Influence of Precipitation Conditions on Properties of Ceria and their Implications on the Mars‐van‐Krevelen Active Surface Area

Quantifying the active surface area of Mars‐van‐Krevelen catalysts is paramount for the elucidation of structure‐property relationships and the knowledge‐based catalyst development. Different cerium oxides were prepared via precipitation‐based techniques. By altering conditions during precipitation, materials with a wide span of material properties were prepared. The influence of temperature during ammonia‐based precipitation was investigated. When using Ce(IV) precursors an increase in precipitation temperature decreases the crystallite size while increasing specific surface area, attributed to an increase in nucleation rate. Sintering stability is also increased for materials precipitated at higher temperature. Measuring the total oxygen storage capacity (TOSC) values of the prepared materials showed that the TOSC is not strictly a function of BET surface area. Our results suggest that crystallites with a domain size under a certain threshold are not reduced via oxygen release but rather hydroxyl formation. A method was proposed with which the redox active surface area for polycrystalline ceria can be estimated on basis of the domain size obtained from Rietveld refinement. These findings were corroborated by CO oxidation light‐off experiments in which the light‐off temperature was found to correlate with the surface that can release oxygen reversibly, the Mars‐van‐Krevelen active surface area, rather than BET surface area.


Introduction
One of the most prominent catalytic processes is the exhaust gas aftertreatment with three-way catalysts (TWCs) of Otto combustion engine cars. [1]The simultaneous removal of NO x , CO, and unburned hydrocarbons leads to a narrow operating window in which conversion is maximized for all components with regard to oxygen content.This operating window is broadened by the introduction of an oxygen storage component, which can store and release O 2 under net-oxidizing and net-reducing conditions. [2]Since the 1980s ceria (CeO 2 ) and more recently ceria-zirconia (Ce 1-x Zr x O 2 ) have emerged as materials of choice due to their high oxygen storage capacity and good thermal stability. [3]nderstanding the link between preparation conditions, material properties and their kinetic relevance is paramount for a rational development of new and improved catalytic systems.One of the most frequently employed methods for the preparation of solid oxide catalysts is precipitation from aqueous media, [4] not only due to its simplicity and scalability but also because the material properties are quite sensitive to the preparation conditions.7] Understanding the link between the material properties and the catalytic activity is imperative for the development of new and improved catalysts and to derive structure-property relationships.However, since area specific site densities and mass specific surface areas are seldomly constant when comparing different catalysts, it is usually not sufficient to solely compare mass specific reaction rates.Therefore, it is common practice to normalize kinetic data to get comparable values.For supported (noble) metal catalysts this is usually done through site density (SD) measurements by selective chemisorption, yielding the number of accessible metal surface sites. [8]In redox catalysis, however, redox active oxides such as CeO 2 , TiO 2 , and Fe 2 O 3 are often used.For such catalytic systems, it is frequently assumed that the entire surface is active.Therefore, it is common practice to normalize kinetic data using the mass specific surface area calculated from physisorption experiments, [5,9,10] which is not site-selective.Oxidation reactions involving ceria and other redox active oxides are often assumed to follow the Mars-van-Krevelen (MvK) mechanism, which comprises typically two steps.The first is the oxidation of the reactant using lattice oxygen of the catalyst, while in the second step the reduced catalyst is subsequently reoxidized via the gas phase oxidant. [11]This is also reported for CO oxidation catalyzed by ceria. [10,12]A normalization of kinetic data using the specific surface area for MvK-type catalysts means that the entire surface area has to be reversibly reducible under oxygen release.It was, however, recently reported by Safonova et al. that in platinum supported ceria catalysts a fraction of the ceria surface acts as spectator Ce(III) species during catalysis. [13][16] This release of oxygen is, however, imperative for oxidation catalysis.The knowledge about the surface that can be reversibly reduced under oxygen release is not only important for plain ceria catalysts but also platinum-loaded systems since these systems rely upon a reverse oxygen spillover from the ceria lattice. [17]Quantifying the active surface of ceria-based materials is very important when trying to derive structure-property relationships.For this purpose materials with different properties were prepare to investigate correlations between preparation conditions such as precipitation and calcination temperature and material properties such as crystallite size, surface areas, and reducibility/reoxidizability and finally their relevance as catalytic properties in CO oxidation.N 2 physisorption, XRD with Rietveld refinement, IR, Raman, H 2 -TPR, and O 2 chemisorption experiments were applied to identify structure-property-activity relationships for CO oxidation on bare CeO 2 catalysts.

