The reaction mechanisms for CO catalytic oxidation by N2O or O2 on the Co3O4(110) surface were studied by DFT slab calculations. CO chemisorbs preferably at a surface Co3+ site. After the Co3+ site is completely covered, CO adsorbs at the neighboring twofold coordinated surface oxygen atom bonded to Co2+ and Co3+ cations, resulting in the formation of CO2 and an oxygen vacancy with a low energy barrier of 0.033 eV, which rationalizes the experimental observation that Co3O4-based systems are active for CO oxidation at low temperatures. N2O or O2 interacts with the oxygen vacancy site to regenerate the surface, leaving N2 or the activated O2− species to be attacked by the second CO to yield CO2 to proceed with the catalytic cycle. The CO oxidation reaction follows the Mars– van Krevelen mechanism.
The spinel-type oxides AB2O4 constitute one of the most interesting and important families of crystalline compounds, with applications in many different areas, such as magnetic materials, ceramics, catalysts, and batteries. The spinel-type transition metal (TM) oxide, tricobalt tetroxide (Co3O4) is of special interest in a variety of technological applications and heterogeneous catalysts. Co3O4 and its nanomaterials have been prepared and used in lithium batteries and gas sensors.1–5 Co3O4 and its doped or supported systems are well known to be active for the oxidation of carbon monoxide at low temperatures.6–13 Recently, Xie et al. reported that Co3O4 nanorods not only catalyze CO oxidation at temperatures as low as −77 °C, but also remain stable in a moist stream of normal feed gas.14 They also found that the Co3O4 nanorods predominantly expose their (110) planes, favoring the presence of active Co3+ species at the surface.
From the experimental results, Jansson8 proposed the following mechanism for CO oxidation over Co3O4/γ-Al2O3: the adsorbed CO reacts with an oxygen atom linked to the active Co3+ site; carbon dioxide is formed and then desorbs quickly. The result is a partially reduced site, which consists of two Co2+ ions or may be regarded as an oxygen vacancy. Using density functional theory (DFT), Broqvist et al. have studied CO oxidation mechanisms on the Co3O4(110)-B surface modeled as a three-layer slab, and proposed another mechanism:15 1) CO adsorbs at a surface Co3+ site; 2) CO then glides towards the oxygen, forming a surface CO complex which bonds to both the surface oxygen and the Co3+ site; 3) a charged OsurfCO species is formed with a bent equilibrium structure; and 4) desorption occurs by surmounting an energy barrier, which was not calculated. The oxygen ion bonded to three Co3+ ions was also referred to as the dominant site in determining the catalytic redox properties.
Jansson et al. also proposed a mechanism of CO oxidation by O2 on Co3O4 at low temperatures using a flow reactor and in situ FTIR studies,9 and suggested that the adsorbed CO reacts with the activated oxygen and desorbs as CO2. However, Pollard et al. described a different mechanism:16 CO adsorbs at a Co2+ site and interacts with an oxygen atom bonded to a neighboring Co3+ site, which is part of a strongly exothermic reaction that yields CO2. The oxygen atom that was lost is then replaced by the oxygen present in the input gas.
Moreover, Co3O4 has been widely investigated and recognized as a promising material, which might provide a technique for nitrous oxide abatement. Pure,17 partially substituted (MxCo1−xCo2O4; M=Ni, Mg, Zn),18, 19 alkali-doped[20–24], and supported25 Co3O4 based-catalysts demonstrate high catalytic activity for N2O decomposition. Zasada et al. reported that the K-doped Co3O4 spinel reveals the beneficial effect on the deN2O reaction by work function measurements and DFT modeling with the KOHsurf, K+surf+OHsurf, and K2Osurf adspecies on the Co3O4(100) surface.26 Piskorz et al. have studied the reaction mechanism of N2O self-decomposition over Co3O4(100) surface using DFT molecular modeling.27 However, little theoretical research has been carried out on the interaction of N2O with the Co3O4(110) surface.
