Development of Highly Stable Low Ni Content Catalyst for Dry Reforming of CH4‐Rich Feedstocks

Highly active and coking‐resistant Ni catalysts suited for the dry reforming of CH4‐rich gases (70 vol %, e. g. biogas or sour natural gas) were prepared starting from a Mg‐rich Mg−Al hydrotalcite support precursor. Calcination at 1000 °C yields two phases, MgO and MgAl2O4 spinel. Complexation‐deposition of Ni with citric acid on the preformed support as well as lanthanum addition yields a catalyst with remarkably low carbon accumulation over 100 h on stream attributed to both high Ni dispersion and preferred interactions of Ni with MgO on MgAl2O4.

Global energy demand is rapidly growing, and over 80 % thereof is covered by fossil fuels in recent time. [1] Natural gas and biogas are emerging feedstocks for the energy market due to their abundance and low specific carbon footprint of the main component methane. [2][3][4] However, most natural gases and biogases contain varying amounts of other compounds. [5] For instance, some major natural gas sources in Vietnam [6] or biogases in Germany [7] contain a high fraction of CO 2 (3 0 vol %). As separation is costly, dry reforming (DRM) is considered to directly convert such CO 2 -rich gases into synthesis gas, a highly preferred starting material in large-scale chemical syntheses. [8] Mostly, DRM has been investigated with stoichiometric mixtures of CH 4 and CO 2 . Processing feeds with an under-stoichiometric concentration of CO 2 via DRM may lead to higher H 2 fractions but encounters high coking rate due to the shortage in CO 2 , which otherwise removes the surface carbon via gasification during the reaction. [9] Nickel is the most commonly studied non-noble metal in DRM [9] because of its low cost and high availability. [5] However, Ni catalysts are prone to fast deactivation by coke deposition.
Consequently, this issue was tackled via applying basic supports, adding dopants/modifiers, controlling the metal loading, or tuning preparation method and thermal pretreatment. [5,10] Another serious problem is the aggregation of Ni species to larger but less active particles.
Several main steps rule the DRM reaction: [11] (1) Dissociative adsorption of CH 4 and CO 2 , (2) first desorption of CO and H 2 , (3) formation of surface hydroxyls and oxygen spill-over and (4) surface hydroxyls and oxygen oxidize CH x species and second desorption of CO and H 2 . The adsorbed oxygen species from CO 2 dissociation react with carbonaceous intermediates (Table S. 1) which otherwise might form carbon. Consequently, we chose several approaches to enhance the activation of CO 2 at high Ni metal dispersion to suppress methane decomposition.
In our previous study on stoichiometric DRM (CH 4 /CO 2 = 1), MgOÀ Al 2 O 3 supported Ni catalyst modified with La and citric acid (CA) (denoted as La.Ni(CA)/Mg 1.3 AlO x ) turned out to be most effective. [12] However, the coking behavior of this sample in CH 4 -rich DRM is still unsatisfactory and limits process development. In this study, the thermal pre-treatment of the Mg-rich MgÀ Al hydrotalcite support precursor was extended from 550 to 1000°C. With this modification, the nature of the support was significantly changed as discrete oxide phases were formed, leading to improved interactions with the subsequently added Ni during DRM. The scheme of the preparation route is shown in Figure S. 1. For comparison, two catalysts were prepared on supports which were initially treated at 550°C. Pathways for carbon formation were studied in order to understand the relationship between catalyst properties and coking resistance.
XRD pattern of the support Mg 1.3 AlO x .1000 (prepared by pre-treating Mg 1.3 AlO x at 1000°C) shows sharp reflections of periclase (the cubic form of MgO, ICDD file No. 01-071-1176) [13] and MgAl 2 O 4 (ICDD file No. 00-021-1152) [14] crystalline phases ( Figure 1a). The corresponding catalysts Ni/Mg 1.3 AlO x .1000 and La.Ni(CA)/Mg 1.3 AlO x .1000 display patterns close to that of the support, suggesting the formation of finely dispersed La 3 + and Ni 2 + species that are not detectable by XRD. [15] Moreover, magnification of the pattern for La.Ni(CA)/Mg 1.3 AlO x .1000 in the 2θ range from 61°to 64°exposes a slight shift of the reflection at 62.3°to higher 2θ values compared to support ( Figure S. 2). This shift suggests the presence of NiOÀ MgO solid solution in La.Ni(CA)/Mg 1.3 AlO x .1000 formed from the Ni precursor and the preformed small MgO particles of 8-9 nm crystallite size. [16] In contrast, in Mg 1.3 AlO x and corresponding catalysts the crystalline MgO (periclase) did not form. Instead, existence of MgOÀ Al 2 O 3 solid solution is suggested which was previously discussed. [12] Regarding the chemical surface states, the Ni 2p 3/2 binding energy ( Figure S. 3) from the XPS measurement of Ni/Mg 1.3 AlO x (856.9 eV) indicates Ni 2 + located at the surface as NiAl 2 O 4 spinel (856.8 eV). [17] The Ni 2p 3/2 binding energy of Ni/Mg 1.3 AlO x .1000 (856.1 eV) is close to that of Ni surface species in NiOÀ MgO solutions (855.7-856.0 eV). [18][19] However, no Ni 2p 3/2 signal was found in Ni/MgO with same Ni load, probably due to the migration of NiO into bulk MgO at high calcination temperature. [20] This confirms the unique structure of Mg 1.3 AlO x .1000, which can stabilize Ni in NiOÀ MgO solid solution but still offers surface Ni species that are beneficial for the catalytic activity.