Preparation of Cerium Oxides -Influence of Precipitation Temperature
After preparation of the materials, structural properties were analysed using N 2 -physisorption and X-ray diffraction (diffraction profiles are shown in the supporting information) to investigate the influence of the preparation conditions on the material properties.Results of BET and Rietveld for the differently prepared materials are listed in Table 1.
All materials exhibited the cubic Fm � 3m fluorite crystal structure (space group 255) with a lattice parameter close to the theoretical value of a = 0.541134 nm (JCDPS 34-394).The observed increase in the cubic lattice parameter a with decreasing crystallite size was already reported in literature findings. [18]The prepared polycrystalline materials exhibited surface areas between 3 m 2 g À 1 to 110 m 2 g À 1 while crystallite sizes between 10 nm and 80 nm were achieved.From Table 1 it is evident that the precipitation temperature plays an important role in governing the material properties during precipitation.When comparing the material parameters of the different AP materials the crystallite size decreases, and the specific surface area increases steadily with increasing precipitation temperatures T prec as shown in Figure 1. Materials prepared by the UP and HUP method do not follow the temperature trend, which is not surprising since the release of precipitation agent is slower in addition to the aging conditions for the HUP sample being longer.
There is little data available on the influence of precipitation temperature on crystallite size and surface area during the ammonia-based precipitation of cerium oxides.Available data published by Chen and Chang and Cui et al. show an opposite trend with precipitation temperature, i. e. a decrease in surface area with increasing precipitation temperature. [19]The difference in our findings compared to the two studies mentioned before can most probably be attributed to a use of Ce(III) precursors in the latter studies compared to the Ce(IV) precursor used in this study.The reason is, as Chen and Chen reported, that precipitation from a Ce(III) precursor requires a preceding oxidation step of Ce(III) to Ce(IV). [7]Therefore supersaturation during precipitation of Ce(III) precursors is limited by the preceding Ce(III) oxidation.Therefore, after initial nucleation, the rate of growth is significantly higher compared to the nucleation rate, since the latter is a function of supersaturation. [20]Newer nucleation models also suggest that formed surfaces act as catalysts during particle growth which would aggravate the influence of lower supersaturation even Comm 0.5411 12 160 [a] Calculated by Rietveld refinement, [b] calculated from N 2 -physisorption using BET method.further. [21]This can be considered as the reason for the formation of larger crystallites when using Ce(III) precursors instead of Ce(IV) precursors.In the case of Ce(IV) precursors, however, supersaturation is increased rapidly through ammonia addition.Thus, nucleation is not hindered.Additionally, the effect of surface tension, which destabilizes nuclei, is known to decrease at higher temperatures. [22]Therefore, higher temperatures would lead to faster initial nucleation rates and thus to smaller crystallite sizes.This explanation is in line with Hirano and Kato's explanation of their investigation of Ce(III) and Ce(IV) precursors during hydrothermal synthesis [23] and can also explain why forming solid solutions with Zr is challenging with Ce(III) precursors as reported by Letichevsky et al.. [24] To elucidate these findings further, we followed the pH of the metal salt solution during precipitation.For this purpose, titration experiments were carried out.The normalized amount of OH À consumed by metal hydroxide formation is shown in Figure 2.
