The Effects of Secondary Oxides on Copper‐Based Catalysts for Green Methanol Synthesis

Abstract Catalysts for methanol synthesis from CO2 and H2 have been produced by two main methods: co‐precipitation and supercritical anti‐solvent (SAS) precipitation. These two methods are compared, along with the behaviour of copper supported on Zn, Mg, Mn, and Ce oxides. Although the SAS method produces initially active material with high Cu specific surface area, they appear to be unstable during reaction losing significant amounts of surface area and hence activity. The CuZn catalysts prepared by co‐precipitation, however, showed much greater thermal and reactive stability than the other materials. There appeared to be the usual near‐linear dependence of activity upon Cu specific area, though the initial performance relationship was different from that post‐reaction, after some loss of surface area. The formation of the malachite precursor, as reported before, is important for good activity and stability, whereas if copper oxides are formed during the synthesis and ageing process, then a detrimental effect on these properties is seen.


Introduction
In recent years the continuingr ise in atmospheric anthropogenic carbon and the ensuing effects that this engenderso n the climateh ave led to increasing efforts towards capture, sequestration, and utilisation of carbon dioxide. Given ag eneral societal trend towards greener energy and fuel sources,t he utilisation of CO 2 as an abundant carbons ource has become more and more relevant. The hydrogenationofCO 2 to produce methanol for use as both af uel and ac hemical precursor is one possible route to this goal. The concept of ac ycle using CO 2 and methanoli ns uch am anner is generally attributed to Olah et al., [1] and is referred to as the anthropogenicc arbon cycle. One of the attractive facets of this method is the possibility of at ruly green fuel;C O 2 can be captured from sources such as powers tations and combined with hydrogen generated from ar enewable source, such as electrolysis of water where the electricityi ss uppliedf rom solar power. The synthesis of green methanol in this way can be regarded as aw ay of storing H 2 ,e ffectively for storing renewable energy chemically, in addition to the productiono fafuel.
Global methanol production is in the region of 80 Mt per annum. [2] Industrially,m ethanol is produced fromamixture of carbon monoxide, carbon dioxide, and hydrogen (syngas) at elevated pressures and moderate temperatures. It hasb een shownt hat carbon dioxide is the carbon source at the molecular scale, producing methanola nd water.C arbon monoxide is present to convert the water produced into CO 2 and H 2 via the water-gas shift reaction. These reactions are represented by Equations (1) and (2): ðmethanols ynthesisÞð 1Þ CO 2 þ H 2 ! H 2 O þ CO ðreverse water-gas shiftÞð 2Þ The current industrial synthesis is from as yngas, deriving from fossil fuels, which has am ix of CO and CO 2, but the main synthesis route is as in Equation (3), with little water production: However,t he reaction studied herein differs from this system in that there is no CO present, since we are testing the possibility of using recaptured CO 2 andr enewable hydrogen, as opposed to fossil fuelg enerated syngas. This meanst hat without CO being presentt oe nable overall water-gas shift, our system will contain as ignificantly higher proportion of water vapour than as ystem running syngas.
The catalysts used for this reaction are composed of copper, zinc, and alumina, and are based on the catalysts originally designedb yI CI during the 1960s. [3,4] The optimisation of these catalysts was performed long before modern techniquesm ade it possible to understand the fundamentals of the actives ites and reactionm echanism. In recent years there have been increasingn umbers of studies into these fundamentala spects of Catalystsf or methanols ynthesis from CO 2 and H 2 have been produced by two main methods:c o-precipitationa nd supercritical anti-solvent( SAS) precipitation.T hese two methods are compared, along with the behaviour of coppers upported on Zn, Mg, Mn, and Ce oxides. Although the SAS methodp roduces initially active material with high Cu specific surfacea rea, they appear to be unstable during reactionl osing significant amountso fs urface area and hence activity. The CuZn catalysts prepared by co-precipitation, however, showed much greater thermala nd reactive stability than the other materials. There appeared to be theu sual near-linear dependenceo fa ctivity upon Cu specific area, though the initial performance relationship was different from that post-reaction, after some loss of surfacea rea. The formationo ft he malachite precursor,a sr eported before,i si mportant for good activity and stability, whereas if coppero xides are formedd uring the synthesis and ageing process,t hen ad etrimental effect on these properties is seen. this catalysis, [5,6] with the focus generally tending towards the simpler binary system of Cu/ZnO. From this has risen the consensus that the methanol productivity is strongly correlated with the specific coppers urface area of ac atalyst, [7,8] and that other factors such as the oxidation state of the copper, [9] and the copper-zinc interaction, [10,11] also have an effect.