Compared to pure NiO which can be significantly reduced at 400°C, [12] all supported Ni samples display poorer reducibility in H 2 -TPR experiments (Figure 1b) as the main reduction peaks appear first above 600°C (β and γ peaks). This behavior is assumed to be caused by the strong metal-support interactions (MSI) at low Ni loading allowing Ni 2 + to disperse easily into stable structures. [21][22] Ni/Mg 1.3 AlO x .1000 discloses poorer overall reducibility compared to Ni/Mg 1.3 AlO x , illustrated by a higher percentage of H 2 uptake in γ peak (Table S. 2) and lower total H 2 consumption (Table 1).
La.Ni(CA)/Mg 1.3 AlO x .1000 consumes less H 2 in TPR than La.Ni (CA)/Mg 1.3 AlO x , but more than Ni/Mg 1.3 AlO x .1000 (Table 1) and even more than the theoretical value for complete Ni 2 + reduction (428 μmol). The latter can be traced back to enhanced oxygen activation from distorted Ni surface structures that will be discussed below in the UV-Vis section. Besides, it should be noted that CA-assisted preparation increases the Ni 2 + dispersion. [23] However, similar to La.Ni(CA)/Mg 1.3 AlO x , sample La.Ni(CA)/Mg 1.3 AlO x .1000 exposes a reduction peak below 600°C (α peak) but with higher intensity (Figure 1b, Table S. 2). This effect is attributed to the presence of La that promotes the oxygen mobility and/or formation of defect structures where MgO is doped with Ni 2 + . [12,24] On the other side,sample La.Ni (CA)/Mg 1.3 AlO x .1000 exposes the β peak at a temperature similar to that of La.Ni(CA)/Mg 1.3 AlO x and an additional split γ peak at very high temperature (980°C) assigned to Ni in strong interaction with the support . [25] These β and γ Ni 2 + species are poorly reduced in usual pretreatment and therefore La.Ni(CA)/Mg 1.3 AlO x .1000 is only partially reduced before DRM. The Ni metal fraction after in situ reduction was evaluated in a two-step experiment: first, the reduction at 700°C was made, directly followed by a regular TPR experiment up to 1000°C. This revealed a H 2 uptake of 292 μmol/g, corresponding to 59 % of the value for the fresh  The UV-Vis DR spectra in the region of 200-350 nm (Figure S. 5) give hints on the Ni 2 + ligand-to-metal charge transfer (LMCT) bands. [26] Ni/Mg 1.3 AlO x .1000 discloses a LMCT band with weaker intensity and blue shift compared to that of Ni/Mg 1.3 AlO x . This reflects a higher Ni 2 + dispersion, probably correlated with the stronger MSI of Ni/Mg 1.3 AlO x .1000. [27] Besides, modifying this material with La and CA-assisted synthesis shows further blue-shift of the UV band, indicating the formation of highly dispersed Ni 2 + in La.Ni(CA)/Mg 1.3 AlO x .1000.
