Pivotal Role of Ni/ZrO₂ Phase Boundaries for Coke-Resistant Methane Dry Reforming Catalysts

Abstract

To identify the synergistic action of differently prepared Ni-ZrO₂ phase boundaries in methane dry reforming, we compared an “inverse” near-surface intermetallic NiZr catalyst precursor with the respective bulk-intermetallic Ni_xZr_y material and a supported Ni-ZrO₂ catalyst. In all three cases, stable and high methane dry reforming activity with enhanced anticoking properties can be assigned to the presence of extended Ni-ZrO₂ phase boundaries, which result from in situ activation of the intermetallic Ni-Zr model catalyst systems under DRM conditions. All three catalysts operate bifunctionally; methane is essentially decomposed to carbon at the metallic Ni⁰ surface sites, whereas CO₂ reacts to CO at reduced Zr centers induced by a spillover of carbon to the phase boundaries. On pure bulk Ni⁰, dissolved carbon accumulates in surface-near regions, leading to a sufficiently supersaturated state for completely surface-blocking graphitic carbon segregation. In strong contrast, surface-ZrO₂ modified bulk Ni⁰ exhibits virtually the best decoking and carbon conversion conditions due to the presence of highly dispersed ZrO₂ islands with a particularly large contribution of interfacial Ni⁰-ZrO₂ sites and short C-diffusion pathways to the latter.

1. Introduction

Carbon capture and utilization (CCU) approaches represent active and sustainable routes to counteract the increasing carbon dioxide concentration in our atmosphere by capturing, converting, and recycling CO₂ back to fuels and valuable chemicals. Among these CCU concepts, the CO₂ reforming or dry reforming of methane (DRM) appears as a promising technology, as it allows the abatement of harmful anthropogenically released greenhouse gases by simultaneously consuming carbon dioxide and methane to produce synthesis gas, i.e., a mixture of H₂ and CO [1,2,3]. Synthesis gas, especially if produced from recycled CO₂ and non-fossil methane sources, offers the opportunity to furnish a broad range of both renewable and environmentally clean fuels and chemicals. Ideally, the DRM reaction yields a stoichiometric H₂/CO ratio (syngas ratio) close to unity. The rather low hydrogen-to-carbon monoxide ratio is desirable for a variety of industrial processes, e.g., carbonylation and hydroformylation of olefins or the production of oxygenate chemicals such as acetic acid, dimethyl ether, and oxo-alcohols [4,5,6]. In combination with steam reforming of methane (SRM) and/or the water-gas shift (WGS) reaction, or also by using membrane reactors, an increased H₂/CO ratio can be obtained for further application as a feedstock for the synthesis of renewable fuels such as gasoline, diesel or jet fuel by Fischer–Tropsch chemistry [6,7,8,9]. However, for industrially relevant applications and the economic feasibility of DRM, not only elevated temperatures (700–800 °C) but also high pressures (20–40 bar) are required. With this requirement, several technological challenges are aggravated, such as catalyst particle sintering, decreased H₂ selectivity due to the water-gas shift equilibrium, and catalyst deactivation by the extensive formation of carbon deposits [2,10,11]. Above all application-oriented obstacles, the deposition of unreactive carbonaceous species at the active catalyst sites, commonly denoted as coking phenomena, display the major impediment of methane dry reforming, especially when less costly non-noble metal-based catalysts are intended to be used.

It is known that for DRM a series of highly active catalyst systems based on noble metals such as Pt, Pd, or Rh exist, exhibiting superior anticoking properties and stability [4,12,13,14]. From a fundamental viewpoint, these materials are of high interest as the findings on those systems can help to elucidate mechanistic insights and help to transfer the particular coking resilience to less costly, non-noble metal catalysts. In an in situ near-ambient pressure (NAP)-XPS study of our group, the anticoking stability and carbon-converting role of the Pd/ZrO₂ phase boundary (in the following denoted as PB) were highlighted [15]. The results showed that carbon-saturated Pd⁰ nanoparticles evolve from an in situ activated PdZr intermetallic precatalyst, leading to the formation of extended C-saturated Pd-metal/Zr-oxide phase boundaries under methane dry reforming conditions. The generation of these PBs results in a highly active state fostering a bifunctional mechanism by effectively activating CO₂ at the oxidic PB sites and CH₄ at the Pd⁰ particles forming Pd_xC, which provides a fast supply of carbon atoms toward the PB for enhanced CO evolution [15,16]. However, even if noble metals such as Pd do actually exhibit superior properties with respect to DRM activity and stability, as mentioned above, they will inevitably remain little attractive for an industrial application simply due to their high costs. It is, therefore, fundamental to develop catalysts based on less costly materials such as Ni or Co, which might also allow for similar metal/metal-oxide bifunctional synergisms as proposed for the noble metal systems.

Regarding the intrinsic catalytic role of Ni⁰ nanoparticles with respect to carbon formation and conversion, it has been shown that certain surface carbon species produced during methane dry reforming on MnO-supported Ni⁰ can act as reactive intermediates instead of only deactivating the catalyst [17,18]. In order to extract an additional co-catalytic contribution of PB sites, we investigated bulk Ni_xZr_y bimetallic catalysts [19], focusing on the formation of extended Ni/ZrO₂ metal-oxide interfaces under DRM conditions and the associated mechanistic details of their carbon chemistry. In particular, coking phenomena in the vicinity of the material-specific Ni/ZrO₂ phase boundaries were scrutinized. We could identify a C_graphite to C_carbide redistribution mechanism at elevated temperatures (700–800 °C) leading to the local carbothermal reduction of ZrO₂ to ZrC. These carbidic zirconium species showed promoted activity towards CO formation with CO₂ and rendered the NiZr-based catalyst highly active under realistic DRM conditions, along with a strongly enhanced anticoking stability [19]

Nickel is interesting not only due to the remarkably lower costs compared to Pd but because it is, in principle, capable of activating both CH₄ and CO₂, which might be particularly attractive if supports are used on which CO₂ activation is not effective [20,21]. However, it appears reasonable to use ZrO₂ as a support as it has been shown to be capable of CO₂ activation [22]. Hence, by combining an already efficient CH₄ activator such as Ni with a support material being capable of activating CO₂, we aimed to trigger an additional phase boundary-related contribution to bifunctionality, as in this case, CO₂ activation and subsequent CO formation are not necessarily limited to the active metal surface itself, but can also occur at the Ni⁰-support interface. Provided that this PB contribution is substantial, empirical development of highly active and stable DRM catalysts should aim at the generation of extended (bi-)metal-oxide interfaces exhibiting superior CH₄ activation capabilities at the metallic sites and enhanced CO₂ activation properties of the oxide support/component.

In the present study, we introduce a new “inverse” system in comparison to the established supported and bulk-intermetallic Ni/ZrO₂ systems, following a novel CVD preparation procedure to engineer the potentially active Ni-ZrO₂ interface in the form of ZrO₂ islands on top of a Ni bulk metal support. Structurally, the resulting “inverse” model catalyst is, therefore, very different from both reference systems. Generally, deposition of oxide islands/particles on top of an active metallic substrate results in structurally “mirrored” systems as compared to conventional catalysts where the active metal is positioned on top of oxidic or other non-metallic substrates or supports. Therefore, our CVD-prepared catalyst based on a Ni foil substrate is denoted as an “inverse NiZr model catalyst” in the following. Usually, conventional oxide powder-supported catalysts do not allow for pinpointing exact local information about the active metal phase and its specific catalytic role due to their structural and morphological heterogeneity. This limits the applicability of surface-sensitive spectroscopic techniques, such as (in situ) near-ambient X-ray photoelectron spectroscopy (NAP)-XPS, but can be overcome by employing model catalyst systems which are particularly suitable for this method. The in situ activation of conductive intermetallic catalyst precursors can provide a quasi-2D region of an active catalyst state via near-surface decomposition, which allows to control conductivity issues and to provide the premises for spectral observation of mechanistic processes taking place at the surface during DRM. Thus, a near-surface intermetallic model catalyst precursor approach, allowing for controlled in situ formation of the aimed “inverse” Ni⁰/ZrO₂ phase boundaries under DRM conditions, was employed. The details of preparation and characterization of the respective pre-DRM states of all three compared NiZr systems—in the following denoted as the NiZr inverse model, NiZr51 bulk intermetallic and Ni10Zr90 supported catalyst, along with a pure Ni foil and ZrO₂ reference sample—are provided in Section 3, which also contains the detailed description of the used experimental Setups 1–5 and of the associated methods.

