Targeted Fe‐Doping of Ni−Al Catalysts via the Surface Redox Reaction Technique for Unravelling its Promoter Effect in the CO2 Methanation Reaction

In promoted catalyst systems, the location of dopants is of very high interest to investigate promoter effects. A Ni/Al2O3 catalyst (wNi=11 wt. %) prepared by deposition‐precipitation and a co‐precipitated NiAlOx (nNi/nAl=1) catalyst are modified with Fe by means of the surface redox reaction technique and tested for activity under differential and integral conditions and for thermal stability (aging at 500 °C, 8 bar, 32 h) in the methanation reaction of CO2. By applying detailed material characterization studies comprising H2 and CO2 chemisorption, ICP‐OES, XRD, STEM‐EDX, FMR and BET, it is shown that the surface deposition techniques can be used to selectively deposit Fe in the vicinity of Ni nanoparticles. Doping with Fe leads to an increase of the catalytic activity, attributed to electronic effects through the formation of surface Ni−Fe alloys, and, for the co‐precipitated Ni−Al catalyst, to an enhancement of the apparent thermal stability at higher Fe loadings, which is assumed to be caused by a dynamic variation of Ni, Fe, and Al interactions depending on the reaction conditions.


Introduction
The CO 2 methanation reaction has recently gained interest for its role in the power-to-gas concept. [1] Synthetic natural gas (SNG) can be stored and distributed in the natural gas grid and therefore serve as a chemical energy storage to buffer fluctuations as well as regional and seasonal dependencies of energy supply by renewables. [2] The highly exothermal character of the CO 2 methanation reaction (Δ R H 298 K = 165 kJ mol À 1 ) leads to a demand for both high catalytic activity to achieve high CH 4 yields at mild operating conditions and high thermal stability to increase catalyst life-time by avoiding excessive catalyst deactivation, e. g. by sintering processes, [3] in industrial fixed bed application. Due to its high abundancy and low costs, [4] as well as its high selectivity to methane formation, [2d] Ni [5] is preferred over other active metals like Rh, [6] Pd, [7] Ru, [8] Pt or Ir. [9] Fe has been claimed to enhance the activity of Ni-based catalyst systems by electronic modification of the active Ni centers, forming NiÀ Fe alloy particles. [10] The effect of Fe on kinetics, however, is not conclusively clarified yet. In literature, the associative and the dissociative methanation pathway are controversially discussed. [11] In associative methanation, CO 2 adsorbs on the catalyst surface on basic sites and undergoes hydrogenation at the interface of the Ni particles, where H ads is supplied. [5e,12] Therein, catalyst performance is critical to the density and distribution of basic sites. [12c] In the dissociative mechanism, both H 2 and CO 2 adsorb dissociatively. [13] In this case, it is generally accepted that CÀ O bond cleavage is ratedetermining. [14] The reported effects of Fe, however, are manifold. Mebrahtu et al. showed that surface basicity can be tuned by varying the Fe loading in NiMgAlO x catalysts. [10f] The Nørskov group showed in an Bronsted-Evans-Polyani relation approach that NiÀ Fe alloys feature improved CÀ O dissociation energies, leading to an improved methanation performance. [15] The computational approach was also transferred to experimental studies [16] and is in line with our findings for coprecipitated NiFeAlO x catalysts [10a,17] and results from Hwang et al., who also claimed that Fe doping to NiÀ Al xerogel catalysts decreases the metal-support interactions. [10e] In addition, beneficial effects of Fe on the reducibility of NiO [17][18] and the Ni dispersion [18] were reported.
Besides the positive effect of Fe on the methanation activity, we recently proved an enhancement of the apparent thermal stability under aging conditions for co-precipitated NiFeAlO x catalysts at sufficiently high Ni/Fe ratios. [10a] However, the reasons for the stability improvement are not clear yet.
When applying conventional catalyst preparation techniques like impregnation or (co-)precipitation for metal doping, the promoter may be distributed on the catalyst surface or within the catalyst structure, and the location of the promoter relative to the active metal centers is usually unknown. In addition, for its redox properties especially true for Fe, the promoter may be present in different oxidation states (depending on its location), complicating conclusive decisions on its effect and structure-activity relationships. Therefore, this work addresses the investigation of the promoting effect of Fe on NiÀ Al catalysts selectively doped at the Ni centers by means of the surface redox reaction (SRR) method to better understand the effect of Fe on NiÀ Al catalysts in the CO 2 methanation reaction exclusively on the Ni centers.
The surface redox reaction (SRR) method is a known material preparation procedure, [19] but rather rarely used and, to the best of our knowledge, has not been applied to NiÀ FeÀ Al systems in the CO 2 methanation so far. It can be applied to selectively replace metal atoms on a material by atoms of a different metal with a lower reduction potential in an appropriate solvent. In this work, Fe 3 + ions dissolved in EtOH are used to oxidize Ni atoms on the activated NiAlO x catalyst. Scheme 1 illustrates the reactions possible on a Ni particle under the chosen conditions. The Fe species are deposited at the location where the electrons are supplied in the form of Fe 0 or Fe 2 + , or maintain in solution as Fe 2 + , while the generated Ni 2 + ions go into solution. The synthesis procedure is very sensitive to the washing process after the surface redox reaction to avoid any formation of clusters of the oxidizing or oxidized ion species by adsorption from the liquid phase or by impregnation during the drying process. Therefore, these catalyst synthesis steps were investigated and reported very detailed in this paper.
A Ni/Al 2 O 3 catalyst prepared by deposition-precipitation and a co-precipitated NiAlO x catalyst were taken as the template catalysts for the surface redox reaction. Besides the approach of doping Fe in a selective manner to the Ni nanoparticles, the comparison of the impact of Fe on two differently synthesized NiÀ Al template catalysts that vary in structure and sorption properties may shine some light on the importance of particlesupport interactions and morphology on the promoter effect of Fe. The Ni loading of the precipitated Ni/Al 2 O 3 catalyst was set to a typical value of 11 wt. %. For the co-precipitated benchmark catalyst, the molar Ni/Al ratio was set to one to provide data comparable to previous studies. [10a] The catalysts are labelled NiY x FeZ, where Y and Z, respectively, denote the metal loadings, subscript x indicates that the catalyst was derived from the co-precipitated NiAlO x catalyst.

