Introduction

Well-defined structure, facile shape tailoring and tuning of its electronic properties by iso- and heterovalent doping, make cobalt spinel oxide a unique model catalytic material, receiving recently a great deal of theoretical and practical attention [1, 2]. Co3O4 exhibits high activity in oxidation of carbon oxide [3], ammonia [4], methane, and other hydrocarbons [5, 6]. It is also well known for its spectacular activity in oxygen reduction reaction (ORR) [7], or low temperature decomposition of N2O [8, 9]. High catalytic performance of cobalt spinel can be attributed to presence of various surface reactive oxygen species (ROS) stabilized on its surface in mild conditions and to easy formation of oxygen vacancies [10, 11]. It has been shown that for catalytic processes such as CO oxidation or N2O decomposition the suprafacial reactive oxygen intermediates play a key role in the catalytic activity [12, 13]. In turn, the oxygen vacancies generated during catalytic reactions [14] or produced spontaneously at elevated temperatures [15] are often involved in the Mars-van Krevelen redox processes over cobalt spinel, exemplified, e.g., by oxidation of methane [16].

To design a catalytic system dedicated for particular application a detailed molecular level understanding of the reaction mechanism is indispensable. One of the crucial functionalities of the catalyst in various oxidation reactions is activation of dioxygen, leading to production of surface reactive oxygen species, and the ability to facile formation of oxygen vacancies, providing a nucleophilic lattice oxygen available for various oxidation processes [17]. Although it is generally accepted that both lattice oxygen and ROS species may be involved in oxidation reactions, there is still a debate about the specific role they play therein. A common method for studying oxygen dynamics in an oxide catalyst is 18O–16O isotopic exchange with gas mass spectroscopic detection. In such experiments it is possible to differentiate between surface exchange, bulk diffusion, and fast diffusion along grain boundaries [18]. From the course of the oxygen exchange reaction, the way in which oxygen takes place in the catalytic processes may also be deduced [19]. Following pioneering work by Klier et al. [20], there are three conceivable parallel pathways for an oxygen molecule to exchange the atoms with the catalyst surface (i) exchange proceeds without direct involvement of lattice oxygen (R 0), (ii) only one of the atoms of a molecule is replaced (R 1), (iii) both atoms of a single dioxygen molecule are exchanged by lattice oxygens (R 2).

According to a generally adopted scheme [17, 21, 22], the interaction of the oxide catalyst with adsorbed oxygen involves electron transfer from the catalyst surface donor sites to the adoxygen, converting suprafacial O2(ads) into lattice O2− anions via a sequence of the consecutive/parallel steps of oxygen reduction, diffusion, and dissociation until final incorporation into the oxide lattice:

$$ {\text{O}}_{{2({\text{g}})}} \leftrightarrow {\text{O}}_{{2({\text{ads}})}} \leftrightarrow {\text{O}}_{{2({\text{ads}})}}^{ - } \leftrightarrow {\text{O}}_{{2({\text{ads}})}}^{2 - } \leftrightarrow 2{\text{O}}^{ - }_{{({\text{ads}})}} \leftrightarrow 2{\text{O}}^{2 - }_{{({\text{lattice}})}} $$

The distribution of isotopic exchange products among various oxygen species depends on the relative rates of the above mentioned events. Hence, a faster incorporation reaction and subsequent diffusion leads to 16O2 as the dominant isotopomer. Faster dissociation of oxygen compared with oxygen incorporation results in cross-labelled 16O18O.

In contrast to CO oxidation or N2O decomposition over cobalt spinel that have recently been investigated quite extensively [13, 2325], much less attention is paid to hydrogen oxidation. Taking into account that oxidative properties of Co3O4 are related to the presence of surface ROS species including surface oxygen vacancies, their elementary electroprotic reactivity can be probed in a neat way by model hydrogen oxidation reaction. In addition to the direct practical relevance of H2 oxidation, it is of great importance to understand fully methane or volatile organic compounds (VOC) oxidation mechanism, where the redox and acid–base functionalities are coupled into electroprotic properties of the catalyst surface. In such reaction hydrogen acts as a source of electrons, accommodated on Co active sites, and protons adsorbed on Lewis base surface O2− oxygen. Another virtue of hydrogen oxidation model studies is that they provide direct insight into surface dehydroxylation processes, which are ubiquitous step for closing catalytic cycles in oxidation of hydrocarbons. Furthermore, such investigations may also help in resolving mechanistic issues that are relevant for preferential oxidation of CO in excess H2 (CO-PROX) [26].

