Single-Atom and Sub-Nano Ruthenium Cluster Catalysts—Application to Biomass Upgrading into Biofuel Additive

Abstract

Sub-nano metal clusters have important physicochemical features that lead to a wide range of applications. Herein, we point out an unfailing reproducible protocol to synthesize ruthenium single-atom catalysts and ultra-small clusters supported on various silica–alumina mixed oxides. The catalysts were synthesized via a dendrimer-free, sonication-assisted route, with ruthenium loadings up to 2 wt%. Raman spectroscopy mapping revealed a wide coverage of the materials’ surfaces by ruthenium, while HAADF-STEM evidenced that 100% of the ruthenium was at the sub-nano scale, with up to 74% of the single atoms and metal clusters having an average size between 0.3 and 0.7 nm, independently of the support or the metal’s loading. These materials exhibited highly selective size-dependent catalytic performances in upgrading biomass-derived furfural into transportation biofuel additive 2,2′-difurfurylether, with turnover frequencies up to 1148 h⁻¹. Ruthenium single atoms and sub-nano clusters showed an exceptional resistance to sintering, with a size variation of ±0.1 nm before and after reaction, and no metal leaching was observed.

1. Introduction

In recent years, single-atom catalysts (SACs) and sub-nanoparticle-supported catalysts (SNCs) have emerged as a fascinating frontier in heterogeneous catalysis, offering unprecedented opportunities for boosting and enhancing both catalytic activity and selectivity. These ultra-small metals, with up to only tens of atoms and high surface-to-volume ratios, exhibit complex and unique electronic structural properties [1], such as a quantum size effect, surface restructuring and atomically precise active sites, and strong metal–support interactions, distinguishing them from their larger counterparts and making them highly desirable for a wide range of catalytic applications [2,3]. Indeed, among the numerous factors influencing nano-catalysis, the size of metallic nanoparticles stands out as a critical and paramount parameter that plays a pivotal role in determining their electronic, geometric and chemical properties [4]. As metallic clusters shrink to the nano-scale and below, their surface-to-volume ratios increase significantly, leading to an abundance of active sites, consequently improving catalytic performances. Moreover, the accessibility to specific crystal facets and surface sites influences the adsorption and desorption of reactants and intermediates, allowing the optimization of catalytic selectivity, a critical aspect in the sustainable production of desired compounds while minimizing unwanted by-products [5]. Tailoring metal particles into tiny clusters brings with it major difficulty in synthesis. Nevertheless, a few authors have reported the successful synthesis of catalysts containing supported sub-nano clusters of platinum [6,7], palladium [8], gold [2,9], silver [1,10] and copper [11], showcasing the direct relationship between size and catalytic performance.

Ruthenium occupies a significant position in organometallic and coordination chemistry, as well as catalysis, because of its unique properties, in addition to the fact that it is the least expensive of the nickel-group metals [12]. Indeed, ruthenium heterogeneous catalysts have emerged as powerful tools exhibiting versatility and efficiency in various chemical processes, such as organic synthesis [13], hydrogenation [14,15], hydrodeoxygenation [16,17], reforming [18], oxidation reactions [19,20], oxygen activation [21], energy conversion [22] and storage [23,24], biofuel production [17,25] and biorefinery [26]. As research in these fields continues, the development of new ruthenium catalysts with enhanced performance is expected, further expanding their applications.

In their review dedicated to ruthenium-based catalysts, Akram et al. [14] thoroughly reported on the catalytic transfer hydrogenation of furfural. The catalysts’ outstanding activities were directly related to the Ru nanoparticles’ size and dispersion, as the reactivity increased with the decrease in Ru nanoparticles’ size (from 12 to 1 nm). This direct size activity relationship was also highlighted by Zhu et al. for supported Ru-Ni bimetallic catalysts in benzene [27] and naphthalene [28] hydrogenation, as well as in the transformation of levulinic acid into γ-valerolactone over Ru/C catalysts by Ruiz-Bernal et al. [29]. The unprecedented catalytic activities were attributed to well-dispersed Ru nanoparticles on which hydrogen was adsorbed and activated, followed by H atom spillover towards Ni(OH)₂ and Ru(O)_xH⁺ sites, respectively, through a highly effective pathway. A similar adsorption–activation–spillover pathway was also evidenced in the case of ¹⁶O/¹⁸O isotopic exchange [21], where oxygen was adsorbed and activated on ruthenium nanoparticles at lower temperatures compared to the other platinum-group metals. Then, oxygen spillover toward the ceria-based supports led to very high oxygen storage capacities, a key parameter in three-way catalysis for automotive pollution abatement [30]. These features are at the origin of the efficiency of small ruthenium-supported nanoparticles in biomass transformation [31], and several authors have demonstrated the crucial role of ruthenium nanoparticles in 5-hydroxymethylfurfural (HMF) oxidation [32,33], furfural hydrogenation [14,34] and guaiacol [35,36], anisole [37] and levulinic acid [38] hydrodeoxygenation.