In due course, the present study focuses on the detailed comparison of the novel “inverse” CVD-prepared NiZr model catalyst with the previously introduced bulk-intermetallic and supported systems [19], using a combination of spectroscopic and structure-determining methods (ex and in situ XPS, scanning electron microscopy imaging, and spectroscopy), and exploiting additionally measured and so far unpublished supporting information data of the latter. Being able to analogously probe the novel “inverse” model catalyst resulting from oxidative in situ decomposition of the CVD-prepared near-surface intermetallic precursor state gives us the unique opportunity to impart the knowledge obtained on a new type of model system to industrially applicable “real” catalyst systems and therefore marks a fundamental step in further attempts of bridging the so-called “materials gap”.

2. Results and Discussion

2.1. Pre-DRM Characterization of the NiZr Inverse Intermetallic, NiZr51 Bulk-Intermetallic, and Ni10Zr90 Supported Powder Catalyst

The upper panels in Figure 1 show the ex situ XPS characterization of the surface of the NiZr inverse intermetallic sample directly after the UHV-based thermal annealing treatment and before introducing it into the batch reactor of Setup 1 for DRM catalysis. The more oxidized state directly after CVD and the changes in the surface Zr redox states induced by in vacuo heating are represented in more detail in Figure S1, Supplementary Information. The Zr3d spectrum of the annealed state qualitatively resembles one of the NiZr51 samples in its pre-DRM state (middle panels, Figure 1), showing a mixture of ZrO₂, reduced ZrO_x components, as well as metallic Zr⁰. No evidence of carbonaceous species was found, and the Ni surface appears 100% metallic.

In due course, the bulk-structural characterization of the initial NiZr51 bulk-intermetallic sample performed in Setup 3 before DRM, as well as its structural phase evolution under DRM conditions, can be derived from Figure S2, panel A (Supplementary Information). The initial bulk phase composition was derived from the room temperature diffractogram in Figure S2, panel B using Rietveld analysis. The precatalyst, therefore, contains 84.5 wt.-% Ni₅Zr and 15.5 wt.-% Ni⁰, resulting in an XRD-detectable atomic ratio of Ni:Zr of ~6.2:1. The deviation from the nominal 4:1 ratio can be rationalized by amorphous Zr-rich intermetallic phases in the catalyst [23,24,25]. A certain contribution of the latter in the phase mixture can be expected considering the experimental quenching process with cooling rates between 10² and 10³ K/s.

The middle panels in Figure 1 show the ex situ surface characterization of the NiZr51 catalyst by XPS (Zr3d, Ni2p, and O1s regions) before DRM and after a sputter/anneal cycle inside the UHV chamber of Setup 1. It can be seen that thermal annealing of the sample under UHV conditions (<5 × 10⁻¹⁰ mbar) leads to the formation of metallic Zr⁰, as well as suboxidic ZrO_x species similar to previously reported species [16,26,27]. Before the transfer of the sample to the batch reactor for the first DRM cycle, the surface shows no traces of carbon (spectra therefore omitted) but Ni exclusively in its metallic state. Furthermore, the signals in the O1s spectrum confirm the presence of zirconia and the associated suboxidic ZrO_x components. The spectra in the lower panels display the pre-DRM state of the Ni10Zr90-supported powder catalyst. After wet impregnation, the powder was calcined and pre-reduced under an H₂ atmosphere leading to ZrO₂ and metallic Ni at the sample surface. Minute traces of suboxidic Zr components, but no carbonaceous species could be detected.

The surface morphology of the three pre-DRM samples was characterized by ex situ scanning electron microscopy (Setup 5), as depicted in Figure 2. Referring to the CVD-prepared NiZr inverse intermetallic sample, it features an uncorrugated, almost structure-less topography (Figure 2A). A rather even element distribution in the EDX images (Figure 2, Panels B and C) suggests that the CVD process leads to the formation of a homogeneously surface-covering ultrathin layer consisting of alloyed NiZr and suboxidic ZrO_x (derived from the XP spectra) on top of the flat Ni foil substrate. Figure 2D shows the NiZr51 sample after the pre-DRM annealing treatment and transfer to the SEM through ambient air. The presence of isolated Ni particles on the Ni₅Zr matrix is supported by the ex situ XRD measurements, which also reveal a low amount of crystalline Ni particles (alongside Ni₅Zr) in the initial sample before DRM. The SEM image in Figure 2, panel E depicts the Ni10Zr90 powder catalyst in its hydrogen-reduced pre-DRM state. According to EDX (not shown), the relatively small amount of deposited Ni (10 at%, corresponding to ~5 wt%) leads to the decoration of the larger polycrystalline ZrO₂ grains with small Ni particles, which appear homogeneously distributed and exhibit a size distribution between approx. 5 and 10 nm.

2.2. Comparison of Catalytic DRM Performance of the NiZr Inverse Intermetallic, NiZr51 Bulk-Intermetallic, and Ni10Zr90 Supported Powder Catalyst

Figure 3 comparatively depicts the temperature-programmed CO₂ conversion profiles of the differently prepared catalyst systems and an ultraclean, Zr-free Ni foil reference. As described in the experimental section, the samples were exposed to a 1:1 mixture of 50 mbar or 15 mbar CO₂ and CH₄ inside the batch reactor cells of Setup 1 and Setup 4, respectively. A linear heating ramp of 10 K min⁻¹ to 800 °C was followed by an isothermal period of 30 min at 800 °C. Regardless of distinct features of the CO₂ conversion profiles, such as variable onset temperatures and profile shapes, it is obvious that all samples exhibit a high initial DRM activity in the first DRM cycle. Starting with the lowest onset temperature, the NiZr51 bulk intermetallic sample shows an exponential increase in CO₂ conversion starting at ~480 °C and enables almost full conversion during the isothermal period at 800 °C. The ratio of CO₂ consumption to CO formation is close to 1:2, and selectivity-wise, the CO:H₂ product ratio remains close to unity. Despite a slightly higher onset temperature, the qualitative performance of the Ni10Zr90-supported powder catalyst closely resembles that of the NiZr51 model catalyst. An exponential increase starting at ~500 °C leads to almost complete CO₂ conversion at 800 °C in the first DRM cycle.

In contrast to the exponential reaction onset on NiZr51 and Ni10Zr90, the pure Ni foil as well as the CVD-prepared NiZr inverse intermetallic sample show a rather spontaneous, step-like activation. Considering the latter, this sample also exhibits a remarkably high DRM activity, as observed for the green CO₂ conversion profile in Figure 3. Starting with a relatively sharp onset at approx. 590 °C, the CO₂ conversion reaches almost 95% during the isothermal period. The sharp, step-wise onset is associated with the sudden breakdown of a passivating NiO layer that is formed on the catalyst surface under DRM conditions between room temperature and approx. 590 °C. The respective quenching experiments and XPS data shown in Figure S3, Supplementary Information, reveal the presence of oxidic Ni at the surface directly before the sharp activity onset, whereas only metallic Ni prevails after the DRM reaction sets in. A similar behavior (c.f. Figure S4, Supplementary Information) can be found for the pure Ni foil reference with a rapid reaction onset at around 640 °C but a lower maximum CO₂ conversion in the isothermal period compared to the other systems.

