J CHEM. SOC'. FARADAY TRANS., 1994. Wj). 773--7XI Reduction of Cerias with Different Textures by Hydrogen and their Reoxidation by Oxygen Vincent Perrichon,* Ahmidou Laachir, Gerard Bergeret, Roger Frety and Louise Tournayan lnstitut de Recherches sur la Catalyse, CNRS,2 avenue Einstein, 69626 Villeurbanne Cedex, France 01ivier Touret Rhdne Poulenc, Centre de Recherches d'Aubervilliers, 52 rue de la Haie Coq,93308 Aubervilliers, France Successive reduction steps of CeO, particles by hydrogen between 300 and 1070 K have been followed by temperature-programmed reduction (TPR) and in situ magnetic measurements on several samples with different BET surface areas. The nature of the phases present in cerias reduced between 670 and 1270 K was determined by X-ray analysis.Finally, reoxidation by oxygen or air was studied at room temperature for all the reduced samples. Magnetic and TPR results show a direct relationship between the degree of reduction and the BET surface area. Indeed, for most of the samples, the degree of reduction at 620-670 K determined by magnetism corre- sponded to the creation of one layer of Ce3+ ions at the surface of the ceria. A similar relationship between the BET surface area and the extent of reduction was established using the area of the low-temperature TPR com- posite peak, the maximum of which was found to be constant at 810 K. When the reduction progresses further into the bulk, two main phases were evidenced: first, an expanded cubic CeO,-, phase derived from the initial ceria by a dilatation of the whole structure and, for deeply reduced samples, the hexagonal Ce,O, phase.A new intermediate phase, cubic Ce,O,, was also observed on samples reduced at 1070-1 170 K. Complete reoxidation by oxygen occurs at room temperature, for all reduction percentages below ca. 60°/,,, i.e. as long as the reduced phase remained in the cubic form. When the hexagonal Ce,O, phase has been formed, the reoxidation cannot be completed at 294 K. The structure of ceria is face-centred cubic (fluorite type). It can be described as a pile of oxygen ions. linked together by the edges, with the cerium atom at the centre of the oxygen cube in octahedral coordination. 'q2 Some oxygen entities in this structure are very mobile and can easily be removed tc give non-stoichiometric compositions, stable in a reducing atmosphere with the general formula CeOz-x, where 0 < x d 0.5.3However, in the presence of oxygen.these com- pounds are easily reoxidized to the stable CeO, . This oxygen lability and the possibility of large deviations from stoichiorn- etry explain the prevalent use of ceria in catalysts for auto- motive post-combustion. In such systems, cerium oxide behaves as an oxygen partial pressure reg~lator.'.~ For this reason, several studies have appeared which have focussed on elucidating the conditions of these redox processes in order to obtain a better understanding of their mechanisms.6 '_' From these studies, it appears that the texture of the initial cerium oxide has a great influence on the interaction between H,and CeO,.However, any quantitative determination of z in the Ce02-, reduced phase cannot be accurate because most of the results are obtained from TPR studies which do not take into account the reduction of impurities (nitrates. carbonates) or hydrogen fixation, all quantities which can be important in dispersed cerias. l6 The control of the reduction extent by oxygen reoxidation is also rarely given. Moreover. although the Ce-0 phase diagram is well established for samples prepared at high temperat~re,~ there is a lack ol' phase determination for the solids obtained by reduction of dispersed cerias, which are those used in catalysis. All these data are necessary to establish a macroscopic scheme of the oxygen transfer in CeO, during reduction-oxidation pro-cesses.In a recent paper," we have studied, by several comple- mentary techniques and particularly by in situ magnetic SUS-ceptibility measurements, the hydrogen reduction processes of two ceria samples having surface areas of, respectively, 5 and I15 m' g-' . The reduction began at 470 K, irrespective of the initial surface. With the low-surface-area ceria, a small reduction plateau at 620 K attributed to the reduction of the ceri$i surface was observed as well as an intermediate state at a higher temperature of reduction (900 K), having the formal composition CeO, 83, r.e. close to that of Ce,O, ,.In the case of the 115 m2 g-' sample. the observation of a plateau corre- sponding to the reduction of the surface depended on the operating conditions, but again a stabilized composition close to CeO, g4 could be obtained at 670 K.The reoxidation by tixygen of the reduced oxides was almost complete at room temperature. Ir the present work, we have generalized these studies to othcr samples of intermediate surface area and to higher reduction extents, in order to see if the same reduction steps can be observed for various ceria types and if the reversibility of the redox process can be effective whatever the solid and redLCtion percentage. In parallel with the quantitative deter- mindion of the reduction percentage by TPR and magne- tism, the evolution of the ceria structure was followed by X-rz y diffraction on samples sealed under hydrogen after redLCtion.The results are presented and discussed in terms of the topology of the reduction and reoxidation processes. together with the structure of the reduced phase. Experimental Materials Twc different HSA cerias were obtained from RhBne-Poulenc I.0.M. with references C1 and C2. They differed by the exis- tence of microporosity in the