Reactive Oxygen Species on the (100) Facet of Cobalt Spinel

Oct 9, 2015 - Maxim Zabilskiy , Petar Djinović , Elena Tchernychova , Albin Pintar. Applied Catalysis B: Environmental 2016 197, 146-158 ...
0 downloads 0 Views 7MB Size
Research Article pubs.acs.org/acscatalysis

Reactive Oxygen Species on the (100) Facet of Cobalt Spinel Nanocatalyst and their Relevance in 16O2/18O2 Isotopic Exchange, deN2O, and deCH4 ProcessesA Theoretical and Experimental Account Filip Zasada,* Witold Piskorz, Janusz Janas, Joanna Gryboś, Paulina Indyka, and Zbigniew Sojka Faculty of Chemistry, Jagiellonian University, ul. Ingardena 3, 30-060 Krakow, Poland S Supporting Information *

ABSTRACT: Periodic spin unrestricted, gradient corrected DFT calculations joined with atomistic thermodynamic modeling and experiment were used to study the structure and stability of various reactive oxygen species (ROS) and oxygen vacancies produced on the most stable terminations of the cobalt spinel (100) surface. The surface state diagram of oxygen in a wide range of pressures and temperatures was constructed for coverage varying from ΘO = 1.51 atom·nm−2 to ΘO = 6.04 atom·nm−2. A large variety of the unraveled surface ROS includes diatomic superoxo (CoO−O2−−CoO), peroxo (CoT−O22−−CoO), and spin paired (CoO−O2−CoO) adducts along with monatomic metal-oxo (CoT−O+, CoO−O2+) species, where CoT and CoO stand for the tetrahedral and octahedral cobalt surface centers, respectively. There are also two kinds of peroxo species associated with surface oxygen ions connected with 3CoO or 2CoO and 1CoT cations ((O2O,1T−O)2− and (O3O−O)2−), respectively). The results revealed that in the oxygen pressure range of typical catalytic reactions (pO2/p° from ∼0.01 to 1), the most stable stoichiometric (100)-S surface accommodates the CoT−O22−−CoO and CoO−O2−CoO adducts at temperatures below 250−300 °C. In the temperature from 250 to 300 °C and from 550 to 700 °C, it is covered by the O species associated with the exposed tetrahedral cobalt sites (CoT− O+) or remains in a bare state. In more reducing conditions (T > 550−700 °C), the (100)-S facet is readily defected due to trigonal oxygen (O2O,1T) release and formation of surface oxygen vacancies. The reactivity of surface ROS was tested in 16 O2/18O2 isotopic exchange, N2O decomposition, and oxidation of CH4 and CO model reactions, carried over Co3O4 and Co318O4 nanocrystalline samples with the predominant (100) faceting revealed by high angle angular dark field STEM examination. The CoO−O2+ adducts associated with octahedral cobalt sites, as well as the peroxo (O2O,1T−O)2− and (O3O−O)2− surface species being thermodynamically unstable are involved in surface oxygen recombination processes, probed by 16O2/18O2 exchange and N2O decomposition. It was shown that at low temperatures CO is oxidized by the suprafacial CoO−O2−CoO and CoT−O2−CoO diatomic oxygen, whereas in CH4 activation, the highly reactive cobalt-oxo species (CoT−O+) are involved. Above 600 °C at pO2/p° = 0.01, due to the onset of oxygen vacancy formation, the suprafacial methane oxidation gradually changes into the intrafacial Mars-van Krevelen scheme. The constructed surface phase diagram was used for rationalization of the obtained catalytic data, allowing delineation of the specific role of the chemical state of the cobalt spinel surface in the investigated processes, as well as the range of the corresponding temperatures and oxygen pressures. It also provides a convenient background for molecular understanding of remarkable activity of Co3O4 in many other catalytic redox reactions. KEYWORDS: Co3O4, catalysis, redox sites, methane combustion, nitrous oxide decomposition, CO oxidation, surface phase diagram, atomistic thermodynamics, DFT

1. INTRODUCTION

various surface reactive oxygen species (ROS) and to facile formation of oxygen vacancies,9 revealed by isotopic exchange between gas phase dioxygen and surface and subsurface oxygen entities of the spinel.10,11 The ROS adspecies are produced by associative and dissociative adsorption modes of dioxygen. They

Cobalt(II,III) spinel is an excellent redox-tunable model catalytic material of widespread fundamental and practical interest. In particular, it has been found to be active in many oxidation reactions involving CO,1 NH3,2,3 CH4, and other hydrocarbons.4,5 Co3O4 is also well-known for its spectacular activity in the oxygen reduction reaction (ORR)6 and in the low temperature decomposition of N2O.7,8 A governing factor of its high catalytic performance can be attributed to presence of © XXXX American Chemical Society

Received: August 26, 2015 Revised: October 7, 2015

6879

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis may be stabilized on the tetrahedral (CoA) and octahedral (CoB) cationic sites or on the anionic (O2−) surface centers.12 It has been suggested that in a number of catalytic processes such as CO oxidation13 or N2O decomposition,14,15 the suprafacial ROS intermediates play a key role in the catalytic activity.16 In turn, the oxygen vacancies generated during catalytic reactions (especially in the reducing conditions17) or produced spontaneously at elevated temperatures18 are implicated in the Mars-van Krevelen redox cycles proposed, e.g. for oxidation of methane over cobalt spinel.19 The redox state of the Co3O4 catalyst (which can be gauged by the Fermi level position probed experimentally by the work function measurements)11,20 plays an important role in governing the charge and spin states and surface dynamics of the ROS intermediates. It can be readily regulated by the spinel stoichiometry adjustment through introduction of cationic/ anionic vacancies by appropriate aliovalent doping or by setting proper redox conditions (controlled by oxygen partial pressure and temperature) during the synthesis. It should also be mentioned that the catalytic activity of cobalt spinel strongly depends on the crystalline structure, grain size, and surface morphological features.21,22 Those factors have to be controlled during the catalyst preparation for sensible comparison of the experimental and theoretical results. Recently, Xie et al. reported that the (110) faceted Co3O4 nanorods can catalyze CO oxidation at temperatures as low as −77 °C.23 Such spectacular activity has driven most of the scientific attention to study of the cobalt spinel (110) facet, and numerous investigations have been focused on elucidation of the structure, relaxation, and reactivity of this particular plane.24−27 However, careful inspection of the available experimental data reveals that vast majority of the Co3O4 specimens prepared by various methods are faceted predominantly on the (100) and (111) planes with only a minor contribution of the (110) termination.28−30 Furthermore, several computational studies have shown that in a wide range of temperatures and oxygen/water partial pressures, the (110) and (111) terminations are less stable than that of the (100) plane.31−33 The (111) termination, due to the presence of highly unsaturated surface Co ions, is covered by water molecules even at elevated temperatures (>400 °C),32 in contrast to (100) that is fully dehydroxylated already at T > 250 °C. Thus, it excludes the prime role of the (111) surface in those catalytic processes where water is present in the feed or is formed in situ as a reaction product, when the low temperature requirement is imposed as a process constraint. Concerning possible catalytic involvement of morphological irregularities such as edges and corners, our previous investigations34 imply that for oxides, in contrast to metals, such sites are not capable to ensure a sustainable turnover. Being very reactive, due to high adsorption energy and a high barrier of oxygen surface diffusion, they are usually self-poisoned already in the induction period of the catalytic reactions. Despite these facts, as far as we know, in comparison to the (110) and (111) planes, the structure and reactivity of the (100) surface has been much less explored until now.33,35,14 First-principles thermodynamics (FPT) is well established as a powerful tool for sensible investigations into the catalyst surface structure, its energetic stability, and interaction with gas phase reactants36−39 The surface equilibrium redox states, revealed by such calculations at various oxygen pressure and temperature conditions, provide a convenient background for an in-depth understanding of the associated catalytic processes involving surface ROS species and oxygen vacancies. Even though the

interaction between dioxygen and the Co3O4 surface plays a critical role in many catalytic processes, a comprehensive description of such interaction is still lacking, and a detailed molecular-level account of oxygen activation, speciation, and stability on the catalytic surface is still to be established. Herein, we use spin resolved DFT+U calculations and firstprinciples thermodynamics to explore the structure and stability of ROS entities on various terminations of the most abundant cobalt spinel (100) facet over a wide range of temperatures and oxygen pressures. The constructed phase diagram of the surface oxygen states was used as a suitable conceptual playground for interpretation of the catalytic data of the 16O2/18O2 isotopic exchange, deN2O, CH4, and CO oxidation reactions, and in discerning the chemical state and reactivity of the cobalt spinel surface in various oxygen environments. It may also help to resolve many past controversies concerning the nature of an exceptional performance of Co3O4 in catalytic redox reactions in comparison to other spinels and oxidation mechanisms and contribute to a better understanding of the surface statereactivity relationships for other redox oxides.

