Surface Oxidation of Supported Ni Particles and Its Impact on the Catalytic Performance during Dynamically Operated Methanation of CO 2

: The methanation of CO 2 within the power-to-gas concept was investigated under ﬂuctuating reaction conditions to gather detailed insight into the structural dynamics of the catalyst. A 10 wt % Ni/Al 2 O 3 catalyst with uniform 3.7 nm metal particles and a dispersion of 21% suitable to investigate structural changes also in a surface-sensitive way was prepared and characterized in detail. Operando quick-scanning X-ray absorption spectroscopy (XAS/QEXAFS) studies were performed to analyze the inﬂuence of 30 s and 300 s H 2 interruptions during the methanation of CO 2 in the presence of O 2 impurities (technical CO 2 ). These conditions represent the ﬂuctuating supply of H 2 from renewable energies for the decentralized methanation. Short-term H 2 interruptions led to oxidation of the most reactive low-coordinated metallic Ni sites, which could not be re-reduced fully during the subsequent methanation cycle and accordingly caused deactivation. Detailed evaluation of the extended X-ray absorption ﬁne structure (EXAFS) spectra showed surface oxidation/reduction processes, whereas the core of the Ni particles remained reduced. The 300-s H 2 interruptions resulted in bulk oxidation already after the ﬁrst cycle and a more pronounced deactivation. These results clearly show the importance and opportunities of investigating the structural dynamics of catalysts to identify their mechanism, especially in power-to-chemicals processes using renewable H 2 .


Introduction
The power-to-chemicals concept is an important strategy for future renewable energy systems based on chemical energy storage. The splitting of water to produce H 2 and the catalytic conversion of CO 2 to methane, alkanes, methanol or higher alcohols are the main steps to generate a chemical energy carrier [1][2][3][4]. Both steps need to withstand fluctuations in supplied electricity from wind and solar plants, which occur temporarily and fluctuate on a time scale of seconds to days. When using a small or even no H 2 reservoir, these fluctuations are transferred to the catalytic reactor imposing a varying supply of H 2 used to hydrogenate CO 2 [5][6][7][8].
Methane (synthetic or substitute natural gas) as one possible chemical energy carrier can be distributed easily and unlimitedly in the existing natural gas grid. The dynamics in the methanation originating from fast load changes of the H 2 supply are a key challenge [5] and are therefore the subject of present research. The catalytic methanation of CO 2 is well-known, and numerous articles can be found in the literature reporting CO 2 conversion and CH 4 selectivity under optimized and steady state conditions [9,10]. Ni-based catalysts generally achieve good activity. In addition, due to the low cost and high activity, Ni catalysts have emerged as the most commonly-used methanation catalysts [11][12][13][14][15][16][17][18], since more active Ru catalysts [19,20] suffer from the high price of the noble metal. The most relevant catalyst support material in industrial methanation reactions is reported to be γ-Al 2 O 3 with a high surface area [19,21,22].
Methanation under transient reaction conditions has been performed to gain a better understanding of the mechanism, investigating the adsorbed species while switching between different gas mixtures [23,24]. Simulations of dynamic methanation reactors have been conducted to gain kinetic data and to address problems such as overheating during the transient process [25,26]. Other systems that have been studied under transient reaction conditions are, e.g., the Fischer-Tropsch reaction, where the influence of unsteady state conditions on the process itself was investigated [7,8,27]. Various spectroscopic, microscopic and diffraction techniques applied under reaction conditions revealed different structural changes of catalysts used in processes operated under changing reaction atmospheres [28][29][30][31][32][33].
Previous studies have shown that during the dynamic methanation of technical CO 2 , a fast partial oxidation of Ni-based catalysts occurred in a less reducing atmosphere after a H 2 dropout, which caused a deactivation in the following methanation sequence due to the presence of inactive NiO on the metallic catalyst [22]. Further deactivation occurred over cycles, but the initial catalytic activity was recovered using reactivation in H 2 at elevated temperatures [34]. In general, interruptions in the H 2 feed must be prevented in methanation processes using CO 2 with traces of oxygen (technical CO 2 ), since otherwise reactivation is required to retrieve high catalytic performance. Hence, methanation reactors have to be kept under a reducing atmosphere during stand-by operation [6]. Within this study, we provide a more detailed insight into the kinetics and mechanisms of the redox processes in methanation applications using CO 2 directly produced from biogas plants or exhaust gas from power plants containing traces of oxygen [35,36]. Based on these considerations, highly time-resolved operando XAS (quick-EXAFS (QEXAFS)) experiments were performed to evaluate the sensitivity of the catalyst towards deactivation. Furthermore, experiments including fast switches of the gas atmospheres were performed to obtain insights into blocking of active sites under fast load changes and thus mechanistic aspects in the methanation of CO 2 .

Preparation, Characterization and Catalytic Performance
A catalyst featuring uniform and rather small Ni particles was essential to determine the structural changes both of the surface and the bulk of the Ni particles during dynamically operated methanation. The homogeneous deposition-precipitation method was chosen to obtain a Ni/Al 2 O 3 catalyst with these properties. Elemental analysis using optical emission spectroscopy with an inductively-coupled plasma (ICP-OES) confirmed a catalyst loading of 10 wt % Ni/Al 2 O 3 . The specific surface area of the catalyst was 200 m 2 /g, and the mean pore diameter was 11 nm. X-ray diffraction (XRD) was not suitable to analyze the catalyst, since the Ni and NiO reflections were either superimposed by the signals of the support, the particles were too small to create reflections or the phases were X-ray amorphous, respectively (cf. the Supplementary Materials, XRD patterns in Figure S1).
As shown in Figure 1, electron microscopy revealed a homogeneous dispersion of the particles on the support resulting in small Ni particles with a diameter of 3.7 ± 1.2 nm, a narrow size distribution and a Ni dispersion of 21%. These values are comparable with those reported in the literature for supported Ni catalysts that were prepared using the same preparation technique [37,38].
Catalysts 2017, 7, 279 3 of 18 distribution and a Ni dispersion of 21%. These values are comparable with those reported in the literature for supported Ni catalysts that were prepared using the same preparation technique [37,38]. As an additional characterization tool, temperature programmed reduction (TPR) experiments were used, which showed H2 consumption in a small temperature range (peak maximum at 565 °C, cf. Figure S2), confirming the uniform particle size and the absence of agglomerates or larger particles. Furthermore, the Ni particles were reduced at lower temperature compared to other Ni/Al2O3 catalysts reported in the literature (peak maxima between 680 and 800 °C [39][40][41]), which suggests a weaker metal-support interaction in the catalyst prepared in this study due to the synthesis method.
