Computational Modeling Study of the Solubility of Cerium at LaCoO3

Jul 23, 2008 - Saira Khan, Richard J. Oldman, C. Richard A. Catlow*, Samuel A. French and Sean A. Axon. Davy Faraday Research Laboratory, The Royal ...
0 downloads 0 Views 2MB Size
12310

J. Phys. Chem. C 2008, 112, 12310–12320

Computational Modeling Study of the Solubility of Cerium at LaCoO3 Perovskite Surfaces Saira Khan,† Richard J. Oldman,† C. Richard A. Catlow,*,† Samuel A. French,‡ and Sean A. Axon§ DaVy Faraday Research Laboratory, The Royal Institution of Great Britain, 21 Albemarle Street, London, W1S 4BS, Johnson Matthey Plc., Technology Centre, Blounts Court, Sonning Common, Reading, RG4 9NH, and Johnson Matthey Plc., Technical Centre, Belasis AVenue, Billingham, TS23 1LB ReceiVed: October 2, 2007; ReVised Manuscript ReceiVed: April 27, 2008

Atomistic computational modeling of the surface structure of the catalytically active perovskite LaCoO3 is reported in order to understand the effect of doping with tetravalent cerium cations, which enhances hightemperature catalytic oxidation processes such as CH4 combustion. In particular, solution energies have been calculated for the important (100) and (110) low index crystal faces. Three reactions for Ce4+ doping have been considered, two of which involve creation of La3+ vacancies or Co3+ reduction in stoichiometric LaCoO3, whereas the third relates to oxygen vacancy filling in reductively nonstoichiometric LaCoO3. We show that Ce4+ is considerably more soluble at the surface than in the bulk to a level of ∼5 atom %, which agrees with experimental estimates by X-ray diffraction. We predict that the nature of the defect compensating reaction will be strongly dependent on oxygen partial pressure, and hence on the preparation conditions. The effect of Ce4+ doping on catalytic oxidation is discussed in terms of the relative redox behavior of Ce4+ and Co3+ and the availability of oxygen vacancies and surface lattice oxygen. 1. Introduction Lanthanum perovskite oxides have received considerable attention for a number of applications, for example, as hightemperature oxygen ion and proton conductors in solid oxide fuel cells, electrochemical applications, hydrogen membranes, and hydrogen sensors.1 These oxides are also catalytically active in the oxidation of CO and the reduction of NO in auto exhaust treatment.2 Other important catalytic reactions include hydrocarbon combustion,3 ammonia oxidation,4,5 and the oxidative coupling of methane to form C2 hydrocarbons, which is important in utilizing the abundant reserves of natural gas.6 Of particular interest here are new high-temperature oxidation catalysts, which take advantage of the outstanding structural and thermal stability of lanthanum perovskites under aggressive reaction conditions. The mechanisms of oxidation at the catalytically active sites in the lanthanum perovskites are, however, still open to considerable debate. Depending on the catalyst preparation and operating conditions, the materials can exhibit different behavior, and the inclusion of dopants in the perovskite lattice adds further complexity. However, it has become clear that two types of oxygen play a vital role in the catalysis. Suprafacial oxygen has been identified as surfaceadsorbed oxygen in equilibrium with gas phase oxygen and is the dominant reactant species at low temperature (550 °C), due to the increase in lattice oxygen anion mobility in these materials, intrafacial lattice oxygen becomes available for catalysis and an ion-redox catalyzed reaction becomes possible.7–9 Within this descriptive framework, there can be a number of catalytic oxidation regimes that vary with temperature depending on the identity of the perovskite B site transition metal, the inclusion * Corresponding author e-mail: [email protected]. † The Royal Institution of Great Britain. ‡ Johnson Matthey Plc., Technology Centre. § Johnson Matthey Plc., Technical Centre.

of dopants, particularly at the La site, and the specific reaction under consideration. Of special interest is LaCoO3, whose catalytic behavior can be modified by both divalent ion (e.g., Sr2+) doping at the La site and tetravalent ion (Ce4+) doping. Although the effect of Sr2+ is relatively well understood experimentally and theoretically in terms of compensating oxygen ion lattice vacancies,10,11 the interpretation of the effect of Ce4+ doping is complicated by the formation of mixed phase materials with the segregation of CeO2.12 Further complications arise when comparing results in the literature, due to different preparation methods, different reaction temperature regimes, and different ways of reporting catalytic activity. Examples of this are the measurement of conversion levels at fixed temperatures or temperatures for specified levels of conversion (frequently 50%) from light-off curves, without the calculation of specific reaction rates. Thus, doping with low levels of Ce (5% atomic, i.e., x ) 0.05 in La1-xCexCoO3) appears to lead to a small drop in catalytic activity for CH4 combustion under conditions of low conversion (50%) shows that, if reaction rates are calculated and normalized to surface area, the derived specific catalytic activity is significantly increased for Ce dopant levels of only 5%, and there is no further increase for higher formulation levels of Ce.13 Furthermore, this work has also shown that the phase-segregated CeO2 would have relatively low activity compared to LaCoO3. Enhancement in activity at high temperatures is therefore dominated by the Ce4+ dopant acting within the LaCoO3 lattice. To understand these effects, we need a detailed knowledge of the defect processes involved in Ce4+ incorporation in the lattice. As an initial step toward this end, in previous work, we have applied atomistic computer simulation methods to study the solution of Ce4+ in bulk LaCoO3.14 Three reactions were

10.1021/jp709638h CCC: $40.75  2008 American Chemical Society Published on Web 07/23/2008

Solubility of Cerium at LaCoO3 Perovskite Surfaces

J. Phys. Chem. C, Vol. 112, No. 32, 2008 12311

TABLE 1: Calculated Solution Energies for Solubility of Ce4+ in Bulk LaCoO3 reaction stoichiometric, lanthanum vacancy creation (1) stoichiometric, Co reduction (2) nonstoichiometric, oxygen vacancy filling (3)

TABLE 2: Interatomic Potentials for Cation Doping Simulations in LaCoO3 Perovskites24–26a

solution energies (eV) 2.89

considered for cerium doping, the first two of which initially assume a stoichiometric host LaCoO3 lattice where the substitution of Ce4+ for La3+ is compensated either by creation of La3+ vacancies (eq 1) or reduction of Co3+ to Co2+ (eq 2) as shown using the Kroger-Vink notation.

