen-lean atmosphere, e.g. pure CO 2 or steam, and the gaseous oxygen produced (mixed with CO 2 or steam) used to burn a solid carbonaceous fuel instead of air. As a result, the dried gaseous products of combustion are largely free of non-condensable gases such as N 2 and consist primarily of CO 2 , which can be subsequently sequestrated with minimal treatment and separation. 1,2 The Cu 2 O produced can be oxidised in air to re-form CuO for repeated use. It is possible to arrange for the combustion of the solid fuel to occur in situ with the oxide releasing oxygen in accordance with the forward reaction of (1) and with the regeneration of CuO taking place in a separate reactor free of fuel. Such a scheme is termed chemical-looping with oxygen uncoupling (CLOU) 3 and has been demonstrated in continuous operation with different fuels at various scales. [4][5][6] Alternatively, the combustion can be conducted in a separate combustor, in which case reaction (1) serves as an air separation step. This is known as chemical-looping air separation (CLAS) 7 and replaces the energetically-demanding cryogenic separation of oxygen commonly proposed for oxy-fuel combustion.
In practice, pure CuO sinters at temperatures as low as 800 C (ref. 8) and must be supported on refractory materials with high melting points to retain its reactivity for repeated redox cycles. Al 2 O 3 is a popular choice of refractory because of its abundance and low cost: many copper-based oxygen carriers have been developed with it. [9][10][11][12][13] Depending on the method of synthesis, it is possible for CuO and Al 2 O 3 to form the spinel CuAl 2 O 4 : The spinel is relatively stable to decomposition. It is therefore important to investigate whether the interaction of CuO and Al 2 O 3 can be inhibited at temperatures around 900 C to preserve the capacity for oxygen transfer due to reaction (1). The current understanding of the Cu-Al-O system appears incomplete. Firstly, no consistent conclusion has been reached with regard to the lowest temperature at which the formation of CuAl 2 O 4 becomes appreciable. Secondly, and importantly, it is still unclear whether CuAl 2 O 4 , once formed, can be decomposed readily into the constituent oxides to restore the capacity for oxygen release. The objective of this paper is to address these two issues, with a particular focus on their relevance to chemical-looping processes for oxygen storage and production.
Amongst previous investigations, Jacob and Alcock 14 showed that CuAl 2 O 4 is thermodynamically stable in air at temperatures above $600 C and remains so until at least 1000 C, depending on whether CuO or Al 2 O 3 is in excess. Misra and Chaklader 15 reported the formation of CuAl 2 O 4 from mechanically-mixed ne powders of CuO and abrasive alumina at 800 C but not at 700 C. De Diego et al. 9 impregnated a porous g-alumina substrate with a solution of Cu(NO 3 ) 2 , followed by decomposition of the nitrate at 550 C for 30 minutes and subsequent calcination in air between 550 C and 950 C for 1 hour. The presence of CuAl 2 O 4 was observed in particles calcined at 850 C or above but not in those calcined at <800 C. Interestingly, aer the particles were subjected at 800 C to 100 cycles consisting of reduction to metallic Cu with CH 4 and re-oxidation with air, CuAl 2 O 4 , as well as a-alumina, were detected in all of them regardless of the calcination temperature used to make the particles originally. Imtiaz et al. 12 found that a signicant amount of CuAl 2 O 4 was formed aer co-precipitates of copper and aluminium hydroxides, produced from their respective nitrates and NaOH, were red at 800 C for 2 hours. On the other hand, the freeze-granulated particles produced from powders of CuO and a-alumina studied by Arjmand et al. 11 only formed a small amount of CuAl 2 O 4 aer being calcined at 950 C for 6 hours, despite having a ratio of Cu : Al much closer to the stoichiometry of CuAl 2 O 4 in the starting mixture. Whilst the spinel can be reduced rapidly to metallic Cu and Al 2 O 3 in the presence of reducing gases such as CH 4 and H 2 (ref. 9-11 and 16-18) at temperatures around 900 C, the decomposition of CuAl 2 O 4 in an inert environment is rather slow 11,12 and therefore unsuitable for the production of oxygen. Interestingly, the equilibrium partial pressure of O 2 (P O 2 ) in is 0.014 bar at 900 C, similar to that of reaction (1), 0.016 bar, at the same temperature. Moreover, if excess CuO is present, the production of oxygen can take place via the solid-solid reaction with an equilibrium partial pressure of O 2 of 0.048 bar. 14 Thus the much slower rate of decomposition of CuAl 2 O 4 in comparison to CuO is probably due to chemical kinetics rather than thermodynamics.
