Enhancing Stability of γ-Al₂O₃-Supported NiCu Catalysts by Impregnating Basic Oxides in the Hydrodeoxygenation of Anisole

Abstract

Basic oxides such as CaO and MgO were added to a γ-Al₂O₃ support in NiCu-catalyzed hydrodeoxygenation of anisole. A commercial CaO-MgO-γ-Al₂O₃ was compared to a benchmark γ-Al₂O₃ and in-house variants with sequential oxide impregnation prior to NiCu impregnation. CaO did not have a significant impact on activity compared to the benchmark, while MgO improved NiCu dispersion, enhancing activity. Co-impregnation of CaO and MgO resulted in intermediate activity. Despite decreased demethoxylation, likely due to moderated support acidity, both CaO-modified and the commercially supported catalysts showed improved stability over 48 h Time On Stream.

1. Introduction

The environmental and economic consequences stemming from the extensive use of fossil resources and their resulting fast depletion represent a growing matter of concern. The main environmental impacts are related to the huge greenhouse gas (GHG) emissions, and are observed under the form of climate change, severe weather anomalies, and biodiversity shifts. On an economic level, the volatility of oil prices, affected by geopolitical tensions and supply chain disturbances, has led to dramatic market fluctuations. Simultaneously, the demand for chemicals and fuels has reached unprecedented levels [1]. Consequently, the urge to transition from fossil-based sources to more sustainable alternatives, ideally renewable ones, has never been more pressing [2,3]. In this respect, lignocellulosic biomass stands out as a particularly promising feedstock for the production of green fuels and specialized chemicals, as it is abundant and does not interfere with food supplies [4,5].

Various potentially viable routes have been explored to convert biomass into value-added products, with options based on pyrolysis being among the most interesting ones [4]. However, pyrolysis typically results in a bio-oil which consists of highly oxygenated compounds [4,6]. The presence of unsaturated bonds and oxygen functionalities renders these bio-oils unstable, with a low energy density, making them inadequate for direct use as biofuel. A hydrodeoxygenation (HDO) step is required to stabilize the components within these bio-oils and make them suitable for use as a biofuel or other value-added product.

One of the main challenges to the feasibility of the process described above is the development of a cheap HDO catalyst [7]. Initially, research efforts mainly focused on noble metal catalysts, exhibiting good performances, apart from being susceptible to rapid deactivation in some cases. This, coupled with their high cost, diminished their economic appeal [8]. Consequently, transition metal catalysts, especially Ni-based ones, have emerged as an alternative [9]. Ni catalysts alloyed with Cu were found to be promising in this respect, exhibiting a high hydrogenation ability [10,11]. These NiCu-based catalysts exhibit considerable potential in efficiently removing oxygen from lignocellulosic biomass-derived compounds [12]. While their high activity and selectivity make them strong contenders for use in lignocellulosic biomass conversion processes, it’s important to note that the NiCu ratio can be utilized to steer the activity and selectivity of these catalysts [11]. However, the NiCu ratio does not significantly impact the stability of the catalysts, which remains an area requiring further improvement.

For Ni-based catalysts, the lack of stability is often related to coking [13,14,15]. In the context of HDO, coking is primarily attributed to polycondensation reactions. Due to the complexity of bio-oil, the use of model compounds in research is often necessary, such as furans, ethers, phenolics, … or their mixtures, to gain more fundamental insights [9,16,17,18]. Methoxyphenyl groups are one of the most prevalent moieties in these oils, making anisole a widely investigated compound for hydrodeoxygenation reactions [19,20]. According to Zhang et al., the carbonaceous deposits formed on spent Ni/HZSM-5, and Cu/HZSM-5 catalysts during anisole HDO mainly comprise chain compounds of aliphatic carbon, aromatics, and carbonyl groups [21]. These deposits are a result of the polycondensation of methyl groups and aromatic compounds during the decomposition of anisole. Alongside aromatic hydrocarbons, other compounds such as carboxylic acids, ketones, and aldehydes have been found to contribute to the formation of coke through condensation reactions. This process is similar to the phenol–formaldehyde polymerization [22].

To mitigate coking, several strategies can be employed. An effective approach involves modifying the catalyst by incorporating metal additives or promoters. These additives play a crucial role in reducing coke formation by stabilizing reactive intermediates, suppressing undesired side reactions, and enhancing overall catalytic activity. Specifically for alumina-supported catalysts, the addition of alkaline earth oxides or lanthanide oxides as promoters has been proposed to minimize coke deposition and to enhance the sintering resistance of Ni- and Cu-based systems [23,24,25]. The addition of CaO and/or MgO, which are basic alkaline earth oxides, to Ni-based catalysts has been studied for steam-reforming and deoxygenation reactions [23,26,27,28,29]. It was demonstrated that the basic additives in CaO–MgO-promoted catalysts can neutralize the acid sites of the a γ-Al₂O₃ support, effectively reducing coking during reaction [26,28,29]. Additionally, their presence increased the dispersion of Ni, leading to a higher activity of the catalyst, and prevented sintering during reaction. Papageridis et al. observed that CaO–MgO-promoted Ni-γ-Al₂O₃ catalysts tend to form relatively more amorphous coke rather than crystalline coke compared to unpromoted Ni-γ-Al₂O₃ catalysts, which is beneficial since amorphous coke is more easily gasified [28]. It was suggested that oxygen atoms from the modified alumina support can be more easily transferred, creating defects in the graphitic crystalline lattice, leading to easier gasification, or that MgO can inhibit the formation of crystalline coke by blocking the pathway towards the nucleation of graphene [28,30].