The second-cycle CO₂ conversion profiles observed on the spent samples are depicted in Figure 4, using otherwise identical experimental parameters. Starting again with NiZr51, a high DRM activity can also be observed for the second cycle, approaching >95% conversion during the isothermal period at 800 °C. Although the 1st and 2nd cycles exhibit similar shapes and maximum conversion values, the temperature of the exponential onset of the second cycle is shifted to ~600 °C. The difference of ~120 °C compared to the first run shown in Figure 3 can be rationalized by pronounced Ni sintering and coking after the 1st cycle, as will be shown in the subsequent Section 2.3. Consequently, we suggest that the in situ activation of NiZr51 proceeds via initially uncoked, nanodispersed Ni⁰ particles with sizes around 10 nm in contact with ZrO₂ between ~300 and ~500 °C (compare Figure 2D), which are held responsible for the earlier conversion onset during the first cycle in Figure 3. The situation before the 2nd cycle, involving pronounced sintering and coking of the segregated Ni⁰, can explain that initial CH₄ and CO₂ activation become kinetically more hindered, resulting in higher DRM onset temperatures. Regarding the Ni10Zr90 powder catalyst, a reproducibly high 2nd cycle DRM activity is observed. The catalytic activity starts at around 600 °C, too, resulting in an almost complete CO₂ conversion of approx. 98% during the isothermal period at 800 °C. For this sample, a similar upshift of the onset temperature of nearly 100 °C can be found when comparing the 1st and the 2nd DRM cycles. Again, the accumulation of carbonaceous deposits and pronounced sintering effects of the originally highly dispersed Ni⁰ nanoparticles (Figure 2E) at realistic dry reforming temperatures around 800 °C are held responsible, as will be shown in more detail in the following Section 2.3. For the NiZr inverse intermetallic sample, only a negligible onset temperature shift can be observed. The catalyst repeats its rapid increase in CO₂ conversion at around 600 °C of the first DRM cycle but exhibits a slightly lowered catalytic activity at 800 °C, as the maximum conversion only reaches 83% in the second DRM run. What appears most remarkable when comparing the CO₂ conversion profiles of Figure 3 and Figure 4 is that basically all NiZr-based systems do exhibit a reproducibly high DRM activity at high temperatures, whereas the pure Ni reference catalyst remains largely deactivated after the first DRM run, as indicated by the orange conversion trace in Figure 4. As will be shown in the following, the Ni domain size/particle size is the most critical factor with respect to surface carbon reactivity and the stability of the DRM performance at high temperatures. What sets the (structurally very much different) model systems apart in terms of low-temperature performance is that the Ni-foil-based, CVD-prepared “inverse” NiZr sample shows a reproducibly different onset behavior compared to the others due to the intermediate passivation of the bulk Ni substrate by a NiO layer. This explains the reproducibility of the step-like CO₂ conversion onset at ~650 °C, which could not be observed on the Ni particle-containing samples. The meso- and nanosized Ni particles/domains are obviously not affected by intermediate passivation.

2.3. Ex Situ Characterization of the NiZr Inverse Intermetallic, NiZr51 Bulk-Intermetallic, and Ni10Zr90 Supported Powder Catalyst after DRM

After the respective 1st cycle DRM experiments, the samples were characterized with regard to carbon chemistry, surface chemical composition and morphology. Figure 5 shows the respective Zr3d, Ni2p, O1s, and C1s regions of the catalysts in their post-DRM states.

The upper panels represent the respective XP spectra of the CVD-prepared NiZr inverse intermetallic sample. After the first DRM cycle, the reduced zirconium species are completely oxidized to ZrO₂ with no indications of hydroxylated components, while Ni remains unaltered in its metallic state. Interestingly, a high dry reforming activity (comparable to the one of the first DRM cycle on the NiZr51 sample is observed, and the catalyst seems to be largely protected against coking under reaction conditions. The C1s spectra show only minor traces of surface-deposited carbon, and the relative C1s peak area is very small in comparison to the spent NiZr51 shown below. Due to the small amount of deposited carbon, hardly any attenuation of the Ni2p signal is detectable.

The middle panels depict the NiZr51 bulk intermetallic sample after the first DRM cycle. It is obvious that the reduced suboxidic ZrO_x and intermetallic Zr⁰ species become fully oxidized toward Zr⁺⁴ species during dry reforming. In addition to ZrO₂, signals at even higher binding energies are assigned to the formation of hydroxylated Zr⁺⁴ species, which appear especially on NiZr51. It has been shown that under realistic dry reforming conditions, significant amounts of (intermediate) water are present in the reactant/product mixture leading to ZrO₂-assisted activation of H₂O and, therefore, to surface ZrO_xH_y components similar to previously reported species [16,27,28,29]. After DRM, the C1s spectrum of NiZr51 shows a strongly dominant feature at 284.4 eV, indicating the formation of relatively large amounts of graphitic carbon at the catalyst surface, which causes pronounced attenuation of the Ni2p signal. Minor C1s intensity at higher binding energies is also detectable and assigned to the superposition of oxidized carbonaceous species (oxygenate, hydrocarbon-type and graphite oxide) ranging from approx. 266–288 eV [30,31]. The O1s spectrum underlines the presence of these C-oxygenates, as well as of the hydroxylated zirconium species. Interestingly, despite the full oxidation of Zr, nickel prevails in its metallic state after the first DRM cycle.

By comparing the Ni10Zr90 spectra in the lower panels of Figure 1 and Figure 5, one can see that the supported powder catalyst does qualitatively not change according to its surface chemical composition since the signals for ZrO₂ and metallic Ni appear to be basically identical before and after DRM. However, in comparison to NiZr51, the C1s region shows a much smaller signal after DRM, meaning that only moderate amounts of graphitic carbon were deposited under DRM conditions. At this point, it is useful for further understanding to keep the strongly different coking propensities of the three samples in mind.

Figure 6 shows the ex situ SEM images of the spent catalysts after DRM up to 800 °C and cooling under vacuum conditions. Compared to its pre-DRM state, the CVD-prepared NiZr inverse intermetallic sample shows a strongly altered surface morphology after DRM. In comparison to the relatively smooth and structurally homogeneous pre-DRM surface morphology visible in the micrograph in Figure 2A, it is obvious that the exposure to DRM conditions led to the formation of mostly interconnected, fractal-like ZrO_x islands on top of the Ni foil substrate. Regarding the pre-DRM state, a rather homogeneous element distribution was observed by EDX and supported the assumption of a surface-covering, ultrathin layer consisting of alloyed NiZr and suboxidic ZrO_x on the Ni bulk material. In contrast, the post-DRM state depicted in Figure 6A–D now suggests that the originally continuous thin film breaks up during DRM to form a heterogeneous distribution of partially interconnected ZrO₂ domains on top of Ni. As a consequence, the Zr-free Ni surface area assumes the form of a local “channel-network”. These channels exhibit lateral dimensions ranging from ~10 nm to ~50 nm and are dispersed over the entire surface, leading to a strongly extended Ni⁰/ZrO₂ interface contribution. Within the Ni channels, finely dispersed ZrO₂ “nano-islands” (10–20 nm) can be observed (marked by arrows), leading to an even further increased phase boundary contribution.