Of course, one strategy to avoid interaction between CuO and alumina would be to utilise other types of support. For instance, MgAl 2 O 4 (ref. 11, 12 and 19-21) or calcium aluminates, either in the form of commercial cement 22 or synthesised from pure materials, 23,24 have shown promising performance. However, given that the impregnation of commercial alumina catalyst support is a very convenient method for preparing copper-based carriers, the formation of the spinel merits attention.

Preparation of materials
Materials with a nominal composition of 70 wt% CuO and 30 wt% Al 2 O 3 were prepared by mechanical mixing. An appropriate amount of a-alumina powder (Sigma-Aldrich, $98%), crushed g-alumina pellets (originally 3 mm spheres, Alfa Aesar) or powdered Al(OH) 3 (Sigma-Aldrich, 50-57% on Al 2 O 3 basis) was mixed with CuO powder (Sigma-Aldrich, $98%) in a planetary ball mill (MTI, MSK-SFM-1) operating at 25 Hz for 2 hours to produce a homogeneous mixture. The resulting mixture was calcined in air for 6 hours at a known, constant temperature between 600 C and 1000 C for various samples, in a box furnace. For some materials, in particular those containing aalumina, the milling and calcination process was repeated up to 4 times. These materials are named in a systematic way for easy identication. For instance, a-Al30Cu70-1000-4 denotes the material prepared from a-alumina (gfor g-alumina and hfor Al(OH) 3 ) and calcined at 1000 C, 4 times.
In addition, CuAl 2 O 4 was synthesised as follows. A 1.0 M solution of NH 4 HCO 3 (Fisher Scientic, 99%) was gradually added to 200 mL of a stirred solution containing 0.10 M Cu(NO 3 ) 2 (Sigma-Aldrich, $98%) and 0.20 M Al(NO 3 ) 3 (Fisher Scientic, $98%) until pH 4.0 was reached. The resulting gel was dried at 80 C overnight and calcined in air at 1000 C for 6 hours in a box furnace. The solid obtained was subsequently crushed and sieved to <50 mm and 50-100 mm for phase iden-tication and thermogravimetric (TG) analysis, respectively.

Characterisation of materials
The phases present in the prepared materials were identied from X-ray powder diffraction (XRD; Empyrean PANalytical, Cu-K a , 40 kV, 40 mA) in the range 2q (tetragonal g-alumina), ICSD-82504 (q-alumina) and ICSD-31545 (a-alumina). The reactivity of various materials was characterised using a thermogravimetric analyser (TGA; Mettler Toledo, TGA/DSC1). The TGA furnace was always purged by two streams of Ar (BOC, >99.998%) at a total owrate of 100 mL min À1 in addition to a stream of reactive gas with a owrate of $50 mL min À1 . The owrates are as measured at 20 C and atmospheric pressure, and this applies to all owrates quoted in the following text, unless specied otherwise. The reactive gas could be switched between air (BOC, >99.995%), N 2 (BOC, >99.998%) or 5 vol% H 2 balanced by N 2 (BOC, >99.999%; referred to simply as H 2 hereaer) using a solenoid valve manifold (Bürkert, type 6011) controlled from a separate program synchronised with the TGA.
Two types of TG experiments were undertaken on fresh materials. In the rst, to quantify the total amount of Cu (on a CuO basis) and CuAl 2 O 4 present in the prepared materials, approximately 20 mg of powdered sample was subjected to isothermal reaction at 900 C with a reactive gas sequence air-N 2 -H 2 -air, each lasting 15 minutes. In the second, to investigate the reactivity of pure CuAl 2 O 4 , approximately 20 mg of the synthesised samples were held at 1000 C, initially in air. Once the temperature and the mass of the sample had stabilised, the atmosphere was switched to a mixture of air and N 2 , resulting in a partial pressure of O 2 ranging from 0-0.02 bar for 50 minutes. It should be noted that when no air was used, the partial pressure of O 2 in the TGA furnace was <5 Â 10 À4 bar. The exact partial pressure of O 2 was determined by measuring the equilibrium temperature of the reversible reaction (1) under the same atmosphere as that used in each experiment with CuAl 2 O 4 . The stabilised CuO-based particles have been characterised in previous work. 24 Some samples of CuAl 2 O 4 with higher masses, $60 mg each, were treated using a combination of the above protocols (with some variations in duration or gas composition, when necessary) to investigate the redox behaviour of the material further. Identication of the solid phases present in some of the recovered samples was also performed using XRD. The details of the changes made to the TG protocols will be specied along with the results in the next section.