Evaluation of Fe Deposition During the Surface Redox Reaction
The metal loadings and the molar element ratios of the catalysts prepared from the template catalysts are listed in Table 1.
For all SRR-modified catalysts, the amount of the Fe precursor substance Fe(NO 3 ) 3 · 9 H 2 O used during synthesis (cf. Table 3) correlates well with the Fe loading on the catalyst. At the same time, a decrease of the Ni loading is observed, owed to the exchange of Ni with Fe. The catalysts originating from the co-precipitated NiAlO x catalyst feature molar Ni/Fe ratios of 9.3, 5.7, and 3.0, respectively, making them comparable to our recent study on co-precipitated NiFeAlO x catalyst. [10a] The exchange ratio ΔN Fe /ΔN Ni describes the number of Fe atoms that are deposited on the catalyst per removed Ni atom. As depicted in Scheme 1, two competing reaction mechanisms need to be considered: Fe 3 + may either be reduced to Fe 0 (ΔN Fe /ΔN Ni = 2/3), which is deposited on top or the perimeter of the Ni particle, or to Fe 2 + , which may either be deposited on perimeter sites (ΔN Fe /ΔN Ni = 2), or stay in solution. The latter pathway seems to be the prominent one in our approach, since, for all catalysts, ΔN Fe /ΔN Ni is lower than the expected minimum value of 2/3. This also indicates that a considerate amount of Fe 3 + is reduced to Fe 2 + , remaining in solution rather than being deposited on the surface. In agreement, the presence of Fe 2 + in the solution was experimentally qualitatively proven by Turnbull's blue formation after adding [Fe(CN) 6 ] 3À (Merck, p.a.). The molar amount of Al in the samples stays constant in all catalysts, no Al 3 + leaching in EtOH could be observed by ICP-OES. Suspending the samples in H 2 O, in contrast, led to significant leaching of Al 3 + as well as γ-AlO(OH) formation (proven by XRD, not shown) for experiment times exceeding 24 h, which is consistent to processes occurring during hydrothermal treatments, however, reported in literature. [20] Therefore, the washing times after the SRR treatment in H 2 O were kept as short as 2 min. Washing the catalysts five times in fresh Scheme 1. Doping of an activated NiÀ Al catalyst with Fe by means of the surface redox reaction, green: Ni, red: Fe, orange: Fe 2 + , grey: oxidic Al-rich phase. degassed water, however, proved to be crucial to wash away redundant Fe n + (n = 2,3) and Ni 2 + species from the liquid phase, and to re-dissolve clusters nucleated on the Al-containing oxide surface. The Fe and Ni contents in the fifth washing filtrate were checked to be below 0.05 mg g cat À 1 by ICP-OES, highlighting that both the amount of Fe species being adsorbed on the surface and the amount of Ni and Fe being re-impregnated on the catalyst surface during drying can be neglected. Besides, the combination of this washing procedure and the degassing at 250°C ensured that no remaining C species originating from EtOH remained on the catalyst, as checked by BET and CHN analysis in pre-studies.
The constant exchange ratios of ΔN Fe /ΔN Ni (0.31 to 0.32) for the catalysts originating from the precipitated Ni/Al 2 O 3 template catalyst and 0.39 to 0.42 for the co-precipitated NiAlO x template catalyst prove that the doping process via the SRR technique is reproducible. The offset of about 0.1 between the template catalysts may result from the differences in the Ni particle size (e. g. accessible Ni sites) and morphological properties.

Scanning Transmission Electron Microscopy/Energy Dispersive X-ray Spectroscopy
Due to their strong ferromagnetic character after reduction (compare also chapter on FMR studies), no STEM images or EDX data of the activated or aged catalyst samples could be collected. The local atomic distributions of Ni, Fe and Al in calcined Ni5Fe2, resolved by STEM-EDX, are exemplarily shown in Figure SI 2, local intensity distributions of NiÀ K α , Al-K α , and FeÀ K α in Figure 1A. As expected for a classical supported Ni/ Al 2 O 3 catalyst, clear NiO clusters in the range of 6 nm can be observed. Fe is not statistically distributed on the surface, but rather located on concentrated spots in close neighborhood to Ni-rich sites (cf. Figure 1A). On the Fe-rich spots, the Ni signal is reduced (e. g. Figure SI 2B, Area 1), consistent with the replacement mechanism proposed in Scheme 1.
As apparent from Figure 1B and Figure SI 3, on Ni27 x Fe9, in contrast, Ni and Al are more homogeneously distributed. The central areas in Figure SI 3 A and SI 3B feature a very homogenous distribution of both Al and Ni. No distinct NiO clusters can be observed, which highlights the different morphologies of a supported Ni/Al 2 O 3 catalyst and a coprecipitated NiAlO x catalyst. Moreover, Figure SI 3C indicates that different phases exist, one rich in Ni 2 + and one that is rich  in Al 3 + and poor in Ni 2 + . Fe, again, seems to be co-localized rather with Ni 2 + than with Al 3 + . Elemental analysis data (by EDX) of selected spots in Figure 1 are shown in the ESI. Based on the STEM-EDX observations on the Ni5Fe2 and Ni27 x Fe9 samples and on the strong correlation of ΔN Fe /ΔN Ni , one can conclude that the replacement mechanism proposed in Scheme 1 is valid.

Structural Characterization
The co-precipitated NiAlO x template catalyst features a hydrotalcite structure after co-precipitation. The structural and morphological properties of this takovite-like co-precipitated [Ni 0.5 Al 0.5 (OH) 2 ](CO 3 ) 0.25 · n H 2 O material have been extensively discussed in previous studies [10a,17,21] and therefore are not repeated in this work. Figure 2  Besides the obvious presence of this crystalline NiO-rich phase, Alzamora et al. proposed the co-existence of a second Xray amorphous Al-rich Ni-containing alumina-like phase, [22] which is consistent to the previously discussed observations from STEM-EDX in Figure SI 3. This structure is common to hydrotalcite-derived materials [22][23] and greatly varies from the one of the precipitated Ni/Al 2 O 3 catalyst. Noteworthy, no bulk NiAl 2 O 4 spinel phase can be found by XRD for any of the catalysts.
The XRD patterns of the reduced reference catalysts Ni11À EtOH and Ni48 x À EtOH are shown in Figure 3. The characteristic fcc Ni peaks evolve at 2θ = 44.50°, 51.85°, and 76.38°. From the Scherrer equation, the Ni crystallite sizes can be estimated to be 6.1 nm for Ni/Al 2 O 3 À EtOH and 3.4 nm for Ni48 x À EtOH. It needs to be mentioned that this low particle diameter for the co-precipitated catalyst is close to the application limit of the Scherrer equation, but nevertheless is consistent with particle size distributions obtained by transmission electron microscopy studies in literature. [2b,24] The catalyst samples are not reduced quantitatively. While the remaining NiO species seem to be X-ray amorphous or too little to be detected by XRD for Ni11À EtOH in Figure 3A, their    3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57 presence in Ni48 x À EtOH is still evident from Figure 3B. Albeit the reflections caused by the NiO-rich mixed oxide shift towards alumina during reduction (cf. Figure 3B, reflections at 36°and 66°), which indicates the partial removal of Ni 2 + during catalyst activation, some Ni 2 + remains in this freshly formed crystalline Al 3 + -rich oxide phase, since its changed reflection signal still does not meet the reflection positions of γ-Al 2 O 3 .
For all Fe-promoted samples, no distinct reflexes caused by Fe species can be observed in the XRD patterns. However, with increasing Fe loading, for both the catalysts derived from the Ni/Al 2 O 3 (cf. Figure 3A) and the NiAlO x (cf. Figure 3B) template catalysts, shifts of the 111, 200, and 220 fcc Ni reflexes to lower diffraction angles can be observed. The corresponding increase of the lattice constant a can be explained by the insertion of Fe atoms into the crystal lattice of fcc Ni, effectively leading to the formation of (γFe,Ni) alloy particles. [25] The bulk composition of the alloy particles can be estimated by line profiling and comparison to tabulated values. [26] From the step width, an absolute error of 0.7 at. % can be assumed for x Fe , detailed data is given in Table SI 1. While for Ni9Fe0.5 the bulk of the particles only contains 2.5 at. % Fe, the Fe content stepwise increases to 4 at.% for Ni7Fe1 and 8 at. % for Ni5Fe2. Comparison to the overall Ni/Fe molar ratios in the catalyst (cf. Table 1) shows that the Ni/Fe molar ratio in the bulk alloy particles is significantly higher, which suggests that the majority of Fe atoms is located on the outer surface in vicinity to or on top of the Ni-rich (γFe,Ni) particles, in the form of a NiÀ Fe surface alloy or Fe 2 + , not contributing to the alloy formation in the bulk of the particles.
A similar picture is found for the catalysts derived from the NiAlO x template catalyst. However, the degree of bulk alloying is significantly higher. For Ni39 x Fe4, the bulk of the (γFe,Ni) particles contains 11 at.% Fe, which further increases to 15.5 at. % for Ni36 x Fe6 to 21 at. % for Ni27 x Fe9. The molar Ni/Fe ratios in the particles are much closer to the overall molar Ni/Fe ratios listed in Table 1.
These observations suggest that for the NiAlO x -derived (coprecipitated) catalysts NiÀ Fe alloy formation under reductive conditions (500°C, H 2 ) seems to be preferred compared to Ni/ Al 2 O 3 -derived (precipitated) catalysts, which might be caused by particle size effects or different stabilization of the particles on the oxidic phases.