The mechanism of CO oxidation on the Co3O4(110) surface is not well-understood. Some issues still remain to be resolved, such as which site (oxygen, Co2+, or Co3+ site) does CO adsorb initially, what are the processes involved and their energy barriers, which oxygen type is removed from the surface during CO oxidation with an oxygen vacancy, and what occurs during CO oxidation in the presence of N2O or O2? This work aims to resolve these questions. The DFT calculations were performed to investigate CO, N2O, and O2 individually adsorbing on the possible sites of perfect and defect Co3O4(110) surfaces, and to further explore the reaction mechanism of CO oxidation by N2O or O2.
Models and Methods
There are two types of surface terminations for the Co3O4(110) spinel orientation: type A, which contains two Co2+ cations (Cotet in the tetrahedral sites), two Co3+ cations (Cooct in the octahedral sites), four O2− anions, and type B, which has two Co3+ cations in the octahedral sites and four O2− anions. The surface terminated by type A (or type B) is denoted as an A (or B) surface (Figure 1). For the (110)-B surface (Figure 1 b), there are twofold and threefold coordinated O atoms in the surface layer. The coordination number of oxygen atoms is denoted by the superscript of the atom. The O2f ion in the outermost atomic layer is bonded to one Co2+ ion and one Co3+ ion, whereas the O3f ion in the third atomic layer has three Co3+ ions as nearest neighbors. The Co3O4(110)-B defect surface with an O2f vacancy at a defect concentration (Θdef) of 1/2 was created by removing an O2f atom in the surface layer from the perfect surface. The partially substituted Co3O4(110)-B surfaces, namely one surface Cooct substituted by Al3+ (Al-sub-Cooct), and one sublayer Cotet by Mg2+ (Mg-sub-Cotet), were chosen as the substituted surface models to explore the role of the Co2+–Co3+ ion pair on the activity of Co3O4(110)-B surface for the CO oxidation reaction.
The perfect, defective, and partially substituted (110) surfaces were modeled using a periodically repeated slab with five layers. The middle layer was fixed and the other four layers were fully relaxed. The vacuum region thickness between the slabs was 12 Å. Surface atomic relaxations and surface energies of the perfect Co3O4(110)-A and -B surfaces can be seen in our recent work.28
All calculations were performed with the DMol3 program package29 in Materials Studio of Accelrys Inc. The DFT slab approach with generalized gradient approximation of Perdew, Burke and Ernzerhof (GGA-PBE) functional30 and a double numerical basis with polarization functions (DNP) were adopted. All electron basis sets were used for C, N, and O atoms. Density functional semicore pseudopotentials (DSPP)31 were used for Co atoms, whereby the outer electrons (3d74s2) of Co atoms are treated as valence electrons and the remaining electrons are replaced by a simple potential including some degree of relativistic effects. A fermi smearing of 0.01 Hartree and a cutoff energy of 4.5 Å were used. The self-consistent field convergence criterion was set to an energy change of 10−6 Hartree. The convergence criterion of optimal geometry based on the energy, force and displacement convergence, were 1×10−5 Hartree, 2×10−3 Hartree Å−1 and 5×10−3 Å, respectively. Throughout calculations, the spin-polarized approach was used. The Brillouin zone integrations were performed using a 3×4×1 Monkhorst–Pack grid for (110) surfaces.
For the surface studies32, 33 of the 3d transition metals, it is necessary to include the valence of at least the 3p6 electrons, which may not be necessary for the description of the solid geometry, but would be crucial for processes involving large electronic changes, such as charge transfer or chemical bonding. The all-electron relativistic method, which includes all electrons explicitly and introduces some relativistic effects into the core as DSPPs, was performed to validate the presently used core potentials comparing CO at Co3+ and O2f sites (Table 1). The same trend is found for the adsorption configuration, the change value of Mulliken populations, adsorption energies between pseudopotentials, and all electron calculations. However, there are some differences in the absolute values, especially with great overestimations in the charge transfer and adsorption energy for CO at the octahedral Co3+ site. In any case, for the theoretical work, the main interest is in qualitative trends, not in absolute values.
Table 1. The equilibrium configuration parameters, change of Mulliken populations, and energies for CO adsorption through carbon at Co3+ and O2f sites on the Co3O4(110) perfect surfaces by different core treatments.
The CO bond length calculated by the all-electron relativistic treatment is 1.142 Å.