The effect of different alumina precursors on the formation of CuAl 2 O 4
The formation of CuAl 2 O 4 was found to occur at different temperatures, depending on the form of alumina (or precursor) used. a-Alumina did not react appreciably with CuO at 900 C, even aer four repetitions of milling and calcination for 6 hours on the same sample, as conrmed by the absence of peaks belonging to CuAl 2 O 4 in the XRD diffractograms shown in Fig. 1. However, when calcined in air at the higher temperature of 1000 C for 6 hours, a signicant amount of CuAl 2 O 4 was formed and further milling and calcining led to the formation of CuAlO 2 with almost all the free alumina consumed by the 4 th calcination. When calcined at 950 C, a small amount of CuAl 2 O 4 but no CuAlO 2 was detected in the material. This observation is in agreement with the nding of Jacob and Alcock that in an atmosphere with P O 2 ¼ 0.21 bar, reaction (4) has an equilibrium temperature of 1003 C. 14 In contrast, Fig. 2 shows that, aer heating in air at 850 C for 6 hours, the formation of CuAl 2 O 4 from g-alumina and CuO was signicant. Since g-alumina is only weakly reecting and therefore difficult to quantify by XRD, the conversion of alumina to CuAl 2 O 4 was determined by measuring the changes in mass of a sample in a TGA at 900 C, exposed successively to air, N 2 , H 2 and nally air. As an example, the result for g-Al30Cu70-900-1 is shown in Fig. 3. It can be seen that the CLOU capacity of g-Al30Cu70-900-1 was $5.3 wt% and the total oxygen capacity of the material was 14.5 wt%, corresponding to a Cu content of 72.5 wt% on a CuO basis, instead of the expected value of 70 wt%. The difference was probably owing to adsorbed moisture in the galumina used during the initial preparation. During the decomposition under N 2 , the material exhibited a fast reaction stage between 900 and 1300 s, followed by a very slow stage until the atmosphere was switched to H 2 . As will be discussed in Section 3.2., the reduction of CuAl 2 O 4 to CuAlO 2 in an inert environment via reaction (3) is extremely slow. Thus it is plausible that the rst  stage is dominated by reaction (1) and the second stage by the slow reaction (3). In fact, the chemical kinetics of decomposition of CuO in N 2 is faster than that being measured in this experiment 24 and here the rate is primarily limited by mass transfer in the TGA. A more detailed discussion is presented in the ESI. † The amount of free CuO present in g-Al30Cu70-900-1, which amounts to 52.8 wt%, can be estimated from the mass change during the fast stage. Assuming all remaining Cu is bound in the form of CuAl 2 O 4 , 91.8% of the g-alumina present initially would have had to have been consumed during the calcination when the material was rst prepared. The same analysis carried out on a sample of a-Al30Cu70-950-4 showed that only 2.4% of the a-alumina had reacted during its preparation, albeit the material having been calcined at a higher temperature of 950 C for a period 4 times longer.