Paramagnetic and Ferromagnetic Resonance Spectroscopy
A further indicator for the composition of the metal particles on the reduced catalysts is the change of the magnetic properties resolved by ferromagnetic resonance spectroscopy (FMR). All activated (reduced) catalyst samples show spectra, line intensities and thermomagnetic behavior typical for ferromagnetic particles, but with varying parameters for the different iron contents.
These differences in the magnetizations (I rel ), ΔB pp , shift of gvalues or anisotropy of the FMR spectra are described in the literature by variations of the Ni particle size and interactions with the support or adsorbed molecules. [27] Ni11À EtOH (cf. Figure SI 4A) features a significantly higher anisotropy (at T = 133 K) of the FMR spectrum compared to Ni48 x À EtOH (cf. Figure SI 5A). For Ni5Fe2 (cf. Figure SI 4B), a severe impact of Fe can be observed. The anisotropy of the FMR spectrum, evident from the decrease of ΔB pp (at 133 K, cf. Figure SI 4C) is significantly reduced, while the magnetization (I rel ) increases. This behavior can be explained by the interference of the ferromagnetic characteristics of Ni and metallic Fe, possibly by Fe atoms located on the surface of Ni particles.
The drastic changes of the ferromagnetic characteristics between the activated Ni48 x À EtOH (cf. Figure SI 4A) and Ni27 x Fe9 (cf. Figure SI 5B) are absolutely analogous to the coprecipitated NiÀ FeÀ Al catalyst [10a] and can only be explained by a substantial ferromagnetic contribution of metallic Fe and the assumption of the formation of NiÀ Fe alloy particles. This is reflected by the strong increase of all relevant criteria as anisotropy of the FMR spectra and Δg, of magnetization (I rel ), and ΔB pp (at 133 K, cf. Figure SI 5C) for Ni27 x Fe9 and is reported in detail in [10a] (and literature cited therein) and in the SI. Figure 4A illustrates the TPR profiles for the Ni/Al 2 O 3 -derived catalysts, Figure 4B the ones of the NiAlO x -derived catalysts. For Ni/Al 2 O 3 , a broad asymmetric reduction signal ranging from 360 to 760°C, caused by two overlapping reduction peaks centered at 520 and 615°C, respectively, and a shoulder at 780°C can be observed. The peak at 520°C can be attributed to the reduction of Ni 2 + weakly interacting with γ-Al 2 O 3 , while the peak centered at 615°C is supposed to be caused by the reduction of Ni 2 + that more strongly interacts with γ-Al 2 O 3 . [28] The third, hightemperature peak at 780°C indicates the presence of Ni 2 + possibly captured in a spinel-like structure. [29] Thereby, the presence of such Ni 2 + species seems to be limited to the catalyst surface, as significant amounts of bulk spinel formation can be excluded on the basis of XRD shown in Figure 2. The NiAlO x template catalyst features three reduction signals. The small low temperature signal at T = 150°C has previously been assigned to Ni 3 + in the outer layers of the catalyst structure. [25] In addition, two overlapping high temperature signals can be observed. The larger reduction signal centered at 585°C originates from the reduction of Al 3 + -containing NiO, while the smaller signal at higher temperature (680°C) is caused by the reduction of Ni 2 + incorporated in the Al 3 + -rich oxide phase. [30] For all Fe-doped catalysts, clear signals that can be assigned to the stepwise reduction of Fe 3 + to Fe can be found. This suggests that the SRR-modified catalysts get re-oxidized in the degassing step by NO 3 À , that probably forms the counter ion of Fe 2 + species located near the perimeter of the Ni(Fe) particles, but also indicates the presence of metallic Fe on the catalysts after reduction at 500°C.

Catalyst Reducibility
The reduction of Fe 3 + on the Fe-promoted Ni/Al 2 O 3 -derived catalysts occurs in the temperature range between 250 to 500°C. The first peak centered at 330°C corresponds to the reduction of Fe 3 + to Fe 2 + , the shoulder at 420°C marks the reduction of Fe 2 + to Fe 0 . The reduction signals of Fe are superimposed by the three reduction signals of NiO previously discussed and increase with increasing Fe loading.
Similar observations can be made for the Fe-doped NiAlO xderived catalysts. The shoulders at 350 and 410°C can be attributed to the reduction of Fe 3 + and Fe 2 + , respectively. The reduction peaks, however, are further superimposed by the reduction of Ni 2 + from the Ni-rich NiÀ Al mixed oxide, that is shifted to lower temperature with decreasing Ni loading. The decrease of the reduction signal from Ni36 x Fe6 to Ni27 x Fe9 is caused by the decrease in Ni loading, superimposing the reduction signal of Fe 3 + .