former (equivalent to almost (5o(J, of the BET surface area), whereas micropores were absent for the latter. Their purity is close to 99.5'%, with lanthanum as the main impurity. By treatment of the initial cerias in various conditions, five new samples were prepared. All the samples studied are pre- sented in Table 1, which also gives the BET surface area and the equivalent microporous surface calculated according to the 't method'.'' Among the samples, C1-400 and C1-850 have already been characterized by means of several tech- niques.In order to compare the samples under the same condi- tions, they were standardized with a pretreatment in situ under oxygen flow (4 1 h-') at 673 K for 1 h (heating rate, 5 K min-I). After that, they were evacuated or placed in an inert gas flow for 1 h at 673 K before cooling to room temperature. Temperature-programmed Reduction (TPR) The hydrogen consumption (1% H, in argon) was followed with a thermal conductivity detector in an apparatus described previously." The catalyst was in a U-shaped reactor. The flow rate was 1.1 1 h-' and the heating rate 8 K min-'.Magnetic Measurements The Faraday microbalance and procedure for calculation of the magnetic susceptibility have been described else-where."." The method allows determination of the para- magnetic Ce3+ ion content (note that Ce4+ in CeO, is diamagnetic). The experimental conditions corresponded to those used previ0us1y.l~ The ceria sample (0.10-0.13 g) was first stan- dardized under oxygen at 673 K, and then evacuated down to 0.1-0.2 mPa at the same temperature. After cooling to 294 K, the sample was placed under a hydrogen flow (4 1 h-I) and heated (4 K min-') with incremental steps of 50 K. For each step, the sample was kept for 2 h at the temperature before cooling to 294 K, always in H,, in order to measure the susceptibility.The latter was found to be nearly the same if the sample was cooled under vacuum instead of hydro- gen." X-Ray Diffkaction (XRD) The X-ray diagrams were obtained using the Debye-Scherrer method with Mo-Ka radiation or Cu-Kcr in some cases. Since the reduced cerias are extremely sensitive to air reoxidation, a special procedure was followed which consisted of sealing the reduced sample in a quartz tube under hydrogen partial pres- sure. Then, in a glove box flushed with oxygen-free dry argon, the sample was transferred into a controlled-atmosphere cell J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 equipped with a beryllium window enabling the X-ray dif- fraction pattern to be recorded. Results Influence of the Initial BET Surface Area on the Reduction Process Study by TPR Three samples were examined for this study and compared with C1-400: C2-600, C2-800 and C1-850.After the stan- dardization treatment, TPR was performed up to around 1100 K. The curves are shown on Fig. 1. Each curve presents a first composite peak beginning at about 500 K with a maximum close to 810 K. The shapes of this peak are not identical for the different samples. Shoulders and slope changes are visible as indicated by the arrows on Fig. 1. Fig. l(b) and (c), relative to the same precursor, present a similar initial profile, whereas for Fig. l(a), the hydrogen consumption is initially rather slow, and then increases quickly between 630 and 660 K. Another minor slope change is also observed at 760-740 K for Fig.l(a) and (b). All these differences have to be connected with a specific chemistry of the oxide surface anions depending on the sample surface. An attempt has been made with FTIR spectroscopy to correlate them with differ- ent types of surface hydroxy groups.20 Nevertheless, although the first peak results from several steps, it can be related to a global process corresponding to the consumption of the surface oxygen species as evidenced bef~re.~~'~' Then, after this qst peak, when the temperature is increased above 900 K, the beginning of a new peak is observed on the curves and must be due to the bulk reduction. For C1-400 [Fig. l(a)] the nature and origin of the nega- tive peak at 965 K were already characteri~ed.'~*~~ This peak arises from the reduction of bulk carbonates" which are present in the solid and are desorbed under vacuum at around 810 K.It can also be due to the desorption of hydro- gen fixed by the solid during the TPR.16 For C2-600 [Fig. l(b)], the negative peak is not clearly observed or, more prob- ably, it occurs at 880 K with a lower intensity, giving rise to the small positive peak centred at 965 K. It must be remarked that this sample was prereduced at 873 K before being reoxidized at room temperature. Such a sample exhibited an almost complete absence of carbonate species, as shown by FTIR spectroscopy.21 This absence of carbonates must therefore explain the absence of the negative peak at 890 K.Assuming that the first peak is due only to the reduction of the various surface oxygen species, it is interesting to note that no major shift is observed at the maximum temperature of this peak. For every sample, the maximum is observed at Table 1 Origin and texture of the different CeO, samples reference c1-400 C1-700 Cl-800 c1-850 C2-600 c2-640 C2-800 origin and treatment RP -C1 C1-400 + air calcination at 973 K C1-400 + vacuum treatment at 1073 K C1-400 + air calcination at 1123 K SBETaImZg-' 115 55 10 5 RP -C2 + H, reduction at 873 K + air at 300 K 78 C2-600 + H, reduction at 915 K + air at 300 K 21 C2-600 + air calcination 1073 K 54 equivalent microporous area/m2 g-55 10 0 0 0 0 -Determined from N, adsorption at 77 K.J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 1.1.1 300 500 700 900 TIK Fig. 1 Hydrogen TPR curves corresponding to the various CeO, samples after standardization at 673 K: (a) C1-400, (b) C2-600, (c) C2-800,(d)C1-850. (H2 : Ar = 1: 99; 8 K min-’) around 810 K. This observation is in disagreement with the results of Johnson and MOO^.