The coked NiZr51 sample in the middle image in Figure 6E exhibits a rather heterogeneous morphology and size distribution of the Ni domains, giving the impression of a bimodal domain size state. In coexistence with small Ni nanoparticles in the range between approx. 5 and 30 nm, large islands with sizes ranging from 100–300 nm are visible.

The SEM image of Figure 6F shows the Ni10Zr90-supported powder catalyst in its post-DRM state. Ni particles ranging from ~5 to ~30 nm, evenly distributed on the ZrO₂ grains, are visible. Compared to the average particle size of approx. 7 nm in the pre-DRM state (Figure 2E), the size distribution in Figure 6F suggests a large fraction of >20 nm Ni particles, which can be assigned to sintering effects upon approaching the maximum DRM temperature of 800 °C.

2.4. Post-DRM Comparison of Surface Carbon Chemistry

As described in Section 2.2 (Figure 3), all three distinct NiZr catalyst systems exhibit comparably high DRM activities, despite completely different preparation routes and resulting interfacial structures. Except for the shift of the onset temperatures between the 1st and 2nd cycle, repeated DRM cycles revealed reproducible CO₂ conversion profiles during many DRM cycles, suggesting considerable structural and chemical stability of the individual catalyst’s active states over time. However, despite the pronounced similarities regarding enhanced catalytic performance with respect to pure Ni, the systems show striking differences regarding the degree of the observed coking phenomena, i.e., the susceptibility towards the formation of graphitic deposits at the surface during and after DRM. In the following, the characterization of the respective active states will be brought into agreement with the already shown structural differences between the systems in order to rationalize the observed coking discrepancies. This leads us to the discussion of mechanistic differences between the involved decoking processes. In particular, we will focus on the question of why reproducible coking stability of catalytic activity during repeated DRM cycles is possible despite largely different degrees of coke deposition.

The key to the understanding comes from the initially bulk intermetallic NiZr51 sample and its preferential coking of the larger Ni domains. The CO₂ conversion in Figure 3 reveals a highly active catalyst in the first DRM cycle. However, the XPS data of the post-DRM state in Figure 5 (middle panels) show an intense signal at 284.4 eV in the C1s region, indicating the presence of large amounts of graphitic carbon. Together with the significantly decreased signal intensity of the Ni2p signal, as compared with the pre-DRM state depicted in Figure 1 (middle panels), this indicates a strongly coked state of Ni on the NiZr51 catalyst surface. However, instead of becoming deactivated for the 2nd and further DRM runs (only a representative 2nd run is shown), NiZr51 exhibits reproducibly high CO₂ conversion (c.f. Figure 4). Intuitively, this contradicts the common view of irreversible Ni site-blocking effects of the relatively large amounts of graphitic carbon deposits. As it is well established that the accumulation of unreactive graphitic carbon species (i.e., “coking”) represents one of the major impediments to the development of long-term stable technical Ni-based DRM catalysts [2,10], it is important to understand why the NiZr51 system retains a stable and high activity despite a high degree of coking [19].

The first question regards, of course, the exact locations at which these extended graphitic deposits accumulate. Complementing the topographic picture of the NiZr51 sample in its post-DRM state (Figure 6E), Figure 7 faces a larger-scale topographic SEM micrograph with the respective EDX data obtained on the zoomed regions. A brighter region in the EDX images (an area marked by a white frame in the topographic overview image) indicates a higher abundance of the respective element. It can be clearly seen that areas with extended carbon deposits coincide with areas where large Ni domains are present. This implies that the major part of the spectroscopically observed coking affects the extended, bulk-like metallic Ni regions. The data also reveal that both the dispersed small particles and the large domains are comprised of metallic Ni and that the rather smooth area in between is Ni-free ZrO₂. An inhomogeneous size distribution of the Ni particles/domains is obvious. From the zoom in the blue frame, it is clear that a part of the Ni resides in nanoparticles of approx. 5 nm up to intermediate sizes of ~20 nm, which are rather evenly distributed on the ZrO₂ support. The visually dominant entities are large Ni islands ranging from approx. 100 nm to ~300 nm, supporting the interpretation of a quasi-bimodal Ni surface state.

In strong contrast to the coked NiZr51 sample, the CO₂ conversion profile of the coked, ZrO_x-free Ni foil (orange 2nd cycle conversion profile in Figure 4) reveals drastically suppressed reactivity. The accumulation of graphitic deposits on pure Ni effectively leads to the blocking of active metallic sites and, therefore, to irreversible deactivation of the Ni reference catalyst already after the 1st cycle, which is clearly not the case on any of the NiZr systems.

In order to clarify the degree of reversibility of the Ni nanoparticle-specific coking phenomena, CO₂ titration experiments were performed inside the high-pressure batch reactor of Setup 1. Figure S5 compares the titration experiments of coked Ni foil and coked NiZr51 in pure CO₂, along with the respective pre- and post-titration C1s spectra. In contrast to the C_graphite-covered Ni foil, on which no or at least only drastically suppressed CO formation can be observed, the carbon deposits on the NiZr51 sample exhibit remarkable activity toward CO₂. The C1s spectral intensities confirm that the direct oxidation process C_graphite + CO₂(g) → 2 CO(g) cannot be initiated and/or proceed on a reasonable timescale on the coked Ni foil, whereas an effective decoking mechanism is immediately active on NiZr51. The almost complete carbon clean-off reaction on the latter clearly highlights the decisive role of the Ni⁰/ZrO₂ phase boundary. To further substantiate the mechanistic function of the latter for decoking, a pure polycrystalline ZrO₂ sample was tested with respect to DRM performance and surface carbon chemistry under otherwise identical DRM conditions, as shown in Figure S6. According to the CO₂ conversion profile (Figure S6B), only little activity can be observed beyond ~570 °C and no signs of coking or carburization of the post-DRM state are detectable, as derived from the unaltered Zr3d and O1s ex situ XP spectra (Figure S6A). This result additionally supports bifunctionality at or close to the phase boundary sites, in the sense that the Ni surface is rather responsible for methane activation toward reactive forms of carbon, but also for coking, whereas ZrO₂ is rather responsible for efficient CO₂ activation. This “active” cocatalytic role can be set apart from predominant textural promoter effects of the support, which were, e.g., reported for the Ni/MnO system. The stabilization of small, strongly support-anchored Ni nanoparticles allows for less refractory forms of carbon deposits, which can be effectively oxidized by CO₂ and do not necessarily lead to catalyst deactivation. Instead, they rather act as a reactive intermediate for methane dry reforming [17,18,19].

For NiZr51, we could show that a quasi-bimodal state of Ni on the catalyst is formed under realistic DRM conditions, consisting of nanoparticulate Ni islands in coexistence with large, bulk-like Ni domains. It is tempting to assume that these extended patches could rather approach the coking properties of pure bulk Ni foil and might thus play only a minor catalytic role. In contrast, sufficiently small Ni nanoparticles have been shown to be less susceptible to coking [32]. Moreover, it appears likely that the reduced dimensions of the nanoparticles allow for enhanced carbon diffusion towards the active phase boundary sites and, thus, faster carbon clean-off. It appears, therefore, logical that these nanoparticles should constitute the more active surface species on ZrO₂. The EDX results of Figure 7, showing preferential coking of the large Ni domains, support this interpretation.