Although the effect of adding CaO and MgO to γ-Al₂O₃ has been investigated before, as outlined in the previous paragraph, a systematic and comprehensive evaluation of the impact of these basic oxide modifiers is yet to be performed for HDO reactions, especially under mild conditions (i.e., temperatures up to 250 °C and H₂ partial pressures below 100 bar) [31]. The present work evaluates this impact for supported NiCu catalysts for the HDO of anisole at 200 °C. CaO and/or MgO was sequentially impregnated onto a γ-Al₂O₃ support, before loading NiCu. The results were compared to a commercially available basic oxide-modified support impregnated with NiCu. Metallic, textural, and acidic properties are probed, and their impact on the catalyst performance for anisole hydrodeoxygenation under mild HDO conditions is mapped.

2. Results and Discussion

First, the textural properties, acidic, and basic characteristics and reduction behavior are determined. Subsequently, the metallic phase properties are analyzed using X-ray diffraction (XRD) and X-ray photoelectron spectroscopy (XPS). Finally, the catalysts’ performance is assessed by comparing their activity, selectivity, and stability.

2.1. Catalyst Characterization

The N₂-sorption measurements were used to determine the BET surface area and pore volume, which are provided in

Table 1. BET surface area (S_BET) with 95% confidence interval, pore volume (V_p), the number of acid sites (C_acid), the number of basic sites (C_basic), the activity with 95% confidence interval, and the selectivity to cyclohexane (S_Cyclohexane) at 75% conversion. The conditions used for the catalyst performance measurements were a temperature of 200 °C, a pressure of 0.5 Mpa, and a H₂/anisole (${\mathrm{m}\mathrm{o}\mathrm{l}}_{{\mathrm{H}}_2}{\mathrm{m}\mathrm{o}\mathrm{l}}_{\mathrm{a}\mathrm{n}\mathrm{i}\mathrm{s}\mathrm{o}\mathrm{l}\mathrm{e}}^{-1}$) of 50. The space time was varied between 3–52 kg s ${\mathrm{m}\mathrm{o}\mathrm{l}}_{\mathrm{a}\mathrm{n}\mathrm{i}\mathrm{s}\mathrm{o}\mathrm{l}\mathrm{e}}^{-1}$. Measurements were consistently taken between 1 h and 2.5 h of Time On Stream.

Material	SBET (m2 g−1)	Vp (cm3 g−1)	Cacid (µmol NH3 g−1)	Cbasic (µmol CO2 g−1)	Activity (molanisole s−1 kgcat−1)	SCyclohexane (%)
γ-Al2O3	240 ± 37	0.60 ± 0.10	427	36	/	/
B-Al2O3	219 ± 21	0.65 ± 0.11	280	87	/	/
NiCu-γ	172 ± 49	0.38 ± 0.11	90	17	0.062 ± 0.011	49 ± 6
CaO-NiCu-γ	121 ± 24	0.26 ± 0.04	33	63	0.072 ± 0.012	1.6 ± 0.3
MgO-NiCu-γ	141 ± 42	0.33 ± 0.09	51	57	0.127 ± 0.034	8 ± 2
CaO-MgO-NiCu-γ	117 ± 20	0.28 ± 0.07	54	58	0.093 ± 0.024	1.5 ± 0.1
NiCu-B	155 ± 12	0.45 ± 0.02	50	56	0.069 ± 0.017	1.3 ± 0.2

and in the Supplementary Materials Figure S1. The pristine γ-Al₂O₃ support and all, correspondingly synthesized catalysts exhibit a type IV isotherm with H2-type hysteresis, indicating their mesoporosity and a poorly defined distribution of pore sizes and shapes [32]. The commercial B-Al₂O₃ support and the NiCu-B catalyst derived therefrom exhibit an H3-loop, stemming from a pore network that consists of meso- and macropores which, during the N₂-sorption measurement, could not be completely filled with pore condensate. This behavior is often observed with aggregates of plate-like particles, giving rise to slit-shaped pores [33]. The BET surface area and pore volume dropped after impregnation when the support materials and the respective catalysts are compared. These variations can be ascribed to the formation of nanoparticles on the surface of the support, leading to pore blockage. Moreover, the impregnation of 20–25% metal particles into the support does not substantially alter the total surface area. However, it does significantly impact the density of the materials, leading to a reduction in surface area per catalyst mass. Furthermore, no notable disparity is observed in the BET surface area and pore volume between the γ-Al₂O₃ catalysts that have been modified with CaO and/or MgO and the unmodified NiCu catalyst.