2. COMPUTATIONAL SCHEME AND EXPERIMENTAL METHODS Electronic Energy and Geometry Computations. All quantum chemical calculations were performed within density functional theory (DFT) as implemented in the Vienna Ab initio Simulations Package (VASP).40 We used the projector augmented plane wave (PAW)41 method for describing electron−ion interactions together with the generalized gradient PW91 exchange-functional.42 All calculations were performed using a standard Monkhorst−Pack43 grid (5 × 5 × 5 sampling mesh for bulk calculations and 3 × 3 × 2 for slab calculations) with a cutoff energy of 450 eV and a Methfessel−Paxton44 smearing with σ = 0.1 eV. For solving the Kohn−Sham SCF equations, the energy change criterion of 10−5 eV between two successive iterations was applied. Following previous literature,45,46 in the case of the cobalt spinel for a proper description of the Coulombic correlation effect, the DFT+U methods were used. In this work, we use a value of U = 3.5 eV since these parametrizations have been found to reproduce well the experimental results.47,48 A systematic validation of the applied calculation scheme against the experimental data has been provided in our previous paper,49 where several bulk properties of Co3O4, such as lattice parameters, band gap, and electronic and magnetic structure, have been computed and discussed. Briefly, the optimized spinel lattice constant, a = 8.149 Å, and the u parameter of 0.2638 compare well with the experimental XRD values of 8.082 Å and 0.2632, respectively. The calculated CoO− O and CoT−O bond lengths, dCoO−O = 1.932 Å and dCoT−O = 1.960 Å, are again in a good agreement with the experimental values of 1.920 and 1.935 Å, respectively, as well as the bandgap at the Γ point, EΓ = 1.72 eV. We can thus reasonably assume that the employed formalism along with the applied parametrization is capable of describing the Co3O4 system with an adequate accuracy. Bulk and Surface Models. The bulk cobalt spinel unit cell was obtained by optimization of the experimental cubic (1 × 1 × 1) unit cell (a = 8.08 Å) containing 56 ions (Co24O32). The optimal cell volume was calculated as a result of the E/V fit (Birch−Murnaghan equation of state)50 with full optimization of all internal degrees of freedom within the error of 10−4 eV/Å. Following our recent work,48 we considered two different terminations of the (100) facet that are stable in typical catalytic 6880

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

Figure 1. Microscopic structure and the shape analysis of the cobalt spinel nanocatalyst. Selected area diffraction pattern (a). HRTEM image of a faceted nanogrian (b1) and its magnification (b2) viewed along the [110] direction together with the corresponding simulation (b3 left) based on a DFT optimized structure (b3 right). HAADF STEM pictures (c1−e1) and their gradient representation (c2−e2) complemented by the superimposed Wulff shapes (c3−e3). The retrieved rombicubooctahedral morphologies of the observed Co3O4 nanograins (c4−e4).

chemical potential. The vibrational entropy contributions to the Gibbs free energy of the bulk and the slabs were neglected as usual, since they cancel out to a large extent (Gbulk ≈ Ebulk and Gslab ≈ Eslab). The μO(T,pO2) term can be factored into the T and p independent total energy of dioxygen molecule (including the zero point contribution), 1/2Etot O2, and into the ΔμO(T,pO2) part covering all the T and p related free energy contributions: μO(T, p) = 1/2Etot O2 + ΔμO(T, pO2). As a result, the chemical state of the spinel surface becomes a function of both the temperature (T) and the dioxygen pressure (pO2) since ΔμO(T, pO2) = μO′(T, p0O2) + kT ln(pO2/pO). The μi′ term was computed using standard statistical thermodynamics.52 Following the literature, the oxygen potential can be varied within the range of μO = 0.0 (oxygen rich limit) and μO = −2.4 eV (oxygen lean conditions).53 The issue of the proper description 1/2Etot O2 and its influence on the thermodynamic diagrams is briefly discussed in the Supporting Information. The temperature and pressure dependence of oxygen surface coverage, Θ, of the cobalt spinel was modeled using a multisite Langmuir isotherm.54,32

conditions: a cation depleted oxidized termination (100)-O and a stoichiometric (100)-S termination. We use a (1 × 1) slab of 17 atomic layers (∼16 Å) with a vacuum separation of 20 Å. The supercell composition was defined as Co44O64 for the (100)-O plane and Co48O64 for the (100)-S one. The top and bottom slab terminations were the same to minimize the spurious electrostatic dipole moment effects. The atomic positions in the four top and four bottom layers were relaxed within the criterion of 1 × 10−3 eV·Å−1. First-Principles Thermodynamics. In order to describe the chemical behavior of the cobalt spinel surface under various oxygen p and T conditions, we used FTP modeling formulated by Scheffler and Reuter.51 The surface free energies, γ(T,p), for the surface containing NO oxygen atoms and NCo cobalt atoms are calculated in the following way γ(T , p) =

NCo bulk 1 ⎡ slab GCo3O4(T , pO ) ⎢GCo O (T , p , NCo , NO) − 2 2A ⎣ x y 3 ⎤ ⎛ ⎞ 4 − ⎜NO − NCo⎟μO (T , pO )⎥ 2 ⎦ ⎝ ⎠ 3

(1)

bulk In the formula above, Gslab CoxOy and GCo3O4 are the free energies of the slab and of the cobalt spinel bulk stoichiometric unit (Co3O4), respectively, A is the area exposed by single termination of the slab, and μO(T,pO2) describes the oxygen

ΘO =

∑ xi i=1

6881

βin 1 + ∑i = 1 βin

(2) DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

Figure 2. Coordination number of the surface ions together with the most important bond lengths and angles for the (100)-O (a) and (100)-S (b) terminations and their defected states with oxygen vacancies (c and d). Color coding: CoO, blue; Co2cT, purple; Co4cT, green; O, red.

