Understanding complete oxidation of methane on spinel oxides at a molecular level

It is crucial to develop a catalyst made of earth-abundant elements highly active for a complete oxidation of methane at a relatively low temperature. NiCo2O4 consisting of earth-abundant elements which can completely oxidize methane in the temperature range of 350–550 °C. Being a cost-effective catalyst, NiCo2O4 exhibits activity higher than precious-metal-based catalysts. Here we report that the higher catalytic activity at the relatively low temperature results from the integration of nickel cations, cobalt cations and surface lattice oxygen atoms/oxygen vacancies at the atomic scale. In situ studies of complete oxidation of methane on NiCo2O4 and theoretical simulations show that methane dissociates to methyl on nickel cations and then couple with surface lattice oxygen atoms to form –CH3O with a following dehydrogenation to −CH2O; a following oxidative dehydrogenation forms CHO; CHO is transformed to product molecules through two different sub-pathways including dehydrogenation of OCHO and CO oxidation. The development of methane oxidation catalysts made of earth-abundant elements is an important challenge. Here, the authors report a cost-effective nickel-cobalt oxide which outperforms precious-metal-based alternatives, due to the combination of transition metal cations and surface oxygen vacancies.

A complete oxidation of hydrocarbons has been a significant topic in the field of heterogeneous catalysis for decades since unburned hydrocarbons must be transformed in power plants or before release of gas exhaust from various engines of vehicles using gasoline, diesel, natural gas or liquid petroleum gas (a mixture of propane, isobutene, n-butane and ethane). Thermal stability of a catalyst (for removal of CH 4 or other hydrocarbons) at a temperature of 1,300°C for applications to power plants or at least 800°C for engines using gasoline or diesel is requested. Ying et al. reported that CeO 2 nanoparticles supported on barium hexaaluminate exhibits high activity at a quite high temperature and remains its structure up to 1,300°C (ref. 1). Recently, Gorte, Fornasiero, and Farrauto and their coworkers reported that a catalyst of Pd nanoparticles caged in a porous CeO 2 shell supported on Al 2 O 3 exhibits exceptionally high activity for a complete oxidation of methane with a high thermal stability up to 850°C (refs 2,3). The activity of Co 3 À x Cu x O 4 , Co 3 À x Zn x O 4 , and Co 3 À x Ni x O 4 for oxidation of hydrocarbons was reported in 1968 and the following studies [4][5][6][7] . A very recent study showed the high activity of these oxides in oxidation of (a) a mixture of liquefied petroleum gas and CO or (b) a liquefied petroleum gas without CO 8 .
Use of natural gas vehicles is a programme launched in Europe and other countries towards diversity of energy sources and utilization of methane, the inexpensive, earth-abundant energy sources [9][10][11][12] . The exhaust of natural gas engines has the following features 9 : low temperature (o500-550°C), low CH 4 concentration (about 1,000 p.p.m.), large amount of water vapour (10%), high O 2 :CH 4 ratio, and presence of NO x or SO x (refs 9,12). A complete oxidation of methane at a relatively low temperature (o500-550°C) is critical for abating unburned methane of the exhaust of engines of natural gas vehicles 9,[11][12][13] . Pt and Pd supported on Al 2 O 3 are the well-studied catalysts 9 . Significant efforts have been made in fundamental understanding of mechanism of the complete oxidation of methane 14 and optimization of catalytic performances of these Pt-or Pd-based catalysts 9 . From reaction mechanism point of view, Neurock and Iglesia et al. first elucidated the reaction mechanism on Pd-based catalysts at a molecular level 14 ; these Pd-based catalysts exhibit different mechanisms of activation of C-H along increase of chemical potential of oxygen.
The high cost of precious metals has driven the search of catalysts made of earth-abundant elements. Spinel oxide, a type of reducible early transition metal oxides, consists of M 2 þ and M 3 þ in its lattice. Co 3 O 4 is one of the spinel oxides. It has almost weakest M-O bonds among all transition metal oxides 15,16 . Oxygen vacancies can be readily generated even at a temperature below 25°C (refs 17,18). In addition, the barrier for hopping oxygen vacancies on the surface of Co 3 O 4 is only 0.23 eV (ref. 17). Due to the high density of oxygen vacancies of surface of Co 3 O 4 (refs 15,19), Ni cations and oxygen vacancies are mixed at atomic scale, while Ni cations are integrated to the lattice of Co 3 O 4 through formation of a doped spinel oxide Ni x Co 3 À x O 4 . Figure 1a schematically shows the integration of Ni cations, Co cations and surface lattice oxygen atoms and vacancies at atomic scale. Although catalytic activity of a complete oxidation of hydrocarbons on doped cobalt oxides was reported as early as 1968 (refs 4-8), there is lack of mechanistic understanding of the complete oxidation of hydrocarbons at a molecular level. Compared with the well-studied complete oxidation of CH 4 on Pt or Pd-based catalysts 9 , the complete oxidation on NiCo 2 O 4 could fellow a different mechanism.