N 2 Physisorption
The catalysts derived from the precipitated Ni/Al 2 O 3 template catalysts feature type VI N 2 adsorption isotherms with H2 hysteresis loops. The BET surface area of the reference catalyst Ni/Al 2 O 3 À EtOH amounts to 182 m 2 g cat À 1 for the Ni11À EtOH reference catalyst. With increasing iron content, the BET surface area stepwise increases to a maximum of 198 m 2 g cat À 1 for Ni5Fe2 (cf. Table 2, index 1: before aging). The pore diameter first stays constant at 7.8 nm, but increases to 8.6 nm for Ni5Fe2. At the same time, for Ni5Fe2 a slightly increased pore volume (0.43 ml g cat À 1 compared to 0.39 to 0.40 ml g cat À 1 ) is observed. Despite the approach of a mild drying process (cf. Experimental), this increase in porosity can be attributed to the rapid evolution of NO x during degassing after the surface redox reaction inside the pores, which may lead to further pore formation or rupture of small pores. The amount of NO x released during the degassing step (which is proportional to the Fe loading) can be correlated to the increase of the BET surface area and the pore volume.
The BET surface areas of the hydrotalcite-derived catalysts are significantly higher. For the reference Ni48 x À EtOH catalyst, a BET surface area of 263 m 2 g cat À 1 is obtained. For all Fe-modified catalysts, the BET surface area stays constant at 300 m 2 g cat À 1 . The lower BET surface area of Ni48 x À EtOH can be explained by the difference in the processes occurring during the thermal pre-treatments. While the Fe-modified catalyst samples are reoxidized during the degassing by NO 2 as stated above, Ni48 x À EtOH stays in its reduced state. In the second reduction treatment, Ni48 x À EtOH seems to undergo some aging, leading to a decrease of the specific BET surface area. For the Femodified catalysts, in contrast, the mixed oxide phase is reconstructed during oxidation, and re-reduction leads to a similar BET surface area as obtained after the first reduction step. In accordance to this hypothesis, the specific BET surface area of the template NiAlO x catalyst after the first reduction amounts to 300 m 2 g À 1 . This indicates that (a) Fe does not have an influence on the structural characteristics of the oxidic phase after catalyst activation and that (b), within the investigated range, the Ni/Al ratio, which decreases with increasing Fe loading from Ni48 x À EtOH to Ni27 x Fe9 according to Table 1, does not have a significant impact on the porosity and the characteristics of the oxidic phase after reduction.
In summary, the results from material characterization are in line with the proposed pathway for the surface redox reaction and support the hypothesis that, after catalyst activation, the Fe species interact with the Ni particles rather than with the oxidic phase.

H 2 Chemisorption
For H 2 chemisorption the metal surface area is set equal to the Ni surface area (cf.  conditions. [25,31] H 2 adsorption at a recommended temperature of 200°C to account for exposed Fe atoms, [32] however, resulted in significant H 2 spill-over, making the determination of the metal surface area impossible. Moreover, CO chemisorption could not be applied, since the adsorption stoichiometry of CO on Fe is known to be structure-dependent [32][33] and therefore is a priori unknown. For Ni11À EtOH, a Ni surface area of 7.7 m 2 g cat À 1 is found. With increasing Fe loading, the Ni surface area decreases, reaching a minimum of 2.3 m 2 g cat À 1 for Ni5Fe2. This trend is consistent with the NiAlO x -derived catalyst samples. The very high Ni surface area of Ni48 x À EtOH (S Ni = 42.5 m 2 g cat À 1 ) strongly decreases with the introduction of Fe as a function of Fe loading. Ni27 x Fe9 features a Ni surface area of 8.2 m 2 g cat À 1 . Based on this trend, three major conclusions can be drawn: first, the samples prepared from the co-precipitated NiAlO x template catalyst show much higher Ni surface areas compared to the classical supported Ni/Al 2 O 3 -based catalysts, which has its reason in the unique characteristics of the hydrotalcite-derived oxide, leading to the stabilization of small Ni particles even at high Ni loadings. [24] Second, with increasing Fe loading, also the amount of Fe species exposed on the surface, blocking H 2 adsorption sites on Ni, is likely to increase for each template catalyst. This trend in the H 2 adsorption capacity is a typical feature for NiÀ Fe alloy particles [25] and therefore consistent with the presence of (γFe,Ni) nanoparticles supported by XRD and FMR. Analogous findings were made for co-precipitated NiÀ FeÀ Al catalysts for a Ni/Fe ratio > 6. [10a,17] Finally, the effect of Fe on the Ni surface area is consistent with the assumption that the Fe atoms were deposited in the neighborhood of the Ni sites during the surface redox reaction.

CO 2 Chemisorption and Temperature-Programmed Desorption
The surface basicity of the catalyst materials as well as the CO 2 binding strength are important parameters in catalyst development for the activation of CO 2 . Especially medium basic sites were proposed to play an important role in CO 2 processing under the assumption of an associative CO 2 methanation pathway via hydrogenation of CO 2 adsorbed on the support material at the particle-support interphase. [12c] In this study, the overall CO 2 uptake of the catalysts is determined by static CO 2 chemisorption, while the binding strength and basic site distribution is investigated by CO 2 -TPD. We would like to mention that the total CO 2 uptake does not completely match the uptake determined by CO 2 -TPD, since some of the weakly adsorbed CO 2 , which is accounted for in static chemisorption experiments, is already removed during the initial purging step prior to the TPD.
As shown in Table 2, the total CO 2 uptake for the catalysts derived from Ni/Al 2 O 3 is approx. 230 μmol g cat À 1 , independent from the Fe loading. Since the total CO 2 uptake primarily mirrors the CO 2 adsorbed basic sites of the oxidic phase, this is a hint that Fe species do not modify the surface basicity of the Al 2 O 3 support, but rather interact with the Ni particles in accordance to XRD, FMR and H 2 chemisorption. Figure 5A illustrates the corresponding temperature-programmed desorption patterns for the Ni/Al 2 O 3 -derived catalysts. Four CO 2 desorption signals can be distinguished: CO 2 bound to weak basic sites (peak centered at 90°), CO 2 bound to medium basic sites as bidentate carbonate (maximum desorption peak at 150°C) and monodentate carbonate (desorption peak at 225°C) as well as CO 2 adsorbed on strong basic sites in the form of "organic-like" carbonates with the maximum desorption signal at 320°C. [17,34] The disturbed desorption signal at 390 to 410°C stems from a negative contribution of CO 2 oxidizing the surface of the metal particles, leading to CO formation.
Within the error range, the TPD patterns of the Fe-doped Ni/Al 2 O 3 -derived catalysts are very similar, which is especially true for the distribution of basic sites. The majority of CO 2 is bound as bicarbonate on weak basic sites. With decreasing Ni loading, the density of monodentate carbonate sites slightly decreases. The formation of these basic sites is known to be promoted by remaining Ni 2 + in the catalyst structure. [35] The declining density of these sites therefore may go hand in hand with the decrease of Ni 2 + remaining unreduced during catalyst activation with increasing Fe loading due to the surface redox reaction.
The CO 2 adsorption characteristics of the NiAlO x -derived catalysts depicted in Figure 5B, in contrast, differ significantly. For the reference Ni48 x À EtOH, the total CO 2 uptake is 261 μmol g cat À 1 . Rising Fe loading and decreasing Ni loading then result in a decrease of the CO 2 uptake, reaching a minimum of 167 μmol g cat À 1 for Ni27 x Fe9. This behavior can be explained from the CO 2 desorption patterns in Figure 5B. For the Ni48 x À EtOH reference catalyst, four distinct CO 2 desorption signals can be found. With increasing Fe loading, one can observe a stepwise decrease of the medium and strong basic sites, while the density of weak basic sites increases only slightly. As stated above, Ni 2 + is known to be responsible for the formation of medium and strong basic sites on AlO x . With an increasing degree of Ni replacement by the introduction of Fe during the surface redox reaction, the effective Ni/Al ratio in the mixed oxide phase after catalyst activation continuously decreases with rising Fe loading, effectively leading to a lower density of medium and strong basic sites. The binding strength of CO 2 on those sites, however, seems to be unaffected. The decrease of the total CO 2 uptake capacity for the hydrotalcitederived catalysts therefore can be considered as an artefact from the surface redox reaction and is supposed to be caused by the decreasing Ni/Al ratio rather than a detrimental effect of Fe.