The isotopically labeled Co318O4 specimen was obtained by the reduction of Co3O4 to metallic cobalt and its further reoxidation with 18O2 back to the cobalt spinel phase. The temperature-programmed oxygen desorption (TPD-O2) measurements were performed in the range of 25−700 °C using a quartz flow reactor with 500 mg of the catalyst (sieve fraction of 0.2−0.3 mm). The flow rate of the feed of 30 mL·min−1 (1% O2 in He, 7000 h−1) and the heating rate of 10 °C·min−1 were applied. Pulse (CO oxidation), isotopic exchange (16O/18O), and thermo-programmed surface reaction (N2O decomposition and CH4 combustion) experiments were carried in the same setup. The progress of the reaction was monitored by a quadruple mass spectrometer (Hiden Analytical HPR20).

where xi = Ni/N is the ratio of the Ni adsorption sites i per total number of the available adsorption centers, whereas βn is the calculated cumulative adsorption equilibrium constant. The latter can be calculated as βn = K1·K2·...·Kn with the component Ki values describing the adsorption equilibrium constant of the subsequent surface states that differ in the number and nature of the oxygen adspecies: ⎛ ΔGi − ΔGi + 1 ⎞ ⎟ K i = exp⎜ − ⎝ ⎠ RT

(3)

Materials and Characterization. The Co3O4 spinel samples were prepared by a pH-controlled precipitation method from cobalt nitrate solution as described elsewhere in more detail.16 In order to obtain good crystalline spinel structure, the precipitate was dried at 100 °C and calcined at 700 °C. The phase composition of the samples was confirmed by X-ray diffraction, using Cu Kα radiation by means of a Rigaku diffractometer (see Figure S1a in the Supporting Information, SI), and by microRaman spectra using a Renishaw InVia spectrometer with an excitation wavelength of 785 nm (see Figure S1b, SI). Transmission electron microscopy ((S)TEM) measurements were carried out using a Tecnai Osiris instrument (FEI) operating at 200 kV. STEM was performed using a high-angle annular dark-field (HAADF) detector, with a nominal spot size of ∼0.5 nm. The camera length was kept in range of 330−550 mm, to maximize the HAADF signal intensity. For high resolution imaging in STEM mode, a nominal spot size of ∼0.18 nm was used. Prior to microscopic analysis, the samples were ultrasonically dispersed in ethanol and dropped on a holey carbon film supported on a copper grid (Agar Scientific, 400 mesh). The simulations of HR-TEM micrographs were carried out using the JEMS simulations package.55 The analysis of thickness and defocus values was performed within the range of 10−80 nm and 50−75 nm, respectively, and the matching imaging conditions were found to be equal to 37 nm of thickens and 67.2 nm of defocus (extended Scherzer defocus). The shape of the nanocrystals was retrieved by an inverse Wulff construction, as discussed elsewhere.29

3. RESULTS AND DISCUSSION 3.1. Morphology of the Spinel Nanocatalyst. The results of structure and the shape analysis of the cobalt spinel catalyst as revealed by selected area diffraction pattern (SAED), HR-TEM, and HAADF STEM are shown in Figure 1. The SAED pattern (Figure 1a) reveals high crystallinity of the synthesized catalyst, which is further confirmed by the HR-TEM imaging of the spinel structure. As an example, a nanocrystal with the (100) and (111) faceting, observed along the [110] direction, is shown in Figure Figure 1b1 and b2 at different magnifications. The indicated interplanar distances, d(400) = 2.01 Å and d(222)= 2.3 Å, corresponding to the (100) and (111) planes, respectively, are in a very good agreement with the literature values. Because in the applied imaging conditions, due to oscillating contrast transfer function, there is no one-to-one relation between the spinel crystalline structure (Figure 1b3, right side) and its HRTEM image (Figure 1b2), the interpretation of the observed picture was univocally confirmed by the simulation (Figure 1b3 left side). Since most of the nanocrystals are randomly oriented, more comprehensive examination of their shape is possible by means of the HAADF STEM. Indeed, in such an imaging mode, the intensity of the pictures depends on the scattering factor and the average atomic number of the projected atomic columns, 6882

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

Figure 3. Geometric structure of the monatomic oxygen adsorbed on different sites of the (100)-S termination: Co2cT (a), Co5cO (b), O2ot1 (c), and O3o (d). Color coding: CoO, blue; Co2cT, purple; Co4cT, green; spinel O, red; adsorbed O, orange.

whereas the diffraction contribution is largely suppressed. As a result, for a given randomly oriented grain, it directly reflects the variation of the nanocrystal thickness along the beam direction. Such variation can be better revealed in image gradient and/or edge detection presentations, which help the shape to be retrieved. The gradient analysis and the edges detection were performed by using the Canny algorithm56 implemented in the FeatureJ software.57 The results for the three typical, yet well exposed nanograins are shown in Figure 1c−e, where the left column corresponds to the original HAADF STEM pictures, the next one to the gradient representation, followed by gradient picture with the superimposed Wulff shape. The retrieved morphologies are collated in the right column. All of the projected contours are convex polygonsnamely truncated squares with the interplanar angles close to 135° (Figure 1c1−e1). According to Steno law, this suggests a basically cubic shape of the nanocrystals truncated at corners and edges along the (111) and (110) planes. The lack of the contrast gradient (Figure 1c2− e2) indicates the constant thickness of the nanocrystals along the view direction, which implies the presence of the (100) facet. It was therefore possible to recover the shape of the investigated nanocrystals by using a reversed Wulff construction (Figure 1c3− e3). The retrieved rhombicuboctahedral shapes (Figure 1c4−e4) show the dominance of the (100) facet with contribution varying from ∼60 to ∼85%, and in the whole population those with the higher (100) facet content (>80%) were much more frequent. These results were used to build up an adequate computational model, based on the most abundant (100) termination, for quantum chemical calculations of dioxygen interaction with Co3O4 nanocrystals. Modeling of the (111) facet, the second most abundant, and evaluation of its role in deN2O and deCH4 reactions is now in progress. Yet, a good agreement between the theoretical predictions and experimental results (vide inf ra) obtained for the investigated cobalt spinel nanocrystals faceted predominantly on the (100) planes confirms that indeed much of the catalytic performance may be firmly associated with this termination. 3.2. Surface Terminations. The cobalt spinel (100) surface was studied in our previous work, where the structure and stability of all possible terminations was elucidated in detail.49

Therefore, in the present study, we considered only a cation depleted oxidized (100)-O termination and a stoichiometric (100)-S termination, which are stable in typical catalytic conditions. Their basic structural characteristics are shown in Figure 2a and b. The surface (1 × 1) unit cell (black square) shows coordination numbers (nc subscript) of the surface cobalt and oxygen ions along with the most important bond lengths and angles. As shown in Figure 2a, the cationic composition of the (1 × 1) cell of the (100)-O termination includes four singly truncated octahedral ions, Co5cO, and two recessed tetrahedral cobalt ions, Co4cT, of full coordination. The anionic counterpart comprises four oxygen ions of 3-fold coordination, which are linked to two CoO and to one CoT ion (denoted hereafter as O2O,1T) and four 3-fold oxygen anions linked to CoO exclusively (O3O). The Bader charge (qB) analysis showed that for the surface oxygen anions the partial charges (qB = −1.03 |e|) are nearly the same as in the bulk (qB = −1.05 |e|), whereas in the case of the surface cobalt ions they are systematically higher with more pronounced changes noted for CoO (qB = 1.45 |e| and 1.54 |e| for the bulk and the surface ions, respectively) than for Co4cT (bulk qB = 1.31 |e| vs surface qB = 1.34 |e|). The reconstruction of the surface is essentially negligible, since both CoT−O and CoO−O bonds are only slightly shorter than in bulk (1.91 vs 1.96 Å and 1.91 vs 1.93 Å, respectively). However, the tetrahedral angle measured for the surface Co4cT is significantly larger than in the bulk (117° vs 109°). In the case of the (100)-S stoichiometric termination (Figure 2b), one protruding tetrahedral Co2cT ion with two dangling bonds is present in the (1 × 1) surface element. The presence of the Co2cT ions changes the coordination of the exposed anions, as the two 3-fold O3O ions transform into the 4-fold O3O,1T species. The Bader charge analysis shows minute changes in the oxidation states of the surface anions and cations comparing to bulk values (with the difference reaching merely ΔqB = −0.02 |e| in the case of Co5cO). Analysis of the reconstruction process showed that Co2cT ions were moved about 0.45 Å toward the bulk, whereas the analogous shift of the Co5cO was much smaller (0.09 Å). Those changes were accompanied by reduction of the CoT−O and CoO−O bond lengths (by 0.14 and 0.03 Å, respectively) and 6883

