The selective activation of C-H bonds for the functionalization of hydrocarbons by oxidation reactions is a major challenge in catalysis. Although a large number of materials show high activity for these reactions, most of them also catalyze undesired side reactions that finally produce deep oxidation products, namely CO₂ and water. A variety of metal oxides, either single or mixed oxides, represent the toolbox of the research field of oxidation catalysis. They are applied as catalysts as bulk materials or deposited as nanoparticles or monolayers at the surface of more or less inert support materials (Hermans et al., 2009; Sun et al., 2014; Schlögl, 2015). For an enhanced activity of the catalysts often a high specific surface is generated by high dispersion of the material. For a high selectivity, a very complex property profile is essential, which is governed mainly by the structural and electronic properties of the oxide material. In a first step, all oxides may be categorized as reducible or non-reducible under the reaction conditions. Reducible oxides provide lattice oxygen (O^2-) from their surface or subsurface for oxygen insertion reactions or sites for the abstraction of a hydrogen atom in oxidative dehydrogenation reactions of hydrocarbons. This oxygen is considered as nucleophilic oxygen and can serve as selective species for the hydrocarbon oxidation reaction. Since the O^2- ion is extracted from the lattice to form a vacancy and the oxygen vacancies need to be replenished with gas-phase oxygen at a sufficiently fast rate to avoid the buildup of oxygen species which may cause undesired side reactions. This relation of the elementary steps of the reaction mechanism, that is assembled in the Mars-van-Krevelen mechanism, results in an overall reaction rate that is often independent of the oxygen partial pressure in the gaseous reaction mixture (Beck et al., 2012). Non-reducible oxides provide adsorption sites for oxygen molecules and control their activity by polarization. This occurs in particular at non-ideal surfaces, offering low coordinated sites e.g., steps and edges as adsorption sites (Schwach et al., 2015a; Schwach et al., 2015b). On the contrary, flat surfaces are almost inactive (Bajdich et al., 2015; Nilius and Freund, 2015; Aljama et al., 2018).

In this situation, different oxidation states of the oxygen molecule are possible, depending on the number and energy of transferred electrons. The resulting oxygen species are considered electrophilic and highly active for C-H activation. This difference in the selectivity behavior of reducible vs non-reducible oxides was nicely demonstrated by J. Haber and M. Witko (Haber and Witko, 2003). Due to the high activity, redox-active catalysts are also suitable for less reactive substrates like methane. For this reason, numerous reducible metal oxides have been tested as catalysts for the OCM reaction, one of the most promising but also most challenging reactions for a future methane utilization strategy (Zavyalova et al., 2011; Nguyen et al., 2020). Many non-reducible oxides also show activity for OCM, but with a rather low selectivity. The main reason for this observation is seen in the presence of different oxygen species with very different reactivity profiles. Therefore, many strategies have been tested for improving their performance in high throughput screening tests, rational approaches, and theoretical studies (Lee and Oyama, 1988; McFarland and Metiu, 2013; Zhang et al., 2020; Ohyama et al., 2021). The most promising approach for controlling the activation mechanism of oxygen and thereby the selectivity of OCM is the doping of the metal oxides with aliovalent ions, generating either a source or a sink for electrons within the material (Voskresenskaya et al., 1995).

Theoretically, wide-bandgap insulators are hardly able to activate oxygen (Bajdich et al., 2015). But in reality, it works due to the natural abundance of impurities (Voskresenskaya et al., 1995; Thum et al., 2019) causing strongly enhanced oxygen-binding on transition metals doped into CaO to form superoxo species. Here the O₂ ⁻ binding energy depends on the oxidation state of the dopant, the lower it is, the stronger is the binding energy. Doping effects depend on the nature of the transition metals, e.g., Mo²⁺/Mo³⁺ has electron donor abilities in the near-surface region (top 3 ML) whereas Cr³⁺ in MgO is charge compensated by the formation of Mg²⁺ vacancies. Also, the cation vacancy formation energy can be drastically reduced by doping or the presence of low valance impurities (McFarland and Metiu, 2013). Mo is an especially good donor that can provide multiple electrons to surface adsorbates due to the high lying states of 4 d and 6s levels. The Cr²⁺→Cr³⁺ transition occurs spontaneously in MgO due to a lattice misfit serving as an intrinsic electron trap (Nilius and Freund, 2015).