Herein we address the problem of suprafacial/intrafacial dynamics and reactivity of surface oxygen species produced on model cobalt spinel catalysts by using pulse 16O2/18O2 isotopic exchange and hydrogen oxidation as a suitable probes. Experimental studies were supported by spin resolved DFT+U modeling of the key molecular surface events on the most abundant (100) termination of cobalt spinel.

Experimental and computational details

Materials and characterization

Cobalt spinel samples were synthesized via pH-controlled precipitation method by addition of 1 M (NH4)2CO3 to 1 M Co(NO3)2 solution in room temperature (to obtain pH = 9). The catalyst precursor was filtered, dried overnight (T = 100 °C), and calcined in air at 700 °C for 3 h to obtain a good crystalline spinel structure. Following our recent work, such treatment leads to formation of the stoichiometric spinel both in the bulk [27]  and the surface regions [28]. Purity of cobalt spinel phase was confirmed by X-ray diffraction, using CuKα radiation by means of a Rigaku Miniflex X-ray diffractometer equipped with a DeTEX detector. Raman spectra were recorded with an inVia Renishaw spectrometer equipped with a Leica microscope and the 785 nm laser excitation.

Isotopic exchange and catalytic studies

The temperature programmed surface reactions (TPSR) were performed in the range of 25–500 °C, using a quartz flow reactor in a fully automated catalytic setup, operating under LabView environment, equipped with a QMS detector (Hiden Analytical HPR20). The experiments were carried out with 100 mg of the catalyst (sieve fraction of 0.2–0.3 mm) with the feed flow rate of 50 mL min−1 (25 mL min−1 of 5% 18O2 in Ar, and 25 mL min−1 5% H2 in He), and the heating rate of 10 °C min−1. The molar flow rates of reactants and products (μmol min−1) were calculated on the basis of the QMS signals calibrated against the post-reactor argon trace of controlled flow. For isothermal 18O2 pulse experiments an automatic Valco sampling valve with a 2.5 cm3 loop was used for injections.

Molecular modeling

DFT+U calculations were performed using the VASP code with the Hubbard parameter U = 3.5 eV, following the literature [29]. The projector augmented plane wave method (PAW) together with the generalized gradient PW91 exchange-functional were employed. We used standard Monkhorst–Pack grid with 5 × 5 × 5 sampling mesh for bulk calculations and 5 × 5×1 for slab calculations. The cutoff energy was set to 500 eV, and the SCF convergence criterion to 10−5 eV. A systematic validation of the applied calculation scheme against the experimental data can be found in our previous paper [30]. Bulk cobalt spinel unit cell was obtained by optimization of the experimental cubic (1 × 1 × 1) unit cell (a exp. = 8.08 Å) containing 56 ions (Co24O32) with full optimization of all internal degrees of freedom within the error of 10−4 eV. The optimal cell lattice constant was calculated via the Birch-Murnaghan equation of state as a DFT = 8.14 Å. Following our recent work [28] for modeling of spinel activity we used a dominant stoichiometric (100) surface. We applied a (1 × 1) slab of 17 atomic layers (~17 Å of thickness) with a supercell composition Co48O64, and a vacuum separation of 20 Å. The atomic positions in the four top and four bottom layers were relaxed with criterion of 10−3 eV Å−1. The nudged elastic band method (NEB) [31, 32] was used to locate transition states (TS) of the investigated surface molecular events. In all calculations, for each identified transition state five to nine NEB images (including the initial and final ones) were used.

Results

Catalyst characterization

The single phase composition and crystallinity of the cobalt spinel catalyst were confirmed by X-ray diffraction, using CuKα radiation and the results (Fig. 1a) are characteristic for the cobalt spinel structure analyzed within the cubic Fd-3m space group (24210-ICSD), revealing that no major structural differences induced by different temperature of calcination are present. The quality of the sample was additionally verified by the micro-Raman techniques (Fig. 1b). The Raman spectra of cobalt spinel sample is characterized by five resolved Raman bands at about 194 (F2g), 480 (Eg), 520 (F2g), 620 (F2g), and 690 (A1g) cm−1, predicted by the factor analysis [33], which also confirms absence of any minor spurious phases.

Fig. 1
figure 1

XRD diffractogram (a) and Raman spectra (b) of the cobalt spinel sample

Further characterization details have been described in our recent papers [28, 30], where the HAADF-STEM pictures show that the Co3O4 nanocrystals expose predominantly the (100) facets. These results were used to develop a computational model, based on the most abundant (100) termination.