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

the magnetic moment of μ = 0.76 μB implies a O− (S = 1/2) moiety attached to the Co2+ (S = 3/2) center constituting a [Co2cTO]+ core with S = 2. Indeed, the spin repartition within the cobalt−oxygen unit and the Co−O bond distance of 1.66 Å agree quite well with the quintet 5Δ state of the gas phase metaloxo CoO+ species with dCo−O = 1.63 Å.58 Analysis of the Bader charges revealed that the actual electron donor center is not constituted by the hosting Co2cT site (the change in its oxidation state ΔqB = 0.07 |e| is quite small) but rather by the adjacent Co5cO ion that is oxidized to a formally tetravalent state characterized by the increase of the atomic charge (from qB = 1.43 |e| to qB = 1.68 |e|) and a significant drop of the magnetization from μ = 1.9 μB to μ = 1.1 μB. The latter is in line with the removal of the electron from the (dxz,dyz)4(dxy)1(dz2)1 configuration typical for truncated octahedral sites with the C4v local symmetry. Such a remote donor site effect is in line with the density of states plot for the (100) termination (see Figure S2, SI), where the top of the valence band in pDOS of the CoT surface cations is located well below the top of the valence band maximum of the CoO pDOS. As a result, the transient hole generated upon tetrahedral cobalt oxidation is spontaneously transferred into the octahedral cobalt site via the Co2cT−O−Co5cO bridge. Since the barrier for the electron hopping via hole mechanism in the octahedral layer is quite small (0.2 eV), the octahedral hole (CoO• = Co4+) is actually itinerant and may serve as a electron acceptor center elsewhere on the catalyst surface. Thus, the adsorption of oxygen on the Co2cT sites can formally be accounted for in the following way:

a substantial increase of the tetrahedral O3O1T−Co2cT− O3O1T angle, from 109° to 120°. For the (100)-O termination among two kinds of surface oxygen vacancies (VO2O,1T, and VO3O described in detail previously),48 the VO2O,1T one produced by extraction of surface oxygen connected to two octahedral and one octahedral cobalt has the lower energy of formation (0.76 eV vs 1.35 eV). As a result, the VO3O vacancy will not influence the surface state diagram in an appreciable way in the investigated temperature range. The resultant relaxed surface is shown in Figure 2c, where the position of the oxygen vacancy is presented as a yellow cube. In the vicinity of the created defect, a contraction of the CoO−O bonds from 1.91 to 1.85 Å occurs, and the electrons left are localized on the adjacent Co5cO cations. This is revealed by the changes in their Bader charges (from 1.46 |e| to 1.23 |e|) and magnetic moments (from 2.0 μB to 2.75 μB). In the case of the (100)-S termination, analogous formation of the VO2O,1T vacancy is more energetically demanding (0.91 eV), and the resultant structure relaxation is also more pronounced (Figure 2d). It entails a shift of the exposed Co2cT cation toward the neighboring octahedral empty site B′. The migrating divalent tetrahedral cation (Co2cT → Co5cB′) that accompanies the reduction preserves its valence and spin state, and the released electrons are accommodated on the pair of the Co5cO sites. 3.3. Surface Oxygen Species on the (100)-S Termination. The formation of the reactive surface oxygen species on the (100)-S termination was investigated by taking into account associative and dissociative modes of dioxygen adsorption, for atomic coverages ranging from ΘO = 1.51 atom·nm−2 to ΘO = 6.04 atom·nm−2 (one atom to four atoms per (1 × 1) surface element, respectively). At low coverage (ΘO = 1.51 atom·nm−2), we considered possible oxygen accommodation on the exposed Co2cT and Co5cO cationic sites, as well as on the O2O,1T and O3O anionic centers. The pathway along which the monatomic species are produced from dioxygen involves O2 adsorption in a top-on form (Figure S2, SI) on cobalt sites that results in an associative adsorption, followed by dissociation of the O2 admolecules and formation of the adjacent CoT−O and CoO− O species (vide inf ra). The kinetic barrier for such a dissociation event was found to be equal to 1.15 eV. The structures of the resultant species are shown in Figure 3, and the corresponding adsorption energies (with respect to gas phase O2 and O) are collected in Table 1. The most stable adsorption (Eads = −0.72 eV) of the O atom, with the Co−O bond length of 1.66 Å, takes place on the Co2cT sites (Figure 2a). The corresponding oxygen covered termination was denoted hereafter as (100)-S-CoT−O. The charge flow from the slab into the oxygen adatom was equal to −0.76 |e|, which together with

T × × (Co2c ) + (CoO 5c) +

Adsorption of the oxygen atom on the Co5cO site (the (100)S−CoO−O termination) results in a Co5cO−O surface adduct with a bond length of dC−O = 1.66 Å (Figure 3b) and a positive value of the adsorption energy, Eads = 0.15 eV. This means that such species are unstable comparing to gas phase O2, which has important mechanistic consequences for, e.g., the deN2O reaction (vide inf ra). The O adatom acquires a negative charge (−0.73 |e|) provided mostly by the Co5cO cation adjacent to the adsorption site, which is oxidized to a formally tetravalent state (ΔqB = 0.35 |e|, Δμ = −0.85 μB). Similarly to the Co2cTO+ species, the oxygen adatom exhibits a magnetic moment of μ = 0.65 μB, which is indicative for O− (S = 1/2) attached to the Co5cO3+ (S = 1) site, giving rise to a Co5cOO2+ core with S = 3/2. Such dipositive metaloxo cations have previously been described in the literature,59 and their electronic and magnetic features are similar. Oxygen adsorption on the O2O,1T sites gives rise to surface peroxy species,60 and such termination is labeled as (100)-SO2O,1T-O. Indeed, the O−O bond distance is equal to 1.49 Å, the magnetic moment is close to 0, and the Bader charge is about half of that of the bulk O2− anion (see Table 2). The electron density transferred to the O adatom is provided exclusively by the O2O,1T adsorption site, as it can be inferred from the drop of its partial charge from qB = −1.01 |e| to qB = −0.62 |e|. The peroxy anion is additionally stabilized electrostatically by the interaction with the nearest Co2cT (dO−Co of 2.15 Å) forming a heterobridged O2O,1T−O−Co2cT adduct with an adsorption energy of 0.42 eV (Figure 3c). In the case of the O3O adsorption sites, attachment of oxygen (Eads = 0.57 eV) leads again to the formation of the peroxy anions, establishing the (100)-S-O3O−O termination. The O−O

Table 1. Energetics, Geometry, and the Electronic and Magnetic Structure of the Oxygen Adatoms Adsorbed on Different Sites of the (100)-S Termination of Cobalt Spinela adsorption site O adatom property

Co2cT

Co5cO

O2O,1T

O3O

Eads./O2/eV Eads./O/eV dCo−O/Å qB(Oads.)/|e| μ(Oads.)/μB

−0.72 −4.07 1.66 −0.71 0.76

0.15 −3.20 1.67 −0.56 0.73

0.42 −2.93 1.49 −0.51 0.02

0.57 −2.78 1.48 −0.48 0.11

1 T O2 → (Co2c − O)′ + (CoO 5c) 2

a

Adsorption energy was calculated versus the molecular (Eads/O2) and atomic (Eads/O) energies of oxygen. 6884

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

Table 2. Binding Energies (Normalized to One Oxygen Atom Per Unit Cell) of the Oxygen Adducts Adsorbed on the (100)-S Termination of Cobalt Spinel at Higher Oxygen Coverages, Calculated versus Molecular (Eads/O2) and Atomic (Eads/O) Oxygen Energies

ΘO/at.·nm−2 Eads/O2/eV Eads/O/eV

CoT−O CoO−O

CoT−O2−CoO

CoO−O2−CoO

CoT−O2−CoO CoO−O

−0.22 −3.56

3.02 −0.61 −3.95

−0.35 3.69

−0.24 −3.58

CoT−O 2CoO−O

CoT−O2−CoO CoO−O2−CoO

−0.10 −3.44

6.04 −0.41 −3.74

4.53

Figure 4. Surface energy (Δγ) of the (100)-S termination covered with atomic oxygen at a low covarage limit (Θ = 1.5 atom·nm−2) as a function of temperature for three different oxygen pressures (a, low; b, medium; and c, high).