The quantification of acid sites was performed using NH₃-TPD. The corresponding results can be found in Table 1 and Figure 1. At temperatures exceeding 600 °C, a pronounced peak is observed for the basic oxide-modified catalysts, not attributable to NH₃ desorption. The CO₂ MS signal in Figure 1 indicates a substantial release of CO₂ at high temperatures, aligning well with the shape of the high-temperature peak. The dissociation of carbonates, leading to the release of components such as CO₂ with fragmentation patterns matching the measured m/z for NH₃, was confirmed using available fragmentation patterns from the National Institute of Standards and Technology (NIST) [29,34,35]. The presence of these carbonates on all basic oxide-modified materials was verified via XRD and XPS, see below. To quantify the number of acid sites present, a log-normal distribution (dashed line in Figure 1) was fitted to the low-temperature peak when CO₂ is detected via MS, as it provided the best fit. Otherwise, the entire desorption profile was integrated for quantification. The pristine γ-Al₂O₃ has a higher number of acid sites compared to B-Al₂O₃, i.e., 427 and 280 µmol NH₃ g⁻¹, respectively. Both supports mainly desorbs NH₃ at temperatures below 500 °C (see Figure 1), suggesting that these acid sites are primarily weak in nature [36]. After impregnation and reduction of NiCu, the acidity of the γ-Al₂O₃ and B-Al₂O₃ support is decreased to 90 and 50 µmol NH₃ g⁻¹, respectively, which can be explained by the metal nanoparticles covering the Lewis acid sites present on Al₂O₃ and the change in density after impregnation. The addition of Ca or Mg to the γ-Al₂O₃ support prior to the impregnation with NiCu leads to an even larger drop in acidity, due to above-mentioned reasons. This decrease is more pronounced for CaO-NiCu-γ, related to the different amounts of modifier added, i.e., 1 wt% MgO vs. 4.5 wt% CaO. Surprisingly, the CaO-MgO-NiCu-γ catalyst exhibits a similar number of acid sites compared to the MgO-NiCu-γ catalyst, even though the amount of loaded basic oxides is higher, indicating the existence of other effects, such as basic oxide dispersion. The maximum of the desorption peak, related to the strength of the acid sites, was similar for all materials, and was thus not further investigated [37,38].

The quantification of basic sites was done via CO₂-TPD and the results can also be found in Table 1 and Figure 1. At high temperatures (>600 °C), an intense peak is observed for the basic oxide-modified catalysts, with the exception of MgO-NiCu-γ. This peak is attributed to the decomposition of formed carbonates, and its shape is in line with the high-temperature peak observed in NH₃-TPD [29,34]. Recognizing that this peak is not related to the basicity of the catalyst, it was excluded from the quantification of the number of basic sites. The γ-Al₂O₃ support has some basicity when probed via CO₂-TPD, with two intense peaks located in the low-temperature region, i.e., at 235 °C and at 315 °C, indicating that the corresponding sites are rather weak. These basic sites are often linked to surface OH-groups and the decomposition of HCO₃⁻ species adsorbed on the surface [29,39,40]. Due to the presence of CaO and MgO in the B-Al₂O₃ support, the number of basic sites is remarkably higher than for the γ-Al₂O₃ support. Impregnating with NiCu causes a decrease in number of basic sites for both supports, explained by a change in density after impregnation and the coverage of the basic sites by the NiCu nanoparticles. Prior impregnation of the γ-Al₂O₃ support with Ca and/or Mg leads to an increase of basic sites compared to the unmodified NiCu catalyst, as has been observed in the literature before [26,41]. The increase in CO₂ desorption is associated with the adsorption onto Lewis basic sites of the basic oxides. All modified catalysts have a similar number of basic sites, i.e., 60 µmol CO₂ g⁻¹, independent from the oxide used and its loading. This observation suggests that factors beyond the type, e.g., CaO is inherently a more basic metal oxide than MgO, and the quantity of basic oxide plays a role [26,42,43]. The strength of the basic sites, indicated by the maximum of the desorption peak, was comparable for all materials, and was thus not further considered [37,38].