^ It indicates that whatever the initial texture, the mean binding energies of the surface oxygen are the same and distinct from that of the bulk. It can also be an indication that the pretreatment procedure at 673 K was efficient enough to leave the surface in a clean reproducible state. In order to make a correlation with the BET surface area, we calculated the hydrogen uptake during this first peak.Two methods were tried. The first one used the experimental curve [Table 2 (exp)]. The second one was undertaken in order to minimize the uncertainty due to the recovering of Table 2 Low-temperature peak hydrogen consumption hydrogen consumptionlpmol g -sample exp. sym. predicted c1-400 423 544 617 c1-400 367 512 617 c2-600 239 354 429 C2-800 155 204 302 C1-850 29 25 29 C1-850 47 56 29 exp., experimental result ;sym., from the symmetrical procedure. the surface and bulk reduction processes at high temperature and also to limit the error caused by the negative peak which might be present in each sample. Since it is dificult to decon-volute the first peak, we have supposed that the surface reduction peak was symmetrical and centred at the maximum temperature (see Fig.1). Thus the hydrogen uptake was cal- culated by doubling the H, consumed between 470 and 810 K, the temperature of the maximum of the low-temperature peak. The corresponding hydrogen quantities are certainly overestimated compared with the real phenomenon. They are presented in Table 2 (sym.). The values of this second series are all higher than those of the first ones. Also included in this table for comparison are the values predicted according to the method proposed by Johnson and MOO^.' These authors assumed a simple model in which the ceria surface is constituted by adjacent oxygen ions (ionic radius, 0.14 nm) the number of which can be easily calculated for spherical or cubic particles.By supposing that the reduction of the surface cerium species corresponds to the elimination of one fourth of the surface oxygen ions (capping oxide ions), it is easy to obtain the corresponding values for hydrogen. The relationships between BET surface area and the differ- ent hydrogen consumptions are shown graphically in Fig. 2. The variations are roughly linear in both cases, which indi- cates that the first overall TPR peak may be related directly to the reduction of the ceria surface. However, the slopes (respectively, 3.1 and 4.2 pmol H, mP2)differ by about 35%, and are well below the predicted one (almost 5.4 pmol H, mP2). This difference can be due to a too simplified model.In particular, it does not take into account the repartition of the crystallographic planes of ceria which may vary when the cal- cination temperature increases. Moreover, the value for the ionic radius of the oxygen ion could be different. One can also think of a specific reason connected with the physico- chemistry of the ceria surface or the experimental conditions. For example, a too low pressure of hydrogen during the TPR could induce a low reduction rate and be in competition with simple oxygen depletion (due to both sintering of the ceria and the lability of some oxygen species) or a decrease of ‘mobile oxygen’ content due to sintering. Consequently, we A / 7 500 Is, 0 1 4005 0.-w n5 300 v) 0 $ 200 Is,2 -0>.= 100 0 25 50 75 100 SBET/m2 9-’ Fig. 2 Variation of the hydrogen consumption in the low-temperature TPR peak as a function of the BET surface area of CeO, : (a) experimental curve, (b) symmetrical procedure, (c) pre-dicted values. See text for more details about the calculation methods. applied the magnetic method to measure the reduction extent of the ceria surface with pure hydrogen. Magnetic Study The reduction percentages estimated from the magnetic sus-ceptibilities at 294 K, are shown in Fig. 3 as a function of the reduction temperature for six samples of Table 1. The pre- vious results for C1-400 and C1-850 are presented again for a complete comparison, Further, the run for C-400 was re-peated and extended to temperatures higher than 673 K, the preparation temperature and the limit of the previous study. In every case, the beginning of the reduction is observed at 470 K.Then, as the temperature increases, the reduction extent increases, and the more so for higher BET surface areas. For all the samples, the curves exhibit an inflexion point or even a plateau in the 620-720 K temperature domain. It must be recalled that on the curves each experi- mental point was obtained after a 2 h isothermal reduction. Thus, the different equilibrium points before obtaining the reduction plateau value correspond to different germination and surface diffusion steps which, in fact, reveal a specific surface chemistry depending upon the sample.For higher temperatures and in the case of the non-microporous solids [Fig. 3(b) and (c)], an important increase in reduction extent is obtained at about 800 K. At 900 K, an 70 I 1 J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 intermediate state seems to be reached, corresponding to a 36-42% reduction percentage, i.e. to cerium oxides having the composition CeO,.,,-CeO,.,, . For the microporous cerias (Cl-400 and C1-700), no real intermediate state was evidenced at about 900 K. The curves show a progressive but irregular increase, with several slope changes. This behaviour can be explained by two main phenomena which probably occur