(1)

x x • ′ + 2La O + O + 4CoCo f 4CeLa + 4CoCo 4CeO2 + 4LaLa 2 3 2

(2) However, as-prepared LaCoO3 after calcination at g900 °C is typically slightly reductively nonstoichiometric.1 Therefore, a third case was considered in the previous work, where cerium is doped into the LaCoO3 lattice containing oxygen vacancies, resulting in filling of those vacancies according to eq 3. x • 2LaLa + 2CeO2 + VO•• f 2CeLa + La2O3 + OOx

(3)

A computational study of the reductively nonstoichiometric material allows us to relate doped LaCoO3 to the material that is normally prepared via high-temperature calcinations for catalytic applications. The previously reported calculated solution energies for the three reactions above are given in Table 1. From the solution energies for the first two reactions it is clear that Ce4+ will have low solubility in stoichiometric LaCoO3. However, the markedly lower solution energy for the third mechanism shows that, if Ce4+ is inserted into the reductively nonstoichiometric and oxygen deficient lattice of LaCoO3, then there is much greater solubility. Thus, the amount of Ce incorporation depends on the level of reductive nonstoichiometry induced under conditions for a typical high-temperature preparation, which has been estimated as ∼5% (x ) 0.05).15,16 Higher formulation levels of Ce lead to segregation of CeO2. The effect of a few percent of Ce in the LaCoO3 lattice on intrafacial oxidation catalysis has been interpreted14 in terms of an increased level of oxygen available via the ion-redox mechanism shown in eq 4,

2Co3+ + OOx S 2Co2+ + VO•• + 1⁄2O2 oxidation.8,9

A/eV

F/Å

1545.21 0.3590 · · ·O Co3+ · · · O2- 24 1329.82 0.3087 O2- · · · O2- 25 22764.30 0.1490 Co2+ · · · O2- 26 3799.3 0.24273 1986.83 0.3511 Ce4+ · · · O2-

La3+

2.69 0.23

x • ′′′ 3CeO2 + 4LaLa f 3CeLa + VLa + 2La2O3

short-range parameters interaction

(4)

which dominates high-temperature However, the Ce4+ solution energy calculations previously reported14 relate to the bulk material, whereas catalysis is normally considered to be a largely surface process, although, in the case of perovskite oxides, subsurface processes relating to oxygen availability for intrafacial catalysis are also important.7,9 Therefore, in the work reported here, we have carried out computer simulations to investigate the solubility of Ce4+ at the surface of LaCoO3, in order to clarify the role of this dopant in the catalytic processes on the material. For this purpose we have chosen interatomic potential-based methods rather than density functional theory (DFT), in line with our previous calculations on oxygen vacancy creation and Sr2+ doping at the LaCoO3 surface,11 as these methods allow us to study a far wider range of systems and reactions; moreover,

a

2- 24

shell model

C/eV Å-6

Y (|e|)

k/eV Å-2

0.0 0.0 43.0 0.0 20.40

-0.25 2.04 -2.38 3.503 7.7

145.0 196.3 42.0 110.5 291.75

Cut-off for short range potentials: 20 Å.

they are of proven reliability for modeling defect surface properties of oxides. 2. Computational Methods The program General Utility Lattice Program (GULP)17 has been used for the simulations of bulk LaCoO3. These simulations incorporate the Born model of ionic solids in which the dominant long-range interactions are Coulombic. Buckingham potentials describe the combination of the short-range repulsion between neighboring electron clouds with van der Waals attraction.18 Because we are simulating defects, which polarize the surrounding lattice, it is necessary to include a representation for ionic polarizability. This effect is taken into account using the shell model of Dick and Overhauser, which treats each ion as a core and shell coupled by a harmonic spring.19 Defects structures and energies were modeled using the Mott Littleton method, which explicitly relaxes ions within an inner spherical region surrounding the defect with polarization of the outer region being calculated via a quasi continuum approximation.20 The sizes of the inner 151 regions were 10 Å in the present 152 calculations. The MARVIN21 program has been used for the simulations of the surfaces of LaCoO3. The program models the system as a two-dimensional (2D) slab in which periodic boundary conditions are applied. The slab consists of one or more blocks, which are split into two regions (I and II). Each region consists of structural units, which may be ions or molecules. Surfaces were reconstructed to quench the dipole perpendicular to the surface plane,22 before energy minimization is performed. The structure and energy of the surface is then obtained by relaxing the ions to their equilibrium positions. To this end, region I includes the surface layer in which all ions are explicitly relaxed to zero net force, and region II includes those atoms more distant from the surface, which are kept fixed at their bulk equilibrium positions. The sizes of regions I and II (10 and 20 Å respectively) were chosen such that the surface energy changes by less than 0.01 J/m2 on expansion of either. All structures were then relaxed using the Broyden-Fletcher-GoldfarbShanon algorithm (BFGS).23 The interatomic potentials used in this study (Table 2) were employed from previous simulations of bulk LaCoO3.24 Interatomic potentials for the Ce4+ dopant are taken from Lewis and Catlow.25 To calculate the solution energies for mechanisms 1–4 discussed above, we need values of the second electron affinity of oxygen26 and of the third ionization potential of Co,27 as well as the lattice energies of CeO2 and La2O3. The values used are those used in our previous study of LaCoO3 surfaces11 and are given in Table 3; lattice energies were calculated using the standard procedure available in the GULP program employing the same potentials as used in the rest of the study; the values reported are obtained after full minimization of cell dimensions and unit cell coordinates.