The reducibility of the catalysts was evaluated via H₂-TPR and is presented in Figure 2. The initial peak, noticeable at lower temperatures (150–300 °C), signifies the transformation of Cu²⁺ to Cu⁰ [44,45]. This transformation is followed by the second and third peak, observed at moderate temperatures (220–450 °C), and attributed to the reduction of CuO interacting more strongly with the support and NiO interacting with Cu [45,46]. The fourth peak, observed at high temperatures (400–550 °C) is attributed to well-dispersed NiO particles exhibiting strong metal-support interactions [45,47]. When Ca is added, a shift towards higher reduction temperatures is observed, as well as a change in the overall reduction profile, implying that the addition of Ca does have an effect on the reduction properties by impacting the interaction intensity of Ni and/or Cu with the support. An intense peak (actually consisting of multiple peaks) at ±250 °C is observed, followed by a raise in intensity compared to the baseline until approximately 600 °C. The addition of Mg does not lead to a major shift in reduction properties compared to the benchmark catalyst. This indicates that the interaction of Mg with Ni and Cu is minimal, most likely due to low amounts added. The NiCu-B catalyst had a similar profile as the CaO-NiCu-γ, suggesting that the interactions of the basic oxides with the reducible metals are similar. Finally, CaO-MgO-NiCu-γ has a reduction profile that is spread out again in multiple distinguishable peaks, but they are clearly different compared to the ones of the benchmark catalyst and the MgO-NiCu-γ catalyst. Correspondingly, the addition of both MgO and CaO leads to an effect where some phases are affected by the addition of CaO, resulting in more intense peak at about 250 °C, and some are not, which have a reduction profile that is more similar to the MgO-NiCu-γ one. As a TCD detector was used during the TPR experiments, the decomposition of carbonates could not be distinguished, since CO₂ and Ar have a similar thermal conductivity. Upon reviewing the TPR results, a reduction temperature of 450 °C was selected for the catalyst before the reaction. At this specific temperature, all metal oxides, except finely dispersed NiO, undergo a reduction. Higher temperatures were avoided to prevent severe sintering during the reduction process.

The crystalline phases were examined via XRD and shown in Figure 3 and in the Supplementary Materials Figure S2. All supports and catalysts exhibit a similar diffractogram. Diffractions observed at 46.26° and 67.22° can be attributed to γ-Al₂O₃. A scattering peak is found at 32.70°, indicating that the used supports also exhibit some amorphous character. The B-Al₂O₃ support has intense CaCO₃ diffractions at, among other angles, 29.42°, 35.96°, and 57.42° which disappeared in its corresponding catalyst, indicating that the carbonates have largely decomposed at the chosen calcination/reduction conditions [48]. Diffractions attributed to the MgO phase were not observed, probably owing to the limited quantity of MgO present in the modified catalyst. Similarly, any CaO phases are not distinguishable, likely due to overlapping with the γ-Al₂O₃ phase. These findings align with earlier research on the modified B-Al₂O₃ [28,29]. The diffraction peaks at 43.3°, 50.43°, 74.13°, and 89.93° indicate the formation of a monometallic Cu phase during reduction. The sharpness of these diffractions suggests that the phase primarily consists of large Cu particles, as is consistent with the literature [25,49,50]. An additional diffraction, seen between 43.3° and 44.51°, indicates the formation of a NiCu alloy phase. This alloy diffraction appears broader in comparison to the pure Cu diffraction peak, suggesting that the alloy particles are smaller than those of Cu. The crystallite size was not determined via the Scherrer equation due to considerable overlap with adjacent phases, rendering an accurate determination impossible via conventional techniques. Furthermore, at 43.27°, there is a small and broad diffraction corresponding to a well-dispersed NiO phase on the γ-Al₂O₃ support [25,51]. This finding is consistent with the expectation in H₂-TPR that reducing the materials at 450 °C does not fully reduce the NiO phase.

The results concerning the surface characterization of the reduced catalysts, performed with XPS analyses, are displayed in Figure 4. The C1s signal generally has 2 distinct peaks for all reduced catalysts, one at 285 eV, related to adventitious carbon, and one at higher binding energies, indicating the existence of carbonates on the surface. The O1s peak is asymmetric and could be deconvoluted into an intense main peak and a shoulder at higher binding energies, which can be assigned to oxygen in the metal oxides and oxygen of superficial carbonates and/or hydroxides, respectively [52]. The MgO-NiCu-γ catalyst also exhibits the presence of carbonate species, despite the absence of a desorption peak at high temperatures in the CO₂-TPD analysis. This suggests a low concentration of carbonates on the catalyst surface, rendering the corresponding TPD signal weak or practically imperceptible. Additionally, a shift in binding energy is observed when the NiCu-B catalyst is compared to others, which signifies different electronic interactions, potentially stemming from variances in synthesis methods compared to the self-synthesized catalysts. Ni2p is split up into three peaks, a small one related to Ni⁰, a pronounced one linked to Ni²⁺, and a satellite peak at higher binding energies. The Ni⁰ peak could not be observed for the CaO-MgO-NiCu-γ and the NiCu-B catalyst due to a low XPS signal, resulting in a high signal-to-noise ratio. For the Cu2p was observed at 933 eV, linked to Cu⁰. The CuLMM peak could not be observed for further distinction between Cu⁰ and Cu⁺. No clear trend could be observed in surface composition when Ca or Mg is added. Ca2s is observed in all CaO-containing samples, except NiCu-B, likely due to the low XPS signal, as mentioned earlier. On the other hand, Mg2p is not detected at approximately 52 eV due to the small amounts added, and because it is a light element, which typically has a higher detection limit in XPS. However, the MgKLL region does indicate that the Mg surface species are oxides, as evidenced by the binding energy being around 306 eV. To summarize the XPS results, the carbonate species are detected across all catalyst samples, manifesting themselves through distinct peaks in the C1s and O1s spectra. While Ca and Mg are identified in the corresponding samples, the NiCu-B sample is an exception due to weak XPS signal strength. Cu predominantly exists as Cu⁰ (with a potential presence of Cu⁺), and nickel is found in both Ni²⁺ and Ni⁰ states, though the Ni⁰ state was not observable in the CaO-MgO-NiCu-γ and NiCu-B catalysts due to low XPS signals.