CO bond length [Å]
In general, theoretical calculations would be more accurate for 3D transition metal oxides by accounting for the correlations resulting from strong on-site Coulomb and exchange interactions of the 3D states. The magnetic structure, band gaps, and the density of states of the bulk Co3O4 were studied and compared with those calculated from the DFT+U calculation and some experimental values, to verify the feasibility of the standard DFT method for the Co3O4 system with our previous work.28 Our theoretical approach gives us a reasonable description of the Co3O4 system. The calculated bond lengths for adsorbate molecules (Table 2) are in good agreement with the corresponding experimental values.34 These results support the computational procedure and methods employed in this work. The adsorption energies (Eads) are calculated using Equation (1):
Table 2. Comparison of the calculated and experimental bond lengths [Å].
where Eadsorbate, Esubstrate, and Eadsorbate/substrate are the total energies of the isolated adsorbate molecule (CO, N2O, or O2), substrate, and adsorbate/substrate system, respectively. By this definition, a positive value, corresponding to an exothermic process, indicates a stable adsorption. One adsorbate molecule per unit cell was included in the calculations for the individual adsorption of CO, N2O, and O2 molecules.
Following the approach used by Ganduglia-Pirovano et al.,35 the average oxygen defect formation energy is defined by Equation (2) or Equation (3):
where Edef, Efree, EO, and E represent the total energies of the defect system, the defect-free system, the free oxygen atom and the molecular oxygen, respectively. Ndef is the number of oxygen defects in the model. Ef (O) and Ef (O2) denote the energies of atomic and molecular oxygen used as references, respectively. A positive value for Ef (O) or Ef (O2) indicates that the vacancy formation is endoenergetic and the difference between them is just half the dissociation energy of the gas-phase O2 molecule, which was estimated to be 5.828 eV.
Vacancy Formation Energy of Co3O4(110)-B Surface
Table 3 shows the calculated defect formation energies Ef (O) and Ef (O2) for O2f-vac formation at full relaxation of 5.641 eV and 2.727 eV for Θdef=0.5, respectively. The defect formation energy of a fully relaxed 5 L slab with an O3f-vac (Θdef=0.5), which was created by removing an O3f atom from the surface, was also calculated. O2f-vac formation is easier due to its lower energy cost than O3f-vac. The implications and problems related to the calculation of O vacancies in TM oxides were discussed in depth and reviewed by Ganduglia-Pirovano et al.35 and Pacchioni.36 The values reported here cannot be taken as true values, due to the intrinsic limitations of pure DFT methods in describing TM oxides and having no available experimental values as a reference.
Table 3. Oxygen defect formation energies (eV) at different Θdef and relaxation values.
The effects of relaxation and defect concentration on the defect formation energy were considered. For fully relaxed O2f-vac surface, the difference of defect formation energy between Θdef=1 and Θdef=0.5 is 0.216 eV, which indicates that the surface with a lower defect concentration is more stable and easier to form. The difference of 0.015 eV between the fully relaxed surface and partially fixed surface with the middle layer in the bulk position confirms that the surface with the middle layer fixed appears to be representative of the Co3O4(110) surfaces.
CO, N2O, or O2 on Co3O4(110) Perfect and Defect Surfaces with an O2f Vacancy
The equilibrium configuration parameters, Mulliken populations, and individual adsorption energies for CO, N2O or O2 molecules on the Co3O4(110) perfect surfaces (A and B) and on the defect B surface with an O2f vacancy are listed in Tables 4–6, respectively. The distances and angles between the adsorption site and the adsorbate are represented by Re and α, respectively. The change of the Mulliken populations of adsorbates, adsorption sites, the surface octahedral and sublayer tetrahedral sites after adsorption are denoted by ΔqCO, Δq, Δq, Δqsite, Δqoct, and Δqtet, respectively.
Several possible surface adsorption sites (Cotet, Cooct, O3f for A surface, Cooct, O2f, and O3f for B surface) were chosen to simulate CO, N2O, or O2 adsorption on the Co3O4(110) perfect surfaces. The CO molecule is selectively chemisorbed through carbon at the Cotet and Cooct sites for the A surface, and at the Cooct and O2f sites for the B surface. For two types of oxygen ion sites, CO adsorbs at O2f sites but not at O3f sites, which is consistent with the fact that the low-coordinated oxygen site at oxide surface is more reactive. N2O (through nitrogen) and O2 selectively adsorb at the Cotet and Cooct sites for the A surface and at the Cooct site for the B surface.