Figure 5. Perspective view of the reactive oxygen species adsorbed on the cobalt spinel (100)-S termination at higher oxygen coverages: CoT−O, CoO− O (a), CoT−O2−CoO (b), CoO−O2−CoO (c), CoT−O, 2CoO−O (d), CoT−O2−CoO, CoO−O (e), and CoT−O2−CoO, CoO−O2−CoO (f) configurations. Color coding: CoO − blue; Co2cT − purple; Co4cT green; spinel O − red; adsorbed O − orange.

conditions with pO2 = 10−8 (achievable, e.g., by pCO/pCO2 ∼ 102.5 buffer), (ii) oxygen pressures of typical catalytic oxidation reactions (pO2 = 10−2 atm), and (iii) strong oxidation conditions (pO2 = 104 atm). Because of the same Co48O65 composition of the corresponding slab models, all lines are parallel, and the differences in their placing reflect different adsorption energies. It is worth noting that only in the case of oxygen adsorbed on the CoT sites (dark red line) is the resultant adduct stable enough to cross the bare surface line (blue). In the low pressure regime, the CoT−O species are stable below 200 °C, whereas above this

distance of 1.49 Å and null magnetic moment are indicative of the O22− species. The O−O2O,1T entities are aligned almost vertically with respect to the (100) plane (Figure 3d), and the top end of the peroxy moiety is stabilized by the interaction with the adjacent Co5cO (dO−Co = 2.11 Å). To evaluate the stability of the monatomic oxygen species under various thermodynamic conditions, we added the temperature and pressure terms to the calculated DFT energies (see eq 1). The results are presented in Figure 4 as a function of temperature for oxygen pressures corresponding to (i) reduction 6885

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

Figure 6. Surface energy, Δγ, of the (100)-S termination covered by oxygen as a function of temperature for three different pO2/p° values: (a) 1.0 × 10−8, (b) 1.0 × 10−2, and (c) 1.0 × 104. The minimum energy lines define the most stable terminations, and the intersections delineate the transition boundaries.

longer than in the case of the cobalt-oxo surface adducts (1.77 Å vs 1.64 and 2.12 Å vs 1.65 Å for dCoT‑O and dCoO‑O, respectively). The O−O bond length, equal to 1.49 Å, is much longer than that calculated for the gas phase dioxygen (1.24 Å), which together with negligible spin density localized on the oxygen admolecule (μ = 0.08 μB) suggests a peroxo nature of this species. Indeed, detailed analysis of the Bader atomic charges revealed a pronounced electron transfer from the cobalt donor centers into the dioxygen moiety (qB(O2) = −0.89 |e|), which is only slightly lower than the framework O2− anion value. One of the electrons is provided by the adjacent Co5cO (ΔqB = 0.31 |e|, Δμ = −0.79 μB), whereas the second is collectively by the octahedral cobalt sublattice. In the case of O2 adsorbed on the [Co5cO, Co5cO] sites, the stabilization energy per one oxygen atom is nearly two times lower (Eads = −0.35 eV, Table 2) than for the [Co2cT, Co5cO] centers. The O2 moiety is attached quite symmetrically with the CoO−O bond lengths of ∼1.9 Å and the O−O interatomic distance of 1.36 Å (Figure 5c). The extent of the cobalt to O2 ligand charge transfer (qB(O2) = −0.59 |e|) and the paramagnetic state (μ(O2) = 0.72 μB) of the admolecule are in line with a superoxo formulation of the CoO−OO−CoO adduct. The electron again is provided by the remote donor center constituted by the adjacent Co5cO (ΔqB = 0.34 |e|, Δμ = −0.86 μB). For the higher ΘO value (4.53 atom·nm−2, tantamount with the presence of three oxygen atoms per (1 × 1) unit cell), there are three conceivable adsorption configurations: [CoT−O, 2CoO−O], [CoT−O2−CoO, CoO−O], and [CoO−O2−CoO, CoT−O]. From the configurations with associatively adsorbed O2, only the [CoT−O2−CoO, CoO−O] species with slightly higher adsorption energy is considered explicitly hereafter (Figure 5e), whereas the [CoO−O2−CoO, CoT−O] configuration is described in more detail in the SI section (Figure S4, Table S2). For the fully dissociative case (Figure 5d), the oxygen adsorption geometries are similar to those calculated for the oxygen adatoms in the low coverage limit (see Figure 3a and b). They all exhibit the Co−O bond lengths only slightly higher than in the diluted case, and the adsorption energy, Eads = −0.10 eV, is lower than the corresponding value (−0.14 eV) for all three adducts calculated separately in the low coverage limit (see Table 1). In the case of the [CoT−O2−CoO, CoO−O] configuration (Figure 5e), the geometry of both entities is essentially the same

temperature the bare surface is thermodynamically preferred (Figure 4a). This threshold is shifted to ∼375 °C in the medium pO2 values (Figure 4b), but under high oxygen pressures, the (100)-S-CoT−O termination prevails in the entire temperature range of the Co 3 O 4 thermal stability. The remaining terminations, i.e., (100)-S-CoO−O (green), (100)-S-O2O,1T−O (dashed orange), and (100)-S-O3O−O (dotted gray), are thermodynamically unstable. They may be significant, however, as useful reference states for mechanistic analysis of catalytic reactions such as decomposition of nitrous oxide. Indeed, a definite residence time of the surface atomic oxygen species (generated by N2O dissociation) may result in their appreciable transient (kinetic) stability (vide inf ra). Moving to higher oxygen coverages, we can consider dissociative and associative modes of O2 adsorption on the cationic sites only, since upon inspection of Figure 4 it is sensible to reject the much less stable peroxy anions. The analyzed monoand dinuclear surface adducts are shown in Figure 4, and the corresponding adsorption energies are collected in Table 2. For Θ O = 3.02 atom·nm−2, there are three possible configurations: one dissociative and two associative (Figure 5a,b,c). The corresponding precursor top-on diatomic CoT−O2 (η1) forms that were much less stable are described briefly in the SI only (Figure S3, Table S1). The dissociative one (Figure 5a) refers to coexistence of the oxygen atoms adsorbed on the Co2cT and Co5cO centers, defining the (100)-S-CoT−O,CoO−O termination. Both oxygen adducts have geometries similar to those in the low coverage limit (cf. Figure 5a with Figure 3a and b). The adsorption energy per oxygen atom, Eads. = −0.22 eV, is significantly lower than the corresponding value (−0.28 eV) for the two adducts calculated separately in the low coverage limit (Θ = 1.51 atom·nm−2). This reflects a sizable lateral Coulombic repulsion between the charged oxygen adspecies, found also on the surface of perovskites.61 In the case of associative adsorption, we tested the top-on and the side-on adsorption modes of dioxygen. In the case of bridging configuration, two different dinuclear adsorption centers of O2 were proposed: [Co2cT, Co5cO] and [Co5cO, Co5cO], and the corresponding terminations are labeled hereafter as (100)-S-CoT−O2−CoO and (100)-SCoO−O2−CoO, respectively. The adsorption energy for dioxygen captured on the [Co2cT, Co5cO] sites is equal to Eads = −0.61 eV (Table 2), and the Co−O2 distances are distinctly 6886

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

Figure 7. Collective 3D plot of the surface free energies versus T and pO2 for various (100) terminations of Co3O4 (a) and the resultant surface redox state diagram (b).