From Figure 2 it is evident that temperature influences the pH at which hydroxide ions are consumed.With increasing temperature, the amount of OH À which does not contribute to a pH increase, and therefore is consumed through precipitation, does increase.Since hydroxide ions are only consumed when nuclei are formed, the higher hydroxide ion consumption at higher temperature points towards a higher nucleation rate, especially since visible precipitation only occurred when the pH exceeded pH > 7. Forming a larger quantity of nuclei at a constant concentration of metal ions would explain the observed smaller crystallites and larger specific surface areas observed with higher precipitation temperatures.
Cerium oxides are known to exhibit comparably low stability against sintering.However, the preparation conditions are also known to be able to influence the temperature stability. [25]When comparing material properties of the prepared materials after high temperature calcination, an influence of the precipitation conditions is observable.7] Although all AP materials showed a very low BET surface area (S BET < 5 m 2 g À 1 ) after high temperature calcination, and although the relative error of nitrogen physisorption is quite large at those low BET surface areas, a trend in crystallite size is evident.Materials precipitated at higher temperatures, exhibiting smaller crystallites retain higher surface area values after high temperature calcination, although smaller particles usually possess a lower sintering stability.Comparing the relative crystallite size growth of the AP materials shows that AP5-500 has an increase of crystallite size by a factor of 3.4 while the crystallite size of AP75-500 only increases by a factor of 2.4 when increasing the calcination temperature from 500 °C to 700 °C.So, not only smaller crystallites are formed during high temperature precipitation but also materials exhibiting a higher sintering resistance.Resistance to sintering can be increased by two different mechanisms.The first one is the homogeneity of particle sizes to limit the driving force of Ostwald ripening and the second one is a reduction of particle growth kinetics.
Figure 3 shows the Raman spectra of the different materials.Compared to XRD, Raman spectroscopy is sensitive to changes in the oxygen sublattice of ceria. [26]It was reported in literature that broadening of the F 2g mode at 464 cm À 1 is related to a decrease in crystallite size, [27] which is in line with the crystallite size trend obtained from Rietveld refinement and shown in Table 1.The feature at 600 cm À 1 is caused by defects in the ceria lattice. [28,29]Thus, the intensity ratio between the 464 cm À 1 main band and the defect related band at 600 cm À 1 is used as a means to assess the defect concentration of ceria related materials. [29,30]From Figure 3 it is evident that precipitation temperature does influence the defect concentration.With increasing precipitation temperature, the defect concentration in AP materials does increase.The AP materials show similar strain values, as represented by the ratio between the Raman bands at 600 cm À 1 and 464 cm À 1 , to UP and HUP materials at high precipitation temperatures.When comparing the difference in crystallite size between the materials calcined at 500 °C and 700 °C D 500 °CÀ 700 °C, a decrease of D 500 °CÀ 700 °C can be observed for the AP materials with increasing defect concentration.This is depicted in the supporting information.However, the UP and HUP materials exhibited greater sintering resistance.Urea-based precipitations are known to produce more homogeneous precipitates because of the homogeneous ammonia release in the entire reaction volume due to the thermal decomposition of urea.Also, for HUP the hydrothermal treatment during precipitation and aging can lead to increased homogeneity through dissolution.This higher degree of homogeneity would lead to a lower driving force for Ostwald ripening.
The observed increased sintering stability can therefore be explained on the one hand through a more homogeneous precipitate because of the faster nucleation rate and on the other hand by the higher defect concentration.Since defects in ceria are mostly related to oxygen vacancies this corresponds to an increase in Ce 3 + concentration.It was suggested that the introduction of larger dopants can limit the diffusion processes necessary for particle growth. [31]

Role Of Material Properties In Redox Properties
Cerium oxides are valued for their capability to store and release oxygen.It is reported that the oxygen storage and release process in cerium oxides below 750 °C is limited to the surface, [3] since vacancy formation is less favourable in the bulk than on the surface below that temperature. [32]TOSC 500 °C of cerium oxides are therefore often described as being a function of the specific surface area [33] This correlation however is not evident in the measurement shown in Figure 4 where the correlation between specific surface area S BET and TOSC values are shown.