12312 J. Phys. Chem. C, Vol. 112, No. 32, 2008 TABLE 3: Energy Values for the Born Haber Cycle for Reactions 1–4 energy values (eV) electron affinity of oxygen third ionization potential of Co3+ lattice energy for CeO2 lattice energy for La2O3

-9.3626 33.5727 -105.64 -127.34

The surface structure of stiochiometric LaCoO3 has been studied by Read et al.10 employing atomistic computational methods. In an earlier study, we reproduced the results of these simulations as part of an investigation of nonstoichiometric surfaces containing oxygen vacancies.11 In all cases, surface simulations of the three low index surfaces, (100), (110), and (111), were carried out in MARVIN using a (2 × 1) expansion of the primitive surface unit cell that was required for the initial reconstruction of the surfaces to remove the dipole moments perpendicular to the cut surface as discussed by Tasker.22 To achieve insensitivity to region size, the sizes for regions I and II required for simulation of the (100) and (110) surfaces were 6 and 12 unit cells and for the (111) plane, 7 and 16 unit cells, respectively. Details of the relaxation processes in the reconstructed surfaces that occur during the energy minimization are discussed in the previous studies,10,11 but to summarize, the relaxations are dominated by the greater energetic constraints required to rearrange CoO5 and CoO6 polyhedra compared to the lanthanum coordination environment. Surface energies have been calculated as in previous studies.10,11 The calculation for nonstoichiometric surfaces, for example, containing La vacancies as in eq 1, raises an interesting and important matter of definition. The definition of surface energy as the excess energy in the surface block compared to the bulk relates to equal numbers of ions in the surface and bulk. The surface simulations as carried out in MARVIN for a system containing defects compares the surface block with undoped, bulk material. Effectively, this represents the surface energy of the doped system compared to a totally segregated system of separate phases. The procedure previously applied11 and repeated here, which involves 2 × 2 × 2 supercell calculations in GULP to correct the bulk energy for the presence of defects, effectively gives a measure of the energy difference between surface-doped and bulk-doped material. This latter method has been adopted throughout the present work because it best meets the definition of surface energy in the present context. 3. Ce Doping in Bulk LaCoO3 In our previous study,14 solution energies for Ce in bulk LaCoO3, both stoichiometric and nonstoichiometric, were derived by a Mott Littleton approach involving the calculation of energies for individual, isolated defects. Here, we have calculated energies following insertion of clusters of compensating defects, which takes into account the interactions between defects. For example three lanthanum ions were replaced by cerium ions, and a lanthanum vacancy was created in the stoichiometric structure according to eq 1. Solution energies were calculated for doping Ce into stoichiometric LaCoO3 as in eqs 1 and 2 and for the reductively nonstoichiometric material, where oxygen vacancies are refilled as in eq 3. Calculations were performed for different dispositions of the compensating defects. The results for stoichiometric LaCoO3 show lower solution energies when compensating defects are closest together, as would be predicted from simple electrostatic considerations. For the nonstoichiometric case, the solution energies are calculated relative to the fully relaxed, nonsto-

Khan et al. TABLE 4: Comparison of Solution Energies for Ce4+ in Bulk LaCoO3 Calculated from Isolated Defect and Cluster Mott-Littleton Models Ce4+ solution energies (eV) reaction

isolated defects

defect clusters

stoichiometric, lanthanum vacancy creation (1) stoichiometric, Co reduction (2) nonstoichiometric, oxygen vacancy filling (3)

2.89

2.70

2.69

2.64

0.23

0.95

TABLE 5: The Surface Energies Calculated for Stoichiometric and Nonstoichiometric LaCoO3 (100), (110), and (111) Planes calculated relaxed surface energies (Jm2-) Miller index (100) (100) (110) (110) (111) (111)

OLa OCoO O LaOCo Co LaOOO

stoichiometric surface

nonstoichiometric surface

2.16 2.69 2.01 3.84 3.69 3.09

1.50 0.79 0.88 1.96 1.51 1.65

ichiometric material. In this example, the most energetically favorable structure has the Ce4+ cations closest to the refilled vacancy and the Co2+ ions as next-nearest neighbors. The solution energy for the model where the cations were closest together is substantially higher. The results reported in Table 4 shows that the inclusion of defect interactions has very little effect on reaction 1 and 2, whereas the energy of reaction 3 is raised as oxygen vacancies in the undoped material are stabilized by interaction with reduced Co2+ ions. However, the solution energy for doping reduced, oxygen vacancy containing, nonstoichiometric LaCoO3 with consequential filling of those vacancies is still considerably more favorable than for doping in the stoichiometric material. 4. Surface Structure of LaCoO3 Surface energies calculated by the MARVIN program for the stoichiometric LaCoO3 (100), (110), and (111) planes for each of the terminations discussed in our earlier work10,11 are presented in Table 5. These include the (100) OLa, the (100) OCoO, and the (110) O terminated surfaces. The results show that the most stable relaxed surfaces are the (110) O terminated and (100) OLa terminated structures. Unrelaxed and relaxed structures for these two surfaces are shown in Figure 1 together with those for the third most stable surface, which is (100) OCoO terminated. In the case of nonstoichiometric surfaces containing oxygen vacancies, the latter were created by removing one oxygen anion from the surface repeat unit of the stoichiometric structures listed in Table 5, accompanied by reduction of Co3+ atoms to Co2+ to maintain a charge balance according to eq 4 above. We calculated the defect energies for creating oxygen anion vacancies as a function of depth below the exposed surface and disposition of the compensating defects (Co2+). Again, the details of the relaxation processes during energy minimizations in the oxygen vacancy containing nonstoichiometric surfaces have been fully discussed previously.11 However, we should note here that in some surface terminations the most stable structures involve creation of an oxygen vacancy immediately

Solubility of Cerium at LaCoO3 Perovskite Surfaces

J. Phys. Chem. C, Vol. 112, No. 32, 2008 12313

Figure 1. The unrelaxed and relaxed structures for the stoichiometric (a) (100) OLa, (b) (100) OCoO and (c) (110) O terminated surfaces (lattice sites are color coded as La3+, yellow; O2-, red; Co3+, green).

below the top layer of ions as in Figure 2. Nevertheless, in general, the creation of oxygen vacancies is energetically more favored at the surface compared to the bulk. Furthermore, as might be expected from simple electrostatic considerations, the compensating defects, O2- vacancies and Co2+ cations, form clusters. In addition, the relaxation process for some of the surfaces is dominated by the transition from an unstable, low oxygen coordination of Co ions, resulting from removal of oxygen anions, toward a more stable, higher coordination. Table 5 lists the calculated surface energies for the three low index, defect energy minimized surfaces, (100), (110), and (111) for each of the possible terminations. The data in Table 5 shows that the most stable oxygen vacancy containing surfaces are the (100) OCoO terminated and the (110) O terminated planes. The relaxed and unrelaxed structures for these two surfaces are shown in Figure 2 together with the (100) OLa terminated surface identified above for the stoichiometric LaCoO3 material.