The catalyst was applied in the methanation of CO2 and showed high conversion and selectivity with CO being the only by-product ( Figure 2). As the carbon balance was almost closed and therefore the calculation of the conversion according to Equations (1) and (2) (cf. Section 3) was similar, the formation of other products in significant amounts can be excluded. In Figure 2, the values obtained from Equation (1) are shown. The formation of CH4 started between 200 and 250 °C, and the conversion of CO2 increased with rising temperature, reaching the highest value of 69% at 400 °C at a maximum selectivity towards As an additional characterization tool, temperature programmed reduction (TPR) experiments were used, which showed H 2 consumption in a small temperature range (peak maximum at 565 • C, cf. Figure S2), confirming the uniform particle size and the absence of agglomerates or larger particles. Furthermore, the Ni particles were reduced at lower temperature compared to other Ni/Al 2 O 3 catalysts reported in the literature (peak maxima between 680 and 800 • C [39][40][41]), which suggests a weaker metal-support interaction in the catalyst prepared in this study due to the synthesis method.
The catalyst was applied in the methanation of CO 2 and showed high conversion and selectivity with CO being the only by-product ( Figure 2). As the carbon balance was almost closed and therefore the calculation of the conversion according to Equations (1) and (2) (cf. Section 3) was similar, the formation of other products in significant amounts can be excluded. In Figure 2, the values obtained from Equation (1) are shown. distribution and a Ni dispersion of 21%. These values are comparable with those reported in the literature for supported Ni catalysts that were prepared using the same preparation technique [37,38]. As an additional characterization tool, temperature programmed reduction (TPR) experiments were used, which showed H2 consumption in a small temperature range (peak maximum at 565 °C, cf. Figure S2), confirming the uniform particle size and the absence of agglomerates or larger particles. Furthermore, the Ni particles were reduced at lower temperature compared to other Ni/Al2O3 catalysts reported in the literature (peak maxima between 680 and 800 °C [39][40][41]), which suggests a weaker metal-support interaction in the catalyst prepared in this study due to the synthesis method.
The catalyst was applied in the methanation of CO2 and showed high conversion and selectivity with CO being the only by-product ( Figure 2). As the carbon balance was almost closed and therefore the calculation of the conversion according to Equations (1) and (2) (cf. Section 3) was similar, the formation of other products in significant amounts can be excluded. In Figure 2, the values obtained from Equation (1) are shown. The formation of CH4 started between 200 and 250 °C, and the conversion of CO2 increased with rising temperature, reaching the highest value of 69% at 400 °C at a maximum selectivity towards Figure 2. Conversion of CO 2 (black), yield of CH 4 (red) and yield of CO (blue), as well as selectivity of CH 4 (green) for the 10 wt % Ni/Al 2 O 3 catalyst; conditions: stainless steel tubular fixed-bed reactor, 150 mg catalyst, H 2 /CO 2 = 4, 75% N 2 , p = 1 atm, T = 200-450 • C, gas hourly space velocity (GHSV) of 26,700 h −1 and weight hourly space velocity (WHSV) of 12,000 mL CO 2 /(g cat ·h).
The formation of CH 4 started between 200 and 250 • C, and the conversion of CO 2 increased with rising temperature, reaching the highest value of 69% at 400 • C at a maximum selectivity towards CH 4 of 95%. At 450 • C, the conversion of CO 2 declined, and the selectivity shifted more toward CO due to the endothermic back reaction and the reverse water-gas shift reaction, which are preferred at higher temperatures [42]. The results are comparable with reports in the literature using Ni/Al 2 O 3 catalysts treated under related reaction conditions [19,43]. The turnover frequency (TOF) at 250 • C was calculated as 0.02 s −1 or as 0.08 s −1 at 300 • C. The TOFs for Ni/Al 2 O 3 methanation catalysts reported in the literature cover a broad range between 0.69 s −1 [11], 0.10 s −1 [18], 0.041-0.097 s −1 [12] and 0.5 × 10 −3 -2.4 × 10 −3 s −1 [17]. The achieved TOFs correspond to typical dimensions; however, it is difficult to compare the results due to the diverse reaction conditions applied in these studies. In conclusion, the catalyst used in this study can be regarded as a representative methanation catalyst, which is well suited for our experiments to evaluate the effects of fluctuating reaction conditions on the structure of the active component and their impact on the catalyst activity.

Operando QEXAFS Studies under Transient Reaction Conditions
Operando X-ray absorption spectroscopic (XAS) studies were performed to gather information on the oxidation state and the atomic structure of nickel during methanation conditions and, especially, during short periods of hydrogen dropout. In previous studies, we observed a significant oxidation of nickel particles, when H 2 was withdrawn from the reaction feed, leading to a lower catalytic activity during the subsequent methanation cycle [22,34]. In those studies, however, the hydrogen feed was interrupted for long periods of time (approximately 1 h) since conventional XAS methods were used, which require relatively long acquisition times per spectrum (approximately 5 min). Therefore, in the present study, we applied the quick-scanning EXAFS (QEXAFS) technique [44][45][46] to obtain data even in the sub-second regime for studying the impact of short hydrogen dropouts (30-300 s).
The methanation of technical CO 2 containing traces of oxygen and the simulation of the H 2 dropouts were performed at 400 • C (highest conversion and selectivity, compare Figure 2) and the same weight hourly space velocity (WHSV) applied in the activity measurements. Before each experiment, the catalyst was reduced in 50% H 2 /He at 500 • C. To simulate a short dropout of H 2 , caused for example by fluctuations in the hydrogen feed or by an unexpected change of the operation mode, hydrogen was switched on and off. The results of the experiment are presented in Figure 3. CH4 of 95%. At 450 °C, the conversion of CO2 declined, and the selectivity shifted more toward CO due to the endothermic back reaction and the reverse water-gas shift reaction, which are preferred at higher temperatures [42]. The results are comparable with reports in the literature using Ni/Al2O3 catalysts treated under related reaction conditions [19,43]. The turnover frequency (TOF) at 250 °C was calculated as 0.02 s -1 or as 0.08 s −1 at 300 °C. The TOFs for Ni/Al2O3 methanation catalysts reported in the literature cover a broad range between 0.69 s −1 [11], 0.10 s -1 [18], 0.041-0.097 s −1 [12] and 0.5 × 10 −3 -2.4 × 10 −3 s −1 [17]. The achieved TOFs correspond to typical dimensions; however, it is difficult to compare the results due to the diverse reaction conditions applied in these studies. In conclusion, the catalyst used in this study can be regarded as a representative methanation catalyst, which is well suited for our experiments to evaluate the effects of fluctuating reaction conditions on the structure of the active component and their impact on the catalyst activity.

Operando QEXAFS Studies under Transient Reaction Conditions
Operando X-ray absorption spectroscopic (XAS) studies were performed to gather information on the oxidation state and the atomic structure of nickel during methanation conditions and, especially, during short periods of hydrogen dropout. In previous studies, we observed a significant oxidation of nickel particles, when H2 was withdrawn from the reaction feed, leading to a lower catalytic activity during the subsequent methanation cycle [22,34]. In those studies, however, the hydrogen feed was interrupted for long periods of time (approximately 1 h) since conventional XAS methods were used, which require relatively long acquisition times per spectrum (approximately 5 min). Therefore, in the present study, we applied the quick-scanning EXAFS (QEXAFS) technique [44][45][46] to obtain data even in the sub-second regime for studying the impact of short hydrogen dropouts (30-300 s).