simultaneously : (i) the elimination of bulk carbonates as evidenced by TPR which induces a transient local dis- order and (ii) the decrease of the BET surface area which is effective on this sample for reduction temperatures higher than 720 K.22 Our previous data on C1-400 and C1-850 have evidenced a rough relationship between the reduction extent at 670 K and the BET surface area.The values calculated here at the 620-720 K plateau for all the samples lead to the almost linear relationship shown in Fig. 4, although the C1-400 point is clearly above the line. The slope is slightly inferior (15%) to the one calculated in the model of Johnson and Mooi. This relatively good agreement supports the model which predicts at the plateau the elimination of one fourth of the surface adjacent oxygen ions. It confirms the idea that for tem-peratures G620-670 K, the reduction process is mainly limited to the uppermost layer of the ceria.The dispersed ceria C1-400 is an exception with a reduction greater than one layer, which is probably due to the more dispersed char- acter of this oxide for which the cubic model of Johnson and Mooi may be very crude. Indeed, the numerous lattice defects must be favourable for a faster diffusion of the oxygen species in the bulk. It could also be connected with the higher germi- nation process rate related to specific surface sites.” By comparison, the reduction values obtained by TPR are smaller than those deduced from the magnetic study. This can be explained by the differences in experimental pro- cedures. During magnetic measurements, longer reduction times and higher hydrogen pressures favour the equilibrium states and a fraction of the bulk oxygen can reach the surface and be reduced. In TPR experiments, although the higher reduction temperature must favour oxygen diffusion, the decrease of the BET area would limit the participation of the bulk oxygen.Thus, it seems normal to obtain lower values for TPR. Another reason can contribute to the lower TPR 0 100 200 300 400 500 600 700 800 900 1000 ternperature/K Reduction percentages of cerias of different specific areas obtained from the magnetic susceptibility as a function of the reduction temperature (2 h at each temperature): (a) C1-400, (b) C1-700, (c) C2-600,(d) C2-640, (e) C1-800, cf)C1-850. (*)C1-400, (0)C1-700, (+) C2-640, (*) C2-600, (x) C1-800, (0)(21-850.200% reduction corresponds to the total reduction of CeO, to Ce,O, . SBET/m2 9-’ Fig. 4 Variation of the reduction percentage calculated at the plateau of the magnetic curve (Fig. 3) as a function of the BET surface area of CeO, .Comparison between the experimental (a) and predicted (b) values. 100% reduction corresponds to the total reduction of CeO, to Ce,O, . J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 values: The TPR data are calculated on the basis of the initial sample weight, i.e. without taking into account the weight loss occurring during the standardization treatment. This is not the case for magnetism. The difference can be esti- mated to about 7% for C1-400 but is much smaller for C1-850 with a specific surface area of 5 m2 8-l.Therefore, even if it is not possible to obtain a complete agreement between the TPR and magnetic results, both sets of data are directly related to the BET surface area of the initial ceria. Accordingly, these methods can be very useful for determining the surface area of peculiar ceria samples, as, for example, the supported cerias. The TPR technique is easy to handle and commonly utilized in research laboratories, in contrast to the magnetic technique. However, the latter has the great advantage of selectively measuring the formation of Ce3+ ions and gives reliable data, irrespective of the content of metallic salts or impurities in the sample, with the excep- tion of strongly paramagnetic or ferromagnetic species.X-Ray Study of the Reduced Phase The composition of the oxide obtained after reduction at 900 K, CeO1.8S1.79, is close to that of the monoclinic phase Ce6011 (CeOla8J or eventually Ce7O12 (CeO,.,,) which have been proposed to explain some diffraction pattern^.,^.,^ The reduced solid could also correspond to a mixture of CeO, and Ce203. Another important point is the nature of the cerium oxide when the reduction is performed at tem- peratures >900 K. In order to characterise these points, we have studied the evolution of the X-ray diagrams of CeO, reduced under pure hydrogen at temperatures up to 1270 K. The starting material C1-400, which has the highest BET surface area, was chosen to obtain the highest reduction per- centage.Additionally, a run was performed with the C1-850 sample. These samples were heated for 15 h under 3.6 1 h-' of hydrogen at fixed temperature, cooled under hydrogen to room temperature and analysed as noted above. Table 3 summarizes the results obtained after reduction at 673 and 873 K. From the position and the intensity of the diffraction peaks, it can be seen that the initial cubic structure of CeO, is maintained but with a slight shift of all the lines. The cell parameters determined from the set of interplanar spacings are 5.433 and 5.509 A, for the Cl-400 samples reduced at 673 and 873 K, respectively. This corresponds to a lattice expansion of the ceria of 0.4% for the sample reduced at 673 K and of 1.8% for the sample reduced at 873 K.No evidence for the Ce6011phase was found. In particular, the (043) Bragg line with spacing d = 2.447 A, characteristic of the Ce6011phase, is absent on all the patterns. The results are similar after reduction at 873 K of the low-surface-area sample C1-850. In this case, the interplanar spacing