Structure and catalytic behaviour of CuO–CeO2 prepared by high-energy ball milling

The aim of this study is to prepare CuO–CeO2 composite by means of mechanical milling and to investigate its characteristics as a catalyst. The structural and morphological features of milled samples are observed by X-ray diffractometry and scanning electron microscopy. The redox property and total OSC (oxygen storage capacity) of the milled sample were measured by using GC-TCD and TG-DTA, which are important parameters to indicate the effectiveness of catalysts. Interestingly, reduction of CuO is repeatedly observed when milling of CuO–CeO2 powder mixture is processed in air. The redox property of milled CuO–CeO2 sample is investigated by H2-TPR, where three reduction peaks are observed for 0 h milling and only one broad peak for various other milling times. The total OSC of mechanically driven CuO–CeO2 catalyst is much higher than that of the CeO2–ZrO2 traditional catalyst system at low temperatures.


Introduction
Nowadays, more than 95% of vehicles produced are equipped with a catalytic converter [1]. The three-way catalysts (TWCs), used for the gasoline-fuelled engine are capable of simultaneously converting CO, hydrocarbon (HC) and NO x , with a stoichiometric air-to-fuel ratio (A/F ¼ 14.7), into harmless CO 2 , H 2 O and N 2 [1]. Oxygen storage capacity (OSC) is one of the crucial factors for the performance of TWCs. The CeO 2 -ZrO 2 composite is well known as an excellent promoter of OSC, where CeO 2 exhibits the oxygen storage/release behaviour by redox variation of Ce ions between Ce 3þ and Ce 4þ , while the introduction of ZrO 2 into CeO 2 improves the reduction temperature of ceria through structural modification of the ceria lattice [2], although the OSCs at low

Catalytic property measurements
The milled sample was subjected to measurement of total OSC according to the method by Tanabe et al. [40] and Morikawa [41]. The weight change of the milled sample was measured with TG-DTA (RIGAKU TG-8120) by the following procedure; the milled sample (about 20 mg) on an alumina container was completely oxidized at 5008C in a N 2 -20vol%O 2 mixed gas flow (dry air, 500 ml min 21 ) for 60 min, followed by cooling to 3008C. Then, the gas atmosphere was switched to an Ar-5vol%H 2 flow (500 ml min 21 ) and the weight decrease due to reduction was monitored until no weight change was observed. Afterward, the gas atmosphere was again switched to the dry air and the weight increase due to oxidation was monitored until no weight change was observed. The procedure was repeated twice.
The dynamic reduction behaviour was measured by TPR, where the milled sample (about 50 mg) was put in a quartz reactor and heated at 4008C for 1 h under a N 2 -20%O 2 gas flow (30 ml min 21 ) and cooled to room temperature (RT). The gas was then changed to Ar-5%H 2 (25 ml min 21 ) and the sample was heated at 15 o C/min for the temperature range of 35-9008C, where the H 2 consumption was measured by GC-TCD (Varian CP-4900). For the second run of H 2 -TPR, the sample (50 mg) after the first TPR was cooled to RT in Ar-5%H 2 gas, re-oxidized at 4008C for 1 h under a N 2 -20%O 2 gas flow (30 ml min 21 ), cooled to RT again, and finally heated under an Ar-5%H 2 gas flow (25 ml min 21 ) at 15 o C min 21 up to 9008C for the repeated H 2 -TPR.

Structural characterization
The XRD patterns of milled samples, having a composition of 50%mol CuO and 50%mol CeO 2 , are shown in figure 1. The reflection peak intensities of the CuO phase are largely reduced with peak broadening after 2 h milling, while no appearance of other phases is detected. Although not detected, some CuO phases may exist in nanostructured or amorphous states. After 7 h milling, the CuO peaks are almost eliminated and replaced instead by the appearance of the faint reflections from a Cu 2 O phase and the clear peaks of fcc Cu. The coexistence of both phases lasts until the milling duration of around 14 h. The Cu 2 O reflections become weaker with milling, while the Cu peaks become more distinguished with milling, a tendency which continues up to 30 h milling. There is no change in the observed phase of CeO 2 up to 30 h milling, although the peak broadening with milling is significant, particularly during the early milling stages regarding the change of the peak shape.