5. Solubility of Cerium at the Surface of LaCoO3 The relaxed stoichiometric and oxygen vacancy-containing structures shown in Figures 1 and 2 (100) OLa, (100) OCoO, and (110)O terminated, which represent the most stable surfaces for stoichiometric and nonstoichiomtric LaCoO3 were selected for further computer simulation using the MARVIN program to calculate the solution energy of Ce substituted at the La site, according to the three mechanisms discussed earlier. We should note here that the atomic concentration levels for the various doping regimes are not the same since we have created models represented by the simplest integer, stoichiometric ratios in eqs 1–3. Thus, in eq 1, three Ce4+ require creation of one La3+ vacancy; in 2, one Ce4+ is balanced by one Co2+; and in eq 3, two Ce4+ are required to balance filling one O2- vacancy with two Co2+ remaining, as in eq 4. Therefore the products of eqs 2 and 3 contain different local concentrations of Ce4+ and Co2+

12314 J. Phys. Chem. C, Vol. 112, No. 32, 2008

Khan et al.

Figure 2. The unrelaxed and relaxed structures for the nonstoichiometric (a) (100) OLa, (b) (100) OCoO, and (c) (110) O terminated surfaces of LaCoO3 containing O2- vacancies/Co2+ cations (lattice sites are color coded as La3+, yellow; O2-, red; Co3+, small, light green spheres; Co2+, large, dark green spheres; O2- vacancy, blue).

TABLE 6: Solution Energies for Doping Cerium in the Bulk and at the Surface for Stoichiometric and Nonstoichiometric LaCoO3 Ce4+ solution energies (eV)

TABLE 7: Surface Energies for Cerium Doped Stoichiometric and Nonstoichiometric LaCoO3 Surfaces surface energies (J/m2) solution mechanism

surfaces solution mechanism

bulk

(100) OLa

(100) OCoO

(110) O

stoichiometric, La vacancy creation (1) stoichiometric, Co reduction (2) nonstoichiometric, filling oxygen vacancies (3)

2.70

0.64

2.27

2.74

2.69

0.99

0.03

0.69

0.95

1.55

1.34

0.97

and will not be geometrically or energetically equivalent. The calculated solution energies per Ce4+ ion are presented in Table 6, and the corresponding surface energies are tabulated in Table 7. 5.1. Stoichiometric LaCoO3. 5.1.1. Creation of La Vacancies. According to eq 1 above, substitution of three Ce4+ ions at La3+ sites results in the creation of one lanthanum vacancy in order to maintain a charge balance. Figure 3 shows Ce doped in the LaCoO3 surface lattice in this way for each of the three surfaces selected for this study,

stoichiometric, La vacancy creation (1) stoichiometric, Co reduction (2) nonstoichiometric, filling oxygen vacancies (3)

(100) OLa

(100) OCoO

(110) O

1.15

1.52

1.63

0.71

0.60

0.64

0.99

0.89

0.37

together with the relaxed structures after energy minimization of the Ce-doped surfaces. The solution energies in Table 6 show that the OLa terminated surface is the most energetically favored for the dissolution of cerium, accompanied by the creation of compensating lanthanum vacancies. This process at the (100)OLa surface is also considerably more favorable than in the bulk. This effect might be expected for the (100)OLa termination as the lanthanum vacancy is created at the exposed surface which is likely to be of lower energy compared to the other terminations where the lanthanum vacancy is below the surface. In addition, calculation of the Ce-doped (100)OLa terminated surface, where the La

Solubility of Cerium at LaCoO3 Perovskite Surfaces

J. Phys. Chem. C, Vol. 112, No. 32, 2008 12315

Figure 3. The inital and relaxed structures for the Ce-doped (a) (100) OLa, (b) (100) OCoO, and (c) (110) O terminated surfaces of stoichiometric LaCoO containing La3+ vacancies (lattice sites are color coded as La3+, yellow; O2-, red; Co3+, green; Ce4+, violet).

vacancy was created below the exposed surface, gives a solution energy of 2.32 eV, which is considerably higher compared to the solution energy where the vacancy is at the surface. Dissolution of Ce by this mechanism therefore preferentially occurs at the surface. Table 7 shows that the Ce-doped (100) OLa termination has the lowest surface energy for mechanism 1, confirming that the (100) OLa termination is the most favored plane for this mechanism. On comparing with data in Table 5 it is clear that the surface energies for the undoped stoichiometric surfaces are higher compared to the surfaces energies for the cerium-doped material. Doping with cerium, in general, stabilizes the surfaces. 5.1.2. Reduction of Co3+. The second mechanism for the dissolution of Ce in stoichiometric LaCoO3 involves reduction of one Co3+ to Co2+ for each Ce4+ substituted at a La3+ site, as in eq 2. Figure 4 shows Ce doped in the LaCoO3 surface lattice for this mechanism for each of the three surfaces selected for this study, together with the relaxed structures after energy minimization of the Ce-doped surfaces. The solution energies in Table 6 show that the energy of this process is lower at the surface compared with the bulk for all

three surface terminations studied, with (100) OCoO being significantly more favored compared to the others. We note, however, for the (100) OLa termination, when cerium was doped at the exposed surface, that the solution energy increased to 1.10 eV and that in the lowest energy structure for the (100) OLa termination cerium doped below the exposed surface. It might be expected that the OCoO termination would be more favored compared to the other terminations as the reduced cobalt is exposed at the surface. For comparison, the defects (cerium dopant and reduced Co2+) were positioned a layer below the exposed (100) OCoO surface. The solution energy increased to 1.97 eV, that is, approaching the bulk value and confirming that Ce doping was compensated by reduction of Co3+ to Co2+, which preferentially occurs at the surface. A comparison of surface energies for reaction 2 in Table 7, with data for undoped, stoichiometric material in Table 5 again indicates a substantial decrease in surface energy, in particular for the (100) OCoO surface. The energy difference between bulk and surface doping may stem from the greater ease of relaxation at the surface to generate a more stable structure.