Amongst the three precursors investigated, Al(OH) 3 was found to be the most reactive, capable of forming CuAl 2 O 4 at temperatures as low as 700 C, as seen in Fig. 2. It should be noted that the Al(OH) 3 used in this work was amorphous, as conrmed by XRD (shown in Fig. S1 in the ESI †), as was the sample dehydrated at 600 C for 24 hours. A small amount of galumina was formed aer Al(OH) 3 was calcined in air at 700 C for 6 hours. Further characterisation of the Al(OH) 3 sample was undertaken using temperature programmed decomposition (TPD) in air in the TGA, with a heating rate of 2 C min À1 . The result is shown in Fig. 4. It can be seen that the hydroxide gradually lost H 2 O until $600 C and a marked decomposition occurred around 800 C. It has been reported that amorphous alumina transforms to g-alumina at 802 C and subsequently to a-alumina at 1087 C. 25 In fact, when a sample of Al(OH) 3 was calcined at 900 C in air for 6 hours, the resulting material was conrmed as g-alumina (as shown in Fig. S1 in the ESI †). These results suggest that amorphous alumina is able to retain some OH groups up to 800 C, which are lost when the phase transition to g-alumina occurs.  Fig. S2 of the ESI, † was found to be slow and was not complete even aer 20 hours. On the other hand, a similar degree of conversion was achieved in 50 minutes when the material was decomposed at 1000 C, and the results are given in Fig. S3 of the ESI. † Based on these limited data of the isothermal decomposition, it appears that a high activation energy is associated with the reaction. Assuming the decomposition is due to reaction (3) alone, an apparent activation energy of the reaction can be estimated knowing the rate of change of conversion of the solid at various iso-conversion points (the procedure is briey described in the ESI †). These values are plotted in Fig. S4 of the ESI, † up to a conversion of 0.8. The values were found to be fairly constant at $370 kJ mol À1 between conversions of 0.25 and 0.75. At lower values of conversion, the apparent activation energy was much Fig. 3 TG curve of the isothermal reaction of g-Al30Cu70-900-1 in different atmospheres at 900 C. The abrupt changes in mass seen between different segments were due to gas switching, disturbing the microbalance.  lower. It may be possible that in the early stage of the decomposition, the rate-determining step is different from that at a later stage but this cannot be ascertained with the current data and further investigation is beyond the scope of this paper. It is worth noting that for a gas-solid reaction like reaction (3), the enthalpy of reaction is a measure of how fast the thermodynamic driving force changes with temperature in an inert gas atmosphere. Here, the apparent activation energy was found to be much higher than the enthalpy of reaction, $140 kJ mol À1 , 14 suggesting that chemical kinetics has a larger effect than thermodynamics on the observed rate of reaction with changing temperature.
Examination of the reduced sample (recovered aer decomposition at 900 C and cooled in Ar) revealed CuAlO 2 , aalumina and g-alumina as the main phases present as seen in Fig. 5. Cu 2 O was not detected in the reduced sample, contrary to the ndings of Arjmand et al., 18 who used CH 4 instead of an inert gas. Interestingly, the reections from g-alumina are more consistent with a tetragonal structure, evident from the splitting of the peak around 2q ¼ 46 , instead of the commonly-adopted cubic structure. 27 The phase obtained from the thermal decomposition of the synthetic CuAl 2 O 4 also appears to be more crystalline than the as-received g-alumina, or that derived from amorphous alumina, the XRD diffractograms of which, given respectively in Fig. 2, and in Fig. S1 of the ESI, † contain much broader peaks. The amounts of various species present in the reduced sample were estimated by quantitative analysis of the XRD diffractogram and the results are shown in Table 1. The results are in good agreement with an independent estimate made from the thermogravimetric analysis, assuming that the unreacted sample consisted of pure CuAl 2 O 4 and the mass loss was due to the evolution of gaseous O 2 according to reaction (3). The fact that a signicant amount of g-alumina was formed from the decomposition of CuAl 2 O 4 raises questions about the thermodynamics of reaction (3) because its standard Gibbs free energy of reaction at 900 C is reported as 41.5 AE 1.3 kJ mol À1 (ref. 14) whereas the Gibbs free energy of transformation from g-alumina to a-alumina is approximately À14 kJ mol À1 at the same temperature. 28 Consequently, the equilibrium P O 2 of reaction (3) would be rather different depending on the polymorph of the product alumina.