CO Adsorption on Co3O4(110) Perfect and Defect Surfaces with an O2f Vacancy
On the (110)-A perfect surface (Table 4), CO adsorbs perpendicularly at Cotet or Cooct sites, with the elongation of CO bonds. The Re values indicate a strong bond interaction between CO and Cotet (or Cooct) sites. The closer Eads values show that competitive adsorption of CO occurs at Cotet and Cooct sites.
Table 4. The equilibrium configuration parameters, change of Mulliken population, and energies for CO adsorption through carbon on the Co3O4(110) perfect and defect surfaces.
The values from reference 15 are indicated in parentheses.
CO bond length [Å]
On the (110)-B perfect surface, Cooct is the most preferred site for CO adsorption, with an adsorption energy of 1.850 eV. The values of Re and Eads are very close to those calculated by Broqvist et al .15 The difference is due to the CO not being adsorbed atop the Cooct site in this work, but along the direction towards a dangling bond (Figure 2 a). The results of Broqvist et al. were obtained from the potential energy scan by single point energy calculations. At the O2f site, CO forms bonds with both O2f and Co3+ sites (Figure 2 b). The CO bond has greatly been elongated from 1.132 to 1.184 Å. For the charge transfer, CO donates 0.335 electrons to the substrate, and the majority of charge (0.224) accumulates at the Co3+ site. The sublayer Co2+ ion bonded to the O2f site loses 0.014 electrons after adsorption, and a charged CO2 surface species forms. It has a charge of −0.064 and a bent equilibrium structure with an O2f-C-O bond angle of 144°, which suggests that a species like CO22− forms. According to the above discussion, we concluded that CO oxidation on the (110)-B surface is a result of the high oxidative activity of Co3+, whether at O2f or Cooct sites. These results are highly consistent with earlier experimental findings,37, 38 which show that the presence of Co3+ in the octahedral sites on the surface of the catalyst is the main contributing factor for the activity of the CO oxidation. Upon CO adsorption, the surface is reduced, due to the electron transfer of Co3++ e−→Co2+.
Focusing on the change of the CO bond length upon adsorption (Table 4), there is an elongation in the range 1.139– 1.184 Å for the CO at the A and B perfect surfaces, compared with the calculated value of 1.132 Å. The elongation of the CO bond should also correspond to a red-shift in the CO vibrational frequency, which is consistent with the experimental red-shift on low-temperature oxidation of CO by a Co3O4 powder surface.39 The typical explanation for the red-shift for CO in transition metal complexes is back-bonding from the metal atom to the 2π antibonding orbital (2π*) of the CO molecule, which is assumed in the Blyholder mechanism.40 In this case however, the analysis of Mulliken populations CO is positively charged, and there is a strong charge transfer from CO to the surface, primary to adsorption sites. It indicates that the highest occupied molecular orbital (HOMO) 5σ of the CO molecule transfers electrons to the substrate. Since the 5σ orbitals are slightly antibonding or essentially nonbonding,41 5σ donation-dominated bonding should cause a slight blue-shift or have no intrinsic effect on the CO vibrational frequency. Does the π*-back-donation takes part in the bonding interactions? If the empty 2π* orbital of CO obtains back-donated electrons from the surface, then the HOMO of the adsorption system should be the interaction of 2π* orbitals and substrates. Thus, without loss of generality, the HOMOs of the CO molecules individually adsorbed at the Cooct and O2f sites on the (110)-B perfect surface (Figure 2), which were used to explain the discrepancy of the CO bond elongation and its positive-charged character upon adsorption. The CO 2π* orbitals were found to interact with the substrate. These analyses revealed that there are two types of interactions for CO adsorption at Cooct and O2f sites, namely σ-donation and π*-back-donation, and the latter is responsible for the observed red-shift. The CO red-shift and the same bonding mechanism have been reported in a number of theoretical (CO/CuCr2O4(100),42 CO/Cu2O(111),43, 44 CO/Al2O345) and experimental (CO/Co3O4,39 CO/Ni-ZSM-546, 47) studies.