The UV-Vis-DR spectra in the region 350-800 nm (Figures 1c and S. 6) reveal the coordination of Ni 2 + . [15,17,28] The Ni catalysts expose mainly absorption bands at 400 nm and 660 nm, which are associated with ν 3 ( 3 A 2g ! 3 T 1g ) and ν 2 ( 3 A 2g ! 3 T 1g ) absorptions caused by Ni 2 + species in octahedral coordination (Oh), similarly to that of Ni/MgO spectra, instead of tetrahedral (Th) geometries. Such lack of Ni 2 + (Th) species reflects the presence of surplus MgO that stabilizes the Ni 2 + species and suppresses the formation of NiAl 2 O 4 spinel. [29][30][31] The XRD pattern of La.Ni(CA)/Mg 1.3 AlO x .1000 proves the presence of the periclase (MgO) structure, suggesting formation of NiOÀ MgO solid solution (Figures 1a and S. 2). Hence, the shoulder at 400-600 nm in the UV-Vis-DR spectrum recorded for La.Ni(CA)/Mg 1.3 AlO x .1000 suggests local defect structures in NiOÀ MgO solid solution [28] that can explain the α reduction peak at 400°C (Figure 1b). [21,24,32] A similar shoulder is found in the UV-Vis spectra for Ni/MgO (Figure 1c The CH 4 -rich DRM tests show that the thermal pre-treatment of the Mg 1.3 AlO x support at 1000°C remarkably improves the stability and coking resistance of the corresponding Ni catalysts in CH 4 -rich DRM (Figures 3 and 4a). The measured conversions were below calculated equilibrium data (Figures S. 11 and S. 12). La.Ni(CA)/Mg 1.3 AlO x .1000 showed improved resistance against deactivation ( Figure 3) and carbon accumulation (Figure 4a), which can be assigned to its reducibility ( Figure 1b) and finely dispersed Ni atoms maintained during the reaction (Figures S. 9b-2, S. 9c-2, S. 13). In contrast, Ni agglomeration occurred with Ni/Mg 1.3 AlO x already after the pre-reduction step and even more seriously during DRM (Figures S. 9b-1 and c-1) and led to bigger Ni particles that may cause serious deactivation and coking. The H 2 /CO ratios reach unity, especially at 750°C, reflecting a low contribution of reverse water gas shift reaction which otherwise would deteriorate the H 2 yield at high temperature with CH 4 -rich feed (Figure S. 14).
At 750°C, La.Ni(CA)/Mg 1.3 AlO x .1000, as the best catalyst among samples with this support (Figure S. 14a-c), performs similarly to La.Ni(CA)/Mg 1.3 AlO x . However, detected carbon amounts on spent samples of both catalysts after CH 4 -rich DRM differ significantly from each other.
The coking pathways were studied in a series of runs with catalyst Ni/Mg 1.3 AlO x at different temperatures (500-750°C) using feeds composed of CH 4 /Ar = 1 or 2 (both without CO 2 ) as well as CH 4 /CO 2 = 1 or 2 at a GHSV of 170 L/(g cat × h) (Figure 4b).
The carbon deposits on spent samples were analyzed after 8 h on stream. At 500°C, the carbon contents on all spent catalysts were negligible, as methane decomposition (MD) and DRM (producing CO as a reactant for Boudouard reaction (BD)) run at higher temperatures. In all tests with Ni/Mg 1.3 AlO x and CH 4 /Ar, raising the temperature and CH 4 concentration caused a proportional rise in carbon deposition, reflecting the impact of methane decomposition and/or metal agglomeration. [34][35] When CO 2 was converted in DRM at 630°C with CH 4 at any portion, the deposition was significantly higher, indicating the extent of BD reaction via CO disproportionation, which outnumbered MD contribution. However, at 750°C, the contribution of BD decreased. [36] Indeed, when Ar is replaced by CO 2 , the carbon deposition changes only slightly in case of CH 4 /CO 2 = 2 but decreases dramatically at CH 4 /CO 2 = 1 (higher CO 2 partial pressure), adapting to the thermodynamically favorable gasification (reversed BD) of C by CO 2 at high temperature. [34,37] These data suggest that La.Ni(CA)/Mg 1.3 AlO x has a structure modification effect that can slow down BD reaction, lowering the coking rate at 630°C in CH 4 -rich DRM (Figure 4a). [12] Nevertheless, at 750°C carbon mainly formed by MD reaction is still observed on all spent Mg 1.3 AlO x -supported Ni catalysts due to the lower concentration of CO 2 and low efficiency in activation of CO 2 . However, such carbon deposition can also be suppressed by applying Mg 1.3 AlO x .1000 supported Ni catalysts ( Figure S. 15).