The key step to obtaining small or ultra-small ruthenium particles is their synthesis. Various preparation routes, aiming to decrease the metal particle size under 5 nm, have been reported in the literature, investigating different strategies to improve metal dispersion. In the case of monometallic ruthenium catalysts, deposition–precipitation and wet and incipient wetness impregnations are widely used. In this approach, supports with high surface areas [23,39,40], Lewis/Brønsted acid sites [35,41], or oxygen vacancies and defects [36,37,42] are required to enhance ruthenium dispersion. A second strategy is to stabilize ruthenium particles to prevent aggregation and sintering. Ruthenium immobilized in a carbon matrix [43] or on nitrogen-doped carbon sheets [44]; three-dimensional alumina-cogelled Ru nanoparticles Ru@Al [25], ruthenium cluster embedded Ni(OH)₂ nano-sheets [33], thermo-regulated ionic liquid-stabilized Ru/CoO [45] or Ru atoms incorporated in the trivalent Fe site of a monolayer NiFe double hydroxide LDH [46] were reported. A third alternative strategy consists of adding a second metal as a promoter. Several authors reported a synergetic effect in bimetallic catalysts, such as Ru-Pd [47], Ru-Cs [24], Ru-Cu [16] and Ru-Ni [48], with very small ruthenium particle sizes ranging from 1 to 5 nm. However, 1 nm seemed to be a threshold hard to go under.

In the present study, aiming to tailor ruthenium catalytic features, we will demonstrate a very efficient and reliable route to synthesize highly stable ruthenium single atoms and sub-nano clusters supported on recalcitrant silica–alumina mixed oxides with Ru loadings up to 2 wt%. Those materials enable innovative biomass upgrading, which involves transportation fuel additive production to boost the cetane number and the ignition characteristics. Indeed, in the literature, difurfurylether (DFE) is mostly reported as a side product formed upon the polymerization of furfuryl alcohol (FA) in the presence of acid sites [49,50,51]. Kim et al. [52] showed that the formation of initial ether bridges from two neutral FA molecules is marginally thermodynamically favorable (−3.8 KJ·mol⁻¹); however, the formation of an ether-bridged dimer from FA and a carbocation is thermodynamically up-hilled by 14.6 KJ·mol⁻¹, which is consistent with the experimental studies that indicate that ether-bridged polymers (H₂C-O-CH₂) result from HO-OH condensation of two FA monomers [53,54]. Difurfurylether was also reported as a side product during the catalytic transfer hydrogenation of furfural in protic solvents [16,55,56,57], where it was formed via the self-etherification of two furfuryl alcohol molecules. Makarouni et al. [58] successfully synthesized difurfurylether from furfuryl alcohol using a mordenite catalyst with strong Brønsted acid sites. DFE yield reached 30% when di-(2-methoxyethyl) ether was used as a solvent, but dropped to only a 5% yield in DFE when a mixture of di-(2-methoxyethyl) ether and ethanol was used. Yang et al. [59] used polyoxometalate catalysts to selectively catalytically dehydrate furfuryl alcohol. The condensation of two FA molecules led to a 31% yield of DFE. In the present work, DFE yields ranging from 86 to 97% are reported.