On the Co3O4(110)-B defect surface with an O2f vacancy, CO adsorbs at the O2f-vac site through carbon. The calculated Eads was 1.780 eV, which was significantly increased compared to the case on the perfect surface. Whereas CO adsorbed at the Cooct site, the Eads value decreased due to the presence of an O vacancy, which implies that CO preferably adsorbs on the Cooct site of the perfect surface rather than the defect surface.
N2O Adsorption on Co3O4(110) Perfect and Defect Surfaces with an O2f Vacancy
For N2O on the (110)-A and -B perfect surfaces (Table 5), the calculated adsorption energies are below 0.510 eV, which indicates a weak interaction between N2O and the surface. There is a slight difference in N2O geometry with respect to the free molecule. The charge flow takes place from N2O to the Co3O4(110)-A and -B perfect surfaces. The self-decomposition of N2O does not occur on the Co3O4(110)-A and -B perfect surfaces.
Table 5. The equilibrium configuration parameters, change of Mulliken population, energies for N2O adsorption on the Co3O4(110) perfect and defect surfaces.
Cotet N end
Cooct N end
Cooct N end
O2f-vac N end
O2f-vac[a] O end
Cooct N end
Cooct O end
[a] The values in parentheses refer to N2O absorption at the O2f-vac site of the Mg-sub-Cotet defect surface. [b] The difference in Mulliken populations of Cooct between the non-adsorbed perfect and defect B surface, and between the nonadsorbed perfect and defect Mg-sub-Cotet B surface are 0.042 and 0.062, respectively. [c] The difference of Mulliken populations of Cotet between the nonadsorbed perfect and defect B surfaces, and between the nonadsorbed perfect and defect Mg-sub-Cotet B surfaces are 0.131 and 0.017, respectively.
NN bond length [Å]
NO bond length [Å]
On the B defect surface with an O2f vacancy, N2O adsorbs favorably at the O2f-vac site, both through N end and O end of N2O. The adsorption energy for the latter is almost three times as large as that for the former (Table 5), which implies CO will preferentially adsorb through the O end of N2O at the O2f-vac site. In this case, N2O dissociates into N2 and surface oxygen (Figure 4). The interaction energy between N2 and the rehabilitated B surface was calculated to be 0.0917 eV, which indicates that there is a very weak interaction between N2 and the rehabilitated B surface. In other words, the produced N2 is unbound. The surface oxygen fills the vacancy site and the (110)-B surface is regenerated, which is consistent with the experimental prediction that the surface vacant sites are responsible for nitrous oxide decomposition.17, 48
O2 Adsorption on Co3O4(110) Perfect and Defect Surfaces with an O2f Vacancy
Two adsorption orientations, the end-on one, involving the O2 molecule axis perpendicular to the surface, and the side-on one, involving the O2 molecule axis parallel to the surface, were investigated for O2 on the Co3O4(110) perfect and defect surfaces. The O2 side-on adsorption mode is more stable than the O2 end-on mode (Table 6). The adsorbed O2 molecule is polarized and negatively charged. The adsorption causes a charge flow and the substrate donates electrons to O2. Interestingly, the majority of electrons is not donated by the corresponding adsorption site. Moreover, some adsorption sites do not donate, but obtain electrons for O2 adsorbed at the Cooct site through the end-on mode on the perfect A surface and the Cooct site on both the perfect and defect B surfaces. The adsorption process leads to charge transfer from the O2 species not only to the adsorption site, but also to the other surface ions. The strong reducing ability of Co3O4(110) surfaces to the O2 molecule is not from the reducing ability of Co3+ or Co2+, but due to the electrons flowing within the surface ions.
Table 6. The equilibrium configuration parameters, change of Mulliken population, energies for O2 adsorption on the Co3O4(110) perfect and defect surfaces.
[a] The values in parentheses refer to O2 adsorption at the O2f-vac site of the Mg-sub-Cotet defect surface. [b] The difference of Mulliken populations of Cooct between in the non-adsorbed perfect and defect B surface and between the non-adsorbed perfect and defect Mg-sub-Cotet B surface are 0.042 and 0.062, respectively. [c] The difference of Mulliken populations of Cotet between the non-adsorbed perfect and defect B surfaces and between the non-adsorbed perfect and defect Mg-sub-Cotet B surfaces are 0.131 and 0.017, respectively.