Catalyst Performance
The catalysts were tested for activity by means of their temperature vs. CO 2 conversion characteristics and for their stability under aging conditions by an artificial aging treatment for 32 h at 500°C. For reasons of comparison, the results are discussed in separate paragraphs.

CO 2 Methanation Activity
The activity for CO 2 methanation was determined under integral as well as under differential conditions. Figure 6 illustrates the CO 2 conversion vs. temperature characteristics (S 1 ) for the catalysts derived from the precipitated Ni/Al 2 O 3 (A) and the coprecipitated NiAlO x (B). For all catalysts, thermodynamic equilibrium gas composition is achieved at 400°C.
All catalysts exhibited excellent selectivity towards methane formation (cf. Figure SI 7 and SI 8). For the Ni/Al 2 O 3 -based catalysts, the maximum selectivities to C 2 H 6 (0.4 %), C 3 H 8 (0.1 %), and CO (2.8 %) were found at approx. 50 % CO 2 conversion. Due to their higher activity, for the catalysts based on NiAlO x , the maximum selectivities (also in the CO 2 conversion range from 50 to 60 %) were 0.9 % towards C 2 H 6 , 0.2 % towards C 3 H 8 and 2.5 % towards CO, merely independent from the Fe content of the catalyst.
To compare the activities of the catalysts under integral conditions, the characteristic temperature necessary to obtain a CO 2 conversion of 50 % can be evaluated. Based on this consideration, the activity rises in the order Ni11À EtOH (321.7°C) < Ni9Fe0.5 (320.9°C) < Ni7Fe1 (317.5°C) < Ni5Fe2 (311.9°C). This order is opposed to the trends of both the Ni loading (cf. Table 1) and the Ni surface area (cf. Table 2). Moreover, since properties like the CO 2 uptake/basic site density and CO 2 binding strength as well as the characteristics of the Al 2 O 3 support stay constant, this trend can only be explained by the effect of the Fe promoter on the active sites during CO 2 methanation in accordance to the theory of (γFe,Ni) nanoparticle formation. Fe may tune the CÀ O dissociation ability of the active sites, as found in computational analyses based on a Brønsted-Evans-Polanyi approach on the most active nanoparticle step sites by Andersson et al. for CO methanation [15b,36] and CO 2 methanation. [36] Thereby, it is assumed, in agreement to the general opinion in literature, that the cleavage of the CÀ O bond (hydrogen-assisted or via direct CÀ O dissociation) is the rate-determining step in the methanation reaction. [2c,14b,15b,36] Their investigations are limited to the assumption of a constant bulk composition of the nanoparticles, which, however, may undergo changes under aging conditions, which is focused on later in this paper.
Interestingly, the behavior under differential conditions suggests a different trend. As shown in Table 2, the apparent activation energy increases with rising Fe loading from 72.5 kJ mol À 1 for Ni11À EtOH to 76.9 kJ mol À 1 for Ni7Fe1 and 84.0 kJ mol À 1 for Ni5Fe2, which is a clear indicator for the modification of the active sites by the introduction of Fe. The unexpected coupling of a rising activity despite an increasing apparent activation energy can be explained by a distinct compensation effect. [37] The increase in the apparent activation energy can thereby be caused by an increase of the change of the entropy of the transition complex according to Eyring's theory, [38] or, more likely, by the simultaneous occurrence of CO 2 methanation on surface centers that involve different activation energies. For the latter case, the determination of the activation energy based on the Arrhenius equation would then yield an average activation energy over all active sites.
The catalytic activities of the NiAlO x derived catalysts can be explained in a similar manner. The activity under integral conditions rises in the order Ni27 x Fe9 (273.5°C) < Ni48 x À EtOH (265.9°C) < Ni39 x Fe4 (262.8°C) < Ni36 x Fe6 (261.2°C). For Ni27 x Fe9, obviously, the beneficial effect of the Fe promoter can no longer compensate the decrease of the Ni surface area and the CO 2 uptake due to the loss of medium basic sites, which are characteristics that we recently proved to be essential for high methanation activity over co-precipitated catalysts. [39] This effectively leads to a decrease of catalyst activity. A picture similar to Ni/Al 2 O 3 can be drawn when evaluating the activity behavior under differential conditions. Here, the apparent activation energy increases from 75.5 kJ mol À 1 for Ni48 x À EtOH to 85.6 kJ mol À 1 for Ni36 x Fe6 and 89.7 kJ mol À 1 for Ni27 x Fe9.
It is noteworthy that, compared to the Ni/Al 2 O 3 -derived catalysts, the activation energies of the NiAlO x -derived catalysts for a constant Ni/Fe molar ratio seem to be systematically increased by approx. 4 kJ mol À 1 , which might have its reason in different characteristics of the nanoparticles depending on the synthesis procedure. Wright et al., for example, showed that in reduced co-precipitated NiÀ Al catalysts Al, possibly in the form of (AlO 2 ) À , is incorporated in the nanoparticles, forming a paracrystalline Ni phase, [40] which might modify the properties compared to crystalline Ni. At this point, we would like to mention that the calculation of TOF values to compare the intrinsic activities of the catalysts was omitted since the number of active sites could not be determined (reasons stated in the experimental section).