Here we focus on mechanistic exploration of a complete oxidation of methane on NiCo 2 O 4 at a molecular level. Multiple in situ techniques are used in experimental exploration of the catalytic mechanism. Surface chemistry of this mixed oxide during catalysis is tracked with ambient pressure X-ray photoelectron spectroscopy (AP-XPS) 20 . Vibrational signature of surface adsorbates is identified with in situ infrared spectroscopy. In addition, computational studies for understanding of the mechanism of the complete oxidation of methane on NiCo 2 O 4 at a molecular scale are performed.

Results
Catalytic performance for complete oxidation of CH 4 . Catalyst NiCo 2 O 4 was synthesized with a coprecipitation method described in the section of Methods. X-ray diffraction (Fig. 1b Catalytic performances of a complete oxidation of CH 4 on these catalysts were measured in a fixed-bed flow reactor. Certain amount of a catalyst given in figure captions was used in different experiments. 10% CH 4 balanced with Ar, and 99.99% O 2 , and 99.99% Ar with certain flow rates were mixed and then flowed through a catalyst bed. The molar ratio of the introduced CH 4 and O 2 is 1:5 for all experiments in Fig. 2. NiCo 2 O 4 is extraordinarily active in combustion of CH 4 to CO 2 and H 2 O. CH 4 is completely combusted at 350°C, while gas hourly space velocity (GHSV) of methane is 24,000 ml 5% CH 4 g À 1 h À 1 (Fig. 2). At the same GHSV, pure Co 3 O 4 and pure commercial 10 (Fig. 3a). Kinetic studies were performed in the temperature range of 270-330°C, while catalysis was in a kinetics control regime by controlling conversions o10%. The measured activation barrier is about 108 kJ mol À 1 (Supplementary Fig. 3). In addition, NiCo 2 O 4 exhibits durability in complete oxidation of methane at 350°C and 550°C for over 48 h (Fig. 3b,c). The thermal stability of catalyst structure was confirmed by the preservation of X-ray diffraction pattern of NiCo 2 O 4 after a complete oxidation at 550°C (Supplementary Fig. 4). In addition, NiCo 2 O 4 remains the original particle size after a complete oxidation at 550°C as shown in Supplementary Fig. 5. This is consistent with the To achieve a deep understanding of the complete oxidation of methane on NiCo 2 O 4 , we performed in situ studies using AP-XPS 15,20,23,24 in the temperature range of 60-400°C by following the reaction conditions used for measurements of catalytic performances in a fixed-bed flow reactor. Surface However, the atomic ratio O/(Ni þ Co) in the gaseous environment of CH 4 and O 2 exhibits a temperature-dependent evolution which will be discussed in the following paragraphs.