OO bond length [Å]
For the perfect (110)-A surface, O2 is most favorably adsorbed at the bridge Cooct site (b-Cooct) and forms bonds with two adjacent Cooct sites, to give the largest adsorption energy of 1.897 eV. The OO bond length is 1.323 Å, elongated by 0.108 Å compared to the calculated value 1.215 Å. For O2 adsorbed at the O2f-vac site on the (110)-B surface, the O atom of O2 occupies the vacancy site. The OO bond has been stretched to 1.345 Å (Figure 4) with a charge of −0.381 e, which indicates that a species like superoxo (O2-) forms. The OO bond was weakened, but not broken. Moreover, CO interaction with the B surface, on which O2 was pre-adsorbed at the O2f-vac site, was considered. As a result, a weak CO2 physisorbed state is formed (CO bond length=1.174 Å, 1.158 Å; Figure 4) and the B surface is regenerated with an adsorption energy of 4.363 eV.
Formation Process of Reduced Co3O4(110)-B Surface with an O2f Vacancy by COads
As discussed above, CO adsorption at the O2f site of Co3O4(110)-B surface might lead to CO2 formation, resulting in an oxygen vacancy and a reduced surface. In this section, the formation process of reduced Co3O4(110)-B surface with an O2f vacancy by COads in the Co3O4(110)-B surface was explored. A search of the transition states (TSs) was carried out by complete linear synchronous transit and quadratic synchronous transit (LST/QST) method.
The adsorption energy for CO at the Cooct site is more than twice that for the O2f site (Table 4). The difference in adsorption energies is large enough to predict that CO can not be adsorbed at the O2f site until the Cooct site is completely covered by CO. Thus, the geometric optimization configuration that CO adsorbs at the O2f site of the pre-adsorbed surface, in which CO has been adsorbed at the Cooct site, was used to model the initial state. The overall energy and geometrical diagram regarding the process of CO oxidation to CO2 with an O2f abstraction are depicted in Figure 3.
In the initial state, the lengths of CotetO2f (2.339 Å) and CooctO2f (2.179 Å) bonds are largely increased, compared with the corresponding lengths (2.054 and 1.991 Å) for the CO adsorbed on the clean surface, respectively (Figure 2 b). In the transition state, the CotetO2f and the CooctO2f bonds are broken. Then, the product CO2 leaves the surface, with the formation of an oxygen vacancy. The low reaction barrier (0.033 eV) indicates that the oxidation of CO to CO2 using lattice oxygen is kinetically favored and can proceed at low temperatures. The large adsorption of heat for CO interacting with O2f and Cooct sites offsets the energy needed for breaking the CooctO bond, resulting in an exothermic process releasing 0.13 eV. This result implies that the adsorptions at the O2f and adjacent Cooct sites are synergistic.
Activity of Partially Substituted Co3O4(110)-B Perfect and Defect Surfaces for CO Oxidation
For N2O and O2 adsorbed at the O2f-vac site, both Cotet and Cooct sites lose electrons and the former dominates (Tables 5 and 6). Interestingly, the loss of electrons from the Cotet site for N2O (0.131) or O2 reduction (0.121) are almost equal to the difference (0.131) of charge of the Cotet site between the nonadsorbed perfect (0.754) and defect (0.623) surfaces. The loss of electrons from the Cooct site for N2O (0.047) or O2 reduction (0.055) are close to the difference (0.042) of charge of the Cooct site between the nonadsorbed perfect (0.679) and defect (0.637) surfaces. These results indicate that the redundant negative charges due to the O2f-vac formation, which disperse at the Cotet and Cooct sites, are released when N2O or O2 is adsorbed at the O2f-vac site, and the loss of the electrons from the Cotet site is the main contributor to the reduction processes (N2O to N2 and O2 to O2−). For the oxidation of CO to CO2, the Cooct site obtaining the electrons is responsible for the high catalytic activity. Looking at the values of Δqoct and Δqtet (Tables 4–4, 5, 6), the role of Cooct and Cotet on the processes of oxidation and reduction can be identified elementarily. The Cooct and Cotet sites play significant roles in the oxidation and reduction processes, respectively.