Stability of the Catalysts under Aging Conditions
To test the stability of the catalysts under harsh methanation conditions at high temperature and elevated pressure, the catalysts were subjected to an aging treatment in thermodynamic equilibrium at 500°C, 8 bar for a duration of 32 h. To evaluate the activity after the aging treatment and to resolve data on catalyst stability, thereafter the CO 2 conversion vs. temperature characteristics were recorded again (S2). Figure 7A-D illustrate the CO 2 conversion vs. temperature characteristics before and after the aging treatment for the catalysts derived from the precipitated Ni/Al 2 O 3 catalyst. With increasing Fe loading, the curve recorded after aging is shifted further to higher temperatures. The difference between the characteristic temperatures necessary to obtain a CO 2 conversion of 50 %, ΔT (X(CO 2 ) = 50 %), can serve as a measure for the apparent stability of the catalyst. ΔT increases from 18.3 K for Ni11À EtOH to 23 K for Ni9Fe0.5, 25.7 K for Ni7Fe1, and 30.8 K for Ni7Fe2. The reference catalyst Ni11À EtOH is the most stable one, and the apparent stability decreases the higher the Fe loading.
The material properties after aging are listed in Table 2 (index 2: after aging). The CO 2 uptakes after aging are in the same order of magnitude and range from 119 μmol g cat À 1 for Ni11À EtOH to 135 μmol g cat À 1 for Ni7Fe1. The BET surface areas decrease to approx. 146 m 2 g cat À 1 for all catalysts, while the total pore volumes stay approximately constant. The mean pore diameter increases to 10.9 nm for all catalysts.
As shown in Table 2, for all Ni/Al 2 O 3 -based catalysts, the Ni surface area decreases significantly during aging. The decrease amounts to 75 % for Ni11À EtOH and Ni9Fe0.5, 70 % for Ni7Fe1, and 60 % for Ni5Fe2. At the same time, evaluation of the XRD patterns after reduction reveals that the bulk composition of the (γFe,Ni) particles is altered during aging, leading to a slightly higher degree of alloying. The Ni/Fe ratio for these samples after aging is closer to the bulk Ni/Fe composition shown in Table 1. For Ni9Fe0.5, the alloy contains 4.5 wt. % Fe (previously 2.5 wt. %), for Ni7Fe1 6 wt. % (previously 4 wt. %), while it increases from 8 wt. % to 10.5 % for Ni5Fe2. The activation energies stay approximately constant; one could, however, suspect a small decrease at high Fe loadings. By a slightly higher degree of alloying, the surface concentration of the Fe-modified active sites might decrease (extent increasing with rising Fe loading), effectively leading to a decrease of the activity after aging, opposite to the trend in initial activity of the Fe-doped Ni/Al 2 O 3 -based catalysts.
The catalysts derived from the precipitated NiAlO x catalysts (cf. Figure 7E-H) feature a significantly different behavior. The differences of the characteristic temperatures necessary for 50 % CO 2 conversion before and after aging amount to 12.6 K for Ni48 x À EtOH, 13.9 K for Ni39 x Fe4, 11.0 K for Ni36 x Fe6 and 4.4 K for Ni27 x Fe9.
Similar to the Ni/Al 2 O 3 based catalyst, the properties of the oxidic phase, BET surface area (approx. 125 m 2 g cat À 1 ) and the total CO 2 uptake (approx. 125 m 2 g cat À 1 ) decrease to the same values for all catalysts. The loss of Ni surface area ascribed to sintering can be calculated to be 56 % for Ni48 x À EtOH and Ni39 x Fe4, 53 % for Ni36 x Fe6, and only 24 % for Ni27 x Fe9. The change in Ni surface area can be ascribed to two different effects: first, particle sintering may occur, as evident from the increase of the (γFe,Ni) crystallite size shown in Table 2, but also to a redistribution of the Fe centers on the metal surface, blocking or releasing H 2 adsorption sites on Ni. The bulk composition of the (γFe,Ni) particles undergoes significant changes during aging: for the Fe-containing NiAlO x -based catalysts, the Ni/Fe ratios within the bulk alloy decreases. XRD analysis suggests that after the aging treatment the bulk of the (γFe,Ni) alloy particles contains 6.5 wt. % Fe in Ni39 x Fe4 (previously 11 wt. %), 9 wt. % Fe in Ni36 x Fe6 (previously 15.5 wt.), and 11 wt. % Fe in Ni27 x Fe9 (previously 21 wt. %), which can be interpreted as partial de-alloying of the metal particles. [41] To confirm the effect of the aging treatment on the composition of the metal particles, FMR was exemplarily carried out on the aged Ni27 x Fe9 catalyst. The aging procedure led to clear changes in the ferromagnetic characteristics, reflected in particular by a strong decrease in anisotropy (ΔB pp ) and Δg, of the main component, while a remaining background indicates residual NiÀ Fe alloyed particles. The ferromagnetic contribution of iron is clearly reduced indicating de-alloying and possibly partial oxidation of Fe. Due to the complex interactions of metal particle sintering (cf. Table 2), particle composition, as well as possible changes of metal particle shape and particle stabilization effects on the oxidic phase over aging, however, conclusive statements or interpretations are not possible at the current state but are part of an ongoing study.
In concordance to the change in the alloy bulk composition over aging, differences in the apparent activation energies can be observed. While the apparent activation energy is constant for Ni48 x À EtOH and Ni39 x Fe4, where the influence of Fe, in accordance to studies using co-precipitated NiÀ FeÀ Al catalysts, [10a] might be too low, it increases by 4.1 kJ mol À 1 for Ni36 x Fe6 and 7.4 kJ mol À 1 for Ni27 x Fe9 over the aging treatment. These drastic changes once more indicate the presence of a compensation effect, caused by the change in the concentrations of exposed Fe sites due to dynamic variation of the (γFe,Ni) bulk composition under aging conditions. At the same time, the segregation process may lead to (a) generation of new active sites and (b) a decrease in the sintering rate of the Ni sites, resulting in an improved apparent catalyst stability.
A decisive statement on the exact composition of the active sites and the role of possible Fe 2 + formation [41] during the segregation process, however, cannot be made. Detailed timeresolved studies on catalyst activity as a function of aging time, coupled with detailed material characterization under inert conditions to further resolve structure-activity relationships are planned in an ongoing study using an industrially more relevant co-precipitated NiÀ FeÀ Al catalyst and might contribute to shine some light on these questions. With respect to this, we would like to note that the selectivities towards CH 4 or any of the byproducts remained merely unchanged over aging, which is an indicator that no isolated Fe clusters on the catalyst surface were formed caused by the (partial) segregation of NiÀ Fe during aging (cf. Figure SI 8).

Effect of Catalyst Aging on (γFe,Ni) Nanoparticle Composition and Fe Surface Enrichment on the Catalytic Activity
It can be concluded that, depending on the nature of the oxidic phase, the composition of the (γFe,Ni) nanoparticles can undergo changes under harsh methanation conditions. These modifications influence the nature and the number of active sites, leading to (a) differences in catalyst activity and (b) changes of the (apparent) stability under methanation conditions. Within the investigated Ni/Fe range, a high surface concentration of exposed Fe atoms, either for the freshly activated catalyst or by (partial) segregation of a NiÀ Fe alloy, leads to an increase of the catalytic activity of the material.
This increase in activity is accompanied by an increase in the apparent activation energy, caused by the modification of the active sites. In fact, for each state, a relation between the apparent activation energy and the bulk (γFe,Ni) nanoparticle composition can be found, which is depicted in Figure 8. With increasing Fe content in (γFe,Ni), the apparent activation energy rises for both the Ni/Al 2 O 3 -and the NiAlO x -based catalysts. The (partial) segregation of the NiÀ Fe particles during aging of the NiAlO x -based catalysts leads to an increase of Fe sites on the surface and consequently to an increase of the apparent activation energy.

Conclusions
The redox surface reaction was successfully applied to selectively dope metallic Ni centers with Fe on a classical supported Ni/Al 2 O 3 catalyst prepared by deposition-precipitation and a coprecipitated NiAlO x catalyst. Activity and (apparent) catalyst stability were found to strongly depend on the surface concentration of Fe species on alloyed (γFe,Ni) nanoparticles formed after catalyst activation.
For the Ni/Al 2 O 3 -derived catalysts an increase of the catalytic activity with increasing Fe loading was found, but aging revealed a decrease of the (apparent) stability under methanation condition, possibly caused by a slightly deeper degree of alloying after aging. For the NiAlO x -derived catalysts, in contrast, a substantial enhancement of the apparent thermal stability upon an aging treatment with increasing Fe loading was found, linked to the (partial) segregation of the previously alloyed NiÀ Fe particles. For all catalysts containing Fe in considerate amounts (Ni/Fe approx. 6), a distinct compensation effect regarding the apparent activation energy was observed, strongly dependent on the surface concentration of Fe and most likely caused by the simultaneous occurrence of the methanation reactions over different active sites.
Further studies will focus on the time resolution of the deactivation behavior and modification of the aging conditions to decouple sintering and de-alloying effects. Material charac-terization studies under inert conditions at different states of deactivation may contribute to further elucidate the structure of the active sites on the (partly) deactivated Fe-promoted NiAlO x catalysts.
Besides the surface redox reaction technique, also organometallic approaches are currently under investigation to allow to draw a comprehensive picture of the deactivation behavior of Fe-promoted NiÀ Al catalysts.