2. Results

2.1. Influence of the Synthesis Route

As stated in the introduction, the main goal of the present study was to tailor the catalytic performances of ruthenium-based catalysts upon the preparation of supported ultra-small ruthenium sub-nano clusters. A total of six Ru/SiO₂-Al₂O₃ catalysts, issued from three different preparation routes with different ruthenium loadings (1 and 2 wt%) supported on Si-Al (1) silica–alumina support, were primarily characterized by electron microscopy to check the influence of the preparation route on the ruthenium particle sizes. The best synthesis route was then extended to the Si-Al (2) and Si-Al (3) supports. Table S1 summarizes the experimental details about each catalyst to evidence the influence of the synthesis route on the ruthenium particle sizes. Those first results show that both catalysts RuSA-1 and RuSA-2, prepared by deposition–precipitation using urea (DPU), exhibit quite large ruthenium particles (14.0 and 15.7 nm, respectively), as illustrated in Figure S1. This preparation route was adapted from Zanella’s protocol describing the synthesis of ruthenium nanoparticles on TiO_2, mirroring the synthesis of gold nanoparticles [60,61]. The authors reported small ruthenium nanoparticles, with sizes ranging from 2.0 to 3.8 nm. In our case, Ru(acac)₃ deposition–precipitation on silica–alumina under a urea-induced gradual increase in pH was harder to accomplish than in the case of RuCl₃, which was used by these authors. Moreover, the intimate interaction between Ru clusters and rutile and anatase TiO₂, as a result of their high degree of lattice matching, is known to be a stabilizing factor that maintains high dispersion and prevents aggregation [62,63]. Similar results were also reported in the case of Ru/CeO₂ catalysts, where the Ce⁴⁺/Ce³⁺ easy redox switch is comparable to that of Ti⁴⁺/Ti³⁺, hence the improved ruthenium dispersion [64]. However, that is not the case with the recalcitrant SiO₂-Al₂O₃ oxides, as both silica and alumina lack redox features to improve the ruthenium dispersion, and the DPU synthesis route led to quite large ruthenium particles. For the catalysts RuSA-3 and RuSA-4, much smaller ruthenium nanoparticles of 2.2 and 2.6 nm, respectively, were obtained. They fit in the lower size range reported in the literature for ruthenium catalysts prepared by incipient wetness or wet impregnation supported on alumina (6.5–18 nm) [25,42,65], zeolites and mesoporous alumino-silicates (1–13.8 nm) [37,40] and silica–alumina mixed oxides (7–11 nm) [35]. On the other hand, RuSA-5 and RuSA-6 catalysts synthesized by a sonication-assisted route (S) exhibit ultra-small ruthenium clusters of 0.4 and 0.6 nm for the 1 and 2 wt% loadings, respectively. Based on these primary results, deposition–precipitation using urea (DPU) and wet impregnation (WI) were ruled out as synthetic routes for sub-nano ruthenium clusters. Further characterizations will only focus on catalysts synthesized by the sonication-assisted route (S).

2.2. Catalysts Characterization

All six catalysts prepared by the sonication-assisted route (S), namely RuSA-5 to RuSA-10, were fully characterized by Inductively Coupled Plasma Atomic Emission Spectroscopy ICP-AES (Table S2), N₂ adsorption/desorption (Table S3), X-ray Diffraction (Figure S2), FTIR spectroscopy (Figure 1), Raman spectroscopy and mapping (Figure S3 and Figure 2), Scanning Electron Microscopy SEM (Figure S4) and High-Angle Annular Dark Field (HAADF) Scanning Transmission Electron Microscopy STEM (Figure 3).

Table S2 regroups the ruthenium loadings deposited on the various supports by sonication-assisted synthesis. Even though ruthenium is known to be recalcitrant and dissolves in the acid solution before ICP-AES characterization, the experimental loadings are very close to the theoretical ones (1 or 2 wt%), indicating an effective deposition of ruthenium on silica–alumina mixed oxides. The textural properties of the supports and Ru-supported catalysts were checked by N₂ adsorption/desorption at 77 K. All the materials present type IV isotherms, with H₁ hysteresis loops of mesoporous materials. When the Ru metallic particles were introduced onto the silica–alumina mixed oxides, the hysteresis loop became narrower, which means that the pores became wider; this is effectively shown in Table S3 and can be explained by the fact that Ru particles have penetrated the silica alumina porosity, hence enlarging and increasing the pore diameter and volume. The specific surface area (ranging from 86 to 140 m²·g⁻¹ for bare oxides) increased remarkably in most cases (up to 190 m²·g⁻¹) after the addition of Ru nanoparticles, which may be indicative of the contribution of very well-dispersed metallic nanoparticles, whose metallic surfaces contributes to the total specific surface area of the materials.