12316 J. Phys. Chem. C, Vol. 112, No. 32, 2008

Khan et al.

Figure 4. The inital and relaxed structures for the Ce-doped (a) (100) OLa, (b) (100) OCoO, and (c) (110) O surfaces of stoichiometric LaCoO3 containing Co2+ cations (lattice sites are color coded as La3+, yellow; O2-, red; Co3+, small, light green spheres; Co2+, large, dark green spheres; Ce4+, violet).

5.2. Nonstoichiometric LaCoO3 Surfaces. Solution and surface energies corresponding to Ce4+ solution according to eq 3 have been presented in Tables 6 and 7 together with the corresponding bulk solution energy data, and Figure 5 shows Ce doped in the relaxed nonstoichiometric LaCoO3 surface lattice according to eq 3 for each of the three surfaces selected for this study, together with the relaxed structures after energy minimization of the Ce-doped surfaces. Upon comparing results for Ce doping in stoichiometric (eq 2) and nonstoichiometric (eq 3) LaCoO3, where Ce4+ is balanced by Co2+, we should recall that the models contain different local concentrations of defect cations and hence different structures as discussed above. The results for solution energies in Table 6 show that doping Ce into the reductively nonstoichiometric surfaces is no more energetically favorable than the equivalent process in the bulk but remains more favorable than for bulk dissolution processes in the stoichiometric material. Comparing the three selected surfaces, the solution energy data in Table 6 shows that dissolution of Ce4+ at the (110) O

termination is the most favored. The solution energy of 0.97 eV is similar to the value obtained for the process in the bulk. There are only small displacements of the ions for the (110) O termination as shown in Figure 5c, as the oxygen vacancy at the surface is filled. In comparison, Figure 5a shows that, for the (100) OLa termination, the most favored relaxed nonstoichiometric surface is created when the oxygen vacancy is situated between the two Co2+ ions. To fill the oxygen vacancy, the oxygen needs to be substituted in the sublayer, which requires more energy. Similarly, for the (100) OCoO termination, in the favored structure, the oxygen vacancy is below the exposed surface, as shown in Figure 5b. When the filled oxygen vacancy is substituted back into the nonstoichiometric relaxed surface it causes the exposed surface Co2+ ion to move toward the bulk, and this process is accompanied by considerable rearrangement of the other ions. Surface energies for nonstoichiometric material doped with Ce are shown in Table 7. The (110) O terminated surface has the lowest surface energy (0.37 J/m2), which corresponds to

Solubility of Cerium at LaCoO3 Perovskite Surfaces

J. Phys. Chem. C, Vol. 112, No. 32, 2008 12317

Figure 5. The inital and relaxed structures for the Ce-doped (a)(100) OLa, (b) (100) OCoO, and (c) (110) O terminated surfaces nonstoichiometric LaCoO3 resulting in filling of oxygen vacancies (lattice sites are color coded as La3+, yellow; O2-, red; Co3+, small, light green spheres; Ce4+, violet; Co2+, large, dark green spheres; O2- vacancy, blue; filled oxygen site, pink).

the lowest solution energy for Ce doping (0.97 eV), as in Table 6. However, the solution energy is no more favorable than for the bulk material. This is not surprising because our earlier calculations have shown that it is very easy to create oxygen vacancies reductively at the surface of LaCoO3;11 therefore, a process in the reverse direction, except under oxygen pressure, is unlikely to be energetically favorable. 6. Crystal Morphology For stoichiometric LaCoO3, the solution energy data in Table 6 indicate that Ce is considerably more soluble at the surface than in the bulk for both processes involving creation of La vacancies, reaction 1, or reduction of Co3+, reaction 2. In particular, for Ce doping compensated by La vacancies, the (100) OLa terminated surface (0.64 eV) appears to be most favorable, whereas for compensation by reduction of Co3+ the (100) OCoO terminated surface (0.03 eV) is energetically most favored. However, for this latter defect compensation process, the (100) OLa and (110) O terminated surfaces also show lower energies for Ce dissolution compared to the bulk. The solution energy data are mirrored in the surface energy data in Table 7.

In reconciling the data for solution energies and surface energies it is important to remember that the surface energy defines which faces will predominate in a crystal, and the solution energy determines how easily Ce will dissolve in a particular surface. For example, the surface energy data in Table 7, suggest that in a Ce-doped material the (100) OLa termination is likely to dominate the crystal morphology if La vacancy compensation dominates, whereas for reduction of Co3+ the (100) OCoO terminated surface is only marginally more stable than the other two faces studied. However, we should note that, in the undoped stoichiometric material, (100) OCoO was the least stable of the three selected surfaces (see Table 5), that is, the high degree of solubility of Ce4+ corresponds to a higher level of stabilization. Predicted morphologies, using the equilibrated morphology method as discussed in our earlier paper,11 for all three mechanisms for doping Ce4+, are shown in Figure 6. In the case of nonstoichiometric LaCoO3, where dissolution of Ce4+ results in filling oxygen vacancies, the solution energy data in Table 6 shows that the surface process is no more favorable than in the bulk. However, this dissolution mechanism

12318 J. Phys. Chem. C, Vol. 112, No. 32, 2008

Khan et al.