The negligible C1s signal intensity of the CVD-prepared inverse NiZr intermetallic catalyst after DRM, shown in Figure 5, suggests that hardly any reaction-induced coking takes place on this system. At first glance, this appears contradictive to the above-mentioned argumentation since this model catalyst basically represents the most “bulk-Ni like” NiZr material in our comparison, and the pure Ni-foil substrate itself shows apparently irreversible deactivation by a graphitic top layer. Due to the absence of sufficient amounts of carbon, analogous in or ex situ C reactivity studies as performed on NiZr51 and Ni10Zr90 were experimentally not possible. However, if the bulk-like nature of Ni was the only reason for irreversible coking, the question arises why the NiZr51 sample becomes strongly coked, whereas the NiZr inverse intermetallic sample, featuring the most extended Ni bulk of all NiZr systems, exhibits almost no carbon deposits after several DRM cycles (compare top and middle C1s spectra of Figure 5), which is also in strong contrast to the fully coke-covered pure Ni foil. Otherwise, NiZr51 and the NiZr inverse intermetallic sample are spectroscopically very similar, both consisting of metallic Ni in contact with ZrO₂, and both exhibit comparable CO₂ conversion activities.

As shown above in Figure 6, the uniform CVD-prepared and UHV-annealed pre-DRM inverse intermetallic state experiences in situ activation under DRM conditions toward an extended Ni-ZrO₂ phase boundary network. Figure 6 Panels A-D, in turn, complement the morphological SEM picture with the respective EDX scans obtained after the 1st DRM cycle. To illustrate the structural development of the CVD-prepared NiZr film, Figure 8 shows that even after three consecutive DRM cycles, the structure and chemical distribution of elements still feature the oxidized ZrO₂ domains present after one DRM cycle (especially visible in the lower right corner; cf. Figure 6A–D), emphasizing the stability of the inverse model catalyst with respect to the distribution of the percolated ZrO₂ patches.

Figure 6, Panel G schematically visualizes the in situ formed phase boundary between the nanoscaled “Ni-channels” and the interconnected ZrO₂ islands. Additionally, potential pathways of carbon atom diffusion are highlighted. The narrow channels, together with the adjacent ZrO₂ nano-islands, provide rather short carbon diffusion pathways in the range of a few nanometers and, hence, facilitate the migration of carbon atoms toward the surrounding phase boundary and their local reaction toward CO at the interfacial sites. We suggest a strong lowering of the C supersaturation of the near-surface Ni layers via this process, which, in turn, shifts the balance of C_graphite growth and decay on the available Zr-free Ni surface toward redissolution in the form of C atoms [33,34]. The scheme also suggests an additional decoking pathway, which is otherwise dimensionally limited, namely diffusional loss of C atoms to the deeper Ni bulk, as the sample substrate consists of a 0.1 mm thick Ni foil and thus represents a quasi-infinite 3D sink for C atoms. This allows a part of the carbon atoms to vanish into deeper Ni regions, thus further lowering C supersaturation near the surface and increasing the coking resilience of the sample. These mechanistic advantages allow explaining the superior coking properties in comparison to both the pure Ni foil, as no “quasi-fractal” metal-oxide phase boundary is present there, and to the large Ni⁰ domains on NiZr51, as no quasi-infinite Ni bulk is available on the latter and relatively long C-diffusion pathways to the phase boundary sites need to be overcome for decoking.

Following the same line of argumentation of combining a high abundance of PB sites with short C-diffusion distances, improved decoking properties and high and reproducible activity can also be expected for the Ni10Zr90 powder catalyst, as the dimensionally limited Ni nanoparticles exhibit an extended phase boundary to the polycrystalline ZrO₂ support along with enhanced decoking properties. This has already been shown by in situ NAP-XPS experiments in Ref. [19]. Regarding C-deposition during and after the 1st DRM cycle, a smaller relative C1s intensity as compared to NiZr51 is detectable in the ex situ C1s spectrum (Figure 5, lower right); however, coking is obviously not fully suppressed, in contrast to the trend on the NiZr inverse intermetallic sample. We assign the respective ex situ C1s spectrum in Figure 5 (bottom panels) to the superposition of DRM-induced graphic carbon at 284.4 eV with traces of adventitious or oxygenate carbon species (285–287 eV) originating from the intermediate contact to ambient air. Yet, the decoking mechanism on the Ni nanoparticle-containing systems (NiZr51 and Ni10Zr90) is clearly related to intermediate ZrC formation at the metal-oxide phase boundary [19], which cannot be verified experimentally for the CVD-based “inverse” model catalyst due to the absence of sufficient post-DRM carbon deposits. This implies that the observed differences can be safely assigned to the absence of strongly C-supersaturated Ni metal domains or regions on the latter. However, why does ZrC form at all on NiZr51 and Ni10Zr90? The direct solid-state synthesis reaction of crystalline bulk ZrC by carbothermal reduction in pure ZrO₂ is moderately endothermic (ZrO₂(s) + 3C(s) → ZrC(s) + 2CO(g), ΔH⁰₂₉₈~+46 kJ/mol), and only proceeds upon heating between 1500 °C and 1800 °C [35]. Therefore, it appears reasonable that small enough Ni⁰ particles and domains play a catalytic key role in local Zr-carbide formation at DRM-relevant temperatures, especially at the metal/metal-oxide phase boundaries. It is known that metallic Ni is capable of redissolving graphitic carbon at temperatures above 600 °C [33,34]. We further suggest for our specific case that the resulting carbon atoms diffuse via the bulk and surface near regions [36,37] toward the phase boundaries, where they form interfacial ZrC_x species by local ZrO₂ reduction around 700–800 °C, as indicated by the Zr-carbidic signals in the respective in situ NAP-XPS spectra in Figure S7 in the Supplementary Information, which were obtained in Setup 2.

This local carbothermal reaction can be simplified as

ZrO₂(s) + 3 C_dissolved →ZrC(s) + 2 CO(g) (CO-forming process I)

and already leads to the first two molar equivalents of the final product CO.

This “interfacial carbide mechanism” could be proven for NiZr51 and Ni10Zr90 [19], but it remains unclear whether it also applies to the inverse model catalyst. Sufficient carbon deposits and diffusion of dissolved C toward the PB appear to suffice only on dimensionally limited Ni⁰ systems to induce detectable amounts of ZrC. The large Ni domains on NiZr51 were found to be responsible for the slower thermal carbon clean-off via the diffusional limited redistribution process C_graphite → Zr_xC_y on NiZr51. This also limits the rate of reaction of C_carbide with CO₂ at the interface to finally form CO (i.e., the final carbon clean-off-step), simplified as

ZrC(s) + 3 CO₂(g) → ZrO₂(s) + 4 CO (g) (CO-forming process II, ΔH⁰₂₉₈~−135 kJ/mol),

which is moderately exothermic and leads to four additional CO molecules. Processes I and II eventually add up to:

3 C_dissolved + 3 CO₂(g) → 6 CO(g).

Hence, the interfacial ZrC_x represents an active intermediate species for efficient CO₂ activation and CO formation. However, its abundance strongly depends on the balance of the build-up of C-supersaturated Ni particles/domains and, therefore, (graphitic) carbon accumulation relative to the diffusion of C_dissolved toward the PB and the clean-off reaction II occurring there. Large Ni particles/domains are, therefore, most unfavorable in terms of coking because their size is apparently small enough for sufficient carbon supersaturation on the timescale of DRM. However, at the same time, they suffer from relatively long C-diffusion paths towards the phase boundaries to form the CO₂-reactive ZrC.