From the XRD results, the reductive valence change of CuO, i.e. Cu 2þ ! Cu 1þ ! Cu or Cu 2þ ! Cu, occurs during milling. This is consistent with a report that there are three ways [12,13] to reduce Cu 2þ to Cu: The presence of Cu is also confirmed by nuclear magnetic resonance spectroscopy in the 7 h-and 18 h-milled samples (not shown here). When the CeO 2 and CuO powder phases are forced to contact at the bounding interphase interface during milling the cations of Ce 4þ (or Ce 3þ for non-stoichiometric sites especially near the surface) could be interchanged with Cu 2þ cations through the vacancy mechanism. The CeO 2 powder, in particular the surface volume, exhibits non-stoichiometric compositions with various defects [42], which can be formed upon the introduction of metal cations with higher or lower valences into CeO 2 . It is also known that the mechanical milling of powders induces the fracture and deformation through high-energy collisions among balls, vial surface and particles [43], leading to a modification of the crystal structure as well as large lattice strain. In this study, the Cu cations can be introduced in the lattice of CeO 2 , replacing the Ce cations as well as producing the extra oxygen vacancies in the CeO 2 lattice. The lattice parameter and the estimated crystallite size of the cubic CeO 2 phase in the (CuO) 0.5 (CeO 2 ) 0.5 composite are shown in figure 3, as a function of milling time. The crystallite size is rapidly decreased during an early stage of the milling periods, less than approximately 5 h, followed by a gradual decrease up to 30 h milling. But the lattice parameter is only gradually increased during an early stage of milling, less than approximately 5 h, followed by a rather rapid increase with the royalsocietypublishing.org/journal/rsos R. Soc. open sci. 6: 181861 increase of the milling time up to approximately 18 h, and then the variation becomes small up to 30 h milling. This indicates that although the change of the crystallite size is not directly related to the lattice parameter variation, the latter is activated only after the former event. This is consistent with the idea that the appearance of Cu and Cu 2 O phases is clearly observed when the milling duration is longer than 7 h, as shown in figure 1, because atomic-order mixing of powders that is prerequisite for the phase change is generally expected, particularly in the vicinity of the interphase interface, only after effective repeated folding of particles during milling, the so-called kneading effect, producing fine layered structures inside powder. It is also suggested that a steady state of the structural variation of CeO 2 is attained after approximately 18 h milling. The decrease of crystallite size and the increase of lattice parameter of CeO 2 with milling time are also reported [44 -48].   The lattice parameters and crystallite sizes of the cubic CeO 2 phase and the specific surface areas (BET) of 18 h-milled samples with various CuO contents are listed in table 1. With an increase of mol% CuO contents, the lattice parameter of CeO 2 is increased, the average crystallite size is decreased and the BET surface area of CuO-CeO 2 powder composite is decreased. Because the smaller crystallite size may cause an elongation of the cubic lattice in the nano-sized CeO 2 phase [47], which is ascribed to the lattice strain from the formation of the Ce 3þ cations and the corresponding oxygen vacancies, the observed tendency is consistent. The slight decrease of the BET area with the CuO ratio is probably due to formation of the soft Cu phase during 18 h milling (figure 2), which may cause agglomeration of powders.

Morphological features
The morphological features and the compositional homogeneity are studied by SEM and EDX. Similar morphology is observed for all the milled samples, where aggregations of packed particles from a few hundred nanometres to micrometres in size exist, as shown in figure 4. Some reports show that the CuO is incorporated and evenly distributed into the CeO 2 structure [49,50]. As shown in figure 5, with the use of each beam size adjusted as 10 mm, the areal average concentrations of each 18 h-milled sample with a different overall composition (20,50,70 or 80 mol% CuO) exhibit high homogeneity in terms of Ce and Cu contents, judging from the EDX analyses at 20 randomly selected squares. The magnification of SEM is fixed at 220 times. This strongly suggests that the dimensional size of the Cu or Cu 2 O phase, as observed in figure 2, is much smaller than 10 mm. Contamination from the milled media (ZrO 2 balls  and a steel vial) is also examined by EDX, and the maximal contamination level is less than 1 wt% for Fe while no contamination is detectable for Cr, ZrO 2 , and so on, after 18 h milling.