2.2. Catalyst Performance

The activity of the catalysts, determined for anisole HDO, is presented in Table 1, at a temperature of 200 °C, a pressure of 0.5 Mpa, and a H₂/anisole ratio (${\mathrm{m}\mathrm{o}\mathrm{l}}_{{\mathrm{H}}_2}{\mathrm{m}\mathrm{o}\mathrm{l}}_{\mathrm{a}\mathrm{n}\mathrm{i}\mathrm{s}\mathrm{o}\mathrm{l}\mathrm{e}}^{-1}$) of 50. The determined order of activity was (including 95% confidence interval): MgO-NiCu-γ > CaO-MgO-NiCu-γ > CaO-NiCu-γ ≈ NiCu-B ≈ NiCu-γ.

Adding 4.5 wt% CaO does not seem to have a noticeable impact on the activity of the catalysts, although it seems slightly improved compared to the unmodified NiCu-γ catalyst. The literature shows that CaO can improve the dispersion of Ni and Cu, resulting in a higher activity, and retard its sintering behavior by strengthening the interaction between the metal and Al₂O₃ [53,54]. It has also been suggested for CaO-modified Pd-γ catalysts that CaO electronically enriches the metallic Pd phase, which could lead to a better activation of hydrogen on the palladium sites, improving its activity [55]. Adding 1 wt% MgO does improve the activity remarkably compared to the benchmark catalyst. In the literature, it is generally accepted that MgO as an additive can improve the dispersion of transition-metal oxide phases on alumina or silica supports and correspondingly lead to a higher specific activity [23,56,57,58,59]. Co-impregnation of CaO and MgO onto the γ-Al₂O₃ support resulted in an intermediate activity compared to its individual oxide-modified catalysts. The NiCu-B catalyst exhibited a similar activity compared to CaO-NiCu-γ, suggesting that the materials are quite similar, which was also already indicated by H₂-TPR. More fundamental conclusions about the structure-activity relationships of the catalysts could not be made based on our data, and would require synchrotron-based methods and post-reaction analyses, which were considered out of scope for this manuscript.

The selectivities of the catalysts are shown in Figure 5 as a function of conversion and in Table 1 at 75% conversion. The carbon-selectivity is provided specifically for methoxycyclohexane and cyclohexane, since all other products exhibit selectivities below 3%. Methanol, while following the cyclohexane trend, exhibited a selectivity six times lower, attributed to the definition of carbon-selectivity and the reaction stoichiometry. In the case of the benchmark NiCu-γ catalyst, it is observed that methoxycyclohexane is primarily formed at lower conversions, whereas cyclohexane is formed at higher conversions. Correspondingly, methoxycyclohexane is the primary product that converts into cyclohexane and methanol in a second reaction step. The hydrogenation of the ring is associated with the presence of metallic sites, while the demethoxylation is linked to the available acid sites. All modified catalysts exhibit, in the measured conversion range, a remarkably lower HDO selectivity, i.e., the selectivity towards cyclohexane, compared to the benchmark catalyst. Catalysts that were modified via CaO all have a cyclohexane carbon-selectivity of roughly 1.5% at 75% conversion, while the MgO-NiCu-γ has a selectivity of 8%. It is clear that the demethoxylation step, linked to the acidity of the catalyst, is inhibited by adding a basic oxide. This inhibition effect is more pronounced when 4.5 wt% of CaO is added compared to 1 wt% of MgO, and is correlated with the decrease in acidity of the modified catalysts compared to the benchmark. The more basic sites present on the surface on the catalyst, the higher the chance that these basic oxide sites will interact with the Lewis acid sites of γ-Al₂O₃. This interaction could weaken the acid sites present, or the basic sites themselves could compete with the acid sites for reactant adsorption, resulting in a reduced demethoxylation [60]. NH₃-TPD, which was used to probe the acidity of the materials, might not indicate this weakening because of the physicochemical differences between methoxycyclohexane and NH₃ as a probe molecule. NH₃ has a lone pair of electrons on the nitrogen atom, making it a Lewis base. This lone pair readily interacts with Lewis acid sites, resulting in a stronger adsorption compared to methoxycyclohexane, which has a lone pair on the oxygen atom. Due to the lower electronegativity of the nitrogen atom compared to the oxygen atom, the lone pair on nitrogen is held less tightly, making it more available for donation and therefore a stronger base. Consequently, the weakened acid sites might still be strong enough for interaction with NH₃, but not for methoxycyclohexane.