is 5.502 A, with an expansion of 1.7%. The X-ray diffraction pattern corresponding to the reduction temperature of 1073 K is shown on Fig. 5(4. In addition to the peaks of the expanded CeO, phase (cell parameter 5.564 A, lattice expansion 2.8%), new well defined peaks appear with low intensity (Table 4). No referenced JCPDS-ICDD phase could be attributed to these peaks which could correspond to a cubic structure.The diffractograms of the solids reduced at 1173 [Fig. qa)]and 1273 K [Fig. qb)] are more complex. The expanded 050,phase is always present (cell parameter 5.555 and 5.520 A, lattice expansion 2.7 and 2.0% for 1173 and 1273 K, respectively) and all the peaks of the Ce,03 phase are observed. After reduction at 1273 K, the concentration of the CezOJphase increases at the expense of the expanded ceria. It must be stressed that the unknown cubic phase evidenced Table 3 Interplanar spacings, d/& calculated from the XRD pat- terns of =ria samples reduced at 673 and 873 K: comparison with the JCPDS-ICDD data for CeO,,Ce,O, ,and &,03a CeO, Ce@, ,' Ce,03f JCPDS JCPDS JCPDS 34-394 32-196 23-1048 c1-4006 c1-400" cubic monoclinic hexagonal 3.38 (20) 3.37 (30) 3.32 (20) 3.152 3.184 3.1234 (100) 3.23 (100) 3.22 (100) 3.20 (100) 3.03 (20) 3.03 (30) 3.01 (20) 2.964 (40) 2.945 (100) 2.7147 2.7556 2.7056 (30) 2.796 (100) 2.773 (100) 2.447 (80) 2.391 (20) 2.274 (20) 2.254 (30) 1.9235 1.9452 1.9134 (52) 1.973 (100) 1.968 (100) 1.960 (loo) 1.945 (35) 1.737 (60) 1.733 (30) 1.730 (60) 1.708 (40) 1.6392 1.6611 1.6318 (42) 1.675 (60) 1.685 (4) 1.637 (30) 1.623 (18) 1.5684 1.5897 1.5622 (8) 1.588 (40) 1.550 (40) 1.516 (2) 1.545 (40) 1.4727 (5) 1.3558 1.3779 1.3531 (8) 1.3823 (2) 1.2444 1.2645 1.2415 (14) 1.2941 (8)1.2113 1.2314 1.2101 (8) 1.2464 (11) 1.1957 (7) 1.1742 (3) 1.1410 (4) 1.1084 1.1244 1.1048 (14) 1.0775 (7) Intensities are given in parentheses.Reduction temperature 673 K;0 :Ce = 1.84 :1. Reduction temperature 873 K;0 :Ce = 1.70:l. dO:Ce=2.00:1. 'O:Ce= 1.83:l. /O:Ce= 1.50:l. The compositions were extrapolated from magnetic measurements performed on samples treated in similar conditions. for the sample reduced at 1073 K is also observed for the sample reduced at 1173 K, but is no longer detected after reduction at 1273 K. In addition, for both samples, small extra peaks appear as shoulders on the expanded ceria peaks, 10 15 20 25 30 35 40 2Bfdegrees Fig. 5 XRD pattern of Ceo, reduced at 1073 K by hydrogen (a) and further reoxidized room temperature (b). Note the Ka,-a, split-ting for the 28 angles >20°. The asterisk denotes the new cubic phase.778 Table 4 Interplanar spacings, d, and relative intensities of the new phase appearing after reduction at 1073 and 1173 K: comparison with the JCPDS-ICDD data (22-0369)for cubic La,O, cerium oxide (our data) La,O, (JCPDS) peaks,"2Oldegrees d/A III, dlA III, (hkl) 8.96 4.539 45 4.62 10 (211) 3.27 100 (222) 2.832 35 (400)15.54 2.623 50 2.668 10 (411) 17.19 2.374 50 2.413 5 (332)18.69 2.184 100 2.220 10 (431) 2.003 40 (440)22.65 1.806 85 1.836 10 (611) 23.84 1.717 95 1.747 5 (541) a Mo-Ka = 0.71 A. on the high-angle side. They can be attributed to the normal cerium dioxide. This beginning of reoxidation was due to a tiny air leak in the measurement cell, as could be deduced from the evolution of the spectrum with time.The increase of the lattice parameter was previously noted by Bauer and Gingerich,' and also by Ray et This lattice expansion is due to the reduction of some Ce4+ ions to Ce3+,the radius of the Ce3+ ion being larger than that of Ce4+ (128.3 pm instead of 111 pm26).However, this increase was associated with the splitting of some peaks suggesting a rhombohedra1 symmetry. Here, no such splitting was observed. The structure remains unchanged and the unit cell varies with the reduction temperature, as shown on Fig. 7. A slight expansion is already detected after reduction at 673 K, but cannot be considered as certain. Then, the increase is evident and reaches a maximum for 1100 K. For higher reduction temperatures, the lattice expansion decreases, prob-ably because Ce3+ ions begin to leave the expanded CeO, phase to form the Ce203phase.The new cubic phase can be tentatively assigned. As shown in Fig. 8 and Table 4, it seems to present some links with the 10 15 20 25 20/deg rees Fig. 6 XRD pattern of CeO, reduced at 1173 K (a) and 1273 K (b). The letters a and b show the expanded CeO, phase and the normal hexagonal Ce,O, phase, respectively. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 ______-0 5.4 I I I 500 750 1000 1250 1500 T/K Fig. 7 Variation of the unit-cell parameter of the CeO, cubic phase us. reduction temperature cubic La203 phase (JCPDS-ICDD 22-0369). The two struc-tures most commonly found for rare-earth-metal sesquiox-ides, M203, are the so-called hexagonal A and cubic C structures.,' A monoclinic B form of some of these oxides has also been recorded.In cases where M,O, crystallizes with more than one structure, A appears to be stable at the highest temperature, B at lower and C at the lowest temperatures. The hexagonal A and cubic C forms are known for lantha-num (JCPDS-ICDD 05-0602 and 22-0369, respectively), but only the A hexagonal form is reported for cerium to our knowledge (JCPDS-ICDD 23-1048). The cubic C structure is closely related to that of CaF, from which it may be derived by removing one quarter of the anions and then rearranging the atoms slightly.26It is interesting to recall that the cerium dioxide CeO, crystallizes with the fluorite structure. If the cubic structure is assumed for our unknown cerium phase, a value of 1113 pm for the unit-cell parameter can be deter-mined from the interplanar spacings of Table 4.This value can be compared to the value of 1121 pm interpolated from the curve of the unit-cell parameter of the cubic C structure for different known lanthanides vs. the ionic radius of 220) I (311)I I I I I I 11 i-..z cQ) c, .