Moreover, two partially substituted surfaces, namely Al-sub-Cooct and Mg-sub-Cotet, were employed to further determine the role of the Co2+–Co3+ ion pair on the redox reaction. Both Al3+ and Mg2+, with the electronic configuration 3s0p0, have lower redox ability compared with Co3+ and Co2+, respectively. The corresponding results for CO adsorption through carbon on the substituted Co3O4(110)-B surfaces and the oxidant molecules (N2O and O2) at the O2f-vac site of the Mg-sub-Cotet Co3O4(110) defect surface are shown in Table 7 and in Tables 5 and 6, respectively.
Table 7. The equilibrium configuration parameters, change of Mulliken population, and energies for CO adsorption on the substituted Co3O4(110) perfect surface.
CO bond length [Å]
For the Al-sub-Cooct surface, weak adsorption with low adsorption energies occurs for CO at the O2f and Aloct sites, which is in agreement with the findings of Broqvist et al. in that CoAl2O4 is inactive for CO oxidation.15 For the Mg-sub-Cotet surface, the equilibrium configuration parameters (Table 7) for CO at the O2f and Cooct sites are close to those at the clean perfect Co3O4(110)-B surface (Table 4). The results indicate that the surface of Co3+ substituted by Al3+ is inactive for CO oxidation, whereas the surface activity of the Mg-sub-Cotet surface remains, due to the presence of Co3+ when Co2+ in the sublayer was replaced by Mg2+.
For the oxidant molecules at the O2f-vac site of the Mg-sub-Cotet defect surface (O2f vacancy values are shown in Tables 5 and 6), N2O (through oxygen) decomposes to N2 and surface oxygen, and O2 forms a superoxo species. CO2 is formed in its physisorbed state (CO bond length=1.172 Å, 1.160 Å) and the surface is restored after CO reacts with O2− on the surface. These results are similar to those on the unsubstituted defect surface. The difference is that the redundant negative charges due to the O2f-vac formation spread out at the Mgtet and Cooct sites (with the latter slightly more) but, for the unsubstituted defect surface, at the Cotet and Cooct sites (primarily at the former). The tetrahedral site is not responsible for the reduction process (N2O to N2, O2 to O2−), which implies that the ion in the tetrahedral site plays a role in accumulating and releasing electrons in the process of formation and disappearance of the O2f-vac site, respectively. Thus, there is little effect on the process of N2O or O2 reduction from substituting Co2+with Mg2+ in the tetrahedral site, although the Mgtet ion has low reductive activity. These results show that the Cooct site is irreplaceable by the ion with low oxidative activity, while the Cotet site can be substituted by an ion with low reductive activity. We conclude that the Co2+–Co3+ ion pair, or more specifically, the strong oxidative activity of the Cooct site and the ability of the Cotet site to store and release electrons during the formation and disappearance of the O2f-vac site, contribute to the activity of the CO+N2O and CO+O2 reactions on the Co3O4(110)-B surface.
The calculated results reveal the sequence of the adsorption energies for CO, N2O, and O2 individually at the Co3O4(110) surfaces: N2O(0.510 eV)<CO(1.167 eV)<O2(1.897 eV) for the perfect A surface, N2O(0.274 eV)<O2(1.175 eV)<CO(1.850 eV) for the perfect B surface, and N2O(2.970 eV)>O2(1.925 eV)>CO(1.780 eV) for the defect B surface. For the B surfaces, the reductant molecule, CO is adsorbed preferentially at the perfect surface, whereas the oxidant molecules, N2O and O2, are preferentially adsorbed at the defect surface.
Proposed Mechanism of CO + N2O Reaction on the Co3O4(110)-B Surface
CO oxidation by N2O with the formation of N2 and CO2, which is an important reaction for N2O abatement and exhaust-gas purification, has been studied in detail on several metal surfaces (Pt,49 Pd,49–52 Rh,53–55 Ir56, 57), zeolites,58–62 a number of atomic cations (M+),63–66 cationic platinum clusters (Ptn+),67 and transition metal oxides Ce0.75−xTi0.25PdxO2−δ.68 Several reaction mechanisms have been proposed with two types being classified. For metal surfaces, zeolites, M+, or Ptn+ acting as catalysts, the reaction involves the following steps: N2O adsorbs dissociatively (leaving oxygen on the catalyst), releases N2, and then CO interacts with the adsorbed O, yielding CO2. For metal oxides, the Mars–van Krevelen mechanism is followed:69 CO interacts with the catalyst, forming oxygen vacancies and CO2, then N2O adsorbs dissociatively, and the resultant O fills the vacancy.