The stability of the catalysts was assessed at the same conditions as the activity and selectivity (see Figure 6). The space time was varied to achieve a roughly equal conversion for all catalysts. The activity loss of MgO-NiCu-γ and CaO-MgO-NiCu-γ is similar or even worse compared to the benchmark NiCu-γ catalyst, i.e., approximately 25% deactivation over 48 h TOS. The NiCu-B and the CaO-NiCu-γ catalyst have an improved stability, with only a loss of conversion of about 10% and 5%, respectively. A similar improved stability behavior has been observed in the literature for the deoxygenation of palm oil employing a Ni-B catalyst, and was ascribed to the ability of the CaO and MgO modifiers to inhibit carbon deposition and/or sintering [28]. Since the CaO-MgO-NiCu-γ catalyst does not exhibit this stabilizing behavior, and, moreover, a different activity was observed compared to the NiCu-B catalyst, it indicates that the results obtained over the self-modified catalyst could not be fully translated to commercial B-Al₂O₃ support, most likely because of a difference in preparation method between the self-synthesized basic oxide support and the commercially available one. In terms of activity, selectivity and stability, the CaO-NiCu-γ catalyst is most similar to NiCu-B, suggesting that CaO is the main acting modifier in the commercial B-Al₂O₃ support. Both catalysts exhibit an enhanced stability for anisole HDO at the expense of the demethoxylation activity, observed as a reduction in cyclohexane selectivity compared to the benchmark catalyst (see Figure 6). This reduction in selectivity could potentially explain their increased stability; catalysts that engage less in acid-catalyzed reactions, such as demethoxylation, are less prone to undergo coking under the conditions applied. This is because the moderated Lewis acid sites on these catalysts are less likely to facilitate coke formation through condensation reactions [13,61,62].

3. Materials and Methods

3.1. Catalyst Preparation

Two types of supports were used to prepare the catalysts: γ-Al₂O₃ (PURALOX SCCa150/200, Sasol, Johannesburg, South Africa) and a commercially available basic oxide modified support (B-Al₂O₃, JR 323 basic alumina, Saint-Gobain Norpro, Stow, OH, USA). The commercial support is a γ-Al₂O₃ support with 5 wt% CaO, 1.0 wt% MgO and 0.5 wt% of SiO₂.

All catalyst preparations started with dissolving the required amount of metal salt precursors, i.e., Cu(NO₃)₂·3H₂O (99.5%, VWR), Ni(NO₃)₂·6H₂O (99.999%, Sigma Aldrich, St Louis, MO, USA), Ca(NO₃)₂·4H₂O (99%, Sigma Aldrich), and Mg(NO₃)₂·6H₂O (99%, Sigma Aldrich) in water, equivalent to the support’s pore volume. The commercial support was loaded in one step with NiCu, while the other catalysts were prepared by modifying the support with basic oxides prior to NiCu loading. The solution was added dropwise to the support under constant stirring. A consistent molar Ni/Cu ratio of 1/1 and NiCu loading of 20 wt% was maintained. The catalysts were subjected to a consistent drying and calcination procedure: dried at 120 °C for 12 h and calcined at 450 °C for 4 h in air with temperature ramps of 2 °C min⁻¹. In case of sequential impregnation, an intermittent calcination was performed under the same conditions. Sequential impregnation of metal oxides was performed on the γ-Al₂O₃ support with 4.5 wt% CaO and/or 1.0 wt% MgO, producing the catalysts CaO-NiCu-γ, MgO-NiCu-γ, and CaO-MgO-NiCu-γ. The loadings of CaO and MgO used are identical to those applied to the commercial B-Al₂O₃ support. It should be noted that there is also 0.5 wt% of SiO₂ present on the commercial support. However, such a small amount is not expected to have any significant effect [63,64].

3.2. Catalyst Characterization

The specific surface area and pore volume of each catalyst were obtained using a TriStar II 3020 device (Micromeritics, Atlanta, GA, USA). Nitrogen adsorption–desorption profiles were recorded at −196 °C, following a 3 h degassing process at 300 °C for approximately 50 mg of the sample, to eliminate volatile surface contaminants. The Brunauer–Emmett–Teller (BET) approach was employed to calculate the specific surface area, whereas the Barrett–Joyner–Halenda (BJH) technique was used for determining the average pore volume. Each measurement was conducted in triplicate to obtain average values, accompanied by their respective 95% confidence intervals.