Figure 7. Idealized model for particles of Ce doped LaCoO3. Figure 6. The relaxed crystal morphologies for Ce-doped stoichiometric LaCoO3 (a) creation of La vacancies, (b) reduction of Co3+, and (c) filling oxygen vacancies in nonstoichiometric material.

results in a significant reduction in the surface energy compared to the nonstoichiometric material, as shown in Table 7. Therefore, it is probable that a significant fraction of the doped Ce will be concentrated at the surface. The resulting predicted crystal morphology is shown in Figure 6c, indicating dominance by the (110) O terminated surface. 7. Implications for Catalysis The data in Tables 6 and 7 represent a complex pattern of results. To interpret these data in terms of the implications for catalysis, three issues require discussion. The first is the extent to which Ce can be concentrated at the surface in order to have a significant effect on catalysis. The second concerns which of the several competing Ce dissolution processes represented by eqs 1–3 is most likely to occur during the preparation of real catalytic materials. Finally, possible reactions involving Ce in catalytic oxidation processes need to be explored. It is instructive, first of all, to quantify the potential levels of surface dopant Ce in real, polycrystalline LaCoO3 catalytic materials. In this respect, Kaliaguine et al. have carried out a series of LaCoO3 preparations as part of a study on oxygen exchange at LaCoO3 surfaces.28 In this work, the authors observed a strong correlation between calcination temperature, the development of crystallinity as expressed by the XRD pattern, the increase in crystallite size derived from X-ray peak widths according to the Scherrer equation, and the decrease in N2 adsorption and BET surface area. However, there was poor agreement between absolute, XRD crystallite size measurements and that derived from BET surface areas, which has been attributed to grain boundary contacts between crystallites making a substantial fraction of the surface inaccessible to adsorption of N2 molecules. It is important, therefore, to take the crystallite size from XRD to estimate surface dopant levels.28 Taking a typical catalytic LaCoO3 material calcined at 900 °C with a BET surface area of ∼1 m2/g29 and interpolating the data of Kaliaguine et al. leads to an estimate of crystallite size of ∼50 nm with a true surface area of ∼17 m2/g. For a crystallite of this size, the outer layer one unit cell deep is equivalent to ∼5% of the volume of the material. In relation to Ce doping, the point at which Ce segregation begins is not known, but the overall solubility with reference to the bulk, as determined by XRD, has been shown to be very low and no higher than 5% atomic,15,16 in accordance with the

bulk dissolution reaction proposed by French et al.14 and confirmed here. If we assume substitution at a level of 5% of the total La sites in stoichiometric LaCoO3 with respect to the bulk, nearly 7% of the total La sites would actually be involved in the situation where defect compensation occurs via creation of La vacancies according to eq 1, because for every three Ce4+ ions dissolved, one La3+ vacancy is created. This level of doping cannot be accommodated in the outer surface extending to a depth of one unit cell but requires a depth of two unit cells. A more realistic model would be an enriched surface layer with a Ce concentration gradient extending into the subsurface. This model, which has not been experimentally verified, is idealized in Figure 7. For comparison, the computational model with Ce substitution compensated by La vacancy creation in Section 5.1 assumes all of the surface La sites are involved. In the case of defect compensation by reduction of Co3+, a substitution level of 5% would involve 5% of the total La sites as in eq 2. This level could be accommodated in a surface layer one unit cell deep, but there have been no careful doping experiments supported by XPS or LEIS measurements to confirm this proposal. Again, a more realistic model would involve segregation of defects at the surface but with a concentration gradient, which extends into the subsurface as in Figure 7. For comparison, the computational model in Section 5.2, based on compensation by reduction of Co3+, assumes a lower level of Ce substitution equivalent to ∼33% of the surface La sites, corresponding to reduction of half the surface Co ions. Now let us consider the nonstoichiometric material, which typically has a composition between LaCoO2.94 and LaCoO3.30 This level of reductive nonstoichiometry corresponds to a loss of ∼1% of the total oxygen anions, whereas ∼6% of the total Co3+ cations are reduced to Co2+ as in eq 4. In an earlier paper, the results of which are summarized in Section 4, we have predicted that oxygen vacancies are preferentially created at the surface. However, the limiting factor in dispersion of the resulting oxygen vacancies is the availability of compensating Co2+ cations. In principle, the nonstoichiometry could be accommodated in the outermost layer to a depth of two unit cells. The presence of such a concentration of Co2+ at the surface has not been verified by careful XPS measurements. However, a more realistic model would be segregation of oxygen vacancies at the surface, compensated by Co2+, but with a gradient of nonstoichiometry extending toward the bulk. For comparison, the model used in the computational studies in Section 5.1.2 corresponds to a level of ∼13% of the surface oxygens removed as vacancies.

Solubility of Cerium at LaCoO3 Perovskite Surfaces

Figure 8. Intrafacial oxidation catalysis over LaCoO3 surfaces (a) non stochiometric; (b) air calcined, Ce doped; (c) stoichiometric; and (d) oxygen calcined, Ce doped.

In the case of Ce doping in nonstoichiometric LaCoO3, containing oxygen vacancies and Co2+ cations, two Ce4+ cations are required to balance the filling of one oxygen vacancy as in eq 3. The end result would represent double the local concentration of Ce4+ and Co2+ ions at the surface compared to the product of eq 2. Surface doping at this level would require a distribution of Ce4+ extending into the subsurface. The above discussion indicates that the formation of a Ceenriched surface on particles of LaCoO3 prepared by typical precipitation and calcination methods is feasible. The details in terms of concentration and dispersion will depend on calcination temperature, which controls crystallite size, that is, surface area, through sintering. For the various Ce doping reactions discussed above, and represented by eqs 1–3, caution is required in deciding which will be most energetically favored from the solution energy data in Table 6, since some of the parameters in Table 3 used to derive solution energies may not be sufficiently precise. The most important consideration in determining which process dominates is the oxygen partial pressure during material preparation. From eq 2 we can see that Ce doping compensated by Co reduction would be strongly favored at lower oxygen partial pressures. In the case of eq 3, where oxygen vacancies are filled in what would otherwise be reductively nonstoichiometric material under standard 900 °C, air calcination, conditions calculations indicate that surface Ce4+ doping is not energetically favored compared to the bulk. However, we would suggest that the presence of a high concentration of Ce4+ at the surface via eq 2, which is very much more favored than in the bulk, would inhibit formation of oxygen vacancies under these preparation conditions. To produce material where Ce4+ doping occurs with defect compensation by formation of La3+ vacancies as in eq 1, calcination under pure oxygen, possibly under pressure, would be required. The role of Ce as a dopant in high-temperature oxidation catalysis has been considered by Fan et al.,31 who have proposed that Ce4+ enhances oxidation of ammonia in a linked redox process with Co where an active oxygen species at the surface reacts with ammonia (or methane), creating an oxygen vacancy, and an oxygen anion is transported from the oxygen capture site (Ce) to the catalytic oxidation reaction site (Co) through the subsurface via the ion redox model for intrafacial catalysis.9 The catalytically active Co site and the Ce site responsible for oxygen capture could, of course, be adjacent to one another. We should also note that the activation processes for reaction substrates such as NH3 or CH4 are not understood. For example, it is not certain whether CH4