Although there is experimental evidence for the beneficial impact of doping of alkaline earth metal oxides (AEMO) by alkaline metal oxides (AMO) or rare-earth metal oxides (REMO) on the oxidative coupling of methane (Lin et al., 1988; Kondratenko et al., 1999; Rane et al., 2006; Papa et al., 2010), the interpretation of such complex experimental results is challenging. For this reason, studies on model systems were done, investigating the properties of the dopants inside their matrices, the interaction of oxygen with the surfaces, and the electronic properties of the doped materials with theoretical and spectroscopic techniques. The electronic properties of oxide surfaces may be investigated by photoluminescence studies that allow monitoring of high energy defect sites (Stankic et al., 2006; Hao et al., 2013). The quenching of the luminescence signal by adsorbed oxygen is a sensitive probe for the interaction of the catalysts with oxygen (Bengoechea-Encabo et al., 2016; Puust et al., 2017). Deeper insight into the doping effects is obtained from theoretical studies that allow a prediction of the key descriptors of the oxidation reactions (Hu et al., 2011; Sun et al., 2013; Bajdich et al., 2015). The study by the Norskov group demonstrates the influence of doping of AEMOs with traces of transition metal ions (TM) on the binding energy of hydrogen atoms and methyl groups as such descriptors. The derived volcano plot shows a strong impact and room for improvement for the activity of the investigated catalysts by TM doping. Although this study was only focused on the activity it also suggests changes in the selectivity caused by doping (Aljama et al., 2018). Another contribution to the understanding of doping effects was published by Freund and co-workers. They showed by STM studies how doping facilitates the dissociative adsorption of oxygen already at ambient temperatures (Nilius and Freund, 2015). To demonstrate the relevance of the model studies for real catalysts, a series of TM doped CaO samples was prepared with different dopant metals and concentrations. Due to the choice of CaO for the model studies, it was also chosen here, although pristine CaO is known to be a poor OCM catalyst. Hence, it should be particularly easy to identify positive effects of doping on OCM performance.

Photoluminescence spectroscopy was applied to gain insight into the effect of doping on electronic properties. EPR measurements allow for the investigation of the environment of the dopant ions within the CaO matrix. The catalytic studies were focused on the selectivity, especially the ratio of selective oxidation to deep oxidation. In order to link these experiments to the model studies the essential research questions of this study were:

How are the dopants dispersed within the CaO matrix?

Where are the dopants located?

Do the dopants change the surface or bulk properties of CaO?

Do dopants contribute to additional redox-active sites at the surface?

What is the impact of doping on catalyst selectivity?

Synthesis. Transition metal-doped calcium carbonates (M_xCa_1-xCO₃, with 0.05–3.2%) were prepared by co-precipitation as oxide precursors in a similar manner as described before for pure CaO.¹⁴ A mixed metal nitrate solution containing 10.7 wt% calcium nitrate (Ca(NO₃)₂ × 4H₂O, Fisherscientific analytical reagent grade) and the appropriate amount of transition metal nitrate (Cr³⁺, Mn²⁺, Ni²⁺, Co³⁺, Zn²⁺, Acros Organics) and an 8.03 wt% ammonium carbonate solution ((NH₄)₂CO₃, Roth, p. a.) were prepared and simultaneously added over a period of 1 hour to a 0.74 wt% ammonium carbonate solution at 25°C under constant stirring using an automated LabMax® reactor (Mettler Toledo). The product was aged in the solution for 1 h up to 72 h at 25°C, filtered, washed multiple times with millipore water, and dried at 65°C for 24 h in static air. The calcination was performed, if not mentioned otherwise, in the corresponding setup, before the experiments. For catalytic measurements the resulting powder was pressed at 200 bar, crushed, and sieved to yield a sieve fraction of 200–300 µm. As a reference sample (ref. CaO), the synthesis was performed with no additional metal nitrate solution and a commercial sample was prepared (pure CaO, Alfa Aesar Puratronic 99.998%).

XRD Analysis. XRD analysis was performed using a STOE STADI P transmission diffractometer equipped with a Cu anode, primary Ge (111) monochromator, and a Dectris MYTHEN 1K position-sensitive microstrip solid-state detector. The samples were crushed before analysis using an agate mortar. The data were fitted using the Bruker TOPAS software.