The results in Fig. 5 can be explained by two possible reaction schemes. In the rst, CuAl 2 O 4 decomposes to both polymorphs of alumina in parallel, each having a different equilibrium P O 2 ; in the second, CuAl 2 O 4 decomposes to g-alumina rst, which then transforms to a-alumina. Further investigations were carried out to distinguish between the two schemes and the results are presented in Fig. 6. From the gure, it can be seen that the initial mass loss (over 160 s) of CuAl 2 O 4 when decomposed at 1000 C in P O 2 up to 0.02 bar showed two distinct linear regimes. The change of the average rate of mass loss with respect to P O 2 was signicantly faster when P O 2 < 0.009 bar. This result supports the hypothesis that the decomposition of CuAl 2 O 4 to the two forms of alumina occurs in parallel. Had the reaction followed the sequence CuAl 2 O 4 / g-alumina / a-alumina, one would not expect such sudden change in the rate of decomposition with respect to P O 2 . As the decomposition proceeded rather slowly, external mass transfer did not have a signicant inuence on the measured rate of reaction: the estimated difference of P O 2 between the surface of the samples and that above the crucible holding the sample (4.9 mm diameter and 4 mm deep) using the fastest rate measured ($0.135 mg over 160 s, as shown in Fig. 6) was approximately 1.5 Â 10 À5 bar. Thus, assuming that the rate of decomposition of CuAl 2 O 4 was proportional to the difference between P O 2 of the gas environment and the equilibrium value, the equilibrium P O 2 of the two parallel reactions can be estimated from the intersection between the two regimes, 8.7 Â 10 À3 bar, and that between the slow regime and the abscissa, 0.034 bar. The difference in the equilibrium P O 2 amounts to a difference in the Gibbs free energy of 14.4 kJ mol À1 of reaction (3), assuming the activities of the solids are unity. However, this value is only $60% of the Gibbs free energy of transformation from g-alumina to a-alumina reported in the literature. 28 The discrepancy could be due to the tetragonal deformation of the g-alumina having a lower Gibbs free energy of formation. In fact, at the temperature concerned, it is possible for g-alumina to start transforming to other transition aluminas, 29 which is probably why a tetragonal deformation, rather than the more common cubic structure, was observed from the decomposition of CuAl 2 O 4 . On a side note, using the value of 0.034 bar, the Gibbs free energy of reaction (3) with a-alumina as the product is estimated as 35.6 kJ mol À1 , slightly higher than the value of 33.0 kJ mol À1 , given by Jacob and Alcock, 14 although the corresponding difference in P O 2 is more than 20%. In addition, two samples of CuAl 2 O 4 were decomposed at 1000 C in the TGA for 3 hours, one with P O 2 ¼ 9.0 Â 10 À3 bar, just above the transition point seen in Fig. 6, and the other in a mixture of N 2 and Ar (where P O 2 < 5 Â 10 À4 bar due to a small leakage of air into the TGA). The XRD diffractograms of the decomposed samples, given in Fig. 7, revealed that in the former case, some CuAl 2 O 4 remained but no g-alumina was detected, whereas the converse is true for the latter. These observations also suggest that the decomposition of CuAl 2 O 4 to different phases of alumina probably occurs in parallel rather than in series.

The oxidation and reduction of CuAlO 2
There are two potential pathways for the oxidation of CuAlO 2 , i.e. the backward reactions of (3) and (4). The equilibrium P O 2 of reaction (4) is signicantly higher than that of reaction (3) in the temperature range of interest, between 900 C and 1000 C. 14 Therefore the two pathways can be examined separately by choosing an appropriate P O 2 so that only the backward reaction of (3) is thermodynamically feasible. Here, a sample of CuAl 2 O 4 was decomposed in a mixture of N 2 and Ar at 1000 C for 50 minutes, followed by oxidation in an atmosphere with P O 2 z 0.10 bar (the equilibrium P O 2 of the reverse reaction (3) is $0.04 bar with a-alumina as the reactant and that of reaction (4) is above 0.9 bar (ref. 14)) at the same temperature for a further 50 minutes and the result is shown in Fig. S6 of the ESI. † It was found that the oxidation of CuAlO 2 via reaction (3) was much slower than the forward decomposition reaction, despite having a slightly higher thermodynamic driving force. The kinetics of the backward reaction (4) were not investigated in this work, as Arjmand et al. have shown previously that complete oxidation of CuAlO 2 via this route was only achieved aer prolonged heating in air. 18 It appears that both routes for the oxidation of CuAlO 2 are slow and the material does not contribute to the oxygen carrying capacity signicantly, bearing in mind that many chemical-looping processes oxidise the oxygen carrier in a circulating uidised bed, which normally has a low residence time.