During the data acquisition of in situ studies, NiCo 2 O 4 catalyst was remained in a mixture of CH 4 and O 2 (molar ratio 1:5) in Torr pressure range. Figure 4, Supplementary Fig. 6 and Supplementary Fig. 7 Fig. 4a,b, the photoemission features of Ni 2p and Co 2p remain constant in the temperature ranges of 60-400°C. Surprisingly, two C 1s peaks at 285.6 eV (peak 1) and 288.3 (peak 2) were clearly observed even at 60°C (Fig. 4d). Figure 5a shows the evolution of the surface atomic ratio, C Ni þ Co of the two C 1s peaks of Fig. 4d along the increase of catalysis temperature. One interesting evolution is a quick increase of species 2 (peak 2 in Fig. 4d), a following decrease, and then a nearly complete disappearance of this species at about 250°C (red lines in Fig. 5a,b) where the conversion of CH 4 starts to increase largely (Fig. 2). The sharp decrease of C peak2 Ni þ Co on NiCo 2 O 4 (red line in Fig. 5a) and a following decrease and then disappearance at 250-300°C along the rapid increase of catalytic The gas composition of the feed gas is 5% CH 4 , 25% O 2 and 70% Ar. Molar ratio of CH 4 to O 2 in the gas mixture is always 1:5. As 500 mg NiCo 2 O 4 catalyst was used, the GHSV of CH 4 is 24,000 ml 5% CH 4 g À 1 h À 1 ; at this GHSV the conversion of CH 4 on NiCo 2 O 4 at 350°C is 100%.   In situ infrared studies of catalysts helped assign the two species corresponding to peaks 1 and 2 in Fig. 4d. As shown in Fig. 6, vibrational peaks at 1,280 1,430 and 1,561 cm À 1 were observed in the temperature range of 50-160°C. Referred to the literature 25 , these vibrational signatures are contributed from formate species (OCHO). The assignment of these vibrational peaks to a formate species is supported by the agreement between C 1s photoemission peak at 288.3 eV (peak 2 in Fig. 4d) and 288.5 eV of the reported formate species 25 . Thus, both in situ photoemission feature of C 1s and in situ vibrational signature of surface species suggest the formation of OCHO species on the surface at a relatively low temperature. C-H vibrational peaks at 2,900-3,000 cm À 1 were identified during in situ studies of CH 4 combustion on NiCo 2 O 4 . They are contributed from C-H stretching of the spectator CH n species formed from dissociative adsorption of CH 4 , supported by the low C 1s binding energy of CH n (peak 1) at 285.9-284.8 eV (refs [25][26][27], and C-H stretching of the intermediate OCHO. To investigate whether that the C 1s photoemission feature at 284.9-285.8 eV (peak 1 in Fig. 4d) is assigned to adsorbed CH n species or accumulated atomic carbon of coke, surface of the active catalyst at 400°C or 140-200°C in the mixture of CH 4 and O 2 (1:5) was tracked on CH 4 was purged but O 2 kept. The in situ AP-XPS ( Supplementary Fig. 8  Ni þ Co , and Cpeak2 Ni þ Co and (b) atomic fraction of carbon species 1 (peak 1), Cpeak1 Ctotal and carbon species 2 (peak 2), Cpeak2 Ctotal at different temperatures. The error of measurement of peak area is ±5% of the corresponding peak area. and but O 2 remained. As shown in Supplementary Fig. 8, C 1s photoemission feature of the left CH n species disappeared within a few minutes due to a quick oxidation by O 2 to form CO 2 . A similar study was designed to test whether the peak 1 of C 1s spectrum observed at 150°C in the mixture of CH 4 and O 2 (1:5) (Fig. 4d) is coke-like carbon or chemisorbed CH n species. As shown in Supplementary Fig. 9, these carbon-containing species formed at 140°C reacted with O 2 to form CO 2 (Fig. 7b,c) along the increase of the catalysis temperature of complete oxidation of methane.
In situ studies using isotope-labelled reactant and catalyst. Both in situ AP-XPS and infrared studies have shown the formation of the two species on NiCo 2 O 4 surface in the mixture of CH 4 and O 2 , formate (species 2) and CH n (n ¼ 0-3; species 1). To elucidate the source of oxygen atoms of the intermediate of OCHO species, we performed two temperature programmed reactions using isotope-labelled reactant and isotope-labeled catalyst: (1) CH 4 and O 18 2 on NiCo 2 O 16 4 (Fig. 8a) and (2) 16 4 À x O 18 x was done in an attached high-pressure reactor of AP-XPS system which is different from the reaction cell of AP-XPS since our AP-XPS system has both reaction cell for AP-XPS studies and high-pressure reactor for pretreatment.  