To explain the full catalytic cycle on all studied NiZr systems, we suggest that the graphitic carbon deposits are continuously fed by methane decomposition at C-depleted metallic Ni surface sites, which, in turn, are permanently re-established due to the solubility of C_graphite in the Ni bulk, which strongly increases at elevated temperatures. The exact onset and rate contribution of the “carbidic interface mechanism” [19] appears now as a matter of local C_graphite antisegregation and C atom diffusion rates. The temperature must be high enough to redissolve C_graphite (T > 600 °C), and the resulting dissolved C-atoms must be mobile enough to diffuse thermally toward the PB to form the reactive C_carbide species (again at T > 600 °C), which then facilitate the kinetics of the final, CO-forming carbon clean-off step II. Taken together, our previous results indicate that, instead of solely deactivating the Ni surface, the graphitic deposits can also be remobilized and provide continuous C-atom supply via thermal antisegregation to drive CO formation at the phase boundaries, provided that the temperature is high enough. In this context, we suggest that the increasing carbon bulk mobility coincides not simply by chance with the onset temperature of CO₂ conversion on the coked catalysts at >600 °C (see reaction onset temperatures in Figure 4).

Only the decoking of the spent Ni foil catalyst by CO₂ appears to be completely blocked on the timescale of our experiments (Figure S5, Supplementary Information) because the direct oxidation process C_graphite + CO₂(g) → 2 CO(g) on the Ni⁰ surface is very inefficient, and no active metal-oxide phase boundary is available.

Summarizing the above-described observations, both the decoking of the Ni surface and the ensuing CO-forming processes via intermediate ZrC_x are strongly particle-size-dependent phenomena. Well-embedded, small Ni nanoparticles with a large phase boundary to ZrO₂, such as on activated Ni10Zr90, feature short carbon diffusion pathways to the phase boundary and are, therefore, much more quickly depleted of dissolved carbon. The opposite is true for the large, bulk-like Ni domains on NiZr51, which can be decoked, but at a much slower rate due to long-distance diffusion limitations. This scenario is experimentally supported by the in situ carbon titration experiments inside the NAP-XPS chamber of Setup 2, which are shown in Figure S7 of the Supplementary Information. The different timescales for decoking NiZr51 and Ni10Zr90 reflect the reactivity difference in the respective carbon species with CO₂. The presence of the larger Ni domains can, therefore, also explain why the coking propensity of NiZr51 is generally much higher than that of Ni10Zr90 (see also C1s spectra in Figure 5).

Considering the prevailing experimental limitations outlined in the experimental section to perform a successful in situ characterization, especially of the very coking-resistant “inverse” CVD-prepared NiZr catalyst, we note that the mechanistic conclusions given above cannot be based on direct observations of the claimed carbon species and their mobilities/reactivates on this system, as they are simply below the detection limit. This imposes the necessity to discuss the analogous mechanistic picture with caution, representing a relatively plausible possibility. Nevertheless, we believe that the application of the same mechanistic picture is not too far off the truth as there are some “hard facts” known about the carbon chemistry in and on bulk Ni from a broad range of graphene-focused studies on different bulk Ni substrates. Fact 1: at and above 700 °C, bulk solubility and mobility of C-atoms in the Ni bulk are greatly improved. Fact 2: graphene/graphitic species can be redissolved/restructured in Ni at these temperatures. Fact 3: methane decomposition at these temperatures leads to supersaturation with dissolved C and resegregation of the latter in the form of graphene/graphite. Fact 4: even in the presence of CO₂, this resegregation leads to irreversible coking of a pure bulk Ni surface after DRM, but once an extended PB between bulk Ni and ZrO_x islands is present, practically no coking occurs at all. For the sake of logic, this can only be due to the presence of the PB between oxide and metal. Our general mechanistic picture presumes these facts but remains intrinsically based on a potential analogy to the in situ proven processes of C_graphite redissolution, C bulk diffusion, and conversion to ZrCx using NAP-XPS on NiZr51 and Ni10Zr90. Practically, the above-discussed mechanistic scenario can explain the largely complete coking suppression on the CVD-prepared NiZr inverse intermetallic catalyst, but it is not the only conceivable explanation. Here, two simultaneous “anti-coking” processes can, in principle, be considered to operate simultaneously. The in situ decomposition of the near-surface intermetallic precursor state under DRM conditions yields a finely dispersed Ni⁰ “channel-network” within the surrounding ZrO₂ islands, as it is shown in the SEM images in Figure 6A–D. This structural situation provides short diffusion paths of a few nm of the adsorbed and dissolved carbon atoms from Ni toward the surrounding metal-oxide PB for the final clean-off step. Furthermore, the Ni⁰ foil substrate represents a “quasi infinite” 3-dimensional bulk where dissolved C atoms can diffuse into deeper regions, being undetectable there both for SEM and XPS. We suggest that the combination of both effects can explain that hardly any graphitic surface carbon accumulation can be observed after several DRM cycles, corresponding to several hours of catalytic operation. Here it should be noted that the decoking mechanism by carbon bulk diffusion might only prevail on the timescale of our experiments. Although the Ni foil substrate represents a relatively large carbon sink, we can neither postulate nor exclude progressive long-term carbon supersaturation of this sink, therefore slowly increasing surface carbon accumulation and coking. As a matter of fact, it remains unclear whether the fractal Ni-ZrO₂ PB structures suffice to suppress surface-near C saturation of the Ni substrate permanently to the extent that no more allows for local nucleation of graphitic deposits by carbon resegregation or not.

3. Materials and Methods

3.1. Synthesis of Specimens

3.1.1. Preparation of Pure Ni Foil and the Respective CVD-Prepared “Inverse” Intermetallic NiZr Precatalyst

As a ZrO₂-free reference catalyst, ultraclean Ni foil (Alfa Aesar, purity 99.994%, 0.1 mm thick, size 18 × 20 mm²) was used. Prior to the DRM experiments, the pure Ni foil was cleaned inside the UHV chamber of Setup 1 (described in Section 3.2.1) by Ar⁺ sputtering (5 × 10⁻⁵ mbar Ar, 2 keV, 1 µA sample current) and thermal annealing to 800 °C under UHV conditions (5 × 10⁻¹⁰ mbar), until XP spectra without evidence for impurities could be obtained.

For the CVD-based preparation of the NiZr inverse intermetallic sample, the same ultrapure Ni foil was used as a substrate, following the same cleaning procedure as the reference foil mentioned above. For the CVD process, zirconium (IV) tert-butoxide Zr(O-t-C₄H₉)₄ (Strem, purity: 99%) was used as a precursor molecule, which was dosed via a leaking valve onto the Ni foil substrate in the ultrahigh vacuum chamber of experimental Setup 1 (described below in Section 3.2.1). The sample was heated to 400 °C and exposed to a pressure of 6 × 10⁻⁶ mbar of ZTB for 240 s, corresponding to approximately 1440 L. This procedure leads to a calculated ZrO₂ adsorbate coverage of ~0.8 mL, as derived from the respective Ni2p and Zr3d XPS intensities on the basis of a non-attenuating ultrathin ZrO₂ adlayer model on a semi-infinite Ni substrate (details are described in refs. [27,38,39]). The sample was then thermally annealed at temperatures between 450 °C and 500 °C for 5 min under excellent UHV conditions (5 × 10⁻¹⁰ mbar) to obtain the near-surface intermetallic state.

3.1.2. Preparation of the NiZr51 Bulk-Intermetallic Precatalyst

The Ni_xZr_y bulk-intermetallic sample was prepared by resistive heating of a stack of alternating small pieces of clean Ni foil (Alfa Aesar, purity 99.994%, 0.1 mm thick) and Zr foil (Alfa Aesar, purity 99,5%, 0.1 mm thick) under high vacuum conditions (1 × 10⁻⁶ mbar) at a nominal ratio of Ni:Zr of 80:20 at% in an Al₂O₃-coated Mo crucible. It is denoted as “NiZr51” in the following, according to its main phase component Ni₅Zr (see Supplementary Information Figure S2). The sample was heated until the melting of both components could be observed. A light flash indicates a spontaneous exothermic reaction between Ni and Zr, leading to the intermetallic Ni-Zr melt. The heating was turned off immediately, and the melt recrystallized within fractions of a second.