The methanation of technical CO2 containing traces of oxygen and the simulation of the H2 dropouts were performed at 400 °C (highest conversion and selectivity, compare Figure 2) and the same weight hourly space velocity (WHSV) applied in the activity measurements. Before each experiment, the catalyst was reduced in 50% H2/He at 500 °C. To simulate a short dropout of H2, caused for example by fluctuations in the hydrogen feed or by an unexpected change of the operation mode, hydrogen was switched on and off. The results of the experiment are presented in Figure 3.  First, the fully-reduced catalyst was exposed to methanation conditions (H 2 /CO 2 = 4; 75% He). After several minutes of operation and a stable mass spectrometer (MS) signal of methane, the catalyst was subjected to a period of fluctuating H 2 feed for 15 min. During this modulation, the feed was switched every 30 s between 5% CO 2 /He and methanation conditions (H 2 /CO 2 = 4). Finally, the catalyst was again exposed to steady state methanation conditions. The applied feed composition expressed as the signal of the valve position is depicted in the upper part of Figure 3. The product gas was monitored during the entire experiment; all analyzed components are shown in the Supplementary Materials ( Figure S14). The methane production was monitored by mass spectrometry (m/z 15), and the resulting signal is presented in the middle part of Figure 3. Throughout the experiment, X-ray absorption spectra were recorded with an acquisition rate of 44 spectra/s. The X-ray absorption near edge structure (XANES, exemplary sequence, see Figure 4) spectra were evaluated with linear combination analysis (LCA, details, cf. the Supplementary Materials) to monitor the Ni oxidation state during transient conditions. The results are plotted in the bottom part of Figure 3. First, the fully-reduced catalyst was exposed to methanation conditions (H2/CO2 = 4; 75% He). After several minutes of operation and a stable mass spectrometer (MS) signal of methane, the catalyst was subjected to a period of fluctuating H2 feed for 15 min. During this modulation, the feed was switched every 30 s between 5% CO2/He and methanation conditions (H2/CO2 = 4). Finally, the catalyst was again exposed to steady state methanation conditions. The applied feed composition expressed as the signal of the valve position is depicted in the upper part of Figure 3. The product gas was monitored during the entire experiment; all analyzed components are shown in the Supplementary Materials ( Figure S14). The methane production was monitored by mass spectrometry (m/z 15), and the resulting signal is presented in the middle part of Figure 3. Throughout the experiment, X-ray absorption spectra were recorded with an acquisition rate of 44 spectra/s. The X-ray absorption near edge structure (XANES, exemplary sequence, see Figure 4) spectra were evaluated with linear combination analysis (LCA, details, cf. the Supplementary Materials) to monitor the Ni oxidation state during transient conditions. The results are plotted in the bottom part of Figure 3. The Ni particles were reduced during the methanation of CO2 (H2/CO2 = 4), and an intense signal assigned to methane was observed in the mass spectrometer data. Short H2 cut-offs, however, strongly affected the catalyst and its activity. As H2 was switched back on the stream after a 30-s period of a less reducing atmosphere without H2 (0-0.5 min), a slightly decreased methane signal was observed. The methane production decreased further with each H2 dropout. However, no significant change of the Ni oxidation state was monitored during the first 5 min of the experiment. During the sixth H2 free period (after 5.5 min), a first slight oxidation of Ni to a fraction of 6% was observed, which was still accounted to be negligible. Fast and significant oxidation of Ni during the seventh H2 dropout (at 6-6.5 min) was observed by XAS, indicated by the increasing intensity of the white line. As discussed later in this paper, the oxidation of Ni in the absence of H2 was caused by the traces of oxygen present in the technical CO2. The oxidation stopped at 29% oxidized Ni according to LCA when H2 was switched back to the feed at 6.5 min. Even though the catalyst was exposed to the H2 containing atmosphere for 30 s (at 6.5-7 min), it did not regain its initial state, and 9% of the Ni atoms remained oxidized. The percentage of oxidized Ni species increased steadily during the subsequent modulation until the 12th H2 dropout, during which approximately 65% of the Ni atoms were oxidized. In the following oxidation and reduction events (between 11 and 15 min), a marginal but no significant increase in the percentage of oxidized Ni was observed, and the fraction of oxidized Ni remained at a constant level altering between 65% and 50% over cycles. Nevertheless, the deactivation of the catalyst continued steadily as shown by the decreasing MS signal of methane. The Ni particles were reduced during the methanation of CO 2 (H 2 /CO 2 = 4), and an intense signal assigned to methane was observed in the mass spectrometer data. Short H 2 cut-offs, however, strongly affected the catalyst and its activity. As H 2 was switched back on the stream after a 30-s period of a less reducing atmosphere without H 2 (0-0.5 min), a slightly decreased methane signal was observed. The methane production decreased further with each H 2 dropout. However, no significant change of the Ni oxidation state was monitored during the first 5 min of the experiment. During the sixth H 2 free period (after 5.5 min), a first slight oxidation of Ni to a fraction of 6% was observed, which was still accounted to be negligible. Fast and significant oxidation of Ni during the seventh H 2 dropout (at 6-6.5 min) was observed by XAS, indicated by the increasing intensity of the white line. As discussed later in this paper, the oxidation of Ni in the absence of H 2 was caused by the traces of oxygen present in the technical CO 2 . The oxidation stopped at 29% oxidized Ni according to LCA when H 2 was switched back to the feed at 6.5 min. Even though the catalyst was exposed to the H 2 containing atmosphere for 30 s (at 6.5-7 min), it did not regain its initial state, and 9% of the Ni atoms remained oxidized. The percentage of oxidized Ni species increased steadily during the subsequent modulation until the 12th H 2 dropout, during which approximately 65% of the Ni atoms were oxidized. In the following oxidation and reduction events (between 11 and 15 min), a marginal but no significant increase in the percentage of oxidized Ni was observed, and the fraction of oxidized Ni remained at a constant level altering between 65% and 50% over cycles. Nevertheless, the deactivation of the catalyst continued steadily as shown by the decreasing MS signal of methane.
During the ensuing 10 min of modulation, the changes of the catalyst (altering oxidation/reduction and continuous deactivation) followed the same trend and are therefore not shown here (full range experiment, see Figure S5). After the period of fluctuating operation (25 min in total), a sequence of steady state methanation was performed during which the oxidation state of the catalyst and structure was further monitored. During a 30-min measuring period, significant reduction of Ni was observed over time. This process, however, was very slow, and even after 30 min (final 5 min of this sequence shown in Figure 3), the initial oxidation state was not regained and 30% still remained oxidized. Consequently, the catalyst was less active and showed a similar methane production and oxidized fraction as observed after the ninth H 2 dropout. This supports a correlation between the oxidized fraction of Ni and the formation of methane [22]. Obviously, the temperature of 400 • C was too low to reduce Ni completely, which would have necessitated a re-activation of the catalyst at elevated temperatures [34].