The optimised industrial catalysti sg ranted excellent copper surfacea rea by way of the structure of the precursor material from which it is derived. Such catalysts are synthesised by coprecipitation of metal nitrate salts with sodium carbonate to produce ah ydroxycarbonatep recursor phase, which is then calcined to form CuO and ZnO. It is known that the most active catalysts are formed from ap recursor consisting predominantly of zinc-substituted malachite phases.W hen properly prepared, the specific structure of the final catalyst is defined by this precursor phase, leading to am aterial with the desired high coppers urface area and good copper-zinc interaction. [12] Whilst copperc omprises the activem etal in the catalysts, the role of the zinc is less clear, [13] and as such there have been many investigations into copper-based catalysts with alternative secondary metal oxidess uch as magnesia, [14] ceria, [15,16] and zirconia. [17,18] Of particular note are studies that show Cu/ MgO catalysts as having higher coppers urface areas,a nd yet having lower methanola ctivity, [14] which runs counter to the accepteds tance of copper surface area being directly linked to such activity.T he issue seems to stem from the nature of the standard catalyst, the synthesis of which has been highly optimised in terms of pH, temperature and ageing times. Such catalysts are generally precipitated in the range of pH 6-7, and have been shown to lose activity if precipitateda th igher pH ranges. [12] However,t he precipitation of magnesiumn itrate requires ap Hi nt he region of 9. Thus it is difficultt od econvolute whether the negative effect on activity is an artefact of the substituted oxide or the pH of synthesis.
Supercriticala nti-solvent( SAS) precipitation presentsa ni nteresting wayt oa pproacht his problem,a st he procedure rapidly precipitates materialw ithout the need for ab ase. It has been shown that this methodc an produce copper-zincc atalysts with high copper surface area, and that these catalysts are active for methanols ynthesis and water-gas shift reactions. [19] Aw ide range of materials can be precipitated in this manner,i na ll cases without the requirement of as pecific precipitating agent. Thisa llows us to sidestep the need for specific pH ranges found in co-precipitation, allowing us to remove it as af actor.
In this study we report the changes in methanol synthesis activity for copperc atalysts synthesised with variouss econdary oxides using both co-precipitation and SAS techniques. The changes were monitored through reactivity measurements, and through assessment of the copper surface area and particle size both before and after exposure to reactionc onditions. Through this we hope to discover what factors result in activity loss in the co-precipitated catalysts, and to investigate which metal oxidesa re capable of producing active, stable catalysts when the negative effects of high-pH co-precipitation are removed.

Results and Discussion
This study was conducted using Cu/M x O y catalysts, where M = Mg, Zn, Mn, or Ce. Based on the results of previouss tudies am olar ratio of 70:30 betweenc opper and the secondary oxide was chosen. This amount has been shown to be close to the limit of incorporation of zinc into am alachite structure, [20] and as such wasu sed as as tandard to whicht he other catalysts were held.

Co-precipitated catalysts
The catalysts werep repareda sd escribed in the Experimental Section. The surface areas of the materials produced are shown in Ta ble 1.