Figure 4 shows that at high BET surface areas S BET the measured values deviate significantly from the theoretically predicted values by eq. ( 4).As the TOSC 500 °C values were measured by means of O 2 chemisorption experiment this finding might be explained by two different reasonings.As Bernal et al. discussed that, besides the classical mechanism of oxygen release through vacancy association, [34] ceria surfaces can also be reduced via the formation of hydroxyl groups. [15,16]he latter reduction does not yield oxygen vacancies, thus a quantification via reoxidation does not account for this type of reduction.Secondly, an irreversible reduction might cause the observed deviation.This could be the case through loss of surface area during the reduction [35] or overreduction that can be observed in other redox active oxides [36] .This would also explain why quantification via the reoxidation leads to smaller values compared to the theoretically predicted values.
We attempted to elucidate the reasons for this deviation using H 2 -TPR and TPD experiments (Figure 5) for sample comm as it showed the largest deviation between calculated and measured TOSC 500 °C.
The TPR profile of the pristine comm material and the material that underwent one redox cycle at 500 °C look similar.Despite a slight change in the shape of the reduction signal, the integral of the surface related reduction (T < 700 °C) did not change significantly.Combined with the negligible change in the physisorption isotherms and S BET before and after one redox cycle means the deviation of the oxygen that is released during reduction between measurement and theory is neither caused by irreversible reduction or overreduction, nor by surface area loss.When comparing the TPR profile to the H 2 -TPD curve in Figure 5, a very intense H 2 desorption feature at around 550 °C was evident.Integration of the desorption signal gives a value of 498 mmol H 2 g À 1 , which in terms of surface oxygens correlates to 249 mmol O 2 g À 1 .This value is close to the difference between the measured TOSC 500 °C of 195 mmol O 2 g À 1 of this material and the theoretical TOSC calculated from the specific surface area with eq. ( 4) of 454 mmol O 2 g À 1 .These results suggest that the difference in theoretical TOSC and measured TOSC 500 °C is caused by the fact that a fraction of the surface is reduced by formation of hydroxyl groups rather than release of oxygen through vacancy association and thus not being active in MvK-type reactions.This was additionally probed by in situ-DRIFTS measurements.The samples were degassed, before reducing  them at 500 °C and subsequently degassed in nitrogen at 500 °C to remove any adsorbed water.According to the H 2 -TPD measurements the H 2 desorption should be low at this temperature and newly formed hydroxyl groups might be detectable (see Figure 5).Difference spectra after reduction and subsequent degassing is shown in Figure 6 while the original spectra are shown in the supporting information.Ceria is known to be active in the hydrogen dissociative adsorption even at low temperature. [37]When comparing the different spectra obtained for the different ceria materials, it is evident that while a substantial difference after reduction and successive degassing at 500 °C respectively was observed for the high surface area samples, little change was observed for the sample with low surface area.This is in line with the H 2 -TPD results described previously where it was shown that hydrogen was released after reduction and successive desorption at temperatures above 500 °C.While this is not really evident for the low surface area material.Thus, the presented DRIFTS results strengthen the assumption that the difference in oxygen storage capacity might stem from the fact that some of the surface is not reducible under oxygen release but rather hydroxyl group formation and thus would not be active in a MvK mechanism.38] Using eq. ( 1) a crystallite surface was calculated from the Rietveld refinement results.At high surface areas, materials that are interesting for catalytic application have a significantly lower calculated S cryst compared to the calculated S BET from N 2physisorption.This finding most probably is caused by small crystallites only contributing very weakly to XRD patterns.Therefore, crystallites below 5 nm would contribute to the BET surface area signfificantly while they would not appear in the diffraction pattern in a quantifiable way. [39]The measured TOSC 500 °C values are plotted as a function of this theoretical crystallite surface in Figure 4.The values obtained in this case are close to the values predicted as a function of surface from eq. ( 4).This result combined with the insensitivity XRD exhibits towards small crystallites suggests that the reduction of ceria with oxygen release is hindered at very small crystallites.Using TEM, the presence of these small crystallites (< 5 nm) could be proven for the materials with high surface area, whereas they were absent for materials with lower surface area.A representative TEM image is shown in the supporting information.For the polycrystalline ceria samples investigated in this study the crystallite surface can therefore be used as a descriptor to assess the surface that can be reduced under oxygen release.This is especially important since oxidation reactions catalyzed by cerium-based materials are known to follow the MvK mechanism in which the reduction of the catalyst with oxygen release is an integral step. [11,40]Therefore, a simple normalization using the BET surface area is not sufficient when looking into structure-property relationships of such catalysts.Instead the MvK-active surface area has to be determined.