Figure 1 presents the FT-IR spectra of silica–alumina mixed oxides with diverse specific surface areas. Tetrahedral SiO₄ units are easily identified by Si-O-Si oscillation bands at 460 and 804 cm⁻¹ and a stretching band at 1100 cm⁻¹ [66,67]. The band at 960 cm⁻¹ is assigned to the silanol groups’ Si-O-H vibrations generated by the presence of structural defects [68]. Due to the weak Al content in all supports, alumina-related bands are very weak. Al-O-Al expected at 670 cm⁻¹ cannot be observed, while the Al-O bond bending vibration is very weak at 870 cm⁻¹ [69,70]. The ɣ-alumina Al-O vibration mode expected at 1050 cm⁻¹ is overlapped by the Si-O-Si band at 1100 cm⁻¹ [71]. The wide band from 3400 to 3700 cm⁻¹ corresponds to surface OH groups either linked to Si⁴⁺ or Al³⁺ cations [66,72].

Upon the addition of ruthenium clusters (S route), no obvious changes could be spotted in the IR spectra of the supported catalysts, which may be due to the supposedly very small Ru particle size. The Ru–O vibration band expected at 490 cm⁻¹ is eclipsed by the intense Si-O-Si band at 460 cm⁻¹ [73]. Nevertheless, a close observation of the Si-OH band at 956 cm⁻¹ reveals its quasi-disappearance upon the introduction of ruthenium on the silica–alumina supports. An intensity decrease can also be observed for the Si-O-Si bands at 460, 804 and 1100 cm⁻¹, suggesting the linkage of ruthenium atoms to Si-O to form a new Si-O-Ru bond. The main Si-O-Ru band generated by asymmetric stretching vibrations would appear at 1160 cm⁻¹, but is overlapped by the Si-O-Si band at 1100 cm⁻¹ [74]. The same linkage might happen with alumina to form a new Al-O-Ru bond, but could not be evidenced as the alumina-related bands are very weak.

Raman spectra for the materials used in the present work (Si-Al and Ru) are displayed in Figure S3. The bands at 504 and 628 cm⁻¹ can be assigned to Ru-O vibration modes out of plane (E_g) and in plane (A_1g), respectively [75,76]. For silica–alumina mixed oxides, we observe two characteristic peaks at 1053 cm⁻¹ and 1328 cm⁻¹, corresponding to Al–O bending and Si-O stretching modes [77,78], and a peak at 715 cm⁻¹ that is attributed to the Si-O-Al vibration mode [79]. In tandem, Raman mapping, an extension of Raman spectroscopy, facilitates the creation of spatial distribution maps of chemical compounds within samples. It enables the visualization of composite structures, the distribution of reinforcing agents and the orientation of fibers, allowing for the identification of specific material components based on their unique spectral signatures. This methodology holds promise for delivering intricate details about material compositions, with potential applications spanning quality control [80,81], failure analysis and detection of corrosion [82] and the development of advanced materials and bio-fouling [83]. Raman mapping experiments were performed to investigate the distribution of Si, Al and Ru in the different synthesized catalysts. The peak at 715 cm⁻¹ will be considered as a reference for Si with a red color, the peak at 1053 cm⁻¹ as a reference for Al with a green color and the peak at 504 cm⁻¹ as a reference for Ru with a blue color. Mappings were performed on images of the ruthenium-supported catalysts, with a ×50 magnification on a surface of 144 µm², as shown in Figure 2 (images [A]). The parts [B] in Figure 2 show the distribution of Si, Al and Ru taken with a mapping of 164 spectra under the same conditions. These mappings evidence the intimate presence of Ru with Si and Al in all analyzed catalysts, with a fairly homogeneous distribution between the three elements. The parts [C] in Figure 2 are a projection of the mappings performed on the small areas in the corresponding images [A] all over the picture, provided by Lab-Spec 6 Horiba Scientific Software. These projections highlight the wide coverage of the catalyst’s surface by ruthenium (blue color), despite the low loadings (from 0.9 to 2 wt%), compared to silica–alumina. This is only possible if the ruthenium is highly dispersed on the surface of the catalysts.