Pulse isotopic 18O2/16O2 exchange and H2 oxidation studies

The isotopic 18O/16O oxygen exchange pulse experiments performed at various temperatures are presented in Fig. 2. Below 300 °C upon 18O2 pulses only small, practically constant, amounts of 18O16O were released. Formation 16O16O species was barely visible, indicating that surface oxygen dynamics is governed mainly by suprafacial processes. In such conditions dioxygen is present in a dinuclear forms stabilized on two vicinal octahedral Co ions (superoxo CoO–(18O18O)–CoO) or on adjacent tetrahedral and octahedral Co (peroxo CoT–(18O18O)2−–CoO) centers, as described in detail in our previous paper [28]. For the sake of clarity in this notation only oxygen charge is given explicitly. The corresponding changes in the cobalt oxidation states are more involved as discussed elsewhere [28]. Whereas the CoO–(18O18O)–CoO) species are stable at ambient temperatures only, the stability of the CoT–(18O18O)2−–CoO species extends over 200 °C. The diatomic species apparently are reluctant to isotopic exchange with the spinel surface, and only small amounts of the cross-labeled 16O18O species were observed in the temperature window of their stability. However, upon increasing of the reaction temperature above 300 °C the contribution of the 18O16O isotopomer gradually increases, reaching 50% of the original pulse intensity (Fig. 2). This effect is also reflected in pronounced enhancement of the 16O16O signal. Such behavior can be interpreted in terms of dissociation of the CoT–(18O18O)2−–CoO species into CoT18O and CoO18O mono-oxo entities:

Fig. 2
figure 2

Isotopic 18O2 pulse experiments over Co3O4 catalyst performed at various temperatures showing formation of different dioxygen isotopomers (16O2, 18O16O and 18O2)

$$ {\text{Co}}^{\text{T}} {-} \left({^{18} {\text{O}}^{18} {\text{O}}} \right)^{2-} {-} {\text{Co}}^{\text{O}} \to {\text{Co}}^{\text{T}} {-}^{18} {\text{O}}^{-} + {\text{Co}}^{\text{O}} {-}^{18} {\text{O}}^{-} $$

In the CoT18O adducts the oxygen adatom is rather strongly covalently bound, and is expected to be reluctant to isotopic exchange. The CoO18O species, however, are labile and the monoatomic oxygen can easily diffuse on the surface via transient peroxy (16O 18lattice O)2− species, as already observed and on alkaline earth oxides [34, 35]:

$$ {\text{Co}}^{\text{O}} {-}^{18} {\text{O}}^{-}\,+\,^{16} {\text{O}}_{\text{lattice}}^{{2{-}}} \to {\text{Co}}^{\text{O}} + \left({^{16} {\text{O}}_{\text{lattice}} {-}^{18} {\text{O}}} \right)^{{2{-}}} $$

Flipping of the both oxygen moieties in such species results in facile isotopic exchange:

$$ \left( {^{16} {\text{O}}_{\text{lattice}} {-}^{18} {\text{O}}} \right)^{{2{-}}} \to \left( {^{18} {\text{O}}_{\text{lattice}} {-}^{16} {\text{O}}} \right)^{{2{-}}} $$

A more detailed molecular account of this process is discussed below. The surface peroxy can undergo recombination with adjacent oxygen adatoms (both peroxy and metaloxo species), leading in both cases to formation of gas phase 16O18O.

$$ \begin{array}{*{20}l} {\left( {^{16} {\text{O}}_{\text{lattice}} {-}^{18} {\text{O}}} \right)^{{2{-}}} + \left( {^{18} {\text{O}}_{\text{lattice}} {-}^{16}{\text{O}}} \right)^{{2{-}}} \to ^{16} {\text{O}}_{\text{lattice}}^{{2{-}}} +^{18} {\text{O}}_{\text{lattice}}^{{2{-}}} +^{18} {\text{O}}^{16} {\text{O}}} \hfill \\ {\left( {^{18} {\text{O}}_{\text{lattice}} {-}^{16} {\text{O}}} \right)^{{2{-}}} + {\text{ Co}}^{\text{O}} -^{18} {\text{O}}^{-} \to^{18} {\text{O}}_{\text{lattice}}^{{2{-}}} + {\text{ Co}}^{O} +^{18} {\text{O}}^{16} {\text{O}}} \hfill \\ \end{array} $$