The active catalyst state was obviously very sensitive towards oxidation, and already, a short exposure to a less reducing atmosphere was sufficient to initiate oxidation of Ni resulting in lower activity. Therefore, the full extended X-ray absorption fine structure (EXAFS) spectra were further evaluated to extract information on the local structural changes of the Ni catalyst during fluctuating conditions. In Figure 5a, Fourier-transformed (FT) EXAFS spectra are presented along with best fitting parameters obtained by fitting the first two coordination shells (Figure 5b). The EXAFS evaluation of the cycles 6-8 is shown in Figure 6; the structural parameters and the fully analyzed dataset are listed in Table 1; more details are shown in the Supplementary Materials. During the ensuing 10 min of modulation, the changes of the catalyst (altering oxidation/reduction and continuous deactivation) followed the same trend and are therefore not shown here (full range experiment, see Figure S5). After the period of fluctuating operation (25 min in total), a sequence of steady state methanation was performed during which the oxidation state of the catalyst and structure was further monitored. During a 30-min measuring period, significant reduction of Ni was observed over time. This process, however, was very slow, and even after 30 min (final 5 min of this sequence shown in Figure 3), the initial oxidation state was not regained and 30% still remained oxidized. Consequently, the catalyst was less active and showed a similar methane production and oxidized fraction as observed after the ninth H2 dropout. This supports a correlation between the oxidized fraction of Ni and the formation of methane [22]. Obviously, the temperature of 400 °C was too low to reduce Ni completely, which would have necessitated a re-activation of the catalyst at elevated temperatures [34].
The active catalyst state was obviously very sensitive towards oxidation, and already, a short exposure to a less reducing atmosphere was sufficient to initiate oxidation of Ni resulting in lower activity. Therefore, the full extended X-ray absorption fine structure (EXAFS) spectra were further evaluated to extract information on the local structural changes of the Ni catalyst during fluctuating conditions. In Figure 5a, Fourier-transformed (FT) EXAFS spectra are presented along with best fitting parameters obtained by fitting the first two coordination shells (Figure 5b). The EXAFS evaluation of the cycles 6-8 is shown in Figure 6; the structural parameters and the fully analyzed dataset are listed in Table 1; more details are shown in the Supplementary Materials.    Table 1. Structural parameters of the local Ni atomic environment extracted from the best fitting EXAFS spectra shown in Figure 6. k range = 3-9 Å −1 , R range = 1.2-3.2 Å, N ind = 6, N var = 5, a = fitted uncertainty 0.01 or lower, f = fixed during fit, N = number of neighboring atoms, R = distance, σ 2 = mean square deviation of interatomic distance, R factor = misfit between experimental data and theory (for details, see the Supplementary Materials). For the first six cycles, no significant changes in the EXAFS spectra were observed. The spectra were dominated by a feature at about 2 Å, predominantly originating from Ni neighbors in reduced Ni. The evaluated interatomic distance of 2.49 Å was similar to the one in metallic Ni. The O contribution determined in the reduced state may originate from the oxide support or incomplete reduction of the Ni particles. Including a Ni-Niox (the Ni-Ni backscattering contribution in nickel  Table 1. Structural parameters of the local Ni atomic environment extracted from the best fitting EXAFS spectra shown in Figure 6. k range = 3-9 Å −1 , R range = 1.2-3.2 Å, N ind = 6, N var = 5, a = fitted uncertainty 0.01 or lower, f = fixed during fit, N = number of neighboring atoms, R = distance, σ 2 = mean square deviation of interatomic distance, R factor = misfit between experimental data and theory (for details, see the Supplementary Materials).

Cycle
Atom N R (Å) a σ 2 ·10 −3 (Å 2 ) E 0 (eV) R Factor (%) For the first six cycles, no significant changes in the EXAFS spectra were observed. The spectra were dominated by a feature at about 2 Å, predominantly originating from Ni neighbors in reduced Ni. The evaluated interatomic distance of 2.49 Å was similar to the one in metallic Ni. The O contribution determined in the reduced state may originate from the oxide support or incomplete reduction of the Ni particles. Including a Ni-Ni ox (the Ni-Ni backscattering contribution in nickel oxide) bond pair led to a worse fit and was therefore not considered in Meth-6. During the seventh H 2 dropout, when significant oxidation was observed for the first time (cf. LCA in Figure 3), changes in the FT EXAFS data were observed, indicating substantial changes in the structure of the Ni nanoparticles. During oxidation, the feature at 2 Å in the EXAFS spectra resulting from reduced Ni-Ni coordination disappeared quickly and was replaced by two peaks related to the O and Ni neighbors in a NiO-like structure. The evaluation of the final EXAFS spectra of each sequence (methanation atmosphere or less reducing atmosphere, respectively) facilitated monitoring the evolution of the respective coordination shells during fluctuating operation. The respective coordination numbers are therefore plotted in Figure 5b for this purpose. The data revealed the transformation from a reduced Ni particle to a partially-oxidized Ni particle. Even though the Ni particles were very small, they did not get fully oxidized due to the short duration of H 2 dropout indicating that a core of reduced Ni remained (coordination number N Ni-Ni,red > 1). After the 12th dropout, predominantly a reversible oxidation and reduction was observed, in accordance with the LCA discussed earlier, and the core of the particles remained reduced.
The following steady state methanation operation (full-range experiment, see Figure S5) slowly transformed the catalyst back into its reduced state. However, even after 30 min in the methanation atmosphere, still, a significant fraction of nickel remained oxidized, also shown by the coordination numbers (N Ni-O = 2.2, N Ni-Ni,red = 2.4, N Ni-Ni,ox = 3.0, cf. Table S3). During this methanation experiment, only slight sintering was observed by an increase in the coordination number of reduced Ni by 0.2. This is within the accuracy of the evaluation; however, the Ni coordination number increased by 0.5 after repeating the whole experiment of 30-s modulation three times further (see Table S2). Note that the evaluated Ni coordination was not adjusted to the fraction of the particles in the Ni/NiO mixture, and no conclusion on the correlation between coordination number and particle size was drawn.
The delayed start of the Ni oxidation as observed in Figure 3 was rationalized when the spatial conditions of the experiment were considered [32,47]. XAS spectra were recorded in the middle of the catalyst bed with a 200 µm-sized X-ray beam and correlated to the integral activity data. Catalyst oxidation therefore probably started at the reactor inlet and propagated through the catalyst bed. We speculate that the oxidation was caused by the O 2 background present in the CO 2 feed (cf. the estimation in the later part of the paper), which oxidized a small amount of Ni at the reactor inlet in the short 30-s period in a less reducing atmosphere and was thereby completely consumed. Significant oxidation by CO 2 was ruled out, as CO 2 was present in excess and should then also have oxidized Ni in the middle of the reactor already at the beginning of the experiment (cf. Figure 3; oxidation started at the sixth switch). Furthermore, an additional experiment in the total absence of O 2 using short and thin stainless steel tubing and a CO 2 (99.998%) bottle directly placed next to the mass flow controllers showed that no catalyst deactivation occurred during 60 s H 2 dropouts (see Figure S17). Catalyst deactivation occurred during 300-s cycles when 300 ppm and 500 ppm O 2 were added to the gas mixture, which was more pronounced with increasing O 2 content ( Figure S18). Finally, the oxidation of Ni by CO 2 is thermodynamically unfavorable (∆G > 0). Estimations of the amount of O 2 present in the spectroscopic experiment based on the formation of NiO showed that an amount of around 1000 ppm was sufficient for the oxidation (for details, cf. the discussion of Figure S7 in the Supplementary Materials). Reduction of the whole catalyst bed was not achieved within the following 30 s during methanation at 400 • C. Therefore, the overall fraction of oxidized Ni increased during fluctuating operation, causing the steady decrease in catalytic activity. The methane production decreased continuously, even after Ni reached a stable oxidized fraction at the monitored position (after the 12th H 2 dropout), since Ni oxidation kept propagating towards the end of the catalyst bed.