Upon the preparation, O 18 2 was purged and a UHV environment (3 Â 10 À 9 torr) was achieved before NiCo 2 O 16 4 À x O 18 x was transformed to the reaction cell of AP-XPS. Then, combustion of CH 4 was performed on this isotope-labeled catalyst in pure O 16 2 in the reaction cell of AP-XPS. The evolution of partial pressure of the three potential products CO 16 (Fig. 8b). It is noted that CO 16 O 18 (M/z ¼ 46) was produced even at a relatively low temperature 130-150°C; However, the partial pressure of CO 16  Catalytic performance for removal of CH 4 in gas exhaust. To examine the catalytic performance of NiCo 2 O 4 in a complete oxidation of CH 4 of gas exhaust of an engine of natural gas, we measured catalytic performance of a complete oxidation of CH 4 on NiCo 2 O 4 in two different mixtures with gas compositions (Fig. 9) similar to the exhaust of lean-burn natural gas engine: (1) CH 4 0.2%, O 2 5%, CO 2 15%, H 2 O 10% balanced with Ar with a flow rate of 200 ml min À 1 (Figs 9a), and (2) CH 4 0.2%, O 2 5%, NO 0.15%, H 2 O 10% balanced with Ar (Fig. 9b). These gas compositions are typically used in the evaluation of catalytic performance of Pd-and Pt-based catalysts for a complete oxidation of CH 4 in the exhaust of engines of natural gas [11][12][13] (Fig. 9b). As the cost of Pd per kg is largely    Fig. 10, three different substitution models of crystals were tested in our calculations, namely substituting two Co 3 þ with two Ni 3 þ (type I, Fig. 10a), substituting one Co 3 þ and one Co 2 þ with one Ni 3 þ and one Ni 2 þ , respectively (type II, Fig. 10b ), and two Co 2 þ with two Ni 2 þ (type III, Fig. 10c). The type I crystal structure was found to be the most stable in our calculations, while the type II was slightly less stable by 0.01 eV per unit cell higher in energy. The type III was most unfavourable thermodynamically, 0.20 eV higher per unit cell than that of type I. In the following calculations, types I and II were considered.  Table 2) suggest that the 110-B of type I crystal are the most active for CH 4 activation on Ni 3 þ with lowest activation energy of only 0.52 eV which is lower than those on Ni(111) (0.88 eV) 30 , Ni(211) (0.61 eV) 30 and Pd(100) (0.79 eV) 31 . Therefore, the low barrier on 110-B of type I crystal is responsible for the high reactivity of NiCo 2 O 4 catalyst.
Active sites for dissociation of the first C-H of CH 4 . To assign the sites for dissociation of CH 4 to CH 3 , the binding strengths of CH 3 on cobalt and nickel sites of NiCo 2 O 4 (110)-B surfaces of both type I and type II crystals were calculated (Supplementary Table 3). The adsorption energies of CH 3 on nickel sites of (110)-B are stronger than those on cobalt sites for both types of crystals. In addition, as shown in Supplementary  Figure 16), respectively. The correlation between atomic fractions of Ni cations in Supplementary Fig. 15 and the catalytic conversion presented in Supplementary Fig. 16 shows that a higher atomic fraction of Ni cations on surface of a Ni x Co 3 À x O 4 catalyst offers higher conversion of CH 4 . It further suggests that Ni cations instead of Co cations are active sites for activating C-H of CH 4 .
Mechanism of complete oxidation of CH 4 on NiCo 2 O 4 . To fully understand the mechanism of CH 4 complete oxidation on NiCo 2 O 4 , a systematic investigation of potential reaction pathways was carried out on the most active surface, (110-B) of type I crystal of NiCo 2 O 4 . After the dissociation of CH 4 to CH 3 , there are two different possibilities for the next elementary steps: a further dissociation of CH 3 into CH 2 (called dehydrogenation in Supplementary Fig. 17) or a coupling of carbon atom of CH 3 with oxygen atom of surface lattice (called oxidation in Supplementary  Fig. 17). In order to determine which step is more favourable, we calculated the activation energies and enthalpy change for the two potential steps. The calculated energy profiles and the corresponding geometries were shown in Supplementary Fig. 17. Upon the first C-H dissociation, CH 3 species is oxidized into CH 3 O binding to a Ni cation. The thermodynamically and kinetically favourable oxidation step forms an intermediate, CH 3 O. This is different from the continuous dehydrogenation of CH 3 on Pd surface 14 . This preference of oxidation to CH 3 O instead of further dehydrogenation to CH 2 or CH species could be understood as follows: the low coordinated species, such as CH 2 or CH, need more than one binding sites to stabilize them; however the singly dispersed nickel cations on surface lattice of Co 3 O 4 are not an adsorption site consisting of continuously packed Ni atoms for CH 2 or CH. Therefore, a further dehydrogenation of CH 3 to CH 2 to CH is a high endothermic step on NiCo 2 O 4 compared with the pathway of oxidation.