Reduction behaviour by TPR
The H 2 -TPR profiles of the (CuO) 0.5 (CeO 2 ) 0.5 samples milled for 0, 4, 7, 14 and 18 h are shown in figure 6. For powders without milling (0 h milling), i.e. by mixing the oxide powders, the reduction of (CuO) 0.5 (CeO 2 ) 0.5 is characterized by rather combined three peaks (a, b, and g) in the range of 220-4208C and one broad f peak in the range of 700 -8808C. After 4 h milling, the former peaks (a, b and g) merge into one large broad peak cantered around 3008C-3208C with some skewed symmetry. The position and the intensity of the peak are not largely changed with further milling, but the skewness is somewhat increased, with the peak top shifting to the higher temperatures with milling time up to 18 h ( figure 6). The progression of the milling process should produce a large amount of bounding interface/interphase between CuO and CeO 2 , and some Cu atoms would even be penetrated into CeO 2 , producing Ce 12x Cu x O 2 solid solutions near the interface vicinity. These would facilitate the mobility of oxygen atoms for both phases, resulting in promoted valence variation of cations and leading to the observed large areas of H 2 consumption peak. As for the latter peak, the broadness of the peak is slightly reduced with milling time, whereas the peak top is around 8008C.      figure 7a,h), there is no distinguished H 2 -TPR peak in the temperature range observed, while the consumption of H 2 steadily exists above 3508C up to approximately 9008C for reducing CeO 2 to Ce 2 O 3 , where the reduction first occurs near the surface defects [51,52] followed by the formation of intermediate CeO2-x and complete transformation to Ce 2 O 3 .
It is reported by Fierro et al. [53] that the TPR characteristics could be affected by mass transfer limitations and experimental operating variables such as the initial amount of reducible species, the initial H 2 concentration, the total gas flow rate, the heating rate, and the activation energy of the reaction. They claim that desorption of H2 attached to the reduced Cu metal surface could exhibit the apparent double-peak behaviour, which may be affected by water vapour produced by the reduction process, and that the H 2 -TPR profile of CuO depends on the particle size and surface area, where the peak top is higher by 2888C for particle sizes of 425 approximately 850 microns compared with less than 100 microns [53]. A similar tendency but with a much larger peak shift (over 3738C) is also reported by Luo et al. [49], who claim that the hydrogen spillover effect is the reason for the difference between CO-TPR and H 2 -TPR.
Kim et al. [54] report that there is an incubation period prior to reduction, which is longer at lower temperatures. The tendency is in agreement with the general theory of nucleation and growth, where the number of newly formed nuclei is copious at lower temperatures, but because the growth is very much limited at lower temperatures, the phase existence is not easily detected by x-ray diffractometry. They claim that CuO reduction is generally easier than Cu 2 O reduction with H 2 -TPR, with the apparent activation energy for Cu 2 O close to twice that of CuO, but when the H 2 flow rate is not high enough for avoiding the rate-limiting of the reduction process, a sequential reduction process such as CuO ! (Cu 4 O 3 !) Cu 2 O ! Cu may happen. Our experimental condition is: 15 o C min 21 , 50 mg and 25 cc min 21 of 5%H 2 flow. This is close to the condition for the appearance of the sequential process due to the 'lean H 2 ' flow. The incubation period and the sequential/simultaneous reduction process are also reported by Kim et al. [54], where, with the lean H 2 condition, the CuO, Cu 2 O and Cu are simultaneously observed after a certain incubation period according to the time-resolved XRD. This is consistent with our thought experiments (not shown here), where the 3 phases are simultaneously observed when the H2-TPR is stopped at 2808C and kept for 15 min, while only CuO is observed when the sample is immediately quenched from 2808C to RT.
Consequently, the a peak for pure CuO reduction, as shown in figure 7, should be mainly attributed to direct reduction of CuO particles into Cu in the surface vicinity, and some simultaneous reduction to Cu 2 O (or Cu 4 O 3 ) should be involved with our experimental operating variables, somewhat contributing to the a peak. Because the rate of nucleation and growth of each reduced phase is different, which involves the shape of nuclei and the nucleation sites besides the spillover effect of H 2 on the metallic Cu, one rate-determining step would gradually dominate the sequential CuO reduction to Cu 2 O and Cu, exhibiting the broad b peak for the slow reduction process, which is related to the larger activation energy for Cu 2 O formation compared with direct Cu formation. Before the end of the sequential reaction, some H 2 molecules would be dissociated from H atoms on newly formed Cu on the powder surface, penetrating into the Cu(/Cu 2 O)/CuO particles, and at certain temperature(s), depending on the surrounding condition, the remaining CuO would be reduced to Cu either directly or sequentially through Cu 2 O (or Cu 4 O 3 ), exhibiting the g peak. The extended TPR peak profile as high as 4708C or even 4808C strongly indicates the difficulty of reduction due to longer the diffusion path for H atoms or H 2 molecules as well as the escape of produced H 2 O molecules. It is acknowledged that the larger the CuO particles, the thicker the reduced Cu phase; this covers the particle and shows the bulk-like behaviour of CuO reduction. Thus, the a and b peaks are probably caused by surface-vicinity reduction with separate locations, whereas the g peak would be due to reduction inside the particles. At the high temperature f peak, it may cause the Cu surface to interact with the SiO 2 quartz reactor [55,56].