Nitrogen adsorption. An Autosorb^® iQ by Quantachrome^® Instruments was used to determine the surface area of the precipitated carbonates, using a 7-Point BET method assuming 0.162 nm² as the nitrogen cross-sectional area. The samples were degassed at 200°C prior to the measurements.

Elemental Analysis. The transition metal contents and impurities were determined by ICP-OES measurements using a Varian ICP-OES 715 ES. Depending on the metal content 50 mg up to 500 mg mixed carbonates were dissolved with 3 ml 69% HNO₃ (ROTIPURAN® Supra, ROTH) and diluted with water (VWR, ultrapure, HPLC grade) to a total volume of 15 ml. A dilution series of an ICP standard (Periodic table mix one for ICP, TraceCERT®) was used for calibration. The main impurity was found to be Mg followed by Sr with 150 ppm for all samples. Sodium impurities were found to be below the detection limit (detection limit 50 ppm).

Thermogravimetric Analysis. Using an STA 449 Jupiter thermoanalyser (Netsch) equipped with the pulse box for PulseTA, 10 mg of the carbonate samples were decomposed under a constant flow of synthetic air (20.5% O₂ in N₂) applying a heating rate of 2°C/min up to 1,000°C.

Catalytic Tests. For the catalytic test 50 mg of carbonate, the precursor was diluted in 750 mg of silicon carbide and fixed with quartz wool in a 4 mm inner diameter quartz tube reactor resulting in a bed length of 40 mm. The catalysts were tested in automated reactor units testing six to eight catalysts in parallel. The analysis of the reaction products was achieved using a GC equipped with TCD and FID detectors. Prior to the catalytic testing, the catalysts were activated in the reactor at 800 °C in synthetic air (N₂: O₂, 4:1, 40 ml/min) for 6 h. Applying a fixed feed composition (3:3:1 N₂:CH₄: O₂, if not indicated otherwise), the total flow was varied between 37.5 ml/min up to 200 ml/min and the temperature between 650 and 800°C. Methane conversion (X _CH4 ) and selectivities (S _i ) were calculated using Eq. 1 and Eq. 2, respectively, with x _i the molar fraction of compound i, ν the stoichiometric factor, and p the index for the reaction products (C₂H₆, C₂H₄, CO, and CO₂). X C H 4 = 1 − x C H 4 x C H 4 + ∑ 1 ν p x p (1) S i = 1 ν i x i ∑ 1 ν p x p (2)

Photoluminescence. Photoluminescence spectra were recorded using a Varian LS-55 Fluorescence Spectrometer. Calcium carbonate samples were placed in a quartz cuvette attached to a vacuum and gas supply line. For preparation the samples were heated under dynamic vacuum to 900°C for 6 h applying a rate of 5°C/min, reaching a vacuum of less than 5 × 10⁻⁶ mbar. For quenching experiments, oxygen was introduced to the sample. For some experiments, the samples could be sealed in the cuvette by melting the quartz tube feed using a torch.

EPR Spectroscopy. The oxide samples for EPR measurements were all prepared similarly to the PL samples avoiding contact with air. 2.9 mm EPR tubes made of high purity quartz were attached to a quartz flange and connected to a vacuum line. The catalyst was loaded into the tubes up to a bed height of 0.5–1.0 cm, degassed, and heated under dynamic vacuum (<5 × 10⁻⁶ mbar) to 900°C for 6 h, applying a heating rate of 5°C/min. After cooling down, the samples were sealed under vacuum with a torch. In addition, the carbonate samples were measured as prepared without any activation process.

Two samples comparing the pre- and post-reaction condition of the catalyst were prepared in a reactor setup suitable for the OCM reaction, equipped with four MFCs (Bronkhorst), providing methane, nitrogen, oxygen, and helium, a GC, and a folding oven. Reactors were prepared from a 2.9 mm outer diameter EPR tube with 6 mm outer diameter quartz tubes attached to fit in the catalytical setup. The carbonate sample was diluted with 0.4 mm quartz spheres and fixed in the reactor with quartz wool. The silicon carbide diluent used in the catalytic tests was found to disturb the EPR measurements. The amount of catalyst was chosen to yield a bed length of roughly 1 cm. Both samples were activated at 900°C for 6 h in synthetic air (20% O₂ in N₂) and cooled to 750°C. For the pre-reaction sample, the feed was switched to helium, cooled down to room temperature, and both ends of the 2.9 mm tube were sealed with a torch. For the post-reaction sample, the feed was switched to an OCM feed of 4:4:1 (CH₄:N₂:O₂) and kept on stream for 12 h. The feed was again switched to helium, the sample was cooled down and sealed, to avoid contact to air.