We note that the co-melting preparation is intended to form an electronically conductive model catalyst interface especially dedicated to UHV-based electron-spectroscopic investigations. As the bulk intermetallic Ni_xZr_y states are subjected to oxidative in situ activation under DRM conditions to yield a PB-rich, thus catalytically active, but still sufficiently conductive “conglomerate” metal-oxide layer, too Ni-rich initial states are less suitable as they lead to excessively large Ni domains without sufficient PB to the underrepresented ZrO₂ phase. On the other hand, too Zr-rich compositions suffer from Zr passivation and conductivity problems/local charging due to lacking Ni percolation upon in situ activation and are therefore equally unsuitable for this purpose. The activated 80:20 intermetallic state represents a compromise, which additionally leads to a quasi-bimodal size distribution of the Ni domains. The latter turned out to be important for the experimental observation of the larger, predominantly coked Ni domains. An additional advantage is that the 80:20 ratio delivered the best initial catalytic performance compared to other ratios (e.g. Ni:Zr 50:50 or 30:70) using the same preparation routine.

3.1.3. Preparation of the Ni-Free Polycrystalline ZrO₂ Powder Reference Sample and the Respective Wet-Impregnated Ni10Zr90 Powder Catalyst

The Ni-free ZrO₂ reference sample was prepared by pressing a pellet of polycrystalline zirconium (IV) oxide powder obtained by drying an aqueous ZrO₂ suspension (Alfa Aesar, zirconium (IV) oxide, 20% in H₂O, colloidal dispersion, and 0.1 micron particles in liquid). The sample was then annealed to 800 °C under UHV conditions. The XP spectra showed no contaminations before it was transferred into the batch reactor of Setup 1 for DRM.

The 10:90 atom% choice for the “Ni10Zr90” supported reference powder catalyst was mainly motivated by particle size and electron microscopy applicability considerations. Moreover, 10 atom%, corresponding to ~5 mass % Ni allows for obtaining small enough Ni nanoparticles for efficient catalysis, which are still abundant and large enough to be detectable by standard microscopic techniques (SEM and TEM).

The “Ni10Zr90” supported powder catalyst was prepared by conventional aqueous impregnation of the same commercial aqueous ZrO₂ dispersion with dissolved Ni(NO₃)₂ to achieve a final Ni:Zr 10:90 atom% ratio in the obtained powder state after calcination at 800 °C in air.

3.2. Characterization Methods and Dedicated Experimental Setups

3.2.1. XPS/LEIS/AES Surface Analysis Combined with Catalytic DRM Rate Quantification (Setup 1)

Setup 1 comprises a UHV system with an attached all-quartz recirculating batch reactor, described in detail by Mayr et al. [40], and is designed for quantitative catalytic/kinetic studies up to 1 bar on polycrystalline foils, detecting products by online MS analysis (HP GC-MS System G1800A) via a capillary leak and/or by conventional GC-MS analysis via column injection. MS signals of CH₄, CO₂ and CO (m/z = 15 + 16, 44 and 28) were externally calibrated and corrected for fragmentation. Ex situ surface analysis was performed using an XPS/AES/LEIS spectrometer (Thermo Electron Alpha 110) and a twin Mg/Al anode X-ray gun (XR 50, SPECS). All DRM reactions were conducted with initial partial pressures of CH₄:CO₂ = 50 mbar:50 mbar. The reaction cell was backfilled with pure He to a total pressure of 1013 mbar in order to achieve efficient gas intermixing via recirculation and fast heat transfer to the sample via improved thermal conductivity. For DRM catalysis, the reactor was heated with a constant linear rate of 10 K min⁻¹ to the final temperature of 800 °C and then kept isothermally at this temperature for 30 min.

For the reactor-specific calculation of the theoretical equilibrium conversion curve in Figure 3 and Figure 4, thermodynamic data from the NIST database were used. The temperature-dependent ΔG⁰_T and associated equilibrium constant values of the stoichiometric reaction CO₂ + CH₄ $\to$ 2 CO + 2 H₂ were calculated using the NIST series expansion for ΔH⁰_T and ΔS⁰_T of CH₄, CO₂, CO, and H₂, respectively. The subsequent gas-phase equilibrium pressure calculation was based on the initial reactant pressures of 50 mbar CH₄ and 50 mbar CO₂ and the assumption of their 1:1 stoichiometric conversion to a fixed 1:1 partial pressure ratio of the products. This appears justified in view of the relatively low total reactant pressure of 100 mbar, favoring high H₂-selectivities throughout the active temperature range of the catalysts. Accordingly, our experiments showed the formation of H₂ and CO at a ratio of close to 1:1 at any temperature, in accordance with a minimized influence of the H₂ selectivity-spoiling reverse water-gas shift reaction.

3.2.2. Near-Ambient Pressure X-ray Photoelectron Spectroscopy (NAP-XPS, Setup 2)

To elucidate the sample surface states in situ under reaction conditions, experiments in a commercial laboratory-based SPECS NAP-XPS system were performed. The UHV chamber is comprised of a µFOCUS 600 NAP monochromatic small spot (300 µm) Al-K_α X-ray source, a hemispherical energy analyzer (PHOIBOS 150 NAP equipped with 1D delay line detector, SPECS Surface Nano Analysis GmbH, 13355 Berlin, Germany) in a vertical configuration and a µ-metal main chamber shielding the system from external electric and magnetic fields. The differentially pumped electrostatic lens system separates gas molecules from the photoelectrons focused toward the hemispherical energy analyzer, which allows XPS experiments during backfilling of the analysis chamber to pressures up to 25 mbar with variable gases and gas mixtures via mass flow controllers. An IR laser (max. 120 W) is attached to the bottom side of the analyzing chamber and allows precise heat of the samples from the backside via an 8 mm hole in the sample holder. The temperature is measured by a K-type Ni/NiCr thermocouple directly spot-welded to the sample. The X-ray source power was set to 70 W and 13 kV, while the analyzer was set to a constant pass energy of 50 eV for all recorded spectra.

We emphasize that dynamic carbon growth chemistry cannot be investigated in CH₄-CO₂ gas mixtures inside the NAP-XPS chamber at present, and, therefore, “pre-coked” samples obtained after DRM catalysis in Setup 1 were used in Setup 2. We do not show in situ NAP-XPS data of the CVD-prepared “inverse” NiZr catalyst under DRM conditions, as it does not coke at all. We tried this several times in our NAP-XPS setup, using pressures in the lower mbar range. However, there is absolutely no carbon C1s signal detectable under the specific reaction conditions prevailing in this system. Thermodynamic considerations imply that one can miss the coking window of DRM exactly under the pressure conditions of a few mbar CH₄ + CO₂ prevailing in this type of experiment at and above ~700 °C. Even beyond this issue, the insufficient gas temperature of CH₄, which adopts the room temperature of the cold chamber walls, cannot mimic the situation in an externally heated tubular reactor, as it leads to a strongly decreased sticking probability of CH₄. This has to be considered, especially in the case of methane, as it has been shown that gas temperature exponentially affects its dissociative chemisorption and sticking coefficient [41,42]. Compensation for this drawback by using a heated gas beam source is currently in the stage of testing. Exactly for this reason, we used the externally heated tubular quartz reactor of our “ex-situ” Setup 1 for “pre-coking” our samples. In this reactor, reactant pressures up to 1 bar and gases with the same gas temperature as the catalyst bed can be provided, which leads to the experimentally observed coking phenomena. In view of the lacking possibility of performing the “full” in situ DRM experiment inside the NAP-XPS, we chose a compromise and precoked our samples in Setup 1 prior to transferring them to the NAP-XPS chamber in order to study the further fate and the reactivity of the deposited carbonaceous species. This precoking step worked well both for the NiZr51 bulk intermetallic and the supported Ni10Zr90 samples but failed for the CVD-prepared inverse NiZr sample, as this sample did not even show a sufficient degree of carbonization inside Setup 1 under the same conditions used for NiZr51 and NiZr90.