CO Oxidation as a Model Reaction
CO oxidation is often used as a model reaction to evaluate the oxygen release properties of cerium-based redox active oxides and is therefore widely studied.CO oxidation catalyzed by ceria catalysts, as previously mentioned, is proposed to follow the MvK mechanism.It was therefore chosen to highlight the implications of the finding in this study, because knowledge about the number of sites that can be reduced under release of oxygen and thus participate in the catalysis is paramount for correlating the CO oxidation activity to the catalyst properties.For this purpose, CO light-off curves for materials exhibiting a wide range of specific surface areas and "MvK-actice surface areas" were measured and are shown in Figure 7.
When comparing the light-off curves in Figure 7 with the reduction profiles (c.f.supporting information) of the materials, the CO oxidation light-off directly correlates with the reduction onset of the oxides.This again points towards the catalysis through reduction of the support, i. e. a MvK mechanism.The light-off curves show significant changes between the different materials.For these experiments a constant weight of catalyst and with it a constant modified residence time τ mod was used for all samples.Figure 7 shows light-off temperatures T 50 as a function of either the BET surface S BET determined by nitrogen physisorption or as a function of S cryst calculated based on crystallite sizes from Rietveld refinement.The latter was previously shown to correlate well with the amount of releasable oxygen from the surface, thus the number of redox active surface sites which can release oxygen reversibly.From this plot a linear relationship between T 50 and the MvK-active surface area S cryst derived from TOSC measurements is visible.Thus, the correct normalization of reaction data allows the investigation of structure-property relationships for MvK type catalysts.
The implications of the findings in this study are not only relevant for unloaded ceria but also noble metal loaded ceria since these materials still rely upon a reverse oxygen spillover from ceria.They are also not limited to the oxidation of CO but are relevant for other oxidation reactions catalyzed by ceria.

Conclusions
We were able to show how the material properties of polycrystalline ceria can be influenced by precipitation temperature.It was shown that with increasing precipitation temperature the crystallite size decreases, while specific surface area and sintering stability increase.This was attributed to a nucleation dominated precipitation process, thus explaining the difference in our findings compared to investigations of the role of precipitation temperature when precipitating Ce(III) precursors.Our results suggest that the higher sintering stability might be related to the higher degree of defects in the system.Catalysts prepared by means of urea-based precipitation showed an even higher sintering stability, probably caused by a more homogeneous particle size distribution, leading to a lower driving force for Ostwald ripening.
Prepared polycrystalline ceria were successively investigated for their redox properties.Results showed that the amount of oxygen that can be released is not correlated with the total surface S BET.Using H 2 -TPD this was traced back to a reduction without oxygen release but hydroxyl group formation.As the values correspond very reasonably with a calculated surface area based on the crystallite sizes from Rietveld refinement, this effect seems to be related to domain size, with smaller crystallites (< 5 nm) having a hinderance of forming associated oxygen vacancies under oxygen release.For polycrystalline ceria samples crystallite size determined by XRD can be used to approximate the surface area which is reduced under oxygen release and therefore active in MvK catalysis.Thus, it can be used as a simple descriptor for the active surface area for MvK catalysts.This was demonstrated by light-off experiments for CO oxidation over various ceria materials exhibiting different MvK-active surface areas.