The quantity of acid sites was probed via temperature-programmed desorption (TPD) of ammonia (NH₃), utilizing an Autochem 2920 apparatus (Micromeritics, Atlanta, GA, USA). Linked to both a thermal conductivity detector (TCD) and a mass spectrometer (MS) to confirm the TCD reading accurately reflected NH₃ desorption. In preparation for the test, around 100 mg of the materials underwent a purge with high-purity helium (50 cm³ min⁻¹) at 120 °C for an hour to eliminate water, followed by a 5 min purge at 200 °C to clear any other possible adsorbates. The catalysts were then subjected to an in situ reduction under a 5% H₂/Ar flow (60 cm³ min⁻¹) at 450 °C over 3 h. Post-reduction, a 4% NH₃/He mixture (75 cm³ min⁻¹) saturated the catalysts for an hour at 150 °C. Afterward, helium (60 cm³ min⁻¹) was flowed across the materials for an hour to extract physisorbed NH₃. The temperature was then increased to 800 °C at a 10 °C per minute ramp with a dwell time of 30 min, while monitoring NH₃ release. To quantify the number of acid sites present, a log-normal distribution was fitted to the low-temperature peak when CO₂ is detected via MS, as it provided the best fit. The high-temperature peak (>600 °C) was lined to a release of CO₂ due to the decomposition of formed carbonates. If no high-temperature peak was present, the entire desorption profile was integrated for quantification.

Similarly, the number of basic sites was evaluated using the same methodology, but with carbon dioxide (CO₂) instead of NH₃. The materials, pretreated under identical conditions, were exposed to a 10% CO₂/He mixture under similar flow and temperature conditions. Also, removal of physically adsorbed CO₂ was ensured by purging with helium at 150 °C. The subsequent temperature ramp-up allowed for the observation of CO₂ desorption, thereby determining the number of basic sites present. At high temperatures (>600 °C), an intense peak is observed for the basic oxide-modified catalysts. This peak is attributed to the decomposition of formed carbonates, and its shape is in line with the high-temperature peak observed in NH₃-TPD. Recognizing that this peak is not related to the basicity of the catalyst, it was excluded from the quantification of the number of basic sites.

The reduction characteristics of the metal phases in the catalysts were investigated through temperature-programmed reduction (TPR) using the identical Autochem 2920 device. Approximately 100 mg of the calcined catalyst was placed in a U-shaped quartz tube reactor, with a thermocouple inserted within the catalyst layer. Before initiating the analysis, the sample was first flushed with argon (50 cm³ min⁻¹) at 120 °C for one hour, followed by a 5 min purge at 200 °C. The reduction process was carried out with a 5% H₂/Ar mixture (60 cm³ min⁻¹), employing a temperature ramp of 10 °C per minute. The H₂ consumption was monitored from 50 °C to 800 °C, with a dwell time of 30 min.

The crystalline phases were analyzed using powder X-ray diffraction (XRD) via a Diffractometer Kristalloflex D5000 (Siemens, Karlsruhe, Germany), employing Cu Kα radiation. Before the XRD procedure, the catalysts crushed into a fine powder and reduced in an Autochem 2920 instrument at 450 °C for 3 h. The reduction environment was selected to mimic the reduction prior to reaction in the catalyst performance testing. To minimize contact with air, samples were preserved and transported in an atmosphere of high-purity argon. The diffractograms were collected in the 2θ range from 10° to 110° with a step size of 0.02 s⁻¹ and 10 s collection time.

To verify the dispersion of the NiCu catalysts, CO and H₂ chemisorption was attempted, but neither of them gave accurate results, as has been seen before in the literature [24,65]. A N₂O-based procedure for Cu catalysts described by Gervasini et al. was also attempted, but also did not lead to accurate data compared to previous work [11,25].

Details regarding the metal surface composition were obtained through X-ray Photoelectron Spectroscopy (XPS) (S-Probe XPS spectrometer, VG Scienta, Uppsala, Sweden) conducted in ultra-high vacuum environments, using a monochromatized 450 W Al Kα source. The analysis chamber maintained a base pressure below 2∙10⁻⁷ Pa. Spectra collection was performed with a source power of 200 W. The analyzer axis made an angle of 45° with the specimen surface. Individual core levels were measured with a pass energy of 40.145 eV, and 140.83 eV was used for survey scans, using a step of 0.01 and 0.02 eV, respectively. The Ni2p, Cu2p, Al2p, Si2p, Mg2p, MgKLL, Ca2s, C1s, and O1s binding energies were recorded. These binding energies were calibrated to the adventitious carbon of C1s component at 285.0 eV. A Lorentzian asymmetric line shape and U 2 Tougaard background were used during the fitting procedure. Prior to the XPS analysis, the catalysts were reduced in an Autochem 2920 instrument at 450 °C for 3 h. To limit air exposure, samples were stored and transferred in high purity argon.