-(mi I I I I I (b ) I (SSO?'400? I i (sn?I I I I 10 15 20 25 2t?/degrees Fig. 8 XRD pattern of CeO, reduced at 1073 K compared with reference data extracted from the JCPDS-ICDD file. (a)CeO, ceria-nite (34-0394);(b)cubic La,03 (22-0369). J. CHEM SOC FARADAY TRANS..1994. VOI.. 90 Shannon and Prewitt" for the valency 111 and a coordi nation number of 6 (Fig.9). The agreement is relatively good considering that the main peaks of the cubic C structure of Ce,O, [peaks (222). (400). (440) and 162211 are not distin- guishable from those of expanded CeOz [peaks (11 I). (200). (220) and (31l)] because these two phases are cubic and thc unit-cell parameter of Ce20, is twice that of CeOz (1 1.13 A for Ce,O, and 5.564-5.555 A for CeO, reduced at 1073 dnc! 1173 K, respectively). Reoxidation by Oxygen of the Reduced Cerium Oxides The reoxidation by oxygen of the reduced samples at rwm temperature was followed by magnetic measurements and !)! X-ray diffraction. The objective here was to ascertain in N hat conditions the reoxidation can be complete and ho\\ the reoxidation process varies with reduction extent..Mugnrric I sotherni jbr O.Yj*yen.A dsorptio 11 Two samples were prepared with a reduction extent hisher than that expected for the reduction of the surface alone. Accordingly, the reduction of C1-400 and C1-850 at 971 K was to 54 and 35"z11, corresponding to CeO, ,73 and CeO ,hZ5. respectively. Their reoxidation was compared with that of CeO,,,, (from C1-400) the reduction of which was close to that of the surface.lS Small doses of oxygen were introduced at room tem-perature. and the variations of the magnetic susceptibilit! were followed as a function of the weight increase expressed in the number of adsorbed oxygen molecules.The results are shown in Fig. 10. The experimental points follow straight lines with slopes equal to -6.8 and -6.5 emu (cgsi-F mol-0,,respectively. They are close to the previous results and correspond to the chemisorption of one oxygen atom on two Ce3+ ions (predicted value. -7.3 x lo-')). The smaller value observed for the slope was shown to be due to a slight excess of adsorbed oxygen. Effectively. after 15 h, the totdl quantity of adsorbed oxygen on CeO,,-, was 820 pmol g- '. a value higher than the 785 pmol necessary for total reoxida- tion. After evacuation, the final magnetic susceptibilit), w;ts 0.02 lo-' emu (cgs) g-'. which confirms the almost complete reoxidation. It indicates that for limited reduction percent- 11.5 11.21 "? 11.0 I I I I I I I 10.5 I 100 110 120 r'pm Fig.9 Unit-cell parameter. ti. of cubic Me20, extracted from the JCPDS-ICDD file rs. the metal ionic radius for diffttrznt lanthanides" (Lalency 111. six oxygen nearest neighbours) 0 200 40C 63'1 800 1000 No2.pmol 9 ' Fig. 10 Magnetic susceptibilitj changes 15 the number of oxqgen nacllrcule\ [,V(O,)] adsorbed at 294 K on reduced CeO, ((11 (e0, ,,.lhlCrO, HZ5.10Ce0,-7 ages. loiver than about 60",o, the reoxidation of the reduced cma to CeO, occurs in the same manner. independent of the rt.duction extent and of the texture of the initial ceria. 3rirdj. 01 the Reosiduriorr Procwv hj. X-Ruj, Dlfrucrion 7 he reoxidation of reduced ceria was also followed by X-ray diflraction.As mentioned above. the sample was reduced by hvdrogen in a separate device. cooled at room temperature under hydrogen and transferred under argon into the X-ray cell. Then. several successive patterns were recorded after it~troductionof increasing pressures of air into the cell. For the cerias reduced at 673 or 873 K, i.e. with a rt,duction degree lower than 60°;1.the introduction of air at room temperature progressively reoxidizes the solids. For p,irtial reoxidation, the patterns of both the expanded and the normal ceria were observed. For the sample reduced at 8'3 K, no modification of the lattice of the expanded ceria ~a?;evidenced. After 24 h under air. only the initial CeOz phse was observed.These results indicate that the migration 0,' oxygen species into the bulk is effective at room tem-p1:rature. We can suppose that the reoxidation occurs pro-gressii.ely into the bulk. with. inside the same particle. bl)undaries between the two phases, CeOz and expanded CeO,. There would not have been a progressive modification oi (he lattice of the expanded ceria during the adsorption of o\!gen. However, our results can also be explained simply by mass-transfer limitation. the inner particles of the bed not being reoxidized owing to a lack of oxygen. The reoxidation of samples reduced at 1073. 1173 and 1.'73 K led to the following observations: (i) The introduction ot air on the sample modifies quick]) and deeply the spectra ol the reduced phases.and after one day under air, the e) panded ceria and the cubic Ce,O, have completely disap- pt.ared. (ii) There is no evidence of a progressive modification ot the lattice of the expanded ceria. (iii) Even after two days utider air at room temperature. the normal hexagonal Ce,O, pliase is always detected. as shown in Fig. 11. N'e can conclude that. when the hexagonal Ce,O, phase is formed during the reduction. it is impossible to obtain com- plete reoxidation at room temperature. On the other hand. the dilatation of the lattice to generate the expanded ceria wrlich maintains the initial cubic structure is very favourable for reoxidation of the solid b) oxJgen. These experimental a b L I I 1 I I 1 I 10 15 20 25 28/degrees Fig.11 XRD pattern of CeO, reduced at 1173 K and then left under air for 48 h at room temperature. The letters a and b corre-spond to the normal CeO, phase and to the hexagonal Ce,O, phase, respectively. facts could be important for the redox processes occurring in the three-way catalytic converters during the rich and lean- burn steps. Discussion The influence of the surface area on the reduction of ceria by hydrogen has been observed previ~usly.