These results show that normalizing kinetic data strictly by BET surface area is not sufficient when trying to derive structure-property relationships.It is however necessary to assess the surface area that can release oxygen reversibly.This not only has implications in ceria-based catalysts but for other redox active oxides as well.It is also not limited to CO oxidation, as this problem is aggravated for oxidation reactions that require successive oxidation steps.

Experimental Material Preparation
Different synthesis methods were employed to prepare materials exhibiting different properties.Ammonia-based Precipitation (AP) was carried out following a widely used method by Hori et al.. [41] In a double-walled 1 l stirred tank reactor, connected to a thermostat, a solution of 0.1 mol l À 1 Ammonium cerium(IV) nitrate ((NH 4 ) 2 Ce-(NO 3 ) 6 ) was prepared with bidistilled water.Subsequently the solution was thermally equilibrated at different precipitation temperatures (T prec = 5 °C, 25 °C, 50 °C and 75 °C) for 1 h under stirring at 300 rpm using a KPG stirrer.Afterwards, 10 wt.% ammonia solution was added using a calibrated piston pump with a constant flow rate of 5 ml min À 1 until a pH of 10 was reached.The precipitate was aged for 4 h separated using a red band filter paper and finally washed with 2 L bidestilled water and 0.2 L of ethanol.The solid obtained was then dried overnight in an oven at 80 °C before calcination at 500 °C and 700 °C for 4 h under air flow (200 l h À 1 ).
Urea-based homogeneous precipitation (UP) was carried out based on a synthesis protocol by Taylor et al. [42] Solutions of a nominal metal ion concentration of 0.1 mol l À 1 were prepared by dissolving ammonium cerium(IV) nitrate in bidestilled water.Afterwards urea was added in a molar ratio of n(Ce 4 + ):n(urea) of 1 : 15.The reaction solution was stirred for 1 h to ensure complete dissolution.Afterwards the reaction solution was refluxed for 4 h before the formed precipitate was filtered and washed with 2 l of water and 200 ml of ethanol.Drying overnight at 80 °C and calcination at 500 °C and 700 °C yielded the final product.
Based on the synthesis proposed by Si et al. urea-based precipitation under hydrothermal conditions (HUP) was also performed. [43]queous metal salt solutions with a metal ion concentration of c(Ce 4 + ) = 0.1 mol l À 1 were prepared before stirring overnight.After adding appropriate amounts of urea to achieve solution compositions n(Ce 4 + ):n(urea) of 1 : 15, 80 ml of the solution was filled into 100 ml teflon liners, which were placed in steel autoclaves.After closing, the autoclaves were placed in a preheated oven equipped with a rotary device at 140 °C at a rotational speed of 10 rpm for 24 h.After cooling down of the autoclaves in an ice bath, the formed solid was separated by centrifugation.Washing was performed 3 times with bidistilled water before the solid was washed with ethanol.The materials prepared from different autoclaves were added and dried overnight at 80 °C, the final material was obtained by calcination under air flow at 500 °C and 700 °C.
The ceria samples synthesized by different methods are in the following referred to as AP-T calc , UP-T calc and HUP-T calc with AP being ammonia-based precipitation, UP being urea-based precipitation, and HUP urea-based precipitation under hydrothermal conditions.For the samples synthesized by ammonia based precipitation, the precipitation temperature T prec is additionally indicated in the sample name in the form APT prec -T cal .
The pH value of the solution during precipitation was followed by means of titration.A double-walled reaction vessel was connected to a thermostat and 200 ml of the metal salt solution were added and thermally equilibrated for 1 h.Afterwards, using an Eppendorf pipette 10 wt.% ammonia solution was added in portions of 0.1 ml portions.After each addition equilibriums was adjusted for 2 min before the pH value was determined using a Mettler Toledo InLab Power Pro-ISM calibrated with NIST standards before every measurement.Blind measurements were conducted using 0.1 M NaCl solution.