J. Phys. Chem. C, Vol. 112, No. 32, 2008 12319 activation occurs via adsorption or directly via interaction with an active adsorbed oxygen species.32 This description implies that Ce enhances the gas phase oxygen capture process and that the oxidation reaction occurs at the Co site. However, Fan et al. have not considered the defect compensation processes required to balance the inclusion of Ce4+ in the LaCoO3 lattice.31 In the case of defect compensation by reduction of Co3+ to Co2+, one redox active species, Co3+, is effectively replaced by another, Ce4+. It can be argued that this process would reduce catalytic oxidation activity rather than enhance it since, in relation to the Ce4+S Ce3+ and Co3+ S Co2+redox reactions, the fourth ionization potential for Ce is 3.3 eV greater than the third ionization potential for Co.33 This difference in redox potential is reflected in the higher enthalpy of reduction per reduced cation for CeO2 reduction to Ce2O3 compared to Co3O4 reduction to CoO. The enthalpies for Ce4+ and Co3+reduction in an oxide environment have been estimated to be 191 kJ/ion and 89 kJ/ion, respectively, from the enthalpies of formation for the individual solid oxides as in the equations below.33

2CeO2 f Ce2O3 + 1⁄2O2

∆H ) 381.2 kJ

(5)

Co3O4 f 3CoO + ⁄2O2

∆H ) 177.3 kJ

(6)

1

These values would indicate that Ce would indeed be more effective at oxygen capture whereas Co would be more active as a catalytic oxidation site. However, the higher energy required to reduce Ce4+ would impact on the oxygen transport process required to reoxidize Co2+. Therefore, it is possible that, rather than operating via a redox process, Ce4+ enhances the surface affinity for oxygen capture by providing a strong electrostatic effect at the surface. In contrast, the alternative defect compensation process, Ce substitution at the La site accompanied by creation of La vacancies in eq 1, could enhance the overall redox capability of the surface. In this second reaction, La3+, which effectively acts as a bystander in the catalytic process, is replaced by the redox active species Ce4+. Other factors, which could enhance catalysis, are an increase in surface disorder, which would promote oxygen anion transport to the reaction site, and the creation of La vacancies, which may influence adsorption processes. The above discussion suggests that Ce4+ may be less effective in terms of redox behavior than Co3+. In this respect, Kirchinerova et al. have shown that phase segregated CeO2 in Ce-doped LaCoO3 formulations would have relatively low activity for CH4 combustion compared to LaCoO3.13 However, the important comparison is not with Co3+-rich stoichiometric LaCoO3 surfaces but between catalytic behavior over undoped, reductively nonstoichiometric LaCoO3 surfaces and Ce-doped surfaces prepared under real conditions. The former material presents a surface rich in oxygen vacancies compensated by Co2+, as in Figure 8a. In contrast, the Ce-doped surface is rich in oxygen, since oxygen vacancies are filled, and some La3+ is replaced by redox active Ce4+ cations as in Figure 8b. This oxygen rich surface is effectively fully charged for high-temperature, intrafacial catalytic oxidation, whereas the oxygen vacancy containing undoped, nonstoichiometric surface requires activation by capture of oxygen from the gas phase or transport of oxygen anions from the subsurface. Interestingly, this model for enhancement of high-temperature, intrafacial oxidation catalysis by Ce4+ incorporation in the LaCoO3 surface suggests that more effective materials might be produced by preparation under more oxidizing conditions, that is, calcinations under pure oxygen. In the case of undoped LaCoO3, preparation under strongly oxidizing conditions would

12320 J. Phys. Chem. C, Vol. 112, No. 32, 2008 result in an oxygen-vacancy-free surface as in Figure 8c, where Co is present as Co3+. For Ce-doped materials, a significant fraction of “inert” La3+ would be replaced by “active” Ce4+ cations, compensated by the creation of La vacancies, as in eq 1 and shown in Figure 8d. These surfaces should be highly active for high-temperature, intrafacial oxidation catalysis such as CH4 combustion and NH3 oxidation. The proposed model for Ce-doped LaCoO3 surfaces in high temperature, intrafacial oxidation may also explain behavior under lower temperature, suprafacial conditions where experimental results indicate that Ce doping can reduce activity. In the suprafacial process, it has been proposed that oxygen is activated as O2- by adsorption of gaseous oxygen at oxygen vacancies. However, the effect of Ce doping for air-calcined materials is to inhibit creation of oxygen vacancies. The surface oxygens only become active at higher temperatures when the lattice anion mobility increases. 8. Summary and Conclusions Because catalysis is a surface process, this paper has focused on doping at the surface for the (100) OLa, (100) OCoO, and (110) O terminated surfaces. Our calculations have shown that Ce is more soluble at the surface in stoichiometric LaCoO3 compared to the bulk. We find for the La vacancy reaction (eq 1) that the (100) OLa terminated surface will be the most energetically favored, and for the Co3+ reduction reaction (eq 2), the (100) OCoO termination is slightly more favored compared to the other terminations. The calculation for dissolution of Ce in nonstoichiometric LaCoO3 surfaces as in eq 3 shows that the surface process is no more energetically favorable than that for the bulk. From XRD and surface area data in the literature, a model has been proposed for air-calcined, Ce-doped, LaCoO3 particles consisting of crystallites of ∼50 nm in diameter with a surfaceenriched Ce layer, as predicted by our calculations, but with a concentration gradient extending into the subsurface. Such a model would accommodate a maximum level of 5% atomic Ce in total with respect to the bulk as indicated by experiment. In considering which of the competing Ce-doping reactions in eqs 1–3 is likely to occur during preparation of real materials, the most important factor is oxygen partial pressure. For aircalcined materials at 900 °C, where reductively nonstoichiometric LaCoO3 containing oxygen vacancies balanced by Co2+ cations would be produced in the absence of Ce, we suggest that Ce doping occurs at the surface via eq 2. This process, which is favored at lower oxygen partial pressure, would produce oxygen-vacancy-free material where Co2+ compensates Ce4+. We predict that Ce4+ doping compensated by La3+ vacancies according to eq 1 would occur under strongly oxidizing preparation conditions, such as pure oxygen, possibly under pressure. The role of Ce as a dopant in enhancing high temperature, intrafacial, oxidation catalysis has been discussed in relation to the model proposed by Fan et al. involving a linked redox cycle for Ce and Co.31 In comparison to undoped, reductively nonstoichiometric LaCoO3 surfaces, Ce doping results in replacement of inert La3+ cations by active Ce4+, which enhances oxygen capture, and the surface (and subsurface) becomes fully charged with oxygen. A consideration of ionization potentials and enthalpies of reduction for pure Ce and Co solid oxides suggests that Ce would be most effective in oxygen capture whereas Co would be most effective in the oxidation process in the intrafacial, ion redox catalytic cycle proposed by Metcalfe et al.9 In addition, we predict that materials prepared