3.2.3. In Situ X-ray Diffraction (Setup 3)

The in situ high-temperature synchrotron XRD experiments in a CO₂:CH₄ = 1:1 mixture as well as in pure CO₂ were performed at the beamline 12.2.2 of the Advanced Light Source (ALS), Lawrence Berkeley National Laboratory, California. Diffraction patterns were collected in angle-dispersive transmission mode with a focused 25 keV monochromatic beam (λ = 0.4959 Å/15 μm spot size). A total of 10 mg of NiZr51 sample powder was heated in a 0.7 mm (inner diameter) quartz capillary under a continuous gas flow (2 mL × min⁻¹) injected through a 0.5 mm tungsten tube. The capillary was heated at a 10 K min⁻¹ heating rate to 800 °C in an infrared-heated SiC tube furnace as described elsewhere [43,44]. Diffraction patterns were recorded by a Pilatus3 S 1M detector (981 × 1043 pixels, pixel size 172 × 172 μm², 30 ms read-out time) every 60 s during the heating cycle. Rietveld refinement was performed using the FULLPROF program [45].

3.2.4. DRM Rate Quantification on Powder Catalysts (Setup 4)

For the characterization of technologically relevant catalyst systems, an ambient-pressure batch reactor cell for powder catalysts was used. It is comprised of a quartz reactor tube with a volume of 13.8 mL placed inside a tubular furnace attached to a QMS for the detection of gases. The calcined Ni10Zr90 powder sample was pre-reduced in 1 bar H₂ for 30 min at 400 °C to obtain metallic Ni⁰ after the calcination step under an ambient atmosphere. For DRM experiments, a reactant mixture of 15 mbar CH₄, 15 mbar CO₂, and 1029 mbar He was used; the reaction cell was heated with a linear temperature ramp of 10 K min⁻¹ to a final isothermal temperature of 800 °C, which was kept for 30 min before cooling.

3.2.5. High-Resolution Scanning Electron Microscopy (Setup 5)

SEM sample characterization was performed using a Zeiss EVO MA25 scanning electron microscope with a LaB₆ cathode. Images were acquired by the use of secondary electrons and a voltage set to 5 kV during the SEM experiments.

3.3. Details of X-ray Photoelectron Spectroscopy (XPS)

All XPS data were analyzed using the CasaXPS software program, version 2.3.24 PR1.0 (Casa Software Ltd.). For peak fitting, a Shirley background was applied to all spectra. Fitting of the suboxidic ZrO_x, ZrO₂, and ZrO_xH_y species was carried out using a weighted sum of Gaussian and Lorentzian peak shapes with a GL(30) contribution. Asymmetric parameters were used for Zr3d fitting of Zr⁰ and carbidic Zr (GL(30)T(2.1)) as well as for C1s of graphitic carbon (GL(30)T(2.1)) and carbidic carbon (GL(30)T(1.2)). Oxidized carbon/oxygenate components occurring in the C1s spectra between 266–288 eV [30,31] were not fully deconvoluted with respect to specific C-C, C-O, or C=O binding energies due to insufficient intensity and spectral resolution. Deconvolution of all Zr3d spectra involved the metallic Zr⁰ component at a BE of 179.1 eV, the carbidic ZrC component at a BE of 179.8 eV and ZrO₂ at a BE of 183.0 eV [46,47,48]. The Ni2p3/2 peak is assigned to a BE of 852.8 eV for the metallic component and the C1s BE of graphitic and carbidic carbon to 284.4 eV and ~282.2 eV, respectively [49,50]. It has to be noted that the presence of traces of carbidic Ni components within our coked catalyst systems cannot be fully excluded. However, meaningful fitting of the respective C1s component is not straightforward, as the C1s signal of the carbidic Ni species (e.g., Ni₃C or interstitially dissolved C) is largely superimposed by the main signal of the Zr carbidic component. Furthermore, the carbidic Ni2p region coincides with the pure metallic Ni⁰ one [51,52]. For fitting the O1s spectra, four different components were considered, i.e., stoichiometric Zr oxide (ZrO₂) at ~530 eV, suboxidic ZrO_x and hydroxylated zirconium species (possessing very similar binding energies) at ~531.4 eV, as well as adsorbed organic compounds (C-O and C=O) at binding energies of ~533 eV [29,31,53].

4. Conclusions

A reproducibly high methane dry reforming activity, together with synergistically enhanced anticoking properties, can be obtained by the combination of an extended metal-zirconia phase boundary approach both on nanoparticulate Ni⁰ and extended bulk Ni⁰ as the respective active catalytic phase for methane splitting. Both the CVD-prepared “inverse” near-surface intermetallic NiZr and the initially bulk-intermetallic NiZr51 model catalyst system become activated via oxidative in situ decomposition under realistic DRM conditions. The chosen in situ activation of intermetallic NiZr states represents an effective way for the generation of extended Ni⁰/ZrO₂ phase boundaries and an enhanced ratio of active interfacial vs. metallic Ni surface sites in either case. This, in turn, creates optimized conditions for bifunctional catalyst operation. In analogy to Ref. [19], CH₄ activation on the DRM-active state of the CVD-prepared “inverse” intermetallic NiZr catalyst also takes place at the metallic Ni⁰ surface sites, thereby decomposing CH₄ to H₂ and adsorbed carbon atoms, which then can diffuse to the nearby phase boundary sites to react off with CO₂ or to subsurface and deeper bulk regions of Ni. On pure bulk Ni, the dissolved carbon simply accumulates in surface-near regions, leading to a sufficiently supersaturated state for ubiquitous graphitic carbon segregation. A fully surface-covering graphitic layer results in irreversible blocking of the metallic Ni sites, leading to virtually complete deactivation of the pure Ni reference catalyst, as the direct decoking process C_graphite + CO₂(g) → 2 CO(g) is kinetically strongly hindered. Only on the Ni-nanoparticle containing active states of NiZr51 and Ni10Zr90, the formation of reactive intermediate ZrC_x species at the Ni⁰/ZrO₂ phase boundaries was observed, resulting from a particle size-dependent C_graphite → C_carbide redistribution process at sufficiently high temperatures, which enables continuous decoking of at least a part of the Ni surface area [19]. Especially small enough Ni⁰ nanoparticles with a large metal-oxide phase boundary contribution are most efficiently decoked, as they provide both enhanced carbon solubility and short diffusion pathways of the dissolved C-atoms toward the surrounding ZrO₂, thereby causing its local carbothermal reduction to form ZrC_x and CO(g). The so-formed carbidic/oxygen-deficient ZrC_x sites assist in reductive CO₂ activation at the phase boundaries, thereby becoming reoxidized and completing the CO-forming catalytic cycle. Due to the absence of sufficient amounts of graphitic surface carbon on the CVD-prepared “inverse” NiZr intermetallic model catalyst after DRM, it is not possible to verify the analogous “interfacial carbide mechanism” that is also on this system. However, this model catalyst features the apparently highest coking resilience of the compared NiZr systems, at least on the timescale of our experiments, due to a unique combination of nanosized Ni domains providing short carbon diffusion paths towards the adjacent fractal-like Ni/ZrO₂ phase boundaries, together with a “quasi infinite 3d” Ni bulk representing a relatively large carbon sink.