Characterization
Specific surface areas were quantified by means of nitrogen physisorption measured with a Quantachrome Autosorb 3b at 78 K.The samples were degassed at 300 °C in vacuo before analysis.The surface area was calculated based on the BET method.
X-ray diffractograms were collected using a Bruker D8 Advance, using Cu Kα radiation and a VANTEC detector in the angular range of 20-100°with an angular velocity of 0.036°s À 1 .Rietveld refinement calculations were carried out using GSAS-II. [44]A certified reference material (SRM660c) was used to create an instrumental profile to account for changes in the diffraction profile caused by the instrument.For Rietveld refinement calculations a cubic crystal system (space group Fm � 3m [45] ) was used as a basis.Lattice parameter a and crystallite size D were extracted from the diffraction patterns.
The crystallite size D is calculated assuming an isotropic polycrystalline material.With the assumption of spherical crystallites the crystallite size D can be used to calculate the surface A cryst and volume V cryst of crystallites and with these the specific crystallite surface S cryst with eq. ( 1).
Temperature programmed reduction (TPR), desorption (TPD) and pulse chemisorption were carried out in an Autosorb iQ from Quantachrome instruments equipped with a thermal conductivity detector (TCD).Samples were pressed, crushed, and sieved to gain a particle fraction of 200 μm-500 μm.250 mg of oxide was placed between two quartz wool pieces into the U-shaped measurement cell.Samples were degassed at 500 °C for 1 h in a continuous flow of 30 ml min À 1 of 10 % O 2 /He mixed with two mass flow controllers (MFCs) from pure gases (Westfalen, 99.999 %).Afterwards the sample was cooled down in a flow of 30 ml min À 1 He to 50 °C.
For TPR experiments, a flow of 10 % H 2 /N 2 (Westfalen, 99.999 %) was used while applying a temperature ramp of 10 K min À 1 to 950 °C.Formed water was condensed in a dry ice/acetone cooling trap.
The TPR signal was calibrated by reduction of known amounts CuO.
Copper content of the standard was verified by ICP-OES, while reduction with 5 % H 2 /N 2 in a STA 449 F5 Jupiter from Netsch was used to verify a complete reduction towards Cu 0 via the mass change.
TOSC values were determined by first reducing the sample in a constant 10 % H 2 /N 2 (Westfalen, 99.999 %) flow of 30 ml min À 1 with a temperature ramp of 10 K min À 1 to 500 °C.After holding 500 °C for 1 h to ensure complete surface reduction (c.f.supporting info), the gas was switched to 30 ml min À 1 He (Westfalen, 99.999 %) for one more hour so any adsorbed hydrogen was desorbed. [15]TOSC was quantified by pulse chemisorption, injecting repeatedly 276 μl of pure oxygen using a sample loop until complete oxygen breakthrough.The injected molar amount of oxygen n O 2 ;inj was calculated assuming ideal gas law as shown in eq. ( 2) with the sample loop volume V SL , the pressure in the sample loop p SL and the temperature at the sample loop T SL .The temperature at the sample loop was constantly measured using a thermocouple while assuming atmospheric pressure.A typical experiment is shown in the supporting information.
The calibration of the signal was carried out by correlating n O2,inj to the mean integral of the injections where total oxygen breakthrough occurred.This one point calibration was used to calculate the detected amount of oxygen n O2,det from the TCD integral.TOSC values can successively be calculated as shown in equation eq. ( 3) using the sample mass m sample after degassing.