Figure 3 features High-Angle Annular Dark Field (HAADF) Scanning Transmission Electron Microscopy (STEM) micrographs for all the reduced catalysts. Ultra-small ruthenium particles can be observed in all samples. The Ru particle size distribution illustrated in Figure 3 is based on advanced image processing with statistics of over 2000 Ru particles per catalyst. The average particle sizes are summarized in

Table 1. Ru particles’ average size determined by STEM for Ru/SiO₂-Al₂O₃ catalysts.

Samples	Synthesis Route	Support	Ru Loading (%)	Ru Particle Size (nm)	Ru Single Atoms (%)
RuSA-5	S	Si-Al (1)	0.94	0.4 ± 0.06	35
RuSA-6	S	Si-Al (1)	1.80	0.6 ± 0.05	24
RuSA-7	S	Si-Al (2)	0.93	0.3 ± 0.07	74
RuSA-8	S	Si-Al (2)	1.83	0.7 ± 0.04	27
RuSA-9	S	Si-Al (3)	0.97	0.4 ± 0.06	35
RuSA-10	S	Si-Al (3)	2.00	0.4 ± 0.06	57

, proving that ruthenium was successfully highly dispersed on the various silica–alumina supports with 24 to 74% of single atoms and 100% of the ruthenium clusters at the sub-nano scale.

All the characterization results mentioned above allow us to state with confidence that highly dispersed ruthenium single atoms and ultra-small sub-nano clusters supported on silica–alumina mixed oxides were successfully prepared via a versatile and reproducible sonication-assisted synthesis route. Even with Ru loadings up to 2 wt%, all particles were below 1 nm, with an average particle size ranging between 0.3 and 0.7 nm, corresponding to single atoms to up to eight ruthenium atoms per cluster [84,85,86], as shown in Figure 4.

As far as we know, very few papers have reported the successful synthesis of ruthenium sub-nano clusters. Ruthenium clusters with an average size of 0.97 nm were reported by Chen et al. [87], while Qian et al. [88] successfully incorporated ruthenium sub-nano clusters of 0.2 to 0.7 nm on the surface of N-doped carbon layer coated SBA-15 but only at a very low metal loading of 0.1wt%.

Table 2. Comparison of the present results with the literature.

Catalysts	Ru Loading (%)	Ru Average Particle Size (nm)	Reference
Ru/CD	n.d	2.3 *	[89]
Ru/Ni(OH)2	n.d	1.5–2 *	[33]
Ru/Ni-MOF-SH	3.5	1.5 *	[90]
Ru/PC	2	1–3.1 *	[91]
Ru SNC/W18O49	very low	1 *	[92]
Ru/Al2O3-SSD	0.98	0.97	[87]
Rux/Mn1−xO2 NWs	1.5	n.d	[93]
Ru/NC-SBA-15	0.1	0.5	[88]
Ru/SiO2-Al2O3	1	0.3–0.4	Present work
Ru/SiO2-Al2O3	2	0.4–0.7	Present work

* Ruthenium clusters are partially at the sub-nano scale. n.d: not determined.

summarizes different comparative studies with the present work.

Based on both the Raman and FTIR spectroscopy results that evidenced the different bonds in our materials, we can suggest the following scheme (Figure 5 and Figure S5) to explain ruthenium anchoring on silica–alumina, leading to ultra-small Ru sub-nano clusters. In the first step, ultrasounds improve mass transport and reduce the diffusion layer thickness. They may also affect the alumina–silica surface morphology, typically enhancing the surface contact area [94]. Those physical phenomena will provide, in a second step, an optimal adsorption of Ru(acac)₃ complexes all over the silica–alumina surface area. In a third step, each octahedral ruthenium complex will lose one of the acetylacetonate ligands, where the two Ru-O-C bonds are replaced by new Ru-O-Al or Ru-O-Si bonds. This reaction produces a surface configuration where each Ru atom, on average, is bonded to two oxygen atoms from the tetrahedral AlO₄ or SiO₄ surface units, as evidenced by DFT calculations performed by Ogino et al. [95]. Finally, upon thermal treatment, the remaining acetylacetonate ligands are removed, and the ruthenium cations are reduced into Ru⁰. During this last step, Ru single atoms undergo a minimal heat-induced dynamic coalescence with adjacent atoms to end up in the form of ultra-small clusters ranging from single atoms to up to eight Ru atoms per cluster.