The apparent increase of the H 2 -TPR starting temperature for the a peak, as shown in figure 6, compared with pure CuO in figure 7b,c is probably due to variation on the mass transfer limitations affected by experimental operating variables such as the initial amount of reducible species and effective total gas flow rate through mixing CeO 2 particles with CuO particles. But, the end temperature for the g peak, as shown in figure 6, is much decreased, compared with pure CuO shown in figure 7b,c. This must be the mixing effect of CeO 2 particles with CuO particles, in that the contact point or the interphase interface plays an important role in the nucleation and growth of the reduced phase such as Cu, or even Cu 2 O. Owing to the valence change of cations requiring high electron mobility as well as the higher oxygen mobility near the contact point, the nucleation event is much more active in the vicinity, and with the fast surface diffusion of Cu atoms the nucleation is royalsocietypublishing.org/journal/rsos R. Soc. open sci. 6: 181861 immediately followed by growth or accumulation of the new phase, always leaving some area of fresh CuO surface contacting with H 2 /Ar gas. In the reduction process, the diffusion paths for H atoms or H 2 molecules as well as the escape path of the produced H 2 O molecules are established, and the reduction is continued until all the CuO is consumed.
There are many reports on the CuO-CeO 2 powder systems and the reduction study. Many researchers report that ceria in the CuO-CeO 2 powder promotes surface species reduction of highly dispersed copper oxide [57 -64] and the CuO particles with smaller sizes are easier to reduce [65]. According to Liu et al. [58], the reduction peak at lower temperature is attributed to copper oxide clusters strongly interacting with ceria, whereas the peak at higher temperature is attributed to larger CuO particles, not associated with ceria. A similar tendency in the H 2 -TPR profile is also reported by Luo et al. [60] and Xiaoyuan et al. [61]. Furthermore, Tang et al. [66] report that the lower temperature peak is assigned to reduction of amorphous CuO clusters closely interacting with the CeO 2 , while Luo et al. [48] claim that with increasing CuO content three peaks appear, with the lowest temperature deriving from reduction of finely dispersed CuO, the intermediate temperature associated with reduction of Cu 2þ ions in Cu x Ce 12x O 22d solid solutions, and the additional highest one due to reduction of bulk CuO.
As observed in figure 6, with increasing milling time on (CuO) 0.5 (CeO 2 ) 0.5 composite samples, the skewness of H 2 -TPR profile is somewhat increased with the peak top shifting to the higher temperatures. Also, in figure 7, as the CuO content is decreased in various 18 h-milled (CuO) x (CeO 2 ) 12x powders, the H 2 absorption at lower temperatures involved with the a peak is gradually diminished, with the peak tops of the main H 2 -TPR profiles around 310-3308C, which is similar to that shown in figure 6. Wang et al. claim that [49] Cu atoms embedded in ceria have an oxidation state higher than those of the cations in Cu 2 O or CuO, where (i) Cu in the doped sample is only partially reduced to metallic Cu, especially at low temperatures as CuO ! Cu 2 O ! Cu, because Cu embedded in ceria is difficult to reduce, (ii) The lattice of the Ce 12x Cu x O 2 systems (fluorite-type) is highly distorted with local order defects and multiple cation-oxygen distances, and (iii) Doping CeO 2 with Cu introduces large strain and O vacancies. This is consistent with our results in that, at the early stage of milling such as before 7 h milling, the ultrafine mixture of CuO/CeO 2 is more effective for H 2 -TPR profile, which corresponds to the rapid crystallite size reduction before 4 h milling, as shown in figure 3. Here, the large area of CuO/CeO 2 interphase interface is extremely efficient for producing the path for H 2 diffusion and H 2 O escape, besides the effective valence change of cations, where the x-ray diffractometry in figure 1 identifies only CuO and CeO 2 phases. But, the longer the milling time, the more intense the formation of Ce 12x Cu x O 2 solid solution systems, where the higher peak temperature for the H 2 -TPR profile should be observed due to Cu embedded in ceria as milling proceeds. The results are also consistent with our x-ray results shown in figure 1, where atomic-order mixing is achieved at least near the interface vicinity of CuO/CeO 2 , producing Cu and Cu 2 O after 7 h milling, and that the clear increase in the lattice parameter is observed (figure 3) after approximately 7 h milling.