Since the 16O18O response is well synchronized with the 18O2 pulses, surface diffusion and the oxygen flipping within the peroxy transient intermediate must be fast within the experiment time scale. Above 700 °C the intensity of singly exchanged 16O18O (R 1) and doubly exchanged 16O16O species (R 2) dominate the intensity of the parent 18O18O pulse, which shows that at these conditions extensive interfacial isotopic R 1 and R 2 exchange between ROS and the surface O 2−surf species takes place. This process should be, however, distinguished form gradual 16O2 release due to oxygen vacancy formation (gray shaded area above 750 °C), which obviously is not paced by the 18O2 pulses (O ×O  → VO + ½ O2), in contrast to distinct response peaks of the 16O16O clearly induced by the 18O18O pulses only. Such behavior indicates that the 16O18O and the 16O16O response peaks are associated with rapid interfacial oxygen exchange involving peroxy intermediates, but not necessarily with explicit vacancy formation, usually implied in such conditions [36, 37]. Fast flipping of both the 18O and 16O moieties in the surface O2 2− leads to efficient isotopic scrambling. Recombination of the atomic surface oxygen into dioxygen may then entail both 18O and 16O moieties leading to 16O18O formation.

The region below 350 °C, where the isotopic exchange of dioxygen in pulse experiment is small, was additionally probed by hydrogen oxidation reaction. The TPSR profiles of H2 oxidation by 18O2 (expressed as F/μmol min−1) are presented in Fig. 3a, where three different stages of the reaction progress may by distinguished. At temperatures below 160 °C, the hydrogen conversion was low (<10%), and the reaction products were dominated by the H 182 O isotopomer. At ca. 160 °C the hydrogen decay profile falls down noticeably, which was accompanied by the concomitant enhanced production of both H 182 O and H 162 O. Above 300 °C as the reaction approaches nearly 100% conversion the reaction progress due to lean conditions is determined by the reactants transport processes. These three stages of the hydrogen oxidation reaction were well reflected in the apparent activation energies, calculated from the associated Arrhenius plot (Fig. 3b). In the first region, the activation energy of hydrogen oxidation was equal to E a = 16 kcal/mol, and it drops down to 9.2 kcal/mol in the second region, whereas above 300 °C the small value of E a = 2 kcal/mol is characteristic of transport controlled process.

Fig. 3
figure 3

The TPRS profile of hydrogen oxidation by isotopically labeled oxygen (18O2) showing the evolution of H 162 O, H 182 O, 18/18O2 , and H2 reagents with temperature (a). Arrhenius-type plot of H2 oxidation (b), and parity plot of water-16 versus water-18 partial pressures (c)

The observed regions can be accounted for by dynamic change in the nature of the reactive oxygen species on passing to higher temperatures, and by various extent of the water retention on the spinel surface as well. As discussed above, in the low temperature region, oxygen is adsorbed in the diatomic form that is resistant to isotopic exchange. Such species exhibit also rather low reactivity toward hydrogen, as implied by the relatively high activation energy of 16 kcal/mol. Upon moving to higher temperatures the dominant surface ROS are monoatomic oxygen entities. They are responsible for enhanced water production activating hydrogen more easily with E a of 9 kcal/mol only. Since the oxygen monoatomic species are stable until 400 °C28 they are responsible for hydrogen oxidation until full conversion is achieved (at 350 °C). In the whole investigated temperature range of hydrogen oxidation, formation of the 16O18O species was one order of magnitude lower than that of H 162 O. Moreover, whereas the 16O18O profile decreased the H 162 O one increased with the temperature. This indicates that the monoatomic ROS species produced on the spinel surface upon oxygen activation are overwhelmingly scavenged by the hydrogen.

The reaction progress was also influenced by retention of water molecules produced upon the hydrogen oxidation, as already mentioned. Following our previous investigations [38] adsorbed water is thermodynamically stable on the cobalt spinel surface at p H2O ~ 0.005 atm. until T ~ 150 °C, thus part of the produced H2O molecules may be retained in dissociative and associative forms. Dissociative forms of water exhibit ΔE ads ranging from 1.18 eV to 0.89 eV, whereas adsorption energies of associatively attached H2O are equal to 0.7 eV. From the thermodynamic diagram published previously [38] it may be thus expected that below 150 °C the most stable water forms are retained on the surface during catalytic reaction, leading to partial self-poisoning of the spinel surface by blocking the octahedral cobalt active sites. This gives rise to temporal stagnation of the catalytic activity around 180–200 °C, clearly seen in Fig. 3, as a deflection in the H2 conversion profile. Above this temperature threshold, water leaves the surface, as manifested by a broad “desorption” peak on the both H2O profiles, and the reaction rate is accelerated by the increasing temperature until 350 °C, where full hydrogen conversion was achieved.