Experimental Section Preparation of Template Catalysts
The Ni/Al 2 O 3 template catalyst was prepared by depositionprecipitation. 200 ml of a 0.02 M aqueous solution of Ni(NO 3 ) 2 · 6 H 2 O (p.a., Merck®) and of 1.3 M ammonia were added to 2.0 g Al 2 O 3 (Sasol) in an open 500 ml Erlenmeyer flask with baffles. The suspension was mixed on a rotary platform shaker (Heidolph) with 150 rpm at room temperature for 48 h. The suspension was decanted and the solid was washed with DI water two times. After drying at room temperature for 18 h, the catalyst precursor was calcined at 450°C for 3 h with a heating rate of 5 K min À 1 . For activation, the Ni/Al 2 O 3 template catalyst was heated from room temperature to 500°C (at a linear heating rate of 2 K/min) in 50 % H 2 in Ar and held there for one hour, before switching to a flow of 100 % H 2 for another hour.
The NiAlO x template catalyst was prepared by co-precipitation at a constant pH of 9. 120 ml of 1 M aqueous solutions of Ni(NO 3 ) 2 · 6 H 2 O (p.a., Merck®) and Al(NO 3 ) 3 · 9 H 2 O (p.a., Sigma-Aldrich®) were mixed and dropwise added to a 3 L double-walled glass vessel containing 1 L of bi-distilled water stirred at 150 rpm with a volumetric flow rate of 2.5 ml min À 1 . Two flow breakers were positioned in the vessel for secondary mixing. The temperature was pre-adjusted to 30°C and kept constant during the synthesis by a thermostat, the pH was pre-adjusted to 9 by adding a 0.5 M mixture of 1 M solutions of Na 2 CO 3 (Sigma-Aldrich®) and NaOH (Merck®). An Alphaline Titrino Plus (Schott) was used to keep the pH constant at 9 � 0.1 by adding the precipitation agent throughout the synthesis. The suspension was aged for 18 h in the mother liquor at pH 9 and 30°C while further stirring. Afterwards, the suspension was vacuum-filtered and the filter cake was washed until the conductivity of the filtrate was similar to DI water. The filter cake was dried at 80°C for 18 h.
The catalysts were calcined in flowing synthetic air at 450°C for 6 h with a linear heating rate of 5 K min À 1 . The catalyst powder was pelletized with a pressure of 450 N cm À 2 , ground and sieved to obtain a particle fraction of 150 to 200 μm. Detrimental effects on the porosity and the surface area of the catalysts at this pelletizing pressure were experimentally ruled out. The NiAlO x template catalyst was reduced in H 2 at 500°C for 5 h with a linear heating rate of 2 K min À 1 .

Doping of the Template Catalysts with Fe
For both template catalysts, three SRR-modified catalysts were synthesized. For doping by means of the surface redox reaction technique, the activated catalyst was evacuated at 10 À 6 mbar at the reduction temperature for 1 h to free the Ni sites from H ads species and cooled down to room temperature at 10 À 6 mbar.
Fe(NO 3 ) 3 · 9 H 2 O (p.a., Merck®) was dissolved in degassed and dried ethanol (p.a., Merck®) before the solution was added to the  Table 3. The suspension was stirred at 300 rpm for 10 min. After filtration under Ar atmosphere, the catalyst was washed five times with degassed DI water. The catalyst was then vacuum-degassed at room temperature for 1 h, at 80°C for 1 h and then at 250°C for another 3 h.
The template catalysts were subjected to the same procedure without Fe being added, (labelled À EtOH) for better comparison. The so modified catalysts Ni11À EtOH and Ni48 x À EtOH serve as the benchmark catalysts throughout the studies. Catalyst testing and material characterization was carried out on the SRR-modified catalysts as well as their benchmark catalysts, respectively. An impact of EtOH on the physiochemical, morphologic and catalytic properties of the NiÀ Al catalysts has been excluded by blank experiments.

Catalyst Testing Procedure
Catalyst testing was carried out in a setup described in a previous work. [2c] 50 mg of catalyst in the particle size fraction from 150 to 200 μm were thoroughly mixed with 450 mg purified SiC (ESK) and placed in the isothermal zone of a 4 mm diameter glass-lined tube reactor. The absence of heat and mass transfer limitations for this specific particle size range under the chosen conditions had been excluded beforehand, both experimentally and by evaluating heat and mass transport criteria. [2c,42] The axial position of the catalyst bed was fixed by 4 mm quartz wool plugs. To track bed temperature during reaction, a thermocouple was placed at the end of the diluted catalyst bed. The catalysts were in situ activated in H 2 (Q = 60 NL g cat À 1 h À 1 ) by heating to 500°C with a linear heating rate of 2 K min À 1 and holding this temperature for 5 h. Initially, the catalyst was subjected to methanation conditions at 8 bar and 250°C at 150 NL g cat À 1 h À 1 (H 2 /CO 2 /Ar = 4/1/5) for 2 h. After this start-up phase, the temperature was varied stepwise from 175°C to 500°C at 8 bar to resolve data on the activity of the catalyst in the form of its CO 2 conversion versus temperature characteristics (labelled S1). In the following, the catalyst was subjected to an aging treatment at 500°C and 8 bar for a duration of 32 h. After this artificial aging treatment, a second temperature variation cycle (S2) was carried out in order to resolve data indicating the apparent thermal stability of the catalyst under reaction conditions. The temperature program is shown in Figure SI 1. Data accuracy was checked in replicate experiments. To provide a clean surface prior the characterization of spent samples, the catalyst bed was heated up to 350°C in Ar (Q = 60 NL g cat À 1 h À 1 ) for 1 h. Furthermore, after cooldown, the sample was removed from the setup under inert atmosphere and vacuum-degassed at 350°C for 1 h.
The activation energies for CH 4 formation before and after aging were determined under differential conditions with the CO 2 conversion ranging from 2 % to 10 % by evaluating the slope of the logarithmic CH 4 formation rate plotted against 1/T. In advance, it was checked that the reaction orders of H 2 and CO 2 do not change in this regime. Experimental errors were calculated by Gaussian error propagation. The calculated and reported errors were higher than the errors observed in replicate experiments.
The purity of all gasses (Westfalen) was 5.0. The gas flow exiting the backpressure regulator was diluted with Ar in a ratio of 1/8. All tubing was heated to prevent water from condensation. An Emerson MTL-4 gas process analyzer was used for online tracking of the molar gas composition (CO 2 , CO, H 2 O, CH 4 , and H 2 ). For each measurement point, the parameters were kept constant for 45 min. Steady-state conditions were reached after 20 min. The actual product gas composition was averaged over 150 s (300 data points). Byproduct analysis was performed on a Perkin Elmer Clarus 580 gas chromatograph equipped with two columns and FID detectors. C, H, and O balances were closed by � 3 %. Conversions X and yields Y were calculated according to Equations (1) to (3), taking volume contraction into account according to Equation (4).
Yields of the hydrocarbon byproducts were calculated from the FID response corrected by the sensitivity factors. [43] Selectivities were calculated according to Equation (5). Enthalpy and entropy data for the calculation of equilibrium data were determined from the Shomate equation on the basis of data provided by the NIST Chemical WebBook. The calculation itself was carried out by the ΔG minimization method.