2.3. Catalytic Tests

The performances of Ru/SiO₂-Al₂O₃ catalysts under study were evaluated in the biomass-derived furfural catalytic etherification. Indeed, furfural is an important platform chemical derived from lignocellulosic biomass [96], and therefore provides a variety of value-added bio-chemicals [97] such as, among others, furfuryl alcohol [34], tetrahydrofurfuryl alcohol [98] and 2-methylfuran [99] by hydrogenation, or succinic [100], maleic [101] and furoic [102] acids by oxidation. The existence of various competing reaction pathways brings, in each case, the difficulty of targeting a desired compound and minimizing by-products. Meng et al. reported how the reaction selectivity can be controlled by regulating the catalyst’s features [103], while Li et al. [55] proved that isopropanol can act as both the reaction solvent and an efficient hydrogen donor for the catalytic transfer hydrogenation of furfural, leading mainly to furfuryl alcohol, 2-methylfuran and etherification and ring-opening side products. In the present study, we provide an innovative application, which is the catalytic transformation of furfural into 2,2′-difurfurylether (DFE), a transportation biofuel additive with a high cetane number and important ignition characteristics [104,105,106]. Difurfurylether is also widely used in the food industry as a flavoring agent to add a coffee-like, nutty taste to food and drinks [59]. The results summarized in

Table 3. Ru/SiO₂-Al₂O₃ catalytic performances in furfural etherification.

Entry	Catalyst	TOF (h−1)	Yield (%)
			DFE	FA	Other
1	Blank test	0	0	0	0
2	Si-Al (1)	0	0	0	0
3	Si-Al (2)	0	0	0	0
4	Si-Al (3)	0	0	0	0
5	RuSA-5	1135	92	0	8
6	RuSA-6	593	97	3	0
7	RuSA-7	1148	95	0	5
8	RuSA-8	583	93	0	7
9	RuSA-9	1100	92	2	6
10	RuSA-10	1067	86	5	8

show that the reaction did not occur without a catalyst (Entry 1, blank test). Similarly, Ru-free materials, SiAl (1), SiAl (2) and SiAl (3), also showed no activity (Entries 2–4). However, once ruthenium was introduced (Entries 5–10), an exaltation of the activity was observed, reaching turnover frequencies TOFs up to 1148 h⁻¹. The reaction products shown in Scheme 1 were separated by gas chromatography and identified by mass spectroscopy (Figure S6).

These results also highlight the great selectivity of Ru single atoms and sub-nano clusters towards DFE under our reaction conditions. The optimal catalytic selectivity on sub-nano catalysts was previously reported in the literature and attributed to the accessibility of specific surface sites, influencing the adsorption and desorption of reactants and intermediates [5]. Based on our characterizations and supported by previous theoretical calculations [107,108], experimental kinetic studies and X-ray crystallographic analysis [108,109], the reaction mechanism featured in Figure 6 can be proposed to explain furfural transformation into difurfurylether. A starting point is the 16-electron complex Ru-O-Si (or Ru-O-Al), whose formation is explained in Figure 5 and Figure S5. In step 1, a dissociative adsorption of gaseous hydrogen H₂ on Ru-O-Si sites leads to an 18-electron complex with two H atoms. A furfural molecule then approaches the active site (step 2), where the C=O bond of the aldehyde is activated (step 3), to form furfuryl alcohol (step 4). This latest molecule, which acts now as a reaction intermediate, is eventually adsorbed on the ruthenium site to activate the OH bond (step 5), and then reacts with another furfuryl alcohol molecule leading to dehydration (loss of a water molecule) to produce difurfurylether (step 6), subsequently liberating the ruthenium active site (16-electron complex) to undergo another reaction cycle.

Moreover, the size–activity relationship is evidenced in Figure 7. Indeed, TOF is at its highest, i.e., 1148 and 1135 h⁻¹, for Ru clusters of 0.3 and 0.4 nm, respectively, and then drops when the particle size increases. Several authors reported similar size–activity behavior under hydrogenating conditions [14,15,27,28] on Ru nanoparticles (from 1 to 12 nm), which is still valid at the sub-nano scale, whereas the reverse phenomenon was observed under oxidizing conditions, where the activity increases with Ru particle size [110,111].