X-ray diffraction (XRD)o ft he precursor phases revealed patterns consistentw ith those of malachite [21] in the cases of all CuZn-CP catalysts, as shown in Figure 1. The other precursors showed broader diffraction peaks. CuMg-CP and CuCe-CP show two major diffraction peaks consistent with CuO, with CuMg-CP having low,b road peaks consistent with malachite. The small peak at 338 in CuCe-CP may be residual malachite, and this material also shows broad peaks at angles consistent with CeO 2 .B ased on these observations it would seem that the hydroxycarbonate phase is formed initially in all cases, as evidenced by the blue coloured materialo ften reportedi n  such cases. The colourc hange towards green in the CuZn and CuMn catalysts can be attributed to the formationo ft he malachite phase, whereas the darkening in colour of the CuMg and CuCe catalysts can be attributed to the formationo fc opper oxide phases. This cannot solely be attributedt ot he effects of the pH, as the CuZn catalysts prepared at higher pH do not show these phases.T herefore, this would seem to be an effect of the oxide, although possibly this is in combination with the elevated pH. The work of Fujita et al. [22,23] showed that ac alcination temperatureo f3 30 to 350 8Ci ss ufficient to form the final catalysts. Based on this, all catalysts were calcined at 330 8Cf or 3h in flowing air,w ith at hermalr amp rate of 5 8Cmin À1 .T he catalysts were uniformly brown after this calcination step, with the exception of CuCe-CP,w hich presented aslightly grey hue.
XRD of the calcined catalysts gave similarp atterns for all the materials except for CuCe-CP ( Figure 2). All had peaks consistent with coppero xide, but with the CuCe-CP material having additional peaks consistentw ith CeO 2 .W ith the exceptiono f the appearance of as mall, sharpp eak at 368,t he precursor and calcined versions of CuCe-CP are very similar.T he precursor and calcined versionso fC uMg-CP are also highlys imilar, with the calcined version losing the small, broad peaks associated with malachite. This is consistentw ith coppero xide already being formed during drying in these materials. The surface areas of the materials after calcination decreased by less than 10 %f rom the values found in the precursors.
The catalyst samples were tested for methanols ynthesis as describedi nt he Experimental Section, andt he only significant products seen were CO and methanol. The CuZn catalysts generally appeared to undergo as trong initial deactivation,b ut were stable after 3h.T he other secondary oxides displayed varying behaviour,w hich will be discussed below.O ft he coprecipitated catalysts, copper-zinc showedt he highest activity towardsm ethanol production, and also showedt he lowest amount of deactivation (Table 2). At rend amongst the copper-zinc catalyst was also evident;i ncreasing pH lowered CO 2 conversion and methanol selectivity,w hile those at lower pH preparation tendedt os how al ower degree of deactivation. CuZn 6.5-CP lost only 5% CO 2 conversion from 1-8 h, compared with 7% for CuZn 9-CP and 10 %f or CuZn1 0-CP. CuMn-CP and CuMg-CP hads imilar,i fn ot higher,c opper surface areas than the CuZn-CP catalysts before reaction, but were not as active. CuMg-CPg ave good CO 2 conversion and excellent methanol selectivity,b ut continued to deactivate after 3h,s tabilising after 6h.C uMn-CP appears to have gained activity over time, but af ullt ime on-line reaction study showedaslightly more complicated effect. CuMn-CP started with the low activity shown above, and appeared to immediately startd eactivating. However, after about 2h,i tb egan to show am arked increase in activity over the next hour,w ith both CO 2 conversion andm ethanols electivity rising rapidlyt o approximately6 .3 %c onversion and 63 %s electivity.I tm aintained thisa ctivity for about 2hbefore undergoing ar apid deactivation.O fp articular interesti st hat the overall CO production rate changed only slightly during this time, implying that the increased CO 2 conversion was primarily drivenb yalarge increasei nm ethanols electivity,a nd that the deactivation occurredi nt he reverse manner.T his would seem to indicate that species or active sites are briefly formed on CuMn catalysts that are highly active, but highly unstable. Ar epeat of this test over al ongert ime period (16 h) showed that the deactivation continued beyond 8h,with the materialhaving apparently stabiliseda fter about 11 h, at which time it displayed CO 2 conversion of < 1%.C uCe-CP deactivated steadily,s tabilisingo nly in the final hour of testing. Whilst it was the only catalystt oi ncrease methanol selectivity steadily,i td oes not seem to be av iable catalystdue to high deactivation and low activity.