Supplementary Figure 18 presents the two potential reaction pathways A and B after the formation of CH 3 O species. DFT calculations show that pathway B is the preferred one. The following section will describe this preferred pathway. On the basis of the experimental results and previous calculations 29,32 , CHO species plays very important roles in the oxidation of methane on many different catalyst surfaces; thus CH 3 O could dehydrogenate to CH 2 O and then CHO with a following oxidation to form product molecules.
For dehydrogenation of CH 3 O to CHO, there are two different dehydrogenation pathways schematically shown in Supplementary Fig. 18: (i) the dehydrogenation of CH 3 O by the OH species of nickel site (pathway A shown in red in Supplementary Fig. 18); (ii) the dehydrogenation by the O species of nearby cobalt site (pathway B shown in black in Supplementary Fig. 18). On the basis of the energy profiles of the two pathways in Supplementary Fig. 19a, pathway B is more favourable than pathway A, which suggests that the dehydrogenation of CH 3 O by oxygen atoms (not shown) bonded to nearby cobalt site (pathway B) is preferred over a dehydrogenation of CH 3 O by OH species bonded to nickel site ( Supplementary Fig. 18) in the overall reaction (pathway A).
Transformation of intermediate CHO to product molecules. As mentioned above, the transformation of intermediate CHO to product molecules could go through a sub-pathway of OCHO dehydrogenation (marked with purple in the right panel of Supplementary Fig. 18) other than the sub-pathway of CO oxidation (marked with black in the right panel of Supplementary  Fig. 18). Supplementary Figure 21a shows the two sub-pathways considered in our DFT calculations. Our DFT calculations revealed that each sub-pathway includes two steps with two transition states and one intermediate. OCHO is the intermediate of the sub-pathway of OCHO dehydrogenation ( Supplementary  Fig. 21).
In the sub-pathway of OCHO dehydrogenation (red in Supplementary Fig. 21a,b), the CHO couples with oxygen atom and thus forms an intermediate OCHO (the first step in transforming CHO to product molecules); then the intermediate OCHO is dehydrogenated to CO 2 (the second step); it is quite different from the sub-pathway of CO oxidation (black line in Supplementary Fig. 21a,b). We calculated the sub-pathway of OCHO dehydrogenation for transforming CHO to product molecules. The calculated energy profiles of the two sub-pathways were presented in Supplementary Fig. 21b. The geometries of transition states and intermediates of the sub-pathways of OCHO dehydrogenation and CO oxidation were presented in Supplementary Fig. 21c and Supplementary Fig. 21d, respectively.
From the calculated energy profiles presented in Supplementary Fig. 21b, several interesting features were found. The barrier of the first transition state in the sub-pathway of OHCO dehydrogenation (red line in Supplementary Fig. 21b) is lower than that of the first barrier in the sub-pathway of CO oxidation (black line in Supplementary Fig. 21b); in addition, the OCHO intermediate in the sub-pathway of OCHO dehydrogenation is more thermodynamically stable than the intermediate of the subpathway of CO oxidation by 0.63 eV. Thus, the formation of OCHO intermediate is thermodynamically and kinetically favourable. However, the barrier for transforming OCHO intermediate to CO 2 is higher than the second barrier of the sub-pathway of CO oxidation (Supplementary Fig. 21b). Therefore, the formed OCHO intermediate could be experimentally observed at a relatively low temperature. In fact, our in situ studies of the catalyst surface in the mixture of CH 4 and O 2 with AP-XPS did identify the intermediate OCHO at a temperature lower than 200°C; the C 1s photoemission feature of this intermediate is at 288.3 eV which is consistent with the OCHO species reported in literature 25 . In addition, the observed vibration signatures of this intermediate at 1,280, 1,430, and 1,561 cm À 1 in situ infrared studies (Fig. 6) at 160°C are consistent with the OCHO species formed in water-gas shift. 25 In the sub-pathway of OCHO dehydrogenation, the second step (from OCHO intermediate to product molecules shown in red line in Supplementary Fig. 21b) is highly exothermic with a free energy change of À 3.50 eV. Therefore, at high temperature the OCHO intermediate can readily overcome the barrier and transform to product molecule CO 2 , driven by the thermodynamics factor although the barrier of the second transition state is relatively high compared with the second one in the sub-pathway of CO oxidation. In fact, the evolution of C 1s photoemission feature at 288.3 eV during catalysis from low temperature to high temperature shows that this intermediate OCHO disappears at about 200°C along the increase of catalysis temperature (Fig. 4d).