~~’*’ ’ The present results give more quantitative information and allow charac- terisation of several points during the reduction processes. When the reduction is performed by isothermal steps, we have evidenced a plateau for the Ce3+ content between 620 and 720 K, as seen in the magnetic susceptibility data.An almost linear relationship was found between the degree of reduction calculated at this plateau and the initial BET surface area. The corresponding slope is very close to that which can be predicted assuming a compact model of capping oxygen ions, with one fourth of these ions eliminated upon reduction. From this result, we have concluded that the first reduction step corresponds to that of the surface alone. The TPR results confirm this conclusion. Two main reduction processes by hydrogen were observed. They corre- spond, respectively, to the surface and bulk oxygen reduction. The corresponding binding energies of these oxygen species in the solid are different enough to lead to well separated phenomena during the reduction.This model is supported by the fact that whatever the initial surface area of the ceria sample, the temperature of the maximum of the first peak remains constant, at about 810 K. An examination of the small shoulders in the first composite TPR peak has shown that the surface reduction is, in fact, more complex and involves the reduction of different surface species, the popu- lation of which may vary from one oxide to another. Thus, the initial reduction process results from germination and surface diffusion steps, which progressively involves the whole surface. This model of reduction corresponds to a cherry-like attack scheme where after the completion of the surface reduction, the following step is the bulk reduction at higher temperatures.One can deduce from our results that the oxygen ion mobility in the bulk is low as long as the initial lattice of CeO, remains non-expanded. In this general scheme, the changes, with texture, of the surface oxygen binding energy distributions are probably limited and do not affect the TPR results. This conclusion is important in the study of modified ceria samples. As noted above, any change observed in the position and the shape of the first TPR peak will be directly attributed to the modification of the chemistry of the ceria surface. This is the case in the presence of anionic species J. CHEM. SOC. FARADAY TRANS., 1994, VOL 90 (after impregnation),28 residual or doping ele- ments in the bulk.Such modifications are also well known when a metal is present on the ceria surface, for which much lower temperatures are observed for the reduction of the surface.Q. 18.29.30 A calculation by two methods of the number of surface oxygen ions using the first TPR peak was presented above. The values are lower than that obtained by magnetic mea- surements or with the m0de1.~ The numerous parameters which influence the validity of the TPR results have been already discussed in detail.16 Another question must also be considered. Indeed, we have observed in the case of highly dispersed cerias that an important sintering occurred after a partial reduction, as soon as the treatment temperature is higher than 720 K2’ As the maximum of the TPR curve is located at 810 K, it is quite probable that, under hydrogen and at this temperature, the actual surface of the catalyst is lower than the initial one.Under these conditions, the quan- tity of hydrogen calculated from TPR for highly dispersed samples may be systematically lower. For low surface areas, this phenomenon is probably limited, but the precision of the measurement is poor because the deconvolution of the peaks is difficult and no clear conclusion can be established. However, in spite of these limitations, it remains that both the TPR and magnetic results of Fig. 2 and 4 can be usefully applied to estimate the initial surface area of an unknown ceria sample. The case of the highly dispersed ceria C1-400 was found by magnetic analysis to be unusual.The degree of reduction observed at the plateau was higher than expected from Fig. 4. As already noted, this may be caused by a higher germination rate and diffusion processes. It could also indicate that the energy required for bulk diffusion is lower in the case of high- surface-area ceria; this seems plausible in view of the defects in dispersed solids. This effect could also be due to the reduction and the elimination of the inner polycarbonate species which may be already effective at this temperature and would deeply modify the top surface layers.” Although the magnetic method shows some stable inter- mediate compositions during the reduction (Fig. 3), no evi- dence of a definite suboxide phase was obtained.For moderate reduction percentages, there is only one phase observed by X-ray diffraction which has the expanded cubic structure of CeO, . No evidence for an initial ceria phase was found, as could have been the case if the bulk reduction had progressed by zones. This suggests that as soon as the expanded ceria is obtained, the mobility of the oxygen ions becomes high in the lattice. This is very well illustrated