The relationship of the coppers urface areas to activity is interesting ( Figure 3a nd Table 2). Initial coppers urface areas appear to match trends in the activity quite well, with the notable exception of CuMg-CP.T his catalyst possesses higher coppers urface area than any of the others, and yet has lower activity.H owever,i fo ne considers the post-reaction copper surfacea reas, there is am ore evident trend. Here, the higher surfacea reas correspond to highera ctivities with the exception of the CuMn-CP sample. However,t he CuMn-CP catalyst was observed to be deactivating rapidly at the termination of the reaction, and the depressurising and cooling steps before recovery of the catalyst take approximately an hour.I tm ay be that the CuMn-CP catalyst continuedt od eactivate through  www.chemcatchem.org this time. It appears that, althoughc onversion has generally diminished, the intrinsic per site activity has increased after 8h running,e videnti nt he data of Figure 3. It is likely that this is due to am orphologyc hange of the Cu particles, perhaps such that the CuÀZnO interaction is not lost as much as the Cu surface area. Thus, the value of the copper-zinc catalysts would appear to be ac ombination of high initial copper surface area and their ability to better retain this surface aread uring reaction conditions.

Supercritical anti-solvent (SAS) precipitation catalysts
Supercritical anti-solventp recipitations were carriedo ut as described below and produced very fine powder,w hich was either blue or green depending on the secondary oxide (Table 3).
XRD analysiso nt he precursor phases showed them to be highly amorphous/nanoparticulate (Figure 4). It is difficult to thus draw any conclusions about the materials formed, but these observations are in line with those reported for the syn-thesis of supercritically preparedg eorgeite [19,24] It is possible that the other oxidesf orm similar amorphous materials as an effect of the extremelyr apid precipitation step found in SAS precipitations.
XRD analysis on the calcinedS AS materials ( Figure 5) showed an umber of similarities to the CP materials. CuCe showsas mall diffraction peak in the region of copper(II) oxide, and shows broad reflectionsc onsistent with the presence of CeO 2 .T he other materials all appear to be copper oxide,a s was observed in the CP materials. However,w hereas the CP materials all displayed ad istinctive doublep eak, the CuZnand CuMg-SAS showasingle, broader reflection. This is indicative of smaller crystal domains in the material, which is likely to be an effect of the highly amorphous precursor being unablet og enerate long-rangeo rder upon calcination. Unlike the co-precipitated catalysts, the SAS catalysts displayed af ar more significant loss of surfacea rea upon calcination, with all    losses being in the region of 30 %. An otable exceptiont ot his is CuMg-SAS, which lost less than 5%. Catalytict esting and coppers urfacea rea measurements were carried out in an identicalm anner to those described for the co-precipitated materials.
The reactivity behaviouro ft he SAS catalystsc an be seen to be significantly different from that of the co-precipitatedm aterials, with the possible exception of the CuCe-SAS material, which in both cases showsd ecreasing activity but increasing methanol selectivity (Table 4). These catalysts all displayed stronger initial deactivation than their co-precipitatedc ounterparts, but were all stable after 5h.T he CuZn-SAS material is of particulari nterest. In keepingw ith evidence that increased coppers urface area is directlyl inked to increased CO 2 conversion ( Figure 6), it is the most activec atalystw hen based on the results taken after 1h.I tr emains active for approximately 4h,b ut loses methanol selectivity as it does so. After this point, it begins to rapidly lose activity andc ontinues to lose selectivity,s tabilising after 5h.C uS SA measurements show as ignificant drop as ar esult of the reaction.
The CuMn-SAS sample once again displayed am ore complex behaviour than is suggested. Initial results shown here are at t = 1h,b ut data recorded before this point show CuMn-SAS to be highly active, more so than CuZn-SAS. However,i ti mmediately deactivates,l osing over 60 %o fi ts activity in the initial hour.A fter 4hit has almost completely deactivated and is predominantly selective towards the production of CO. The coppers urface area measurements show as ignificant loss of surfacea rea throughout the reaction, which is likelyt ob et he cause of this deactivation. It is of note, though, that the initial activity is significantly higher than the pre-reaction Cu SSA of the CuMn-SAS sample would suggest.I ti sp ossible that this is as imilare ffect to that which was seen for CuMn-CP,b ut without the induction period. Theh ighly mixed anda morphous structure of the CuMn-SASc ould be very active initially,b ut then undergoes severe deactivation for the same reasons as before.T his would seem to be borne out by the dramatic drop in coppersurfacea rea.