Discussion
In situ studies using ambient pressure photoelectron spectroscopy, vibrational spectroscopy, and isotope-labelled experiments show that (1)  The mechanisms of methane oxidation on NiCo 2 O 4 (100)-B are different from those on metal surfaces 14,30,32 . For instance, CH 4 on metal surfaces sequentially and continuously dehydrogenates to CH species and then is oxidized to an intermediate CHO, while our studies suggest that after dissociation of first C-H in methane, the CH 3 species will couple with oxygen atom of the lattice oxygen, forming CH 3 O species. The lack of direct dehydrogenation of CH 3 to CH 2 and then to CH on NiCo 2 O 4 likely results from the low binding energy of CH 2 or CH species on isolated Ni cations anchored on surface lattice of Co 3 O 4 . In other words, the separated Ni cation sites could not stabilize the low coordinate species such as CH 2 and CH. In these calculated pathways on the NiCo 2 O 4 (100)-B, the surface lattice oxygen atom in the nearby cobalt sites was found to be very important in the dehydrogenation of CH 3 O.
Computational studies suggest that the pathways from transferring CHO to CO 2 are different from that on metal surface 14,33,34 . Sub-pathway of CO oxidation and sub-pathway of OCHO dehydrogenation were both proposed for the transformation of CHO intermediate to CO 2 . CO oxidation sub-pathway was found more kinetically favourable in the second elemental step (Supplementary Fig. 21b). Furthermore, the OCHO intermediate (red in Supplementary Fig. 21 4 nanocatalysts during catalysis were performed on the in-house AP-XPS system. Monochromated Al Ka line was used as X-ray source. Unlike traditional UHV studies, in AP-XPS studies reactant gases are introduced to continuously flow through the catalyst at a desired temperature in the reaction cell while data acquisition is going on. Thus, the in situ study is defined to be a study in which the measurement of catalyst surface or adsorbate is being performed, while a catalyst is buried in gas reactant (s). The in situ characterization of AP-XPS 22,35,36 could be different from the in situ studies defined in other places since sometimes in situ study is defined as a characterization of a sample without changing location/position after a sample preparation or vapour deposition.
The reaction cell was incorporated into the UHV chamber of AP-XPS system. Gas flows through the cell and exits through the exit port and an aperture that interfaces the gaseous environment in the reaction cell and vacuum environment of the prelens. Flow rate in the reaction cell was measured through a mass flow metre installed before the entrance of the reaction cell. In this study, the typical flow rate of pure gas is in the range of 3-5 ml min À 1 . As the gas source used in AP-XPS studies is pure gas, 3-5 ml min À 1 equals to 60-100 ml min À 1 of 5% reactant gas used in catalytic measurements in a fixed-bed flow reactor. The total pressure of the mixture gas at the entrance is measured with a capacitance gauge, while the pressure at the exit is measured with another capacitance gauge. An average of the pressures at entrance and exit is defined to be the pressure above the catalyst in the reaction cell. An external heating source in the UHV environment heats the catalyst in the environment of reactant gases in the reaction cell. More detailed description of the in-house AP-XPS system can be found in our previous publications 20 .
To avoid surface charging, we use Au, Ag, or HOPG to load catalysts for AP-XPS studies and control the thickness of catalyst layer on a surface 19,37,38 .
In this work, we use Au foil as the substrate to physically load a thin layer of nanocatalysts. The Au foil is deliberately roughened using a SiC knife to increase the adhesion for loading. The apparent binding energies of Au 4f, Ni 2p, Co 2p, and O 1s of sample prepared on the Au substrate are very close to their real values, suggesting there is no identifiable surface charging. All XPS spectra are calibrated to their corresponding Au 4f 7/2 binding energy which is at 84.0 eV. Analyses of XPS peaks are performed with Casa XPS program.
In situ infrared studies are performed in a reaction cell filled with 1 bar reactant gas at Oak Ridge National Lab. This reaction cell can be heated up to high temperatures. The same catalytic condition as measurements of catalytic performances was used in the infrared studies.