Khan et al. under strongly oxidizing conditions, that is, undoped fully stoichiometric LaCoO3 or Ce-doped, La vacancy containing material, will also be effective in high-temperature, intrafacial oxidation. The proposed model for Ce-doped LaCoO3 could also explain behavior under low-temperature, suprafacial oxidation conditions, where Ce reduces activity. In the suprafacial reaction, oxygen is believed to be activated as O2- by adsorption in oxygen vacancies. The absence of anion vacancies in Ce doped materials would inhibit this process, and surface lattice oxygen would only become active at higher temperatures when anion mobility increases. Acknowledgment. We are grateful to Dr F.Cora´ for many useful discussions. We also thank EPSRC and Johnson Matthey Plc. for financial support for a CASE studentship. EPSRC is also thanked for the funding of computer resources at the Royal Institution, and we are grateful to Accelrys for the provision of molecular graphics software. References and Notes (1) Haung, K.; Lee, L. Y.; Goodenough, J. B. J. Electrochem. Soc. 1998, 145, 3220. (2) Nishhihata, Y.; Mizuka, J.; Akoa, T.; Takala, H.; Uenishi, M.; Kimura, M.; Okamoto, T.; Hamado, H. Nature 2002, 418, 164. (3) Forni, L.; Rossetti, J. Appl. Cat B 2002, 38, 29. (4) Wu, Y.; Dou, B.; Wang, C.; Xie, X.; Yu, Z.; Fa, S.; Fan, Z.; Wang, L. J. Catal. 1989, 88, 120. (5) Yu, Z.; Gao, L.; Yuoh, S.; Wu, Y. J. Chem. Soc., Faraday Trans. 1 1992, 88, 3245. (6) Lin, Y. S.; Zeung, Y. J. Catal. 1996, 164, 220. (7) Voorhoeve, R. J. H. AdVanced Materials in Catalysis; Academic Press: New York, 1977. (8) Aria, H.; Yamada, T.; Guahi, K.; Seiyama, T. Appl. Catal. 1986, 26, 265. (9) Petrolekas, P. D.; Metcalfe, I. S. J. Catal. 1995, 152, 147. (10) Read, M. S. D.; Watson, G.; King, F.; Hancock, F. E. J. Mater. Chem. 2000, 10, 2298. (11) Khan, S.; Oldman, R. J.; Cora, F.; Catlow, C. R. A.; French, S. A.; Axon, S. A. Phys. Chem. Chem. Phys. 2006, 8, 1. (12) Oliva, C.; Forni, L.; Ambrosio, A. D.; Novarrini, F.; Stepanov, A. D.; Kagramanov, Z. D.; Mikhailichenko, A. I. Appl. Catal. A 2001, 201, 245. (13) Kirchinerova, J.; Alifanto, M.; Delmon, B. Appl. Catal. A 2002, 231, 65. (14) French, S. A.; Catlow, C. R. A.; Oldman, R. J.; Rogers, S. C.; Axon, S. A. Chem. Commun. 2002, 2706. (15) O’Connel, M.; Norman, A. K.; Hutterman, C. F.; Morris, M. A. Catal. Today 1999, 47, 123. (16) Forni, L.; Oliva, C.; Vatti, F. P.; Kandala, M. A.; Ezerets, A. M.; Vishniakov, A. V. Appl. Catal. B 1996, 7, 269. (17) Gale, J. D. J. Chem. Soc. Faraday Trans. 1997, 629. (18) Jensen, F. Introduction to Computational Chemistry; John Wiley and Sons: Chichester, 1998. (19) Dick, B. G.; Overhauser, A. W. Phys. ReV. 1958, 112, 90. (20) Mott, N.; Littleton, M. Trans. Faraday Soc. 1938, 34, 485. (21) Gay, D. H.; Rohl, A. L. J. Chem. Soc, Fraday Trans. 1995, 91 (5), 925. (22) Tasker, P. W. J. Phys. C 1979, 12, 4977. (23) Cerius2 Forcefield-Based Simulations; Molecular Simualtion Inc.: April 2000. (24) Cherry, M.; Islam, M. S.; Catlow, C. R. A. J. Solid State Chem. 1995, 118, 125. (25) Lewis, G. V.; Catlow, C. R. A. J. Phys. C 1985, 18, 1149. (26) Catlow, C. R. A. J. Phys. Chem. Solids 1977, 38, 1131. (27) Shackelford, J.; Alexander, W. Material Science and Engineering Handbook; C.R.C.Press; Boca Raton, 2001. (28) Royer, S.; Duprez, D.; Kaliaguine, S. J. Catal. 2005, 234, 364. (29) Axon, S. A. Wolfindale, B.;. Johnson Matthey, PriVate Commuincation, 2006. (30) Nitadori, T.; Misono, M. J. Catal. 1985, 93, 459. (31) Fan, S.; Wang, Q.; Dou, B.; Yu, Z. Cuihua Xuebao 1991, 12, 199. (32) Zhu, J. Ph.D. Dissertation, University of Twente, 2004. (33) Lide, D. R. Handbook of Chemistry and Physics, 83rd ed.; C.R.C Press; Boca Raton, 2002.

JP709638H