On the other hand, the reduction of CuAlO 2 in H 2 was fast. For instance, a sample of CuAl 2 O 4 rst decomposed in an atmosphere with P O 2 ¼ 9.0 Â 10 À3 bar for 10 hours at 1000 C was reduced completely in 5.5 minutes once the reactive gas was switched to 5% H 2 . The average rate of loss of mass during this time was approximately 0.6 mg min À1 . This is equivalent to 38 mmol min À1 of oxygen atoms, and is comparable to the rate of transfer of H 2 to the solid estimated in previous work for the same experimental arrangement, i.e. the reaction was limited by mass transfer. 23 XRD analysis of the reduced sample showed that the phases present were metallic Cu, q-alumina and aalumina, as annotated in Fig. 8. Several studies [30][31][32] suggest that the reduction of CuAlO 2 at around 1000 C results in interspersed Cu and q-alumina without any a-alumina being formed. The a-alumina present in this work is probably the result of the decomposition of CuAl 2 O 4 only and this is supported by a subsequent TG experiment, which is shown in Fig. 9 and will be discussed later. However, no attempt was made to synthesise pure CuAlO 2 to conrm this, it not being the primary focus of the current work.

Discussion
The results presented in this work concur with previous ndings that the generation of gas phase oxygen by the decomposition of the spinel phase, CuAl 2 O 4 , is slow up to 1000 C (ref. 11 and 12) and therefore unsuitable for oxygen production schemes such as CLOU or CLAS. Thus, the formation of the spinel must be prevented to retain the high reactivity of CuO with regard to oxygen release. Since the typical operating temperature for the CuO/Cu 2 O system is around 900 C, it is viable to use a-alumina as a support material but not amorphous alumina or g-alumina, which react with CuO at much lower temperatures. Usually the latter two are preferred as Fig. 7 XRD diffractograms of CuAl 2 O 4 decomposed at 1000 C in different atmospheres. Only a section between 2q ¼ 40 and 50 is shown to distinguish the peak due to CuAl 2 O 3 at 45.2 and that due to g-Al 2 O 3 at 45.6 . The reflection peaks are annotated as a-Al 2 O 3 (À), CuAl 2 O 4 (B), CuAlO 2 (Â), g-Al 2 O 3 (>). A separate graph containing the diffractograms between 2q ¼ 10 and 80 can be found in the ESI (Fig. S5 †). support materials since they possess much higher internal surface areas compared with a-alumina. However, because the operating temperature is sufficiently high, a high surface area is not necessary to secure a high reaction rate and this has been demonstrated in previous work for various metal oxide oxygen carriers. 6,23,33 On the other hand, when high surface area is desired, e.g. for catalysis at intermediate or low temperatures, amorphous alumina and g-alumina can be used as long as the operating temperature is below 600 C and 700 C, respectively, where the formation of CuAl 2 O 4 from CuO and respective alumina does not occur, as seen in Fig. 2. In any case, alumina supports should be avoided when the operating temperature is constantly above 950 C since even the relatively-stable aalumina starts to form CuAl 2 O 4 at this temperature.