CuMg-SAS displayed behaviour entirely contrary to the CP equivalent, provingt ob et he most stable of the SAS catalysts in terms of both CO 2 conversion and selectivity.I tdoes not appear to display such rapid deactivation, nor the switching to CO production of its CP counterpart, and after 8his comparable in activity to CuZn 6.5-CP,w hich is due to its higher selectivity.W hilst it is shown to lose as ignificant amount of copper surfacea rea,t he loss is nowhere neara ss evere as in the cases of the other SAS materials.

Conclusions
An umber of conclusions can be drawn from the results herein.O ne of the initial questions was to what extent,i nc oprecipitated catalysts, is he activity of copper-based catalysts determined by the pH of precipitation and to what extent is it affected by the secondary oxide. Based on the resultss hown, we can say that both play ar ole in the activity of the catalyst. CuZn-CP catalysts were more active than all of the other secondary metals at the equivalent pH. From the resultso ft he CuZn materials we see that increased pH leads to decreased activity.Adifference in precipitation behaviour was seen as well;w hereas the CuZn-CP catalysts form zincian malachite at all three pH values, when the zinc is replaced with magnesium or ceria at elevated pH it leads to the direct formation of coppero xides duringt he ageing step of the synthesis, as confirmed by XRD. The better performance of the Zn materials is probablyd ue to the formation of this phase.
CuMn-CP showed interesting behaviour,i nt hat there appearedt ob ean induction period where the activity increased, reachingaplateau for at ime before rapidly deactivating.I n many ways this mirrorst he findings of Helveg et al., [25] who showedt hat copper-zinc catalysts will displays imilar tendencies depending on the oxidising or reducing nature of the atmosphere. Although the gas mixture is highly reducingd ue to 60 %H 2 ,t he oxidising nature of the atmosphere increases with increasing steam content. [26] As H 2 Oi saby-product of both the methanol and reverse water-gas shift reactionsw hich occur,i tw ould seem that the catalystg enerates an active  phase which is then adversely affected by the increased water content that this improved activity engenders.T his then leads to the severe deactivation seen.
Once the results of the SAS catalysts are factored in, more conclusions can be drawn. Whilst for the CP materials increasedp Hl eads to lower activity,t he SAS results show that this is not the sole determinant. The activities still do not correlate exactly with the Cu SSA,a st he CuMg has ah ighera rea than CuZn. This showst hat there is indeed an additional effect from the secondary oxide beyond the simple improvement of the active metal surface area, and that the relativelyl ower activity of the CuMg catalysts is not only ar esult of the higher precipitation pH required in CP.
Further investigation of these effects are required to ascertain which properties of the secondary oxidesa re affectingt he Cu. H 2 -TPR could be usefult oi nvestigate the reducibility of the catalysts, and CO 2 -TPD can be used to assess changes in the basicity of the catalysts.
The CuZn-SAS catalysts hows as imilar deactivation to that reportedb efore, although it does not show the initial induction period. Interestingly,t he CuMn catalyst appearst oh ave very similarbehaviour,althoughitdeactivates even more swiftly.B oth CuZn-SASa nd CuMn-SAS suffer ap articularly pronounced loss of Cu surface area during the reaction, with CuMn-SAS falling to the lowest value of any tested catalyst. Based on these overall results, it would seem that Mn and Zn behavei nabroadly similar manner when paired with Cu, but that Zn is the better choice due to increased stability of the supported Cu metal.