To provoke a faster oxidation through the catalyst bed and to investigate the oxidation and its kinetics in a longer cycle duration, the catalyst was exposed to H 2 dropout events to an extended period of 300 s in a subsequent experiment using the same catalyst sample. In this experiment ( Figure 7), oxidation was observed immediately during the first dropout event. However, there was a significant delay (approximately 80 s; details, see Figure S7) between the switch of the gas atmospheres and the oxidation observed in the XANES spectra, due to the gradual oxidation of the Ni along the catalyst bed. Again, oxidation was found to be fast, resulting in 91% of NiO at the end of the 300-s cycle. Switching to methanation conditions after 300 s led to a re-reduction of the Ni particles and methane formation. About 38% of oxidized nickel remained on the catalyst and changed the overall catalyst composition resulting in a decreased catalytic activity in the second methanation cycle. Reduction occurred approximately 20 s after the gas switch and, therefore, seemed to propagate faster than the oxidation due to the much higher concentration of the reducing agent in the feed. In the following H 2 dropout cycles, the maximum oxidized fraction in the less reducing atmospheres remained at approximately 95%. Therefore, a 300-s H 2 dropout led to a (nearly) complete oxidation of the catalyst. A slowly-progressing reduction of the catalyst was observed during the 30-min period of steady state methanation at the end of the experiment, reaching a final value of a 48% reduced fraction. Ni along the catalyst bed. Again, oxidation was found to be fast, resulting in 91% of NiO at the end of the 300-s cycle. Switching to methanation conditions after 300 s led to a re-reduction of the Ni particles and methane formation. About 38% of oxidized nickel remained on the catalyst and changed the overall catalyst composition resulting in a decreased catalytic activity in the second methanation cycle. Reduction occurred approximately 20 s after the gas switch and, therefore, seemed to propagate faster than the oxidation due to the much higher concentration of the reducing agent in the feed. In the following H2 dropout cycles, the maximum oxidized fraction in the less reducing atmospheres remained at approximately 95%. Therefore, a 300-s H2 dropout led to a (nearly) complete oxidation of the catalyst. A slowly-progressing reduction of the catalyst was observed during the 30-min period of steady state methanation at the end of the experiment, reaching a final value of a 48% reduced fraction. Similar to the 30-s modulation, the estimated amount of O2 based on the formed NiO showed a value around 1000 ppm. Additional experiments with a quantified amount of O2 ( Figure S18) showed a full Ni oxidation after a few 300-s H2 dropouts applying 500 ppm O2. Consequently, we assumed an O2 content between 500 and 1000 ppm present in the QEXAFS experiments. This small quantity is ideal to oxidize the most reactive sites and block in fact those Ni sites for methanation.
The FT EXAFS spectra collected during the 300-s modulation and the coordination numbers from the best fits of the EXAFS spectra are shown in Figure 8a,b, respectively. As already observed during the 30-s modulation, the FT EXAFS spectra in the reduced state were dominated by one signal at 2 Å, which was assigned to Ni-Ni coordination in reduced Ni. Note that the NNi-Ni,red of 5.1 was slightly higher than the one determined from the EXAFS evaluation during the 30-s modulation (NNi-Ni,red = 4.4). This can be explained by sintering of the Ni particles since the same sample was used for all experiments and three additional 30-s modulation experiments (not shown here) were performed before the 300-s modulation experiment (see Table S2). Applying the 300-s modulation, the maximum of the Ni-O path was already reached during the first H2 dropout. A bulk oxidation of the Ni particles occurred during the absence of H2 indicated by the NNi-Ni,red, which decreased to zero. The oxidation state reached its steady state after the first cycle, while the amount of receded Ni during the methanation periods decreased until the fourth cycle.
Reduced Ni particles with a NNi-Ni,red of 5.1 were transformed into oxidized Ni particles with a NNi-Ni,ox of 7.6. This can be either explained by incomplete reduction of the Ni particles contributing  Figure S18) showed a full Ni oxidation after a few 300-s H 2 dropouts applying 500 ppm O 2 . Consequently, we assumed an O 2 content between 500 and 1000 ppm present in the QEXAFS experiments. This small quantity is ideal to oxidize the most reactive sites and block in fact those Ni sites for methanation.
The FT EXAFS spectra collected during the 300-s modulation and the coordination numbers from the best fits of the EXAFS spectra are shown in Figure 8a,b, respectively. As already observed during the 30-s modulation, the FT EXAFS spectra in the reduced state were dominated by one signal at 2 Å, which was assigned to Ni-Ni coordination in reduced Ni. Note that the N Ni-Ni,red of 5.1 was slightly higher than the one determined from the EXAFS evaluation during the 30-s modulation (N Ni-Ni,red = 4.4). This can be explained by sintering of the Ni particles since the same sample was used for all experiments and three additional 30-s modulation experiments (not shown here) were performed before the 300-s modulation experiment (see Table S2). Applying the 300-s modulation, the maximum of the Ni-O path was already reached during the first H 2 dropout. A bulk oxidation of the Ni particles occurred during the absence of H 2 indicated by the N Ni-Ni,red , which decreased to zero.
The oxidation state reached its steady state after the first cycle, while the amount of receded Ni during the methanation periods decreased until the fourth cycle.
Reduced Ni particles with a N Ni-Ni,red of 5.1 were transformed into oxidized Ni particles with a N Ni-Ni,ox of 7.6. This can be either explained by incomplete reduction of the Ni particles contributing to the apparently lower coordination in reduced Ni and a higher coordination after oxidation. Another explanation for the difference in the coordination numbers can be a wetting effect of the flat reduced particles, increasing the contact with the oxidic support material, whereas the oxidized particles are more spherical with a higher average Ni-Ni coordination [30]. to the apparently lower coordination in reduced Ni and a higher coordination after oxidation.