On-line observation of products at different temperatures. Our in-house AP-XPS system also allows us to monitor the change of the gas composition during catalysis with a quadrupole mass spectrometer. Three differential pumping stages were installed between the reaction cell and the energy analyser. The aperture mounted on the reaction cell separates the high-pressure environment of the reaction cell from the vacuum of the first differential pumping stage (prelens), and also allows for leaking gases on the reaction cell into the differential pumping stages. The analyses of gas composition at different temperatures during catalysis were done by a quadrupole mass spectrometer, which is mounted at the second differential pumping stage of the AP-XPS system. To exclude the possibility that gas expansion at a high temperature or fast diffusion of molecules from reaction cell of AP-XPS to UHV chamber of mass spectrometer results in an increase of partial pressure of product molecules, Ar gas was introduced as a reference gas (background) to the reaction cell. No change in the partial pressure of Ar along with the increase of temperature of catalysts shows that the increase of partial pressure of CO 2 and H 2 O is not due to thermal expansion of gas at a relatively high temperature.
Such an on-line analysis of gas composition (reactants and products) in the reaction cell of AP-XPS allows for building a simultaneous correlation between the products and real-time surface chemistry of catalysts at different catalysis temperatures. The on-line mass spectroscopy experiments allow for tracking evolution of different products at different temperatures formed through isotopelabeled catalysts NiCo 2 O 16 4 À x O 18 x or through using isotope-labelled reactant O 18 2 during data acquisition of in-situ AP-XPS studies.
Measurements of catalytic performance. The catalytic performance was measured in a fixed-bed tubular quartz micro flow reactor (inner diameter ¼ 6 mm, length ¼ 300 mm), at atmospheric pressure. The catalyst (25-500 mg) was sieved (40-60 mesh) and loaded into the reactor between two layers of quartz sands, packed in between two quartz wool plugs. The catalyst was heated in a furnace equipped with a proportional-integral-derivative temperature controller. The temperature of the catalyst was measured with a K-type thermocouple inserted into the reactor and touching the catalyst bed. The reactant gas composition was controlled by varying the flow rates of CH 4 mixture (10% CH 4 balanced in Ar) and O 2 (pure O 2 ). The flow rate of CH 4 mixture (10% balanced in Ar) and O 2 (pure) was precisely controlled through their own mass flow metres (Dakota Instruments, Inc.). When vary the flow rate of CH 4 mixture and O 2 , the ratio of pure CH 4 to pure O 2 was always keeping at 1:5. The typical flow rate of CH 4 mixture (10% balanced in Ar) is 100 ml min À 1 , while the flow rate of O 2 (pure) is 50 ml min À 1 . The reliability of the fix-bed reactor was checked through a blank experiment with the same experimental parameters including reactant gases, flow rate and reaction temperature. In the blank experiment, only silica supporting material was loaded in the reactor. No appreciable conversion of CH 4 was found in the blank experiment in the temperature range of 25-800°C.
The effluent gas was connected to a gas chromatograph equipped with a HayeSep D (6 0 Â 1/8 00 ) packed column, a molecular sieve 13 Â (6 0 Â 1/8 00 ) packed column, and a thermal conductivity detector (TCD) for analysis of both reactants and products. The CH 4 conversion X% was defined as: X % ¼ mol CH4;inlet À mol CH4 ;outlet À Á =mol CH4 ;inlet Â Ã Â100. Herein, mol CH4;inlet and mol CH4 ;outlet are the amounts of pure CH 4 at the inlet and outlet of the reactor, respectively. In the CH 4 combustion reaction, mol CH4;inlet and mol CH4 ;outlet can be directly represented by the peak area of CH 4 in gas chromatograph before reaction and after reaction, respectively.
Computational studies. In this work all the DFT calculations were carried out with a periodic slab model using the Vienna ab initio simulation program 33,[39][40][41] . The generalized gradient approximation was used with the Perdew-Burke-Ernzerhof 42 exchange-correlation functional. Due to the strong correlation effect among the partially filled Co 3d states, we used the Hubbard parameter, U, for the Co 3d electrons to take the on-site Coulomb interaction into account, which is the well-known DFT þ U method 34 . According to previous work 29,43 , the value of U-J of 2 eV was applied. The projector-augmented wave method 44,45 was utilized to describe the electron-ion interactions, and the cut-off energy for the plane-wave basis set was 400 eV. Brillouin zone integration was accomplished using a 2 Â 2 Â 1 Monkhorst-Pack k-point mesh. All the adsorption geometries were optimized using a force-based conjugate gradient algorithm, while transition states were located with a constrained minimization technique [46][47][48] .