Continuous-wave (cw) EPR measurements at room temperature were conducted at 9.8 and 34 GHz microwave (mw) frequencies employing a Bruker B-ER420 spectrometer upgraded with a Bruker ECS 041XG microwave bridge and a lock-in amplifier (Bruker ER023M) using a Bruker SHQ resonator for 9.8 GHz and an ER051QT resonator for 34 GHz measurements. Echo-detected pulsed EPR measurements at 34 GHz were conducted on a Bruker ElexSys 580 setup with a home-built cavity (F. Lendtzian, TU Berlin). Field swept echo (FSE) spectra at 34 GHz were obtained at 20 K applying a two-pulse mw sequence (16-250-32 ns). Transient nutation (TN) measurements at 34 GHz were conducted applying a PEANUT (Phase-inverted Echo-Amplitude detected Nutation) pulse mw sequence with a π/2 pulse length of 32 ns, a delay time τ of 250 ns, and a high turning angle (HTA_x) pulse of 2048 ns incrementing the phase inversion time within the HTA_x pulse by 2 ns starting with an initial inversion after 32 ns (Stoll et al., 1998).

Phase pure metal-doped calcium carbonates as oxide precursors were synthesized and no indication for phase separation of the dopants was observed using XRD analysis, neither in the precursors nor in oxide materials, however, this cannot be excluded due to the low total abundance of metal dopants. The doping of metal in the case of Ni appears to have an influence on the crystallization of calcium carbonate during the precipitation process, prolonging the lifetime of the metastable vaterite phase during aging. Whether the Ni²⁺ ion is directly exchanged against a Ca²⁺ in the host lattice, however, remains unclear. In the case of Mn-doped CaO, the EPR measurements could show, that Mn²⁺ occupies high symmetry Ca sites (see also Supplementary Figure S4). Following the Hume-Rothery rule, none of the dopants is likely to form a stable solid solution, due to the significant difference (>15%) in atomic dimension. Due to their low concentration in the host lattice, verification is also very challenging and remains to be shown. In the case of low Fe³⁺ doped CaO, a clustering of the transition metal is known to occur (Biasi and Grillo, 2003), whereas in the case of Mn-doped CaO this appears not to be the case, as indicated by the EPR measurements (Biasi and Grillo, 1999) and also by the lattice contraction observed with XRD.

The addition of transition metal dopants to CaO has a significant influence on the photoluminescence of the sample, which, by oxygen quenching, could be proven to exclusively arise from surface sites. While alkaline dopants have been found to enhance the luminescence intensity of CaO and Ce³⁺ can be used to shift the emission spectrum (Hao et al., 2013), the doping of CaO with transition metals mainly results in a loss of luminescence. Already with the addition of small amounts (0.04 atom% Mn), the overall photoluminescence is strongly reduced, and the luminescence caused by edge sites becomes more prominent (feature at 290 nm, Figure 4 left) due to site isolation (Garrone et al., 1980). The introduced transition metals could offer new energy states favoring the non-radiative recombination for surface excitons. The observed behavior is independent of the nature of the dopant (Supplementary Figure S9) and thus cannot be used to explain the altered selectivities in OCM performance. However, the significant decrease in luminescence intensity confirms the change in the electronic structure of the material surfaces that accompanies the marked change in catalytic activity.

Doping of CaO with Ni and Mn changes the catalytic performance and its spectroscopic properties in a similar way. While low dopant concentrations have a strong impact on both phenomena the doping effect gets less pronounced with increased loading. In Figure 12 (Supplementary Figure S22 for Ni), the change of the reaction rate with the metal loading vs. the metal loading (in relation to the rate of the pure CaO) is plotted, showing that the scarcest dopants have the most significant effect on the reaction rate. Increasing dopant concentrations cause strong PL quenching and significant signal broadening in EPR, indicating changes in the interactions with the adsorbed oxygen as well as among the dopants themselves. With increasing dopant concentrations, the average distance between two dopant atoms decreases down to the range of dipole exchange interactions which is known for Mn²⁺ to be about 0.59 nm, which is well in the region of the here applied doping amounts and might be an explanation for the observed behavior (Biasi and Grillo, 1999). Further decreases of the distance could cause clustering of the metal atoms leading to a cancellation of the doping effect, indicated by the again increasing activation energies. For other metals this distance can be even larger, e.g., 0.83 nm for Gd³⁺, and 0.68 nm for Cu²⁺, where in the case of Fe, it is known for clustering at distances below a similar range (Biasi and Grillo, 2002a; Biasi and Grillo, 2002b; Biasi and Grillo, 2003).