CuMg catalystsp rovedi nteresting, as they weret he only instance in which the SAS material was more stable than the CP material. This is seen in both the activity data and the copper surfacea rea data, and could be down to an umber of factors. The CuMg-CP materials howede videnceo fC uO formation during the initial precipitation, and whilst this materialh ad ah igh copper surface area it swiftly deactivated under reaction conditions. This behaviour was not observed in the SAS material, implying that the formation of the CuO phase was not conducive to retention of the high copper surface area even thoughi tg enerated ah igh initial value. The amorphous SAS precursor,h owever,l ed to am aterialm ore stable than its CP counterpart or anyo ther SAS prepared material. This may be due to the properties of MgO itself, which is not reported to form strong interactions with Cu (unlike zinc) and does not have av ariety of possible oxidation states (unlike manganese and ceria).
When taken as aw hole, the results strongly imply that whilst the initial copper surface area is important, the ability to retain this surfacea rea whilst under reactionc onditions would appear to be key.F urther,t he idea that coppers urface area is directly correlatedt omethanolactivity may not be easily applicable to materials using different secondary oxides. An excellent example of this lies in the CuMg catalysts. CuMg-CP has ah igher initial copper surfacea rea than its CuZn-CPe quivalent, but its rapid deactivation means that the post-reaction area value shows at ruer measure of its activity.T he same is true of CuMg-SAS and CuZn-SAS. In this instance the CuMg-SAS has the lower initial surface area,b ut proves to be the more active catalyst in the long run due to its stability.T his focus on stability appears to be as trength of the co-precipitated CuZn materials, which displayed the lowest amounto fd eactivation.
Thus, the stability of the materials, and their effectiveness as catalysts, can be attributedt oan umber of factors beyond initial coppers urfacea rea. Thef ormationo ft he malachite phase seems to be especially important in coprecipitation;C uZn and CuMn-CP catalysts form this phase, and were significantly more activet han their amorphous SAS counterparts. This phase appears to grant ag reater degree of stability to the resulting catalysts. Where materials did not form this phase, they were all found to be less stable. This effect cannot be attributed to the presence of zinc as the secondary oxide, as the CuZn-SASc atalyst was highly unstable. By contrast, the formation of the CuO phase during precipitation was indicative of ap oor catalyst.
The results obtained using the SAS-prepared catalysts help to back up the benefits of the malachite phase, but also show that for some materials the pH is as ignificant factor.C uMg is ag ood example of this;n either the CP nor the SAS catalyst form the malachite phase, but the elevated pH led to the formationo ft he undesirable CuO phase during co-precipitation. Where this phase was not observed, in the SAS material, the catalystw as far more effective. This wasn ot the case for the CuCe materials, which were less effective regardless of preparation method. This indicates that the choice of oxide is highly relevant.
Overall, the results seem to show that when considering coprecipitation,C uZn catalysts appear to be significantly better due to an umber of benefits granted by the precursor phase. CuMn catalysts behave in as imilarm anner,b ut deactivate more rapidly.W hen the materials were prepared by am ethod which leads to ah ighly amorphous precursor,o ther oxides become viable. CuMg seems in particulart ob eh amperedb y the high pH neededf or precipitation. Once this limitation was removed, it proved to be an effective catalyst. Thisc ould potentially be of use as other precipitation methods are investigated.
Another important conclusion is the apparent confirmation of the work of Hadden et al., [27] who suggested that the correlation between coppers urface area and activity was only valid betweenf amilies of catalysts prepared with similar method. This is borne out in our results, as the highers urface area materials do not always prove to be the most active, and nor is the coppers urfacea rea across the range of oxides alwaysd irectly proportional to the activity.W ec an extendt hesec onclusions to accountf or the post-reaction surfacea rea losses. It seemst hat different preparation conditions, methods, and secondary oxides strongly influence the rate of initial deactivation of the catalysts, whichi sakey factor in their activity after stabilisation.

Co-precipitated catalysts
The co-precipitated catalyst precursors were synthesised by co-precipitation of metal salts using aT oledo Metrohm autotitrator. As mall aliquot (20 cm 3 )o ft he mixed metal solution was added to the reaction vessel, which was stirred continuously.T he amount of liquid was chosen such that it was sufficientt oc over the pH probe. This initial aliquot was brought to pH 6.5 by the addition of the base solution until the target pH was reached.