The fact that CuAl 2 O 4 decomposes to different polymorphs of alumina depending on the P O 2 of the environment can be exploited to regenerate the active CuO phase should the spinel be accidentally formed. The scheme is proposed as follows: rst, the material containing the spinel phase would be maintained in an atmosphere with appropriate P O 2 (e.g. slightly above 9.0 Â 10 À3 bar if treated at 1000 C, the transition point shown in Fig. 6) to slowly decompose the spinel into a-alumina and the delafossite phase, CuAlO 2 , while avoiding the formation of galumina. Once the decomposition is complete, the material would be further reduced by a fuel gas (e.g. CH 4 or syngas) at 900 C or below, to minimise the sintering of the metallic Cu formed (the melting point of Cu is lower than its oxides, or the aluminates, at 1084 C). The reduced material would then be reoxidised to regenerate the CuO phase, preferably at low temperature and P O 2 , e.g. $700 C and 0.01 bar, because particles containing Cu agglomerate easily when being oxidised in air at high temperatures. 9 To demonstrate the effectiveness of this procedure, a comparative study of CuAl 2 O 4 with different pre-treatments was performed in the TGA and the results are shown in Fig. 9. It can be seen from Fig. 9 that as-synthesised CuAl 2 O 4 (sample a) released a limited amount of oxygen in the rst 15 minutes of the experiment and reducing the sample in H 2 and re-oxidised once (sample b) did not bring signicant improvement. However, the sample decomposed to CuAlO 2 before being reduced in H 2 (sample c) was able to release much more oxygen during the inert stage. All three samples were capable of being reduced completely in H 2 and re-oxidised fully in air within minutes. The rates of reduction and oxidation were again limited by external mass transfer and did not reect the reactivity of the materials. The slightly slower rates of reduction and oxidation observed for (sample c) were due to a slight increase in the ow rate of the purging Ar during this particular experiment, resulting in a higher dilution of the H 2 , rather than the deactivation of the material. This was conrmed by identical rates of decomposition of the CuO observed for all materials during the decomposition in N 2 , which is not affected by dilution of gases. Using the same analysis as undertaken in Fig. 3, it was estimated that approximately 6% of the Cu was present as CuO in sample (a) and the corresponding percentages in samples (b) and (c) were 23% and 55%, respectively. Assuming the additional CuO present in sample (c) arose entirely from the formation of a-alumina (which does not reform CuAl 2 O 4 on subsequent oxidation), and that the decomposition of CuAl 2 O 4 to CuAlO 2 produces a-alumina exclusively, the expected amount of CuO in sample (c) would be 53% of the total Cu, very close to the value measured experimentally. In conjunction with results shown in Fig. 8, it can be inferred that the alumina formed from the reduction of CuAlO 2 should be mostly, if not all, composed of the q-phase. Furthermore, q-alumina was able to react with the interspersed Cu to form CuAl 2 O 4 directly on oxidation without forming CuAlO 2 as an intermediate-otherwise the oxidation would not have been complete. Owing to the presence of q-alumina, it would not be possible to recover all the Cu as CuO following the proposed protocol but 50% of the Cu bound in CuAl 2 O 4 could be recovered and if repeated several times, most of the Cu might be re-activated into the CuO phase.   H 2 and air). The samples were kept in air before the start of the experiment to maintain a fully oxidised state. The abrupt changes in mass seen between different segments were due to gas switching, disturbing the microbalance. The dotted and dashed horizontal lines indicate, respectively, the theoretical mass change if all the Cu 2+ present is reduced to Cu + and metallic Cu. The 3 samples were: (a) CuAl 2 O 4 as synthesised (60.89 mg), (b) CuAl 2 O 4 fully reduced in H 2 and re-oxidised in air at 900 C prior to the experiment (61.37 mg) and (c) CuAl 2 O 4 first decomposed in 0.90% O 2 at 1000 C for 10 hours, then reduced by H 2 (i.e. the same as in Fig. 8) followed by re-oxidation at 900 C prior to experiment (61.11 mg).
The series of interactions between CuO and alumina investigated in this work is summarised in Fig. 10.

Conclusion
The formation of CuAl 2 O 4 from CuO and alumina was investigated, together with its decomposition in an oxygen-lean environment. The oxidation and reduction characteristics of the related copper aluminate, CuAlO 2 , was also studied. It was found that (1) The solid-solid reaction between CuO and different polymorphs of alumina occurs at different temperatures. The ease of reaction follows the order amorphous alumina > galumina > a-alumina.
(2) It is possible to support CuO with a-alumina for oxygen storage and production up to 950 C without forming the undesired CuAl 2 O 4 phase, whereas aluminium hydroxide (and its derivative amorphous alumina), g-alumina and q-alumina are unsuitable for this purpose.
(3) CuAl 2 O 4 can decompose slowly to form both a-alumina and g-alumina. By controlling the P O 2 of the environment, it is possible to bias the selectivity towards 100% a-alumina so that on subsequent oxidation of the material, CuO could be formed instead of CuAl 2 O 4 .
(4) CuAlO 2 resists oxidation in air but can be easily reduced in H 2 . q-Alumina is formed in the reduction process.
(5) Based on the reactivity of CuAl 2 O 4 and CuAlO 2 , a possible procedure for the regeneration of CuO from copper aluminates is proposed, which could recover up to 50% of the copper originally bound in the form of CuAl 2 O 4 in a single treatment.