conditions. Between 50 °C and ∼175 °C, the (100)-S-CoT−O surface, associated with the monatomic oxygen attached to the Co2cT sites, dominates (orange line), whereas the bare (100)-S surface is expected at T > 175 °C. Other terminations (dashed lines) do not contribute to the bottom envelope of the plot, being thermodynamically less favored; however, as already mentioned, they may appear as transient states during the catalytic reactions. Moving to higher pressures of oxygen, the overall sequence of the surface phase transitions is preserved, but the stability of the involved oxygen adspecies extends to higher temperatures. At medium pO2 values (pO2/p° = 10−2, Figure 6b), the bare (100)-S surface is the most stable above ∼350 °C, being progressively defected by formation of oxygen vacancies with the increasing temperature (vide infra), whereas in the high oxygen pressures (pO2/p° = 104, Figure 6c), even at the highest considered temperatures (800 °C), the (100) surface of cobalt spinel is always covered by oxygen. To summarize tersely the obtained results, the most stable terminations of the (100)-S surface oxidized by oxygen adsorption are presented in the form of a 3D plot: γ vs T and pO2 (Figure 7a). At low O2 pressures and in high temperatures, the oxidized terminations are situated above the stoichiometric (100)-S surface (coded blue), and since their slopes are quite steep, they shall not coexist. In the opposite corner of low T and high pO2 (the most oxidizing conditions), the (100)-S-2·Co− O2−Co termination (purple) is clearly the lowest in energy and is followed by the less covered (100)-S-CoT−O2−CoO (red) and the (100)-S-CoT−O (orange) terminations. The corresponding surface redox state diagram of the (100) facet can be at best examined using the 2D projection of the lowest surface free energies on the (T, p) plane (Figure 7b). The largest area covering the lower right corner, with pronounced extensions into high pressure and low temperature regions, corresponds to the (100)-S-2·Co−O2−Co termination (ΘO = 6.04 atom·nm−2). Moving along the diagonal direction (toward lower pO2 and higher T), a sequence of terminations with the gradually decreasing oxygen coverage appears until the bare (100)-S surface is reached. They correspond to the (100)-S-CoT−O2− CoO and (100)-S-CoT−O planes with ΘO = 3.02 atom·nm−2 and 1.51 atom·nm−2 respectively. Upon further increase of temperature under low oxygen pressures, the stoichiometric surface

as in their separate states (see Figures 3b and 5b). The charge distribution reveals that the dioxygen moiety has the peroxy nature, whereas the oxygen ligand in the CoO−O adduct is of the metaloxo type. The highest considered coverage (ΘO = 6.04 atom·nm−2, equivalent to four atoms per unit cell), taking into account the energetic constraints, was modeled only as an associative adsorption of two O2 molecules to produce the [CoT−O2−CoO] and the [CoO−O2−CoO] surface adducts (Figure 5f). Whereas the O−O bond length in the [CoT−O2− CoO] adduct is slightly abridged (1.48 Å), for the [CoO−O2− CoO] species it is significantly shorted (1.28 Å) in comparison to the corresponding values calculated for the isolated forms (see Figure 5b and c). From analysis of the qB values, we found that the [CoT−O2−CoO] configuration exhibits a peroxy character, whereas in the [CoO−O2−CoO] adduct, the O2 ligand is almost neutral (qB = −0.15) but diamagnetic (μ = 0.09), which implies a unique attachment via double spin pairing, as on both Co5cO adsorption sites the significant drop of the magnetic moment is observed (Δμ ∼ − 0.8 μB) with no substantial change in their charge. The adsorption energy, equal to Eads = −0.41 eV per oxygen atom (Table 2), is lower by ∼0.1 eV, compared to the corresponding energies calculated for each dioxygen adduct separately. The results of the thermodynamic calculations for the oxygen covered (100)-S termination are presented in Figure 6, where the solid lines and the color coded areas correspond to the thermodynamically most stable surfaces, whereas the dashed lines refer to the “virtual” terminations of higher surface energies. For the sake of clarity, the equilibrium oxygen coverage values are also indicated. The corresponding surface free energies are plotted as a function of the temperature for three different values of the oxygen partial pressure (pO2/p°). Since in each case the Co/O ratio is lower than 0.75, the slopes of the γ(T,p) lines are positive. In the low pressure regime (pO2/p° = 10−8, Figure 6a) at the lowest temperatures, the surface is highly covered by the associatively adsorbed O2 molecules (ΘO = 6.04 atom·nm−2), and the corresponding (100)-S-2·Co−O2-Co termination (navy blue) is stable until about −100 °C, when the oxygen released from the CoO−O2−CoO adducts lowers the coverage to the ΘO = 3.02 atom·nm −2 . This makes the (100)-S-CoT −O 2−CoO termination the most stable one (red line) under such 6887

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

Figure 8. Geometric structure of the reactive oxygen species adsorbed on the cobalt spinel (100)-O termination: CoO−O (a), 2CoO−O (b), CoO−O2− CoO (c), and 2CoO−O2−CoO (d) configurations. Color coding: CoO, blue; CoT, purple; spinel O, red; adsorbed O, orange.

Although both adducts have geometry similar to that found in the low coverage limit (Figure 8a), the adsorption energy per oxygen atom, Eads. = 0.41 eV, is lower than the corresponding value (0.23 eV) calculated separately in the low coverage limit (see Table 3), reflecting sizable lateral Coulombic repulsion between the charged oxygen adspecies.61 In the case of associative adsorption, a dinuclear adsorption center [Co5cO, Co5cO] was proposed (Figure 8c), and such termination was labeled as (100)-O−CoO−O2−CoO. Stabilization energy per oxygen atom was equal to Eads. = −0.31 eV, and the O2 moiety was attached in a symmetric fashion with the CoO−O bond lengths of ∼1.98 Å and the O−O interatomic distance of 1.31 Å. The cobalt to dioxygen charge transfer (qB(O2) = −0.52 |e|) and the magnetization on the admolecule (μ = 0.65 μB) are in line with the superoxo formulation of the CoO−O2−CoO adduct. The highest coverage, ΘO = 6.02 atom·nm−2, was modeled with two associatively adsorbed molecules (Figure 8d). In this case, the stabilization energy per one oxygen was equal to −0.24 eV, and both O2 species were again bound in a symmetrical way. The cobalt to the ligated O2 charge transfer was equal to qB(O2) = −0.47 |e| per O2, and the magnetic moment of both admolecules equals ∼0.7 μB. This implies again a superoxo nature of the CoO− O2−CoO adducts. The results of the thermodynamic modeling of ROS adspecies on the (100)-O termination are presented in Figure 9. In the low pressure regime (pO2/p° = 1.0 × 10−8), the surface is covered with the associatively adsorbed O2 molecules (ΘO = 6.04 atom·nm−2) only at T < −125 °C (Figure 9a), giving rise to the (100)O(2Co−O2-Co) termination (red line). Above −125 °C, desorption of O2 takes place, leading to the appearance of the bare (100)-O plane (orange). The latter prevails until ∼50 °C, where it crosses the reference (100)-S line of the stoichiometric termination. Other terminations (dashed lines) do not contribute the bottom envelope in Figure 9a. Moving to higher oxygen pressure, the general sequence of the phase crossing is

becomes defected by gradual formation of anionic vacancies (see Figure 2d), corresponding to the Co3O3.88 and then to Co3O3.75 nonstoichiometry (dark and light green areas). Those terminations are denoted as (100)-S-VO2O,1T, and (100)-S-2VO, respectively, and were discussed in our previous paper in more detail.49 3.4. Surface Oxygen Species on the (100)-O Termination. In the case of the (100)-O termination containing the Co2cT vacancies, which may appear when cobalt spinel is calcined at low temperatures and high oxygen pressures,49 the ROS speciation is strongly abridged. The structures of the conceivable oxygen adducts are shown in Figure 8, and the corresponding adsorption energies are summarized in Table 3. At low coverage Table 3. Binding Energies (versus Molecular (Eads/O2) and Atomic (Eads/O) Oxygen) of Different Oxygen Adspecies on the (100)-O Termination of Cobalt Spinel ΘO/at.·nm−2 Eads/O2/eV Eads/O/eV