Subsequently,t he mixed metal solution was added to the vessel at ar ate of 5cm 3 min À1 with continuous stirring. Concurrently,b ase solution was added at as ufficientr ate to ensure that the reaction mixture maintained ac onstant pH of 8. Once all of the mixed metal solution was added, the pH was monitored and controlled for afurther 10 min to ensure complete precipitation of the material. Thereafter,t he precipitate was allowed to age in solution at 65 8Cf or 3h.
This precipitate was filtered under suction and washed with water to remove excess sodium salts. This material was then dried at 110 8Cf or 16 hb efore being calcined at 325 8C( thermal ramp rate 5 8Cmin À1 )i ns tatic air for 3h to produce 2.5-3 go ft he catalyst.

Supercritical anti-solvent precipitation catalysts
Am ixed solution of Cu(OAc) 2 ·H 2 O( 4.1561 mg mL À1 )w ith either Zn(OAc) 2 ·2 H 2 O( 1.9584 mg mL À1 ), Mg(OAc) 2 ·4 H 2 O( 1.9132 mg mL À1 ), Mn(OAc) 2 ·4 H 2 O( 2.1866 mg mL À1 ), or Ce(acac) 3 ·x H 2 O (3.9026 mg mL À1 )w as prepared in a5vol %H 2 O/ethanol mixture (1000 mL) to give an ominal Cu:X molar ratio of 70:30. SAS preparation was performed using apparatus manufactured by Separex. Liquefied CO 2 was pumped to give af low rate of 6.5 kg h À1 and the whole system was pressurised to 110bar and held at 40 8C. Initially,p ure solvent (5 vol %H 2 O/ethanol) was pumped through the fine capillary into the precipitation vessel, with af low rate of 6.5 mL min À1 for 15 min, in co-current mode with scCO 2 in order to obtain steady state conditions inside the vessel. After this initial period, the flow of liquid solvent was stopped and the mixed metal solution was delivered at af low rate of 6.5 mL min À1 .T his gave as cCO 2 /mixed metal solution molar ratio of 22:1. The system pressure and temperature were maintained and the preparation conditions were carefully controlled. Leak checks were also periodically carried out throughout the procedure using snoop solution. When all the mixed-metal solution had been processed, ad rying step was carried out. This was achieved by pumping pure ethanol at 6.5 mL min À1 co-currently with scCO 2 for 30 min, before leaving with just scCO 2 to pump for af urther 60 min. This was to wash the vessel in case residual solvent condensed during depressurisation and partly solubilised the prepared materials. When the drying step was complete the scCO 2 flow rate was stopped, the vessel was depressurised to atmospheric pressure and the precipitate was collected. Experiments were conducted for approximately 5h which resulted in the synthesis of ca. 2.5-3 go fs olid.

XRD
Powder X-ray diffraction measurements were performed using aP ANalytical X'pert Pro diffractometer with Ni filtered CuKa radiation source operating at 40 kV and 40 mA. Patterns were recorded over the range of 10-808 2q using as tep size of 0.0168.A ll patterns were matched using the ICDD database.

Surfacearea measurements
Cu surface area analysis was carried out on aQ uantachrome ChemBET chemisorption analyser equipped with at hermal conductivity detector (TCD). Calcined samples (50 mg) were reduced to catalysts using 10 %H 2 /Ar (30 mL min À1 )w ith heating to 140 8Ca t 10 8Cmin À1 ,a nd then to 225 8Ca t1 8Cmin À1 .F or Cu surface area analysis, catalysts were cooled to 65 8Cu nder He for N 2 Op ulsing. 12 N 2 Op ulses (113 mle ach) were followed with 3N 2 pulses for calibration. The amount of N 2 emitted was assumed to amount to half am onolayer coverage of oxygen and that the surface density of Cu is 1.47 1019 atoms m À2 . formed again to minimise the effects of passivation caused by contact with air during transit.