The more detailed insight into water formation mechanism was provided by the pH2(16)O versus pH2(18)O parity plot presented in Fig. 3c. The detailed analysis shows that in the initial stage of the TPSR experiment (T < 140 °C) the H 182 O/H 162 O ratio was equal to f H2O 18/16  = 40 and this value may be explained by suprafacial oxidation of hydrogen by diatomic oxygen species that are immobile; therefore, their enrichment in 16O via oxygen flip within the peroxy intermediate is virtually impossible. Above 140 °C, once monoatomic oxygen species become dominant, the oxygen isotopic exchange through peroxy flipping mechanism becomes feasible, which is reflected in dramatic decrease in the H 182 O/H 162 O ratio to f H2O 18/16  = 4.

From the catalyst mass and its BET surface we calculated the number of oxygen ions in first layer of Co3O4 (100) termination as 8.3 × 1018. From the flow rates of 16O containing products and the time of experiment we estimated the number of exchanged 16O as 3.2 × 1020 in single TPSR run (50–350 °C). This results show that in given conditions, as the reaction proceeds the spinel surface may be gradually enriched in 18O isotope to full coverage, making the interfacial oxygen-18 exchange into oxygen-16 less effective. This is reflected in the gradual increase of f H2O 18/16 value with the progress of the reaction (increasing temperature).

Molecular modeling

The isotopic results of facile 18O2/16O2 exchange and H2 oxidation were accounted for by corroborative DFT modeling. Following our recent paper [30], we selected the most abundant stoichiometric (100) termination as a playground for the investigated molecular events.

Thermodynamic considerations reveals that in oxygen rich conditions, this surface may be covered by reactive oxygen species of different nature [28], and among monoatomic species the metal-oxo form stabilized on the tetrahedral cobalt ion are the most stable (adsorption energy calculated versus O2 molecule equals, E ads = −0.72 eV). In the case of diatomic forms the peroxy nature CoT_OO–CoO configuration is the most stable (E ads = −1.22 eV). In typical conditions of catalytic experiments (p O2 = 0.01÷1 atm) the former species is thermodynamically stable below 200 °C, whereas the latter from 200 °C to 400 °C. Above 400 °C the bare (100) surface of cobalt spinel is thermodynamically preferred. Since the monoatomic reactive oxygen species are essential in the investigated reactions (see below), we modeled the processes of dioxygen adsorption and subsequent activation leading to such adoxygen forms. The energetic profile and detailed molecular level description of dioxygen activation mechanism is presented in Fig. 4, where the initial configuration of the topmost layer (1 × 1) surface element of Co3O4 and gas phase O2 is shown panel a. The surface cationic composition includes the octahedral pentacoordinated Co O5c ions (coded blue), topmost dicoordinated tetrahedral Co T2c (violet) and two concealed tetrahedral cobalt ions of full coordination (Co T4c , green), which are located in the first sublayer. The surface anionic composition (oxygen ions are marked in red) contains four truncated 3-fold coordinated oxygen ions (denoted as O2O,1T) that are linked to two CoO and to one CoT cations, two 3-fold coordinated anions (O3O) that are linked to the CoO ions exclusively and two fully coordinated (fourfold) O3O,1T species. In transition state geometry (Fig. 4a#) the CoT–O bond is nearly formed (d Co–O = 1.83 Å), whereas the second bond (CoO–O) is far from its optimal length (d Co–O = 2.51 Å) and oxygen molecule is stretched from optimal 1.25–1.29 Å. Such geometrical strain is responsible for energy barrier for this step calculated as 0.47 eV. The resulting bidentate CoT–OO–CoO conformation (Fig. 4c), with the Co–O bonds of 1.77 Å and and O–O bond of 1.49 Å, is more stable than initial gas phase molecule by 1.22 eV. From this stable configuration, a step of the O–O bond breaking is required to produce monatomic oxygen species, and the transition state for this mechanism is presented in Fig. 4b#, where significant elongation of O–O bond (from to 1.49 to 1.75 Å) is visible. After the complete O–O bond breaking, the system optimize the geometry (Fig. 4e) and the resulting Co–O bond lengths of 1.65 Å (CoT–O) and 1.67 Å (CoO–O), indicates metal-oxo character of both species. The resulting monoatomic species are less stable than CoO–OO–CoO by 0.71 eV, but in the temperatures of the oxygen exchange and H2 oxidation reaction they can easily diffuse on the surface, with formation of transient peroxy gropus. Such peroxy species constituting one reactive (top) and one lattice (bottom) oxygen are the most probable forms for straightforward isotopic exchange. Indeed, the 18O–16O2O,1T peroxy groups, which appear as transient intermediates during oxygen adatom diffusion along the CoO and O 2−2O,1T sites:

$$ ^{18} {\text{O}}{-}{\text{Co}}^{\text{O}} {-}^{16} {\text{O}}_{{ 2 {\text{O,1T}}}}^{{2{-}}} {-}{\text{Co}}^{O} \to {\text{ Co}}^{O} {-}\left( {^{18} {\text{O}}^{16} {\text{O}}_{{ 2 {\text{O,1T}}}} } \right)^{{2{-}}} {-}{\text{Co}}^{\text{O}} \to {\text{ Co}}^{\text{O}} {-}{\text{O}}_{{ 2 {\text{O,1T}}}}^{{2{-}}} {-}{\text{Co}}^{O} {-}^{18} {\text{O}} $$

may constitute suitable platform for rapid oxygen isotopic exchange.

Fig. 4
figure 4

The energetic profile of the adsorption and activation of dioxygen on the (100) plane. Each individual step is characterized by energy barrier (marked in red) and reaction energy (green), stable structures are labeled with letters, and the corresponding TS with additional # sign. The color coding is as follows: spinel oxygen: red, oxygen adatom: orange; exposed tetrahedral cobalt: violet, tetrahedral cobalt in first subsurface: green, octahedral cobalt: blue. (Color figure online)

Indeed, once the (18O–16O2O,1T)2− peroxy entities are formed during the surface diffusion process, the top and down oxygen moieties can undergo easy flipping, exchanging the mutual positions of the suprafacial (top) and intrafacial (bottom) oxygen atoms (18O–16O 2−2O,1T  → 16O–18O 2−2O,1T ). The energetic profile of this process on the (100) surface is presented in Fig. 5. For each individual step taken into account the energy barrier (marked in red) and the reaction energy (green) were calculated. The consecutive stable structures appearing along the diffusion pathway are labeled with a, b, c, and d letters, and the corresponding transition states (TS) with an additional # sign. The starting structure (a) shows oxygen adatom (coded yellow) stabilized on the octahedral cobalt sites as metaloxo CoO18O entity with the bond length of 1.66 Å, which has been described previously [28, 39].

Fig. 5
figure 5

The energetic profile of the diffusion mechanism of reactive oxygen species on the (100) plane between two adjacent octahedral cobalt sites. Each individual step is characterized by energy barrier (marked in red) and reaction energy (green), stable structures are labeled with letters, and the corresponding TS with additional # sign. The color coding is as follows: spinel oxygen: red, oxygen adatom: yellow; exposed tetrahedral cobalt: violet, tetrahedral cobalt in first subsurface: green, octahedral cobalt: blue. (Color figure online)

The oxygen adatom can be relocated to nearest O2O,1T anionic site through transition state (a # ), (energy barrier of 0.72 eV), which can be described by slight elongation of the Co O5c –O bond (Δd Co-O = 0.1 Å) and shortening of the O–O2O,1T distance (from d O-O = 2.71 Å to 1.52 Å). Further evolution of the adoxygen leads to the formation of stable peroxy (O–O2O,1T)2− anion (b), characterized by the O–O bond length of 1.41 Å. In this species one of the oxygen atoms is exposed above the surface plane (suprafacial oxygen), whereas the second O atom is located at the surface (intrafacial oxygen). The formed peroxy intermediate being less stable than the parent metal-oxo adduct by 0.41 eV may return to Co O-5c (with E a of 0.31 eV), restoring the starting configuration (a). It can also undergo a flip shown in the b → b # → c sequence with quite small energy barrier of 0.39 eV. The latter scenario, resulting mainly from rearrangement of the cobalt–oxygen and the peroxy O–O2O,1T bonds by 0.2 and 0.1 Å, respectively, exchange the positions of the top and the surface oxygen atoms, and is responsible for the facile suprafacial(18O)/intrafacial(16O) isotopic oxygen exchange. The isotopically scrambled peroxy anion can next be transformed into CoO18O metal-oxo adduct along the c → c # → d sequence with activation energy of 0.31 eV, which closes the elementary path of the oxygen surface diffusion. As a result the isotopic exchange preferentially occurs via easy flipping of the oxygen atoms within the surface peroxy intermediate without explicit formation of energetically more costly oxygen vacancies.