3.3. Catalyst Performance Testing

The catalyst performance of the materials was assessed in a high-throughput screening set-up, described in detail elsewhere [66]. On-line analysis was conducted through GC-FID assessments on a TraceGC1310 (Thermo Fisher Scientific, Waltham, MA, USA), outfitted with an Rtx-PONA column (L = 100 m, i.d. = 0.25 mm). Ethane served as the internal standard added after the reactors (comprising 20 ± 0.5 mol% ethane in N₂, supplied by Air Products, Allentown, PA, USA) to calculate flows and confirm mass balances within a 5% range.

Before loading the catalysts into the reactors, they were crushed, sieved, and pelletized until a particle size of 100–125 µm was achieved. It was verified, both theoretically as well as experimentally, that intrinsic reaction rates are measured under the considered operating conditions using these particle sizes (see the Supplementary Materials Table S1) [66]. Preceding the catalytic evaluations, an in situ reduction of the materials was executed based on the subsequent methodology. The catalysts were initially dried at 120 °C for an hour, then at 200 °C for another hour in an N₂ atmosphere at 0.5 mPa, utilizing temperature ramps of 5 °C min⁻¹. This phase was followed by a reduction stage where a H₂ stream was introduced to the reactors at 450 °C and 0.5 mPa for a duration of 3 h. Upon completion, the temperature was lowered to initiate the data collection under specific reaction conditions: a temperature of 200 °C, pressure maintained at 0.5 mPa, and a H₂/anisole molar ratio set at 50. The space time was adjusted between 2.5–10 kg_cat s mol⁻¹ anisole by modifying either the catalyst weight or reactant throughput, aiming to gather data across a conversion range of roughly ±15–100%. Pristine supports were tested prior to the catalytic tests and were found to be inactive. Catalyst deactivation was studied as a function of Time On Stream (TOS). The start of these experiments (i.e., TOS = 0) was defined as the moment at which the anisole feed to the reactors was applied. To avoid cross-contamination amid different experiments, data collection was started roughly 1.5 h after the start of an experiment, guaranteeing the downstream segment was adequately purged by the reactor’s effluent.

The conversion of anisole ($X_{anisole}$) was calculated on molar basis and is defined as in Equation (1):

$$X_{anisole}=\frac{F_{anisole}^0-F_{anisole}}{F_{anisole}^0}$$

(1)

where $F_{anisole}^0$ and $F_{anisole}$ are, respectively, the inlet and the outlet molar flow rate of anisole. The carbon selectivity towards a product k ($S_k$) was calculated by Equation (2):

$$S_k=\frac{n_k{F}_k}{7(F_{anisole}^0-F_{anisole})}$$

(2)

where $F_k$ is the outlet molar flow rate of product k and $n_k$ is the number of carbon atoms present in a molecule of product k. The activity is expressed as space time yield (STY, in mol_Anisole s⁻¹ kg_cat⁻¹):

$$STY=\frac{\mathsf{\Delta }{X}_{anisole}}{\mathsf{\Delta }\left(\frac{W_{cat}}{F_{anisole}^0}\right)}$$

(3)

4. Conclusions

This work investigated the impact of basic oxide-modified γ-Al₂O₃ supports on the performance of NiCu catalysts for anisole HDO. Depending on whether CaO and/or MgO were sequentially added prior to NiCu incorporation onto the γ-Al₂O₃ support, the catalysts were labeled CaO-NiCu-γ, CaO-MgO-NiCu-γ, MgO-NiCu-γ, and NiCu-γ, with the latter used as a benchmark catalyst.

The activity decreased in the order of MgO-NiCu-γ > CaO-MgO-NiCu-γ > CaO-NiCu-γ ≈ NiCu-B ≈ NiCu-γ. The presence of MgO improves the metal dispersion. When CaO and MgO are co-impregnated onto the γ-Al₂O₃ support, the resulting catalyst exhibits intermediate activity compared to the individually modified catalysts. The NiCu-B catalyst exhibits similar activity to the CaO-NiCu-γ.

However, all modified catalysts demonstrate a significant decrease in HDO selectivity compared to the benchmark catalyst. This suggests that the addition of basic oxides hinders demethoxylation by moderating the support’s acidity. An increased basic oxide loading enhances the likelihood of interaction between basic oxide sites and the Lewis acid sites on γ-Al₂O₃, moderating these acid sites or competition with them for reactant adsorption, ultimately reducing demethoxylation.

Both CaO-NiCu-γ and NiCu-B catalysts exhibit improved stability over a 48 h TOS period, unfortunately at the cost of selectivity towards deoxygenated products, suggesting that CaO addition is crucial to obtain a higher stability. In terms of activity, selectivity, and stability, the CaO-NiCu-γ catalyst is most similar to NiCu-B, suggesting that CaO is the main acting modifier in the commercial B-Al₂O₃ support.

In short, this work found that incorporating basic oxides can enhance hydrogenation activity and stability under mild HDO conditions (200 °C, 5 bar), albeit at the expense of demethoxylation activity, due to the moderating effect on the support’s acid sites. Future investigations should consider exploring adjusting the loadings of basic oxides and exploring harsher conditions, potentially leading to improved selectivity in the HDO process.