Further coking tests with CH 4 /Ar = 2 (no CO 2 ) at 750°C (Figure 4c) elucidate the specific influence of MD on carbon deposition. While Mg 1.3 AlO x supported catalysts show similar or higher carbon contents after CH 4 -rich DRM (CH 4 /CO 2 = 2) than after MD (CH 4 /Ar = 2), Ni/Mg 1.3 AlO x .1000 and especially La.Ni(CA)/Mg 1.3 AlO x .1000 form lower amounts of carbon in CH 4rich DRM compared to MD. This contrast points to the high potential of the latter catalysts in CO 2 activation, even at low partial pressure, which cannot be achieved with Ni catalysts supported on Mg 1.3 AlO x . The CO 2 activation, in this case,    [38][39][40] In the first step of the reaction, this structure is supposed to accelerate the CO 2 dissociation at the metalsupport interface as well as the CH 4 cracking on Ni atoms, [41][42] thereby releasing the first CO and H 2 molecules. [11] With La 2 O 3 / La 2 O 2 CO 3 enriched around these metal atoms (Figure 2), defective sites (e. g. oxygen vacancies) are generated (Figure 1c), promoting dissociative CO 2 adsorption which subsequently forms surface oxygen species. [41,[43][44][45][46] The increased availability of these oxygen species transforms carbonaceous intermediates from CH 4 -rich mixture and releases second CO and H 2 (Table S. 1). This behavior enhances both the reaction rate and the carbon resistance of the catalyst even at low CO 2 partial pressure. Such solid solution was recently also supposed to be beneficial for CH 4 -rich DRM by suppressing carbon formation. [47] However, concerning the loading of Ni and the catalyst productivity in that investigation (Table S. 3), the catalyst in the present study is superior. It should be noted that the mentioned NiOÀ MgO interaction is expected in Ni/MgO as well. However, due to low Ni loading, this sample showed almost no activity due to low Ni surface concentration.
La.Ni(CA)/Mg 1.3 AlO x .1000 was employed in a long-term test over 100 h at typical conditions to evaluate the application potential ( Figure 5). In known literature, [9] a long-term CH 4 -rich DRM run with such high feed rate was not reported.
Complete avoidance of carbon accumulation at high CH 4 / CO 2 ratio with Ni catalysts was previously considered infeasible. [48] However, over 100 h of CH 4 -rich DRM in the present study, the H 2 /CO ratio remained constantly near unity whereas the conversions slightly decreased, but low carbon amount (~5 wt %) was found on the spent sample. This fraction was predictably higher than the values in DRM tests over 8 h, but not proportional to total run time. Interestingly, while carbon accumulation on spent La.Ni(CA)/Mg 1.3 AlO x .1000 was observed in STEM annular bright field (ABF) image after 8 h on stream, carbon was hardly found on the spent sample after 100 h ( Figure S. 13), highlighting the exceptionally stable coking resistance due to gasification. This behavior is in accordance with the stable dispersion of small Ni particles (5-10 nm) which are also partially attached to the support ( Figure S. 13). Besides, STEM-HAADF images prove that the mentioned preferred localization of Ni in the MgO-enriched structures is preserved during the reaction (Figures S. 16 and S. 17). These factors are crucial for both carbon removal by CO 2 and stably high DRM performance with high H 2 yield of La.Ni(CA)/Mg 1.3 AlO x .1000. In terms of active metal price and loading, productivity and stability against coking, La.Ni(CA)/Mg 1.3 AlO x .1000 is one of the most promising candidates for DRM under CH 4 -rich conditions (Table S. 3). [47,[49][50][51] We conclude that catalysts supported on Mg 1.3 AlO x .1000 possess improved coking resistance without losing the DRM activity and are therefore suitable for the reaction with CH 4 -rich feed. Citric acid induces a high dispersion already during the catalyst preparation. The NiOÀ MgO solid solution domains excellently stabilize small Ni particles throughout all catalyst pre-treatment steps and DRM. Highly dispersed Ni activates CO 2 as an oxidant for carbon gasification, thereby reducing the coking rate in CH 4 -rich DRM even at low CO 2 partial pressure. La generates additional oxygen vacancies that help to activate CO 2 as well. La.Ni(CA)/Mg 1.3 AlO x .1000 appears to be the best catalyst as it has high and stable activity over at least 8 h on stream and its coking rate is lowest at both 630 and 750°C. Moreover, this catalyst exposes quite stable activity in CH 4 -rich DRM over 100 h on stream with little coking. Such exceptional performance is certainly ruled by high Ni dispersion and enhanced reducibility.

Experimental Section
MgÀ Al mixed oxide supports were prepared from MgÀ Al hydrotalcite (Pural MG50, Sasol). The default precursor Mg 1.3 AlO x was obtained by calcining the MgÀ Al hydrotalcite at 550°C. This material was thermally pre-treated at 1000°C with a rate of 2 K/min to prepare Mg 1.3 AlO x .1000 support.