Based upon the cubic crystallite system and its lattice parameter a, Madier et al. calculated a theoretical oxygen storage capacity TOSC theo from specific surface area S BET and the theoretical oxygen density of the surface as shown in eq. ( 4). [33]SC Hydrogen temperature programmed desorption (H 2 -TPD) curves were measured by employing the same reducing treatment as for the TOSC samples (reduction at 500 °C for 1 h).Afterwards the sample was cooled down to 40 °C under hydrogen flow.The gas was switched to 30 ml min À 1 pure nitrogen (Westfalen, 99.999 %) before applying a temperature ramp of 10 K min À 1 to 950 °C.Using a dry ice cooling trap, any water formed was condensed from the gas stream before detection.Any influence of other impurities was taken into account by subtracting a blind measurement, which was obtained by subjecting a fresh sample to the TPD profile without preceding reduction.
Raman spectra were recorded on a Horiba LabRamHR using a 50x objective with an incident laser wavelength of l ¼ 532 nm and a maximum power of 100 mW.To account for the high intensity of the F 2g mode of cerium oxide a filter was applied to circumvent saturation of the detector.Measurements were taken from 100 cm À 1 -1000 cm À 1 with an acquisition time of 5 s and the spectra were averaged over 20 scans.The measurements were carried out at the German Aerospace Center (DLR).

CO Oxidation Experiments
CO oxidation light-off experiments were measured in a home-built flow-through experimental setup equipped with an online gas chromatograph coupled with a mass spectrometer (GC-MS).The flow chart of which is shown in the supporting information.The GC-MS (Agilent 7890B, Agilent 5977B) was modified with a methanizer and a column switching system, custom built by Teckso GmbH.Permanent gases were separated using two Plot-Q precolumns (1.5 m, d I = 1 mm) and a packed molecular sieve column (5 Å, L = 1.5 m, d I = 1 mm).After separation the gases were detected by a TCD before flowing over the methanizer and through a flame ionization detector.For quantification of nitrogen and oxygen the TCD signal was used whereas CO and CO 2 were quantified using the FID signal.
Using different MFCs the feed gas mixtures were generated.Total gas flow was set to 50 ml min À 1 (STP).The prepared catalysts were pressed, crushed, and sieved to get particle sizes between 100 μm and 200 μm. 100 mg of catalyst was diluted with 3 g of acid washed glass beads before filling it into a stainless-steel reactor using a paper funnel to limit buildup of static charge.A thermocouple was placed inside of the catalyst bed to monitor the catalyst temperature during the reaction.Before reaction an oxidation and degassing step was applied.50 ml min À 1 of 20 % O 2 in N 2 was employed while applying a temperature ramp of 10 K min À 1 to 300 °C which was subsequently held for 3 h before cooling down the reactor to 50 °C.Afterwards the feed composition was changed to 0.5 vol.% carbon monoxide and 10 vol.% oxygen in nitrogen with a flow rate of 50 ml min À 1 , giving a modified residence time of 120 kg s m À 3 .Bypass measurements prior to the light-off experiments were done to determine an average value of the feed composition x i,0 from five bypass measurements.Subsequently, the experiment was started.After equilibration of the catalysts for 5 h at 50 °C the temperature was slowly increased to 400 °C with a temperature ramp of 1 K min À 1 .The final temperature was held for 2 h.During the duration of the experiment the outlet compositions x i,t were analyzed continuously.
CO conversion X CO was calculated using the CO molar fraction during bypass measurements x CO,0 and after reaction x CO according to eq. ( 5): X CO ¼ x CO;0 À x CO x CO;0 (5)

Figure 1 .
Crystallite size D (black) and BET surface area S BET (red) as a function of the precipitation temperature T prec .

Figure 2 .
Figure 2. Normalized reacted amount of OH À as a function of the pH of solution for different precipitation temperatures T prec = 5 °C (black), 25 °C (red) and 50 °C (blue).Full pH profile is shown in the supporting information.

Figure 4 .
Figure 4. Measured TOSC 500 °C values as a function of S BET (black) and S cryst (red).Theoretical values TOSC theo calculated with equation (4).