Whether these observations support the model postulated by Freund et al., meaning a dopant-adsorbent interaction (Nilius and Freund, 2015), remains unclear. Doping aimed to alter the electronic structure should result in a significantly changed photoluminescence spectrum, such as significant shifts, or the creation of bulk new phosphors (Hao et al., 2013). However, proving with oxygen quenching, that the PL only derives from the material surface, the only observation made was a change in intensity, as well as a shift of the intensity distributions already existing in CaO, indicating a change in the surface structure. The observed effect can also not be explained by the “healing” of defects on the CaO surface. Defect sites, in general, are more reactive compared to pristine surfaces, which in the case of CaO is even postulated to be inactive (Nilius and Freund, 2015; Aljama et al., 2017, Aljama et al., 2018). The healing of defects should mainly result in an altered selectivity, as well as reduced activity, which is not observed in contrast, activity, as well as selectivity, are altered simultaneously. A more compelling explanation for the altered catalytic activity could therefore be the change of the crystallization due to the presence of foreign atoms with a significantly different ionic radius. An indication of this can be seen in the X-S plots (Figure 7). The Mn series appears to continue the C2-selectivity trend of the undoped CaO, implying the presence of more of the same active sites. As the PL measurements suggested, the ratio of the lower coordinated site compared to higher coordinated sites is higher in the doped samples. With highly coordinated sites, such as terrace sites being considered inactive in the OCM reaction, the increase of the lower coordinated sites explains the observed higher activity of the Mn-doped CaO. It is also quite reasonable that this effect is limited since highly isolated foreign ions can have a large impact on the surface structure by introducing lattice distortions, strain, stress (Khranovskyy et al., 2014), when the foreign atoms are getting close to each other, shown by the signal broadening in the EPR spectra (Figure 6B), the effect might be canceled out. Seeing a similar behavior with the Ni and Mn series also supports this hypothesis. A definitive proof, however, can unfortunately not be provided and more research is needed. With this, the initial research questions can be answered.

In the case of Mn doping, EPR experiments could verify Mn²⁺ in low concentrations positioned in high symmetry sites, most likely to substitute Ca²⁺ ions in the lattice.

No indication could be found for the dopants to introduce new energy states; however, their presence significantly alters the surface structure of the catalyst, indicated by the PL.

In the case of Mn doping, the post-reaction EPR measurements showed no indication for a change of the oxidation state of Mn, The catalyst selectivity can be enhanced via transition metal doping, with not all chosen elements being suitable. Especially ions with a stable 2+ oxidation state can enhance the catalyst selectivity if introduced in quantities up to 0.1 atom%.

Phase pure precursor materials could be synthesized by a co-precipitation method while avoiding disturbing contaminants such as halides and alkalines. Upon activation to CaO, no indications for phase segregation could be observed with the applied methods. Mn dopants probed by EPR exhibit O_h site symmetry as expected for substitutional insertion and no indication for clustering was found for Mn concentrations up to 0.36 mol%. In catalytic tests, metal dopants with a preferential oxidation state of two were found to have a beneficial effect on the OCM reaction (Mn, Ni, Zn), whereas other dopants, promoting the formation of cationic vacancies, were found to be unfavorable (Co, Cr) leading to increased methane combustion. Synthesizing a series of Mn and Ni-doped CaO, a minimum in the activation energy and was found at doping concentrations around ∼0.04%. Increasing the amounts of doping results in a reduction of efficiency of the individual doping atom. These findings are in line with observations made in PL and EPR, showing that increased doping results in loss of intensity (PL) and signal broadening (EPR), which indicates a dopant-dopant interaction not beneficial for the catalyst optimization. In conclusion, transition metal doping in low quantities can be applied to improve the catalytic performance of CaO, but it has to be noted, that the overall effect, in this case, is limited.