In order to prepare the final catalysts, both supports were treated with Ni(NO 3 ) 2 · 6H 2 O (99 %, Alfa Aesar) and La(NO 3 ) 3 · 6H 2 O (99 %, ABCR GmbH) by wet impregnation (nominal Ni content 2.5 wt %). Citric acid (> 99 %, Alfa Aesar) was added simultaneously in some cases. The molar ratio of La and Ni was set to 0.8, and the CA/metal molar ratio was fixed at 1.5. The calculated amounts of Ni, La precursors and CA were dissolved in deionized water and the solution was stirred for 4 h at 50°C. The MgÀ Al supports were then added and the slurry was stirred at 60°C for 15 h. Water was gradually removed by a rotary evaporator for 4 h and the samples were dried overnight at 120°C and calcined at 400°C for 3 h and then at 800°C for 6 h both in air with a rate of 2 K/min. MgO (FLUKA) as well as its corresponding Ni-loaded samples La.Ni(CA)/MgO and Ni/MgO served as reference materials. Pure NiO was prepared by calcining Ni(NO 3 ) 2 · 6H 2 O at 800°C.
XRD powder patterns were recorded on a Panalytical X'Pert diffractometer equipped with a Xcelerator detector using automatic divergence slits and Cu Kα1/α2 radiation (40 kV, 40 mA; λ = 0.15406 nm, 0.154443 nm). Cu beta-radiation was excluded using a nickel filter foil. The samples were mounted on silicon zero background holders. The obtained intensities were converted from automatic to fixed divergence slits (0.25°) for further analysis. Peak positions and profile were fitted with Pseudo-Voigt function using the HighScore Plus software package (Panalytical). Phase identification was done by using the PDF-2 database of the International Center of Diffraction Data (ICDD).
The low-temperature N 2 adsorption was performed on a Micromeritics ASAP 2010 apparatus at À 196°C. The samples were degassed at 200°C in vacuum for 4 h before the analysis. The carbon deposition on spent catalysts was analyzed using a TruSpec Micro CHNS analyzer (LECO Corporation). Up to 10 mg of the investigated sample were catalytically burned with oxygen in a helium stream at 1100°C. The resulting gas was analyzed with an infrared detector and a thermal conductivity detector. H 2 -TPR experiments were performed with a Micromeritics Autochem II 2920 instrument with a thermal conductivity detector. A 300 mg sample was loaded into a U-shaped quartz reactor and heated in 5 %O 2 /He (50 ml/min; r.t. to 400°C with 20 K/min, 30 min hold, then cooled to r.t. in Ar flow). TPR run was made up to 1000°C in 5 %H 2 /Ar (50 ml/min; 10 K/min, final hold 30 min before cooling to r.t.). A TPR for pre-reduced sample was also conducted. After oxidation with 5 %O 2 /He, the sample was pre-reduced in pure H 2 (50 mL/min) at 700°C for 1.5 h to imitate the pretreatment in DRM setup. Then the system was cooled to r.t. and a regular TPR experiment was appended. The H 2 consumption peaks were recorded using a thermal conductivity detector.
XPS measurements were carried out with an ESCALAB 220iXL instrument (Thermo Fisher Scientific) with monochromatic Al Kα radiation (E = 1486.6 eV). Samples were prepared on a stainlesssteel holder with conductive double sided adhesive carbon tape. The electron binding energies were obtained with charge compensation using a flood electron source and referenced to the C1s peak of adventitious carbon at 284.8 eV (CÀ C and CÀ H bonds). The peaks were deconvoluted with Gaussian-Lorentzian curves using the software Unifit.
UV-Vis-DR spectra were measured over 200-800 nm using a Cary 5000 spectrometer (Varian) equipped with a diffuse reflectance accessory (praying mantis, Harrick). BaSO 4 was used as a white reference standard and diluted material was used for the measurement with pure NiO because of its high Ni content. STEM measurements were performed at 200 kV with an aberrationcorrected JEM-ARM200F (JEOL, Corrector: CEOS). The microscope is equipped with a JED-2300 (JEOL) energy-dispersive X-ray-spectrometer (EDXS) for chemical analysis. The aberration-corrected STEM imaging (High-Angle Annular Dark Field (HAADF) and Annular Bright Field (ABF)) was performed under the following conditions: HAADF and ABF both were done with a spot size of approximately 0.13 nm, a convergence angle of 30-36°and collecting at semiangles for HAADF and ABF of 90-170 mrad and 11-22 mrad, respectively. The sample was dry deposited without any pretreatment on a holey carbon film supported by a Cu-grid (300 mesh) and transferred to the microscope.