The quantum chemical modeling of H2 molecule activation and subsequent surface dehydroxylation was performed for the bare and ROS covered (100) termination. Following the available data on hydrogen interaction with oxide surfaces [40, 41], a H2 dissociative adsorption on the lattice oxygen anions or the suprafacial ROS species was considered. For H2 activation on bare surface we selected two vicinal anionic O2O,1T sites (Fig. 6a) with the interionic distance of 2.95 Å, suitable for sizeable overlap between the corresponding H2 and O2− orbitals. In the proposed scenario the hydrogen molecule with the initial equilibrium H–H bond length of d H–H = 0.75 Å, while approaching the O2O,1T surface sites is stretched to 1.23 Å upon reaching the transition state geometry (Fig. 6b). In a slightly asymmetric TS significant interaction between the surface oxygens and the both hydrogen atoms takes place, with the incipient bond lengths of 1.14 and 1.17 Å. The energetic cost of this step was equal to 1.11 eV, and the final product with two surface hydroxyls pointing toward each other (Fig. 6c) was significantly more stable than gas phase hydrogen (ΔE reac. = –1.14 eV). Analysis of the charges and the magnetic moments revealed that both hydrogen adatoms moieties exhibit total Bader charge of about 0.5 |e| and null magnetic moment, which is characteristic of H+ cation. The energetics of the H2 activation step was summarized in the corresponding energy diagram (panel 6d).

Fig. 6
figure 6

The topmost layer atomic composition of the (1 × 1) surface elements of the consecutive steps of hydrogen activation on bare (ac) and ROS covered (a 1 c 1 ) cobalt spinel (100) surface. Left panels shows initial structure, middle panels transition state geometries, whereas right panels final structures. The reactions energetics is shown in panel d, d 1 . Color coding as follows: hydrogen: white, cobalt oxide oxygen: red, oxygen adatom: yellow; exposed tetrahedral cobalt: violet, tetrahedral cobalt in first subsurface: green, octahedral cobalt: blue. (Color figure online)

Analogous hydrogen activation for oxygen covered surface was modeled with previously described CoO18O metal-oxo adducts (Fig. 6a1) with the shortest distance to the adjacent lattice oxygen (d O–O = 2.71 Å), which is beneficial for H2 molecule activation. The transition state for hydrogen attachment (Fig. 6b1) is not symmetrical. As a result, the O–H bond formed between the hydrogen and the oxygen adatom is essentially formed (d O–H equals 0.99 Å), whereas the distance between the second H atom and the lattice oxygen is quite large (d O–H = 1.68 Å), showing rather small interaction. Both H–H and Co–O bond lengths are stretched comparing to their initial state (by 0.36 and 0.30 Å, respectively), but the energetic effort for this deformation is largely compensated by the energetic gain due to O–H bond formation. This gives rise to quite low activation energy of 0.69 eV (nearly twice lower than the corresponding value for the surface O2− sites). The transition state can be described as H2 activated on the O adatom, and its slight bending toward the lattice oxygen is responsible for appearance of final double hydroxyl structure shown in Fig. 6c1, which is 1.75 eV more stable than the gas phase H2. Both suprafacial and intrafacial stabilized hydrogens exhibit the charge of about +0.5 \( \left| e \right| \), consistent with their protonic nature. The energetics of this reaction activation step was briefly summarized in the diagram (Fig. 6d1).