Elemental Analysis
Elemental analysis was carried out via inductively coupled plasmaoptical emission spectroscopy (ICP-OES) on an Agilent 700. For sample preparation, approximately 50 mg of the catalyst were dissolved in 50 ml of 1 M H 3 PO 4 by sonication for 2 h at 60°C. The samples were cooled down and diluted in a ratio of 1/10 with bidistilled water. The solutions were filtered using 0.45 μm syringe filters (Pall). The multi-element standard IV (Merck) was used to prepare metal standard solutions for 1, 10 and 50 ppm metal ion concentrations. Matrix interactions and metal signal interference were excluded. The wavelengths tracked for quantification were 230.299 nm (Ni), 396.152 nm (Al), 238.204 nm (Fe), and 568.263 nm (Na). All data were averaged over five measurements. The Na signal in all samples was below the detection limit (corresponding to a Na loading w Na < 10 À 3 wt.%), meaning that Na poisoning by the coprecipitation agent can be excluded.

X-ray Powder Diffraction Analysis
Ambient X-ray powder diffraction (XRD) was carried out on a Philips X'pert equipped with CuÀ K α radiation and a monochromator. The powders were scanned with 0.017°step À 1 and 83 steps min À 1 . XRD on reduced and spent catalyst samples was carried out on a STOE Stadi P diffractometer using CuÀ K α radiation, a Ge(111) monochromator and a Dectris MYTHEN 1 K detector. Approximately 5 mg of catalyst was transferred into glass capillaries (outer diameter 0.5 mm) under Ar atmosphere. Diffractograms were taken in the range of 2Θ = 5-90°with 0.015°step À 1 and a stepping rate of 45 steps min À 1 . The mean particle diameters were calculated by line profiling (Pseudo Voigt function) using Highscore 3.0d, evaluating the reflex caused by X-ray diffraction on the (020) plane at 2Θ = 51.5-51.8°of Ni or (γFe,Ni) crystallites, respectively. Nomenclature was taken from Swatzendruber et al. [26] Estimation of the (γFe,Ni) alloy particle composition was carried out by comparing the calculated cell parameter a of the fcc crystal lattice, determined from the reflection caused by the (020) plane at 2Θ = 51.5-51.8°, to tabulated values. [26] From the XRD step-width, the absolute error in the molar Ni/Fe composition of the (γFe,Ni) crystallites can be estimated to be � 0.7 at. %.

Scanning Transmission Electron Microscopy/Energy Dispersive X-ray Spectroscopy
To evaluate the relative positions of Fe and Ni on the oxidized SRRmodified catalysts, energy dispersive X-ray (EDX) spectroscopy in scanning transmission electron microscopy (STEM) mode was carried out at 300 kV on a FEI Titan Themis microscope equipped with a Super-X EDX detector. 1 mg of catalyst powder was dispersed in bi-distilled H 2 O and sonicated for 10 min. After sedimentation of the larger particles, 3 μl of the suspension were dropped onto a carbon film coated copper grid. The droplet was removed after an adsorption time of 10 s using filter paper.

Temperature-Programmed Reduction
Temperature-programmed reduction (TPR) profiles were recorded by thermal gravimetric analysis/mass spectrometry (TG-MS) on a NETZSCH ST 409. The parameters were chosen in accordance to Malet and Caballero [44] and Monti and Baiker. [45] 50 mg of catalyst were heated in a flow of 60 ml min À 1 Ar to 350°C with a linear heating rate of 5 K min À 1 . After cooling down to room temperature, the sample was heated to 850°C in 5 % H 2 in Ar with a total volumetric flow rate of 70 ml min À 1 and a linear heating rate of 5 K min À 1 . To gather the TPD patterns, the H 2 O signal at m/z = 18 was evaluated. Data was smoothed using a Loess filter with a span of 0.03. For determination of the reduction temperatures the signals were deconvoluted by Gaussian peak fitting.

N 2 Physisorption
N 2 physisorption experiments on activated and spent catalyst samples were carried out at 77 K samples on a Quantachrome NOVAtouch. For the determination of the BET surface area, the p/p 0 range between 0.05 and 0.3 was taken for evaluation. For the catalysts derived from the precipitated Ni/Al 2 O 3 catalyst, the total pore volume was taken from the data point at p/p 0 = 0.995. For the Al 2 O 3 -based catalysts, the pore size distribution was determined applying the BJH method on the adsorption branch since the samples exhibited type IV isotherms featuring a H2 hysteresis. The N 2 physisorption characteristics of the samples derived from the coprecipitated template catalyst can be classified into type IV isotherms featuring a H3 hysteresis, which is a hint for plate-like particles or slit-like pores. Therefore, the conventional theories on the pore volume and pore sizes determination cannot be applied. [46] For this reason, reporting of the total pore volumes and the pore size distributions for the NiAlO x -based catalyst is omitted.

H 2 and CO 2 Chemisorption
H 2 and CO 2 chemisorption experiments were conducted on an Autosorb 1 C (Quantachrome). For the pre-treatment, the fresh catalysts were activated in H 2 at 500°C for 5 h (linear heating rate 2 K min À 1 ). Adsorption equilibration time was set 2 min (H 2 ) and 10 min (CO 2 ), respectively. A dissociative adsorption mechanism of H 2 on Ni was applied for the calculation of the specific metal surface area. [47] As generally accepted in literature, [25,31] it was assumed that under the chosen conditions H 2 exclusively adsorbs on Ni and not on Fe. Furthermore, in preliminary studies it was ensured that the adsorption of CO 2 at the chosen conditions was not kinetically hindered on our samples.

Paramagnetic and Ferromagnetic Resonance Spectroscopy
Paramagnetic and ferromagnetic resonance (EPR/FMR) spectra of the activated catalysts were recorded on a JEOL JES-RE 2X at Xband frequency at temperatures between 113 and 473 K, a microwave frequency of 9.4 GHz, a microwave power < 0.2 mW, and a modulation frequency of 100 kHz. The microwave frequency was measured with a microwave frequency counter Advantest R5372. The catalyst samples were transferred into glass capillaries (diameter 0.5 mm) after activation (fresh catalyst samples) without contact to air. The integrated intensity was determined by double integration of the resonance signals of a weighed catalyst sample calibrated to a known standard (Mn 2 + /MgO). Conclusions on FMR data of the aged catalyst samples are difficult to interpret due to factors like particle size and shape [27a] (owed to possibly different susceptibilities to sintering), possible adsorbates on the catalyst surface [27c] as well as modified particle-support interactions [27b] during aging, influencing ferromagnetic characteristics like magnetic intensity and anisotropy. Investigations on aged samples by FMR and their interpretation are part of an ongoing study.