CoO−O

2CoO−O

1.51 0.23 −3.11

0.41 −2.93

CoO−O2−CoO 3.02 −0.31 −3.72

2CoO−O2−CoO 6.04 −0.24 −3.64

(ΘO = 1.51 atom·nm−2, i.e., one adatom per unit cell), the Co5cO−O adduct, with the Co−O bond length of 1.64 Å (Figure 8a) and Eads = 0.23 eV, constitutes the (100)-O−CoO−O termination. The charge flow from the cobalt donor center into the oxygen moiety equals −0.65 |e|, which along with the magnetic moments on the adoxygen and cobalt moieties (0.76 μB and 1.91 μB, respectively) corresponds to the metaloxo CoO2+ species (observed also on the (100)-S surface). For higher coverage (3.02 atom·nm−2), two possible modes of O2 adsorption, a dissociative and an associative one, were taken into account (Figure 8b and c). The first one refers to the coexistence of two oxygen atoms adsorbed on the different Co5cO centers, constituting the (100)-O-2CoO−O termination. 6888

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis

Figure 9. Surface energy, Δγ, of the (100)-O termination covered by oxygen as a function of temperature for three different pO2/p° values: (a) 1.0 × 10−8, (b) 1.0 × 10−2, and (c) 1.0 × 104. The minimum energy lines define the most stable terminations, whereas the intersections delineate the transition boundaries.

preserved. Obviously, the stability of the oxygen covered terminations extends now to higher temperatures. In particular, at a medium pO2 value, the bare (100)-S surface becomes the most stable above ∼50 °C (Figure 9b) and at high oxygen pressure above ∼300 °C (Figure 9c). 3.5. Comparison to Experiment and Relevance for Catalysis. The obtained thermodynamic results are summarized in a concise way in Figure 10a, where the temperature sequence of the most stable terminations is shown for the oxygen pressure

(p/pO2 = 0.01) representative for typical catalytic conditions. Three principal regions (marked in color) of the surface state diagram include the (100)-S surface covered by ROS (T < 375 °C, pink), the bare (100)-S surface (from 375 to 475 °C, light blue), and the oxygen vacancy defected (100)-S surface (>475 °C, light green). The surface free energies of the terminations shown in Figure 10a were then used to calculate the total oxygen coverage, ΘO(T), in the ROS region (Figure 10b, black line) and development of the surface and bulk vacancies due to oxygen release (expressed as the Co/O ratio) at high temperatures (Figure 10b, gray dotted line). In the former case, the ΘO(T) plot was calculated following a multisite Langmuir isotherm (eq 2), which takes into account the coexistence of various oxygen adspecies, leading to smooth variation of the surface coverage with temperature instead of the unrealistic abrupt changes implied by the simplified lowest Gibbs energy criterion only. At the lowest temperatures, the surface coverage equals ∼6 atom· nm−2 and corresponds to (100)-S-2·Co−O2−Co termination with the associatively adsorbed O2 molecules. Around −150 °C, the [CoO−O2−CoO] adsorbed molecule leaves the surface, liberating two Co5cO sites. This process, better seen as a distinct peak with a maximum at 0 °C in the dθO/dT plot (green line), ends at ∼125 °C with the coverage falling to ∼3 atom·nm−2. Next, dropping of the coverage occurs around 175 °C, where oxygen discharge from the [CoT−O2−CoO] adducts takes place, resulting in ΘO = 1.51 atom·nm−2. Above this temperature, the surface is covered exclusively by the atomic oxygen stabilized on the Co2cT centers. They are released at ∼375 °C, leaving the bare (100)-S surface. Further increasing of the temperature triggers gradual formation of anionic vacancies which results progressively increasing in the spinel nonstoichiometry (gauged by the Co/O ratio). The theoretical evolution of the cobalt spinel (100) surface state in equilibrium with oxygen as a function of the increasing temperature remains in excellent agreement with the thermoprogramed oxygen release experiments shown in Figure 10c. Indeed the observed desorption peaks α, β, and γ can be associated with the oxygen liberated from the [CoO−O2−CoO], [CoT−O2−CoO], and [CoT−O] suprafacial adducts, respectively. The peak δ, in turn, can be attributed to the surface trigonal oxygen (O 2O,1T) release, whereas the peak δ′ corresponds to the onset of the immense discharge of lattice oxygen at high temperatures ( 550−700 °C). The oxygen desorption peaks at 0, 180, and 380 °C can be associated with the [CoO−O2−CoO], [CoT−O22−− CoO], and [Co2cT−O+] suprafacial species, respectively. Superoxo, peroxo, and metal-oxo species are produced by surface to oxygen electron transfer involving remote donor centers constituted by octahedral Co5cO sites. Isotopic 16O2/18O2 exchange and N2O decomposition were used to probe recombination of surface ROS on the bare surface, whereas CO and CH4 oxidation were used to examine their reactivity. The suprafacial [CoO−O2−CoO] and [CoT−O22−−CoO] adducts are involved in low temperature CO oxidation, whereas the cobaltoxo [Co2cT-O+] species may activate the C−H bond in methane. The reaction starts at 350 °C following the Eley−Rideal scheme and, above 600 °C, evolves toward the Mars-van Krevelen mechanism due to gradual engagement of lattice O3c oxygen. These mechanistic turning points remain in very good agreement with the calculated temperature thresholds for changes in the surface oxygen dynamics on cobalt spinel.





ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acscatal.5b01900. Comment on the EO2 error on the accuracy of the FPT plots; spinel sample characteristic: Raman spectra and XRD diffraction patterns; cobalt spinel density of states diagrams; top-on adsorption modes of dioxygen on the (100)-S termination of cobalt spinel; the [CoO−O2−CoO, CoT−O] adsorption modes of reactive oxygen species (PDF)



REFERENCES

(1) Liotta, L. F.; Wu, H.; Pantaleo, G.; Venezia, A. M. Catal. Sci. Technol. 2013, 3, 3085−3102. (2) Petryk, J.; Kolakowska, E. Appl. Catal., B 2000, 24, 121−128. (3) Schmidt-Szałowski, K.; Krawczyk, K.; Petryk, J. Appl. Catal., A 1998, 175, 147−157. (4) Finocchio, E.; Busca, G.; Lorenzelli, V.; Escribano, V. S. J. Chem. Soc., Faraday Trans. 1996, 92, 1587−1593. (5) Liotta, L. F.; Di Carlo, G.; Pantaleo, G.; Deganello, G. Appl. Catal., B 2007, 70, 314−322. (6) Liu, Y.; Higgins, D. C.; Wu, J.; Fowler, M.; Chen, Z. Electrochem. Commun. 2013, 34, 125−129. (7) Chromčaḱ ová, Z.; Obalová, L.; Kovanda, F.; Legut, D.; Titov, A.; Ritz, M.; Fridrichová, D.; Michalik, S.; Kuśtrowski, P.; Jirátová, K. Catal. Today 2015, 257, 18−25. (8) Asano, K.; Ohnishi, C.; Iwamoto, S.; Shioya, Y.; Inoue, M. Appl. Catal., B 2008, 78, 242−249. (9) Sokolovskii, V. D. Catal. Rev.: Sci. Eng. 1990, 32, 1−49. (10) Takita, Y.; Tashiro, T.; Saito, Y.; Hori, F. J. Catal. 1986, 97, 25−35. (11) Maniak, G.; Stelmachowski, P.; Kotarba, A.; Sojka, Z.; Rico-Pérez, V.; Bueno-López, A. Appl. Catal., B 2013, 136−137, 302−307. (12) Stelmachowski, P.; Zasada, F.; Piskorz, W.; Kotarba, A.; Paul, J.-F.; Sojka, Z. Catal. Today 2008, 137, 423−428. (13) Broqvist, P.; Panas, I.; Persson, H. J. Catal. 2002, 210, 198−206. (14) Piskorz, W.; Zasada, F.; Stelmachowski, P.; Kotarba, A.; Sojka, Z. Catal. Today 2008, 137, 418−422. (15) Omata, K.; Takada, T.; Kasahara, S.; Yamada, M. Appl. Catal., A 1996, 146, 255−267. (16) Stelmachowski, P.; Maniak, G.; Kaczmarczyk, J.; Zasada, F.; Piskorz, W.; Kotarba, A.; Sojka, Z. Appl. Catal., B 2014, 146, 105−111. (17) Liotta, L. F.; Ousmane, M.; Di Carlo, G.; Pantaleo, G.; Deganello, G.; Marci, G.; Retailleau, L.; Giroir-Fendler, A. Appl. Catal., A 2008, 347, 81−88. (18) Li, J.-B.; Jiang, Z.-Q.; Qian, K.; Huang, W.-X. Chin. J. Chem. Phys. 2012, 25, 103−109. (19) Bahlawane, N. Appl. Catal., B 2006, 67, 168−176. (20) Inger, M.; Kowalik, P.; Saramok, M.; Wilk, M.; Stelmachowski, P.; Maniak, G.; Granger, P.; Kotarba, A.; Sojka, Z. Catal. Today 2011, 176, 365−368. (21) Ding, Y.; Xu, L.; Chen, C.; Shen, X.; Suib, S. L. J. Phys. Chem. C 2008, 112, 8177−8183.