The hydroxylated surface states produced according to both activation mechanism (Fig. 6c, c1) were chosen as starting points for molecular modeling of surface dehydroxylation. In the interfacial mechanism, formation of H2O involves extraction of the lattice oxygen, thus the resultant dehydroxylated surface contains an oxygen vacancy. Mechanistic details of this process are shown in Figs. 7a–c, with the transition state presented in the middle panel. During the reaction one of the OH groups (right in Fig. 7a) is shifted towards the second hydrogen maintaining the initial O–H distance of 0.98 Å. The O–Co bonds are strongly elongated from 1.98 to 2.21 Å (CoT–O) and from 2.01 to 2.42 Å (CoO–O). The second hydrogen is only slightly moved from its initial Osurf site (d O–H = 1.1 Å), and the formation of H–O bond is not launched at this point (d O–H = 1.69 Å). Such reorganization of initial bonds is energetically costly and requires 1.71 eV. The resultant atomic configuration of the relaxed surface after final formation of H2O is shown in Fig. 7c, where the vacancy is indicated as a yellow cube. In the vicinity of the created defect, shortening of the Co–O bonds from 1.98 to 1.89 Å took place, and the energy is enhanced to 1.35 eV with respect to initial hydroxylated state of the surface. A dominant part of this value is contributed by the energy required for the oxygen vacancy formation (\( E_{vac}^{{O_{2} }} \) = 0.91 eV) [28]. The energy diagram of this step is presented in Fig. 7d.

Fig. 7
figure 7

The topmost layer atomic composition of the (1 × 1) surface elements showing subsequent frames of surface dehydroxylation from initially bare (ac) and initially ROS covered (a 1 c 1 ) cobalt spinel (100) surface. Left panels shows initial structure, middle panels transition state geometries, whereas right ones final structures. The reactions energetics is shown in panel d, d 1 . Color coding as follows: hydrogen: white, cobalt oxide oxygen: red, oxygen adatom: yellow; exposed tetrahedral Co: violet, tetrahedral cobalt in first subsurface: green, octahedral cobalt: blue. (Color figure online)

For the (100) surface of cobalt spinel covered by the suprafacial oxygen atoms, the dehydroxylation process does not involve lattice oxygen atoms since water can be produced by association of surface hydroxyl and surface stabilized H+ species (Fig. 7a1). Transition state for this mechanism (Fig. 7b1) shows that only slight shift of the hydrogen stabilized on the O3O centers together with small elongation of the H–O3O and Co–O bonds (from 0.99 to 1.19 Å and from 1.90 to 2.14 Å, respectively) are required to pass the energy barrier (ΔE act = 1.15 eV). The final state of this route is shown in Fig. 7c1, and the reaction energy equals 1.01 eV (Fig. 7d1), being substantially smaller than in the previous case. As a result the suprafacial dehydroxylation being kinetically and thermodynamically much more favorable than the intrafacial one may operate even at relatively low temperatures (below 300–350 °C). The appearance of significant amount of H 162 O in the reaction products is thus associated with the fast isotopic 18O/16O scrambling in the transient peroxo intermediates, and is not diagnostic of involvement of the Mars–van Krevelen mechanism. The intrafacial dehydroxylation, in turn, is energetically costly due to the oxygen vacancy formation. When such mechanism dominates production of H 162 O and H 182 O is not expected to be commensurate, but the presence of H 162 O is now indicative for the Mars-van Krevelen mechanism.

Conclusions

Pulse 16O2/18O2 isotopic exchange and TPSR of hydrogen oxidation on the cobalt spinel nanocatalyst were investigated in detail and corroborated by periodic spin unrestricted DFT-PW91+U molecular modeling. Detailed analysis of the experimental and theoretical results revealed that catalytic activity of the cobalt spinel surface is strongly associated with the presence of various oxygen species such as superoxo (CoO–(O2)–CoO, peroxo (CoT–(O2)2−–CoO), monoatomic metaloxo (CoT–O, CoO–O), and oxygen vacancies (VO), which appear in different temperature windows. The diatomic species stable below 160 °C are reluctant to isotopic exchange, and exhibit lower reactivity toward hydrogen (E act = 16 kcal/mol) in comparison to monoatomic species, highly reactive in isotopic exchange and H2 oxidation (E act = 9 kcal/mol). Surface diffusion of monoatomic oxygen along the CoO and O 2−2O,1T sites is responsible for formation of transient peroxy (18O–16O2O,1T)2−, and easy flipping of the suprafacial (top) and intrafacial (bottom) oxygen atoms. Small activation energy of 0.39 eV leads to facile isotopic exchange, without explicit formation of energetically more costly oxygen vacancies. It was also shown that monoatomic species are not only more active in H–H bond activation (E act of 1.15 eV) than lattice oxygens (E act = 1.71 eV), but also give rise to more facile dehydroxylation, as in the latter case dehydroxylation is not accompanied by oxygen vacancy formation. The appearance of significant amount of H 162 O in the reaction products is associated with the facile isotopic 18O/16O scrambling in the transient peroxo intermediates, and is not diagnostic of involvement of the Mars-van Krevelen mechanism.