by the facile reoxidation of such samples upon introduction of oxygen at room temperature. When the reduction extent is increased, the hexagonal Ce,O, phase is nucleated, with the intermediary formation of a cubic Ce203 phase. The hexagonal lattice of Ce203 is more stable. It presents many fewer anionic vacancies than the expanded cubic lattice of CeO, -x.Consequently, the diffu- sion of oxygen ions will need a much higher activation energy and the reoxidation process will not be completed at room temperature as soon as the hexagonal Ce,O, phase is formed. This fact could be of some importance in the three- way catalysts in which the gas composition changes with high frequency (0.5-5 Hz), as soon as the redox processes are not limited to the surface. Conclusions The interaction of hydrogen with cerias of different surface areas was followed by TPR, in situ magnetic susceptibility measurements and X-ray diffraction. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 781 By the magnetic method, a stable reduced state is found between 620 and 720 K.The degree of reduction obtained at this plateau varies linearly with the BET surface area. The corresponding slope is very close to that which can be calcu- lated assuming a compact model of capping oxygen ions with one fourth of the surface ions eliminated upon reduction. It is concluded that the formation of one Ce3+ ion layer on the surface corresponds to a stable intermediate state. The TPR data confirm this conclusion. Although the agreement with the predicted value is poor, a linear relationship could be obtained between the BET surface area and the hydrogen consumption during the low-temperature reduction compos- ite peak. These results open the possibility of determining the contribution of the ceria surface in some complex systems containing cerium dioxide.For temperatures between 720 K and about 900 K, which correspond to reduction extents probably lower than 50-60%, the bulk reduction preserves the initial structure of the 5 6 7 8 9 10 11 12 13 14 15 P. Loof, B. Kasemo and I(.E. Keck, J. Catal., 1989,118, 339. H. C. Yao and Y. F. Yao, J. Catal., 1984,86,254. M. F. L. Johnson and J. Mooi, J. Catal., 1987, 103, 502; 1993, 140,612. S. H. Oh and C. C. Eickel, J. Catal., 1988, 112, 543. J. Barrault, A. Allouche, V. Paul-Boncour, L. Hilaire and A. Percheron-Guegan, Appl. Catal., 1989,46,269. K. Otsuka, M. Hatano and A. Morikawa, J. Catal., 1983, 79, 493. K. Otsuka, M. Hatano and A. Morikawa, Znorg. Chim. Acta, 1985,109, 193. J. G. L. Fierro, J. Soria, J. Sanz and J. M. Rojo, J. Solid Stare Chem., 1987,66, 154. S.Bernal, F. J. Botana, R. Garcia, Z.Kang, M.L. Lopez, M. Pan, F. Ramirez and J. M. Rodriguez-Izquierdo, Catal. Toda,y, 1988,2, 653. C. Li, Y.Sakata, T. Arai, K. Domen, K. Maruya and T. Onishi, J. Chem. SOC., Faruday Trans. I, 1989,85,929; 1451. A. Laachir, V. Perrichon, A. Badri, J. Lamotte, E. Catherine, J. C. Lavalley, J. El Fallah, L. Hilaire, F. Le Normand, 13. Quemkre, G. N. Sauvion and 0.Touret, J. Chem. SOC.,Faraday cubic ceria but an appreciable expansion of the lattice is observed. Under these conditions, complete reoxidation can be effective at room temperature. For higher reduction per- centages, the hexagonal Ce203 phase is formed and its pres- ence decreases the reversibility of the redox process since most of the Ce,O, phase remains stable under air at room 16 17 18 19 Trans., 1991,87, 1601.F. M. Z. Zotin, L. Tournayan, J. Varloud, V. Perrichon and R. Frtty, Appl. Catal. A: Gen., 1993,98,99. B. C. Lippens and J. H. de Boer, J.Catal., 1965,4,319. L. Tournayan, N. R. Marcilio and R. Frety, Appl. Catal., 1991, 78, 31. J. P. Candy and V. Perrichon, J. Catal., 1984,89,93. temperature. Finally, a new intermediate phase, cubic Ce203, was observed on the ceria samples reduced at 1070-1 170 K. 20 21 J. C. Lavalley and M. Waquif, personal communication. J. C. Lavalley, A. Badri, E. Catherine and J. Lamotte, personal communication. The authors are indebted to F. Luck and E. Quemere for helpful discussions. Prof. S. Bernal is acknowledged for numerous discussions, as well as Dr. G. A. Martin and Mrs. F. M. Z. Zotin. Part of this research was done with the finan- 22 23 ,24 A. Laachir, Doctoral Thesis No 241-91, University of Lyon I, 1991. 0.T. Sorensen, J. Solid State Chem., 1976,18,217. S. P. Ray, A. S. Nowick and D. E. Cox, J. Solid State Chem., 1975,15,344. cial support of the M.R.E.S. The authors also thank P. Moral for his technical assistance. 25 26 G. Brauer and K. A. Gingerich, J. Znorg. Nucl. Chem., 1960, 16, 87. R. D. Shannon, C. T. Prewitt, Acta Crystallogr., Sect. B, 1969, 25, 295; 1970, 26, 1046; R. D. Shannon, Acta Crystallogr., Sect. References D. A. Johnson, in Advances in Inorganic Chemistry and Radio- chemistry, ed. H. J. Emeleus and A. G. Sharpe, Academic Press, London, 1977, vol. 20, p. 1. M. P. Rosynek, Catal. Rev. Sci. Eng., 1977, 16, 111. R. Korner, M. Riken, J. Nolting and I. Riess, J. Solid State Chem., 1989,16,136. B. J. Cooper, W. D. J. Evans and B. Harrison, Catalysis and Automotive Pollution Control, ed. A. Crucq and A. Frenet, Else- vier, Amsterdam, 1987, vol. 1, p. 117. 27 28 29 30 A, 1976,32,751. A. F. Wells, Structural Inorganic Chemistry, Oxford University Press, London, 3rd edn., 1962, p. 464. J. Barbier, Jr, F. Marsollier and D. Duprez, Appl. Catal., 1992, !m,11. S. Bernal, J. J. Calvino, G. A. Cifredo, J. M. Rodriguez-Izquierdo, V. Perrichon and A. Laachir, J. Catal., 1992,137, 1. A. Trovarelli, G. Dolcetti, C. Leitenburg, J. Kaspar, P. Finetti and A. Santoni, J. Chem. SOC., Faruday Trans., 1992,88,1311. Paper 3/05314H;Received 6th September, 1993