AUTHOR INFORMATION

Corresponding Author

*Phone: +48 12 663 20 73. E-mail: [email protected]. 6891

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892

Research Article

ACS Catalysis (22) Feng, J. H.; Zeng, C. Chem. Mater. 2003, 15, 2829−2835. (23) Xie, X.; Li, Y.; Liu, Z.-Q.; Haruta, M.; Shen, W. Nature 2009, 458, 746−749. (24) Chen, J.; Selloni, A. Phys. Rev. B: Condens. Matter Mater. Phys. 2012, 85, 085306−1−085306−9. (25) Chen, J.; Selloni, A. J. Phys. Chem. Lett. 2012, 3, 2808−2814. (26) Xu, X.-L.; Chen, Z.-H.; Li, Y.; Chen, W.-K.; Li, J.-Q. Surf. Sci. 2009, 603, 653−658. (27) Xu, X.-L.; Yang, E.; Li, J.-Q.; Li, Y.; Chen, W.-K. ChemCatChem 2009, 1, 384−392. (28) Kapteijn, F.; Rodriguez-Mirasol, J.; Moulijn, J. A. Appl. Catal., B 1996, 9, 25−64. (29) Zasada, F.; Gryboś, J.; Indyka, P.; Piskorz, W.; Kaczmarczyk, J.; Sojka, Z. J. Phys. Chem. C 2014, 118, 19085−19097. (30) Liu, X.; Qiu, G.; Li, X. Nanotechnology 2005, 16, 3035−3040. (31) Montoya, A.; Haynes, B. S. Chem. Phys. Lett. 2011, 502, 63−68. (32) Zasada, F.; Piskorz, W.; Cristol, S.; Paul, J.-F.; Kotarba, A.; Sojka, Z. J. Phys. Chem. C 2010, 114, 22245−22253. (33) Shojaee, K.; Montoya, A.; Haynes, B. S. Comput. Mater. Sci. 2013, 72, 15−25. (34) Stelmachowski, P.; Zasada, F.; Piskorz, W.; Kotarba, A.; Paul, J.-F.; Sojka, Z. Catal. Today 2008, 137, 423−428. (35) Shojaee, K.; Haynes, B. S.; Montoya, A. Appl. Surf. Sci. 2014, 316, 355−365. (36) Reuter, K.; Scheffler, M. Phys. Rev. B: Condens. Matter Mater. Phys. 2001, 65, 035406−1−035406−11. (37) Inoǧlu, N.; Kitchin, J. R. ACS Catal. 2011, 1, 399−407. (38) Añez, R.; Sierraalta, A.; Martorell, G.; Sautet, P. Surf. Sci. 2009, 603, 2526−2531. (39) Piskorz, W.; Gryboś, J.; Zasada, F.; Zapała, P.; Cristol, S.; Paul, J.F.; Sojka, Z. J. Phys. Chem. C 2012, 116, 19307−19320. (40) Hafner, J. J. Comput. Chem. 2008, 29, 2044−2078. (41) Kresse, G.; Joubert, D. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 1758−1775. (42) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865−3868. (43) Monkhorst, H. J.; Pack, J. D. Phys. Rev. B 1976, 13, 5188−5192. (44) Methfessel, M.; Paxton, A. T. Phys. Rev. B: Condens. Matter Mater. Phys. 1989, 40, 3616−3621. (45) Selcuk, S.; Selloni, A. J. Phys. Chem. C 2015, 119, 9973−9976. (46) Qiao, L.; Xiao, H. Y.; Meyer, H. M.; Sun, J. N.; Rouleau, C. M.; Puretzky, A. A.; Geohegan, D. B.; Ivanov, I. N.; Yoon, M.; Weber, W. J.; Biegalski, M. D. J. Mater. Chem. C 2013, 1, 4628−4633. (47) Wang, L.; Maxisch, T.; Ceder, G. Phys. Rev. B: Condens. Matter Mater. Phys. 2006, 73, 195107−1−195107−6. (48) García-Mota, M.; Bajdich, M.; Viswanathan, V.; Vojvodic, A.; Bell, A. T.; Nørskov, J. K. J. Phys. Chem. C 2012, 116, 21077−21082. (49) Zasada, F.; Piskorz, W.; Sojka, Z. J. Phys. Chem. C 2015, 119, 19180−19191. (50) Murnaghan, F. D. Proc. Natl. Acad. Sci. U. S. A. 1944, 30, 244−247. (51) Reuter, K.; Scheffler, M. Phys. Rev. B: Condens. Matter Mater. Phys. 2001, 65, 035406−1−035406−11. (52) MOPAC Manual. http://nova.disfarm.unimi.it/manual/pdf/ Mopac7.pdf (accessed Oct. 2, 2015). (53) Montoya, A.; Haynes, B. S. Chem. Phys. Lett. 2011, 502, 63−68. (54) Piskorz, W.; Gryboś, J.; Zasada, F.; Cristol, S.; Paul, J.-F.; Adamski, A.; Sojka, Z. J. Phys. Chem. C 2011, 115, 24274−2428. (55) Stadelmann, P. Ultramicroscopy 1987, 21, 131−145. (56) Canny, J. IEEE Transactions on Pattern Analysis and Machine Intelligence 1986, 8, 679−698. (57) Meijering, E. FeatureJ. http://www.imagescience.org/meijering/ software/featurej/ (accessed Oct. 2, 2015). (58) Shiota, Y.; Yoshizawa, K. J. Am. Chem. Soc. 2000, 122, 12317− 12326. (59) Michel, C.; Baerends, E. J. Inorg. Chem. 2009, 48, 3628−3638. (60) Piskorz, W.; Zasada, F.; Stelmachowski, P.; Kotarba, A.; Sojka, Z. J. Phys. Chem. C 2013, 117, 18488−18501. (61) Fleig, J.; Merkle, R.; Maier, J. Phys. Chem. Chem. Phys. 2007, 9, 2713−2723.

(62) Li, W.; Gibbs, G. V.; Oyama, S. T. J. Am. Chem. Soc. 1998, 120, 9041−9046. (63) Boreskov, G. K. Catalytic Activation of Dioxygen. In Catalysis: Science and Technology, 3rd ed.; Anderson, J. R., Bourdard, M., Eds.; Springer-Verlag: New York, 1982; pp 39−137. (64) Xie, X.; Li, Y.; Liu, Z.-Q.; Haruta, M.; Shen, W. Nature 2009, 458, 746−749.

6892

DOI: 10.1021/acscatal.5b01900 ACS Catal. 2015, 5, 6879−6892