Furthermore, stability is a crucial parameter in heterogeneous catalysis, particularly for size-dependent reactions. Indeed, at the nano-scale and below, under heating conditions, metal particles are unstable and tend to aggregate upon dynamic coalescence or surface/sub-surface atoms diffusion (Ostwald ripening) [112,113,114]. In our materials, ruthenium sub-nano clusters showed exceptional resistance to sintering and agglomeration. High-Angle Annular Dark Field (HAADF) Scanning Transmission Electron Microscopy (STEM) showed still ultra-small ruthenium clusters after a catalytic reaction, with the same average size ±0.1 nm compared to the fresh catalysts before the reaction. Moreover, ICP-AES analysis revealed negligible metal leaching under the catalytic conditions ($∆x$ < 0.002, where $x$ is the ruthenium loading). The enhanced stability against sintering and leaching results from the strong anchoring of ruthenium to the silica–alumina surface via Ru-O-Si and Ru-O-Al bonds, as evidenced above in Figure 5 (Section 2.2), a behavior similar to that of strongly immobilized single-atom catalysts SACs [44,46].

3. Materials and Methods

3.1. Materials Synthesis

All the products described in the Supplementary Materials were used without any further purification or treatment. Home-made silica–alumina mixed oxides (SiO₂-Al₂O₃) were synthesized via wet impregnation and annotated as follows: Si-Al (1), Si-Al (2) and Si-Al (3).

As one of the goals of the present study was the preparation of supported ultra-small ruthenium sub-nano catalysts, different preparation routes were evaluated. The experimental conditions for each preparation sequence are fully described in the Supplementary Materials and summarized in Table S1. At first, a total of six Ru/SiO₂-Al₂O₃ catalysts, with 1 or 2 wt% Ru loading, were prepared by deposition–precipitation by urea DPU (RuSA-1 and RuSA-2), wet impregnation WI (RuSA-3 and RuSA-4) and sonication-assisted route S (RuSA-5 and RuSA-6) using the Si-Al (1) as support and primarily characterized by electron microscopy to check the ruthenium particle sizes. The best synthesis route was then extended to the Si-Al (2) and Si-Al (3) supports with four more samples (RuSA-7 to RuSA-10). All the dried powders resulting from the different synthesis routes were thermally treated under H₂ at 300 °C for 3 h to reduce the Ru³⁺ cations to Ru⁰, as it was reported that the calcination of ruthenium catalysts favors sintering and metal loss due to the volatility of ruthenium oxides [115].

3.2. Catalytic Activity

The catalyst’s activity was checked in the furfural etherification reaction. The tests were carried out in a Parr-5400 continuous flow under pressure tubular reactor under the following conditions: 6.04 mmol of freshly distilled furfural in 15 mL of pure isopropanol as solvent, 100 mg of catalyst, under 10 bars of H₂, at 180 °C for 6 h. Catalysts were reduced in situ under H₂ at 300 °C for 3 h and then cooled down to room temperature before each test. The recuperated reaction mixture was analyzed via gas chromatography coupled with mass spectroscopy (GC-MS) on a Bruker (SCION GC-MS System SQ) equipped with a capillary column (DB-5, with methylphenyl siloxane as the stationary phase). Workstation-8 software with the NIST-11 MS database was used for product identification.

The catalytic performances were calculated using the following formulae:

$$\mathrm{Y}\mathrm{i}\mathrm{e}\mathrm{l}\mathrm{d}\left(\%\right)={\displaystyle \frac{\mathrm{m}\mathrm{o}\mathrm{l}\mathrm{e}\mathrm{s} \mathrm{o}\mathrm{f} \mathrm{p}\mathrm{r}\mathrm{o}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{t}}{\mathrm{m}\mathrm{o}\mathrm{l}\mathrm{e}\mathrm{s} \mathrm{o}\mathrm{f} \mathrm{f}\mathrm{u}\mathrm{r}\mathrm{f}\mathrm{u}\mathrm{r}\mathrm{a}\