According to the results of CO or N2O on the perfect and defect Co3O4(110)-B surfaces, adsorption energies for CO on the perfect B surface are significantly higher than those found during N2O adsorption, which indicates that CO would be adsorbed primarily when these two gases approach the surfaces simultaneously. N2O self-decomposition does not occur on the perfect Co3O4(110)-B surfaces. Thus, a model for N2O reduction by CO was proposed (Figure 4). CO is preferentially adsorbed at the Cooct site; after the Cooct sites are fully covered, CO reacts with the neighbor O2f ion, which leads to the formation of CO2 and an oxygen vacancy reduced site. N2O is then dissociatively adsorbed on an oxygen vacancy site, generating N2 and Oads. The N2 molecule leaves the surface and Oads fills the oxygen vacancy. As a consequence, the Co3O4(110)-B surface is regenerated. The key step is the formation of an oxygen vacancy, which makes the CO+N2O reaction recyclable and sustainable. On the contrary, the CO concentration on the surface determines the formation rate of the oxygen vacancies and not the N2O decomposition to N2. This redox mechanism is in accordance with Mars–van Krevelen mechanism, which is also what was followed for the CO oxidation over metal oxide catalysts.
Proposed Mechanism of CO+O2 Reaction on the Co3O4(110)-B Surface
On the basis of the results for CO or O2 on the perfect and defect Co3O4(110)-B surfaces, we have proposed the mechanism of CO oxidation by O2 (Figure 4). As in the CO+N2O reaction, CO is adsorbed initially, which leads to the formation of an oxygen vacancy; O2 is then adsorbed at this vacancy site. The second CO interacts with the adsorbed O2− species, leading to the formation of a CO2 physisorbed state and a recovered surface. The latter two steps are exothermic and have no energy barriers. The reactions can then proceed in a recyclable fashion. Comparable to the reaction of CO+N2O, the key step is the formation of an oxygen vacancy. A similar mechanism was found for the CO+O2 reaction over the RuO2(110) surface.70
The mechanisms for the catalytic cycles of CO+N2O and CO+O2 reactions on the Co3O4(110) surface were investigated by density functional calculations. The established mechanism for CO oxidation on the Co3O4(110)-B surface at low temperatures is a stepwise process where the surface is first reduced by CO causing the formation of an oxygen vacancy, followed by a reoxidation of the surface by gas phase N2O or O2. The metal atoms must be able to change oxidation states during the repeated reduction and oxidation of the surface in order to facilitate this process. The easy removal of the lower-coordinated O2f atom and the Cooct site, which can change oxidation states and have strong redox activity, ensure that CO oxidation reactions by N2O or O2 are sustainable.
In conclusion, CO oxidation by N2O or O2 on the Co3O4(110)-B surface follows the Mars–van Krevelen mechanism. The reaction of CO alone with Co3O4 initially yields CO2, and then the rate of this reaction quickly drops. This rate decrease is because active surface sites (active surface O) tend to deplete with the formation of an oxygen vacancy, and the surplus CO molecules fill the vacancy, binding through carbon with the formation of carbonaceous species Csurf to deactivate the catalyst completely. These results can be used to explain the experimental results71 of Jansson et al. that the rate of Co3O4 catalyst deactivation for CO oxidation at low temperatures increases with increasing CO concentration. In the presence of N2O or O2, the rate of CO oxidation can be maintained because they can replenish the O vacancy and rehabilitate the surface. With the help of the O2f vacancy as a bridge, the CO+N2O and CO+O2 reactions can proceed with a catalytic redox cycle on the Co3O4(110)-B surface.
We kindly acknowledge financial support from the Natural Science Foundation of China (20673019, 20303002), the Key Project of the Fujian Province (2005HZ01-2-6), the Foundation of State Key Laboratory Breeding Base of Photocatalysis of Fuzhou University (K-081001), and the Foundation of State Key Laboratory of Structural Chemistry of Fujian Institute of Research on the Structure of Matter at the Chinese Academy of Sciences (20090060).