It is thus quite reasonable to consider that, at the early stage of milling, the major H 2 -TPR peaks caused by milling of CuO/CeO 2 are attributed to reduction of Cu 2þ ions in the interfacial vicinity of CeO 2 , which exist in the ultrafine mixture of CuO/CeO 2 or some Cu 2 O/CeO 2 . The CuO particles are highly dispersed, producing many active sites for Cu 2þ reduction on the interphase interface with CeO 2 . With the increase of milling time, some Cu atoms are even incorporated into CeO 2 , causing formation of Ce 12x Cu x O 2 solid solutions. The large intensity and the broad profile of the H 2 -TPR peaks would be observed due to the dual effect. Figure 8 shows the repeated H 2 -TPR cycles of milled (CuO) 0.5 (CeO 2 ) 0.5 composites (50 mol% CuO milled for 0, 4, 7, 14 and 18 h) in figure 6, each after the first TPR. The peak intensity for the 0 h milled sample is largely decreased, when compared with the first run, with high temperature peak(s) remaining, suggesting agglomeration/recrystallization of CuO particles that produce bulk-like CuO after heating to 9008C. When increasing the milling time, the CeO 2 particles are finely distributed into the CuO particles, or some Ce-Cu -O solid solutions are produced, which prevents agglomeration/ recrystallization of CuO, leading to the slow decrease observed in the peak intensity. The temperatures of the reduction start as well as the peak top of the H 2 consumption are lowered compared with the first run. On considering that the pure CuO is reduced in the wide range of temperatures (200-4708C, figure 7b,c) regardless of milling or without milling under this particular experimental condition, some CuO is heterogeneously reduced (lean H 2 -TPR) at low temperatures (200-2508C) to produce Cu and Cu x O simultaneously. But, because the (CuO) 0.5 (CeO 2 ) 0.5 composite powders are finely mixed with milling, even after the H 2 -TPR to 9008C, the elemental mixture is maintained, and the CuO reduction is activated by surface contact with CeO 2 .
royalsocietypublishing.org/journal/rsos R. Soc. open sci. 6: 181861 Figure 9 shows the total OSC at 3008C of various milled samples (50 mol% of CuO) after 0, 4, 7, 14, 18 and 30 h milling. The total OSC is increased with milling time, with the rapid increase from 0 to 7 h milling followed by the gradual increase from 7 h to 30 h. This is consistent with figure 1 (x-ray results) and figure 3 in that the rapid atomic-order mixing is effective during the early stage milling of 7 h, evidenced by both solid state reactions and crystallite size variations. It is also noted that, because the lattice parameters are largely varied after 7 h milling, the CuO/CeO 2 interphase interface creation and the corresponding atomic contact would be more important for the effective OSC than formation of the Ce-Cu -O solid solution, although both contribute to the OSC promotion. That is, the activated valence change of Cu 2þ /Cu þ /Cu assisted by neighbouring Ce-O bonds is extremely important, and the oxygen transport path is also critical for the OSC. The enhanced oxygen storage/transport would    figure 10, from 0 to 7094 mmol-O 2 /g-catalyst with increased CuO content from 0 to 80 mol%, but is decreased to 3919 mmol-O 2 /g-catalyst for pure CuO (BET surface area of 15 -16 m 2 g 21 ). After ambient ageing at 10008C for 5 h, however, the total OSCs at 3008C are significantly reduced (figure 10), probably ascribed to the agglomeration/recrystallization of fine particles.
As shown in figure 11, the total OSCs of the CeO 2 -20 at %ZrO 2 and the CeO 2 -50 at %ZrO 2 at 3008C are 185 and 258 mmol-O 2 /g-catalysts, respectively, while those of the CuO-CeO 2 catalyst system prepared with the same condition are at least one order greater. It clearly shows that the mechanically driven CuO/CeO 2 system exhibits the high OSC property, and this should contribute to improving the performance of TWCs at lower temperatures.

Conclusion
Mechanical milling was applied to the CuO -CeO 2 powder system to produce mixed-oxide catalysts. The milled sample was characterized by the use of XRD, SEM, TG-DTA, GC-TCD and BET analyses. Milling of powder mixtures of CuO and CeO 2 showed the reduction of CuO when milling was processed in air. The crystallite size of ceria was rapidly decreased at the early stage of milling, followed by an increase of the lattice parameter, indicating formation of Ce 12x Cu x O 2 solid solutions after the rapid crystallite size reduction. The redox property of milled CuO-CeO 2 samples was investigated by H 2 -TPR. Three reduction peaks were observed for 0 h milling and only one broad peak for various milling times, where the valence change of Cu ions enhanced the redox activity. The higher OSC for the CuO-CeO 2 system was observed with increased milling time. The total OSC of the CuO -CeO 2 catalyst was much higher than that of the CeO 2 -ZrO 2 traditional catalyst system at low temperature.
Data accessibility. The datasets supporting this article have been uploaded as part of the electronic supplementary material.