Another explanation for the difference in the coordination numbers can be a wetting effect of the flat reduced particles, increasing the contact with the oxidic support material, whereas the oxidized particles are more spherical with a higher average Ni-Ni coordination [30]. The direct comparison of the MS signals of both modulation experiments ( Figure S16) showed that the deactivation was more distinctive during 300-s switches, where a significantly lower CH4 signal was detected after the second cycle up until the end of the experiment. Furthermore, in the experiments with 30-s modulation, the catalyst was much more active during steady state methanation after the H2 dropout sequence. This observation clearly emphasized the correlation of the Ni oxidation state and the methanation activity. Comparing the results of the LCAs from both experiments, full oxidation was interrupted during the 30-s cycles due to the fast switching. During the 300-s modulation, the oxidation state reached a plateau indicating the complete oxidation process. Thus, the impact of the short cycles on the catalyst deactivation was less severe. Figures 9 and 10 show schemes of the Ni particles during the dynamic methanation of CO2. Short-term H2 dropout (Figure 9) first led to an oxidation of the most reactive sites such as low coordinated Ni atoms located at step sites, defect sites or at the interface to the support material [48][49][50][51][52]. Particularly, the interface sites appear important for the methanation of CO2, since the key steps of the hydrogenation are speculated to take place at the interface of the metal particle and support [53,54]. Furthermore, step sites are more reactive in CO dissociation compared to plane facets [51,55]. NiO islands may form and grow to NiO structures of several layers [56,57]. Diffusion of oxygen into the bulk did not occur due to the short period of less reducing atmosphere. Full oxidation of the Ni particles was interrupted, and NiO planes were reduced partly in the methanation atmosphere. The direct comparison of the MS signals of both modulation experiments ( Figure S16) showed that the deactivation was more distinctive during 300-s switches, where a significantly lower CH 4 signal was detected after the second cycle up until the end of the experiment. Furthermore, in the experiments with 30-s modulation, the catalyst was much more active during steady state methanation after the H 2 dropout sequence. This observation clearly emphasized the correlation of the Ni oxidation state and the methanation activity. Comparing the results of the LCAs from both experiments, full oxidation was interrupted during the 30-s cycles due to the fast switching. During the 300-s modulation, the oxidation state reached a plateau indicating the complete oxidation process. Thus, the impact of the short cycles on the catalyst deactivation was less severe. Figures 9 and 10 show schemes of the Ni particles during the dynamic methanation of CO 2 . Short-term H 2 dropout (Figure 9) first led to an oxidation of the most reactive sites such as low coordinated Ni atoms located at step sites, defect sites or at the interface to the support material [48][49][50][51][52]. Particularly, the interface sites appear important for the methanation of CO 2 , since the key steps of the hydrogenation are speculated to take place at the interface of the metal particle and support [53,54]. Furthermore, step sites are more reactive in CO dissociation compared to plane facets [51,55]. NiO islands may form and grow to NiO structures of several layers [56,57]. Diffusion of oxygen into the bulk did not occur due to the short period of less reducing atmosphere. Full oxidation of the Ni particles was interrupted, and NiO planes were reduced partly in the methanation atmosphere. The highly active lower coordinated Ni sites remained oxidized under these reaction conditions due to their higher stability. Hence, the methanation of CO2 could only take place at the reduced Ni plane sites resulting in lower activity. This supports earlier studies in the literature [48][49][50]58] that selective blockage of the low coordinated Ni edge or corner sites leads to differed or deteriorated catalytic properties. Similar structure sensitivity due to easier oxidation of low-coordinated sites was also observed during CO oxidation on Pt catalysts [59,60]. For example, oxidation of low coordinated Pt atoms located at edge and corner sites during CO oxidation was found by XAS resulting in an irreversible CO adsorption on PtO [61]. Cluster models show that metal particles of our size or dispersion contain 4% edge sites [62][63][64][65]. This value correlated with the remaining oxidized fraction (10%) in the methanation sequence after the first oxidation during the seventh H2 dropout when additional sites at the interface, as well as corner and defect sites, would have been taken into account.
The schematic structural changes of Ni particles during 300-s modulations are illustrated in Figure 10. Compared to the short-time H2 dropouts, the longer 300 s H2 dropouts resulted in a less reducing atmosphere enabling a complete oxidation of the Ni particles, which then suffer from insufficient re-reduction during the reducing methanation atmosphere. Figure 10. Scheme of the structure of the supported Ni particles exposed to H2 dropout modulations at a duration of 300 s.

Catalyst Preparation
A 10 wt % Ni/Al2O3 catalyst was prepared via homogeneous deposition-precipitation using urea as the precipitating agent [66][67][68]. Al2O3 support (Alfa Aesar, Karlsruhe, Germany, 1/8 inch pellets, crushed to fine powder, calcined at 600 °C (5 K/min) for 4 h) was suspended in a solution of 0.03 mol/L Ni(NO3)2·6H2O (Merck, Darmstadt, Germany, ≥99%) in water. Five equivalents of urea (Carl Roth, Karlsruhe, Germany, ≥99.6%) were added to the mixture and stirred for 1 h at room temperature, where a pH value of 5 was observed. The suspension was then heated to 90 °C and The highly active lower coordinated Ni sites remained oxidized under these reaction conditions due to their higher stability. Hence, the methanation of CO 2 could only take place at the reduced Ni plane sites resulting in lower activity. This supports earlier studies in the literature [48][49][50]58] that selective blockage of the low coordinated Ni edge or corner sites leads to differed or deteriorated catalytic properties. Similar structure sensitivity due to easier oxidation of low-coordinated sites was also observed during CO oxidation on Pt catalysts [59,60]. For example, oxidation of low coordinated Pt atoms located at edge and corner sites during CO oxidation was found by XAS resulting in an irreversible CO adsorption on PtO [61]. Cluster models show that metal particles of our size or dispersion contain 4% edge sites [62][63][64][65]. This value correlated with the remaining oxidized fraction (10%) in the methanation sequence after the first oxidation during the seventh H 2 dropout when additional sites at the interface, as well as corner and defect sites, would have been taken into account.
The schematic structural changes of Ni particles during 300-s modulations are illustrated in Figure 10. Compared to the short-time H 2 dropouts, the longer 300 s H 2 dropouts resulted in a less reducing atmosphere enabling a complete oxidation of the Ni particles, which then suffer from insufficient re-reduction during the reducing methanation atmosphere. The highly active lower coordinated Ni sites remained oxidized under these reaction conditions due to their higher stability. Hence, the methanation of CO2 could only take place at the reduced Ni plane sites resulting in lower activity. This supports earlier studies in the literature [48][49][50]58] that selective blockage of the low coordinated Ni edge or corner sites leads to differed or deteriorated catalytic properties. Similar structure sensitivity due to easier oxidation of low-coordinated sites was also observed during CO oxidation on Pt catalysts [59,60]. For example, oxidation of low coordinated Pt atoms located at edge and corner sites during CO oxidation was found by XAS resulting in an irreversible CO adsorption on PtO [61]. Cluster models show that metal particles of our size or dispersion contain 4% edge sites [62][63][64][65]. This value correlated with the remaining oxidized fraction (10%) in the methanation sequence after the first oxidation during the seventh H2 dropout when additional sites at the interface, as well as corner and defect sites, would have been taken into account.
The schematic structural changes of Ni particles during 300-s modulations are illustrated in Figure 10. Compared to the short-time H2 dropouts, the longer 300 s H2 dropouts resulted in a less reducing atmosphere enabling a complete oxidation of the Ni particles, which then suffer from insufficient re-reduction during the reducing methanation atmosphere. Figure 10. Scheme of the structure of the supported Ni particles exposed to H2 dropout modulations at a duration of 300 s.

Catalyst Preparation
A 10 wt % Ni/Al2O3 catalyst was prepared via homogeneous deposition-precipitation using urea as the precipitating agent [66][67][68]. Al2O3 support (Alfa Aesar, Karlsruhe, Germany, 1/8 inch pellets, crushed to fine powder, calcined at 600 °C (5 K/min) for 4 h) was suspended in a solution of 0.03 mol/L Ni(NO3)2·6H2O (Merck, Darmstadt, Germany, ≥99%) in water. Five equivalents of urea (Carl Roth, Karlsruhe, Germany, ≥99.6%) were added to the mixture and stirred for 1 h at room temperature, where a pH value of 5 was observed. The suspension was then heated to 90 °C and Figure 10. Scheme of the structure of the supported Ni particles exposed to H 2 dropout modulations at a duration of 300 s.

Catalyst Preparation
A 10 wt % Ni/Al 2 O 3 catalyst was prepared via homogeneous deposition-precipitation using urea as the precipitating agent [66][67][68]. Al 2 O 3 support (Alfa Aesar, Karlsruhe, Germany, 1/8 inch pellets, crushed to fine powder, calcined at 600 • C (5 K/min) for 4 h) was suspended in a solution of 0.03 mol/L Ni(NO 3 ) 2 ·6H 2 O (Merck, Darmstadt, Germany, ≥99%) in water. Five equivalents of urea (Carl Roth, Karlsruhe, Germany, ≥99.6%) were added to the mixture and stirred for 1 h at room temperature, where a pH value of 5 was observed. The suspension was then heated to 90 • C and stirred under reflux for 18 h, cooled to room temperature and stirred for another hour, resulting in a final pH value of 7-8. The solid was filtered, washed with deionized water, dried over night at 110 • C and calcined for 4 h at 500 • C (5 K/min). Typically, 2 g of the catalyst were prepared in one batch containing 1.80 g Al 2 O 3 , 1.00 g Ni nitrate, 1.04 g urea and 115 mL water.

Characterization
The chemical composition was determined via optical emission spectroscopy with an inductively-coupled plasma (ICP-OES) using an Agilent (Waldbronn, Germany) 720/725 emission spectrometer. The specific surface area of the catalyst was determined using N 2 physisorption at −196 • C according to the Brunauer-Emmet-Teller (BET) method via multipoint measurements using a BELSORP-mini II (MicrotracBEL, Osaka, Japan). X-ray diffraction (XRD) patterns were recorded in the range 2 θ = 20-80 • (step size 0.017 • , 0.51 s per scan step) using a rotating sample holder containing catalyst powder and a PANalytical (Almelo, The Netherlands) X'pert PRO diffractometer with Ni-filtered Cu-Kα radiation (1.54060 Å). Scanning transmission electron microscopy (STEM) images were recorded using a FEI (Hillsboro, OR, USA) Titan 80-300 microscope operated at 300 kV and a Fischione Model 3000 high-angle annular dark-field (HAADF-STEM) detector. The measurements were conducted using reduced catalyst powder dispersed on a gold grid covered with holey carbon film. The Ni particle size was determined with ImageJ software (1.48v, National Institutes of Health, Rockville, MD, USA, 2016) measuring the mean diameter of 965 marked particles using ellipsoid shapes. The Ni dispersion was calculated using the mean diameter of the Ni particles, the area occupied by a Ni surface atom and the volume occupied by a Ni atom in the bulk metal assuming spherical particles [69]. A Micromeritics (Aachen, Germany) AutoChem II 2920 chemisorption analyzer was used for temperature programmed reduction (TPR) experiments. One hundred milligrams of the catalyst sample (100-200 µm) were placed in a U-shaped quartz tube and fixed with quartz wool plugs. First, the catalyst was heated to 500 • C (10 K/min) in 10% O 2 followed by the TPR between 40 and 900 • C (10 K/min) using 10% H 2 . The H 2 consumption was recorded using a thermal conductivity detector (TCD).

Catalytic Performance
The catalytic activity in the methanation of CO 2 was tested in a continuous flow laboratory setup with a stainless steel reactor filled with a fixed-bed of 150 mg catalyst sample (300-450 µm, diluted in SiC) operated in down-flow. The reactor was heated by a custom-made oven (HTM Reetz, Berlin, Germany), and the temperature was controlled using a thermocouple placed in front of the catalyst bed inside of the reactor and regulated by a Eurotherm 2416 PID controller. Gas dosing was adjusted using individual mass flow controllers (Bronkhorst, Ruurlo, The Netherlands). First, the catalyst was reduced in situ in 50% H 2 /N 2 (600 mL/min) at 500 • C (10 K/min) for 2 h. Then, the reactor was cooled to 200 • C, and the gas atmosphere was switched to H 2 /CO 2 = 4 in 75% N 2 (600 mL/min total flow), resulting in a gas hourly space velocity (GHSV) of 26,700 h −1 (concerning total gas flow and catalyst volume) or a weight hourly space velocity (WHSV) of 12,000 mL CO 2 /(g cat ·h) (concerning gas flow of CO 2 and catalyst mass), respectively. The methanation reaction was performed between 200 and 450 • C in 50 • C steps to determine the optimum reaction conditions. Reactant and product gas compositions were analyzed by a Thermo Scientific (Waltham, MA, USA) C2V-200 micro gas chromatograph (microGC) equipped with a molecular sieve (5 Å) and a QS-BOND column and a thermal conductivity detector (TCD). Conversion, yield and selectivity were calculated as follows: Conversion: X(CO 2 ) = 1 − CO 2,out CO 2,out + CH 4,out + CO out × 100% (1) or including N 2 as the internal standard:Conversion: Yield: Y(CH 4 or CO) = CH 4,out or CO out CO 2,out + CH 4,out + CO out × 100% (3) The turnover frequency (TOF) was calculated as moles of CH 4 produced per moles of Ni surface atoms per second using the inlet flow of CO 2 , the molar gas volume V(m) and the catalyst mass: The number of Ni surface atoms per gram catalyst N(Ni, surf ) was determined using the dispersion from TEM analysis [69].

Operando XAS Experiments
Quick-scanning X-ray absorption spectroscopy (QEXAFS) [70,71] was performed at the SuperXAS beamline at the Swiss Light Source (SLS) synchrotron facility (Paul Scherrer Institute, Villigen, Switzerland), which operates in top-up mode at 400 mA and 2.4 GeV. The measurements were performed at the Ni K-edge (8333 eV) in transmission mode. The polychromatic X-ray beam was collimated using a Si-coated mirror at 2.5 mrad located in front of the monochromator. The monochromatized beam was focused to a beam size of 200 µm × 200 µm at the sample position using a Rh-coated mirror at 2.5 mrad. A channel-cut Si(111) monochromator oscillating at 22 Hz was used together with N 2 -filled gridded ionization chambers both dedicated for fast data acquisition, where a Ni foil was measured simultaneously with the data for absolute energy calibration. In each oscillation of the monochromator, two spectra were recorded, one with increasing energy, another one with decreasing energy. With this system operando X-ray absorption spectroscopy (XAS) data were recorded with 44 spectra/s. For the operando XAS experiments the catalyst sample was diluted with Al 2 O 3 (1:1) and placed into a 1.5-mm quartz glass capillary (100-200 µm sieved fraction, 10 mg, 10 mm catalyst bed) and fixed with quartz wool plugs [72]. Spectra were acquired at the middle of the catalyst bed. The capillary-based reactor was connected to the gas dosing system equipped with individual mass flow controllers (Bronkhorst, Ruurlo, The Netherlands) and heated using a hot air blower (FMB Oxford GSB-1300, Oxford, UK) [73]. The total gas flow was adjusted to 20 mL/min using similar compositions as in the laboratory experiments (reduction: 50% H 2 /He, methanation: 25% reactants (H 2 /CO 2 = 4) in He). The same WHSV of 12,000 mL CO 2 /(g cat ·h) as in the laboratory was used resulting in a higher GHSV of 71,700 h −1 due to the different dimensions of the reactor. Fast switches of the gas atmosphere were realized using a micro-electric VICI (Schenkon, Switzerland) 4-way valve. The gas composition was analyzed by a Pfeiffer Vacuum (Aßlar, Germany) ThermoStar TM GSD 320 mass spectrometer (MS) equipped with a quartz capillary and a C-SEM/Faraday detector.
The measurement of 44 spectra/s allowed detection of changes with a time resolution of 23 ms. However, it was found that a 1-s time resolution was sufficient to monitor the structural changes of the catalyst, and thus, the spectra could be averaged for 1 s each to improve data quality. Matlab ® (v8.6, The Mathworks Inc., Natick, MA, USA, 2015) routines were used for merging QEXAFS data to 1 s/spectrum, energy calibration to a metallic Ni foil, normalization and for conducting linear combination analysis (LCA). For the latter, a linear combination of reference spectra was fitted to the sample spectra in the XANES region (from −30 eV to +50 eV). The spectrum of the reduced catalyst in 50% H 2 /He atmosphere and of the oxidized catalyst in 5% O 2 , both acquired at 400 • C, were used as reduced and oxidized reference, respectively (see Figure S3). EXAFS data analysis was performed using Athena and Artemis of the IFEFFIT software package (v1.2.11, Chicago, IL, USA, 2008) [74] on spectra collected during steady state conditions and on selected spectra during fluctuating operation. In this case, 220 spectra were averaged, as this was sufficient to monitor the structural changes and improve the data quality (5 s/spectrum). The theoretical data was adjusted to fit the experimental spectra by a least square method in Fourier-transformed (FT) EXAFS spectra (with k 1 -, k 2 -and k 3 -weighted EXAFS functions, FT EXAFS spectra, k and R range, as well as fixed and varied parameters are shown in the Supplementary Materials) considering metallic Ni (ICSD: 64989) and NiO (ICSD: 9866). Corresponding backscattering amplitudes and phases were calculated with Feff 6.0 [75]. Amplitude reduction factors (S 0 2 ) were obtained on metallic and NiO bulk references (more details, cf. the Supplementary Materials). Coordination number (N), bond distance (R) and energy alignment between theoretical and experimental data (∆E 0 ) were refined for each spectrum. The mean square deviation of interatomic distances (σ 2 ) was obtained by simultaneously refining spectra of the sample in similar conditions and at the same temperature (cf. the Supplementary Materials).

Conclusions
Experiments with periodically changing atmospheres during the methanation of technical CO 2 simulated as H 2 dropouts were performed and resulted in detailed insights into the catalyst deactivation mechanism. During H 2 interruptions, catalyst deactivation occurred due to oxidation of the most active surface Ni sites in the presence of O 2 impurities (500-1000 ppm) in technical CO 2 . Thirty-second short-term H 2 dropouts showed surface oxidation and reduction occurring after the seventh modulation cycle monitored in the center of the catalyst bed. The core of the Ni particles remained reduced, since full oxidation was interrupted. Nevertheless, catalyst deactivation occurred already after the first dropout since the oxidation of the active sites propagates stepwise through the catalyst bed starting from the reactor inlet. This gives important insight into the mechanism of the methanation reaction and, obviously, even short interruptions in the H 2 feed, and thus, slight oxidation of the most active Ni sites cannot be tolerated by the catalyst. Therefore, it is important to prevent H 2 dropouts in industrial methanation applications, where traces of O 2 might be present. This can also be transferred to other power-to-chemicals processes using renewable H 2 . A higher impact on the deactivation was observed during 300-s modulation, where a bulk oxidation was observed even in the first H 2 dropout, which resulted in a higher fraction of oxidized Ni remaining on the catalyst during the methanation sequence and, thus, an even lower CH 4 production.
The presence of small amounts of oxygen in the CO 2 stream can already block a small fraction of most active Ni sites, which had a strong influence on the CO 2 -methanation activity. This "titration" of active sites may be further used in future to understand the methanation of CO 2 , but also other CO 2 hydrogenation mechanisms in more detail.
Supplementary Materials: The following are available online at www.mdpi.com/2073-4344/7/9/279/s1, Figure S1: XRD patterns of the γ-Al 2 O 3 support and the 10 wt % Ni/Al 2 O 3 catalyst in its oxidized and reduced state. Figure S2: H 2 -TPR profile of the 10 wt % Ni/Al 2 O 3 catalyst. Figure S3: Reference XANES spectra of the bulk materials (recorded at room temperature) and in situ oxidized and reduced spectra of the catalyst (measured at 400 • C) for linear combination analysis (LCA). Figure S4: XANES and FT EXAFS spectra of the reduced catalyst in H 2 /He at room temperature (RT) and at 400 • C and in He atmosphere at 400 • C. Figure S5: Full-length experiment of the 30-s H 2 dropout modulations during methanation of CO 2 . Figure S6: Zoom of the first 6 min during the 30-s modulation. Figure S7: Zoom of the first 14 min during the 300-s H 2 dropout modulation. Figures S8-S13: Experimental and simulated EXAFS data. Figures S14 and S15: Full MS dataset recorded during the 30-s and 300-s H 2 dropout modulations, respectively. Figure S16: Comparison of the MS signals of CH 4 (m/z 15) during the 30-s and 300-s H 2 dropout modulations. Figure S17: Sixty-second H 2 modulation in an oxygen-free experiment. Figure S18: Series of 300-s H 2 dropout periods during the methanation of CO 2 containing 300-500 ppm O 2 . Tables S1-S5: Structural parameters of the local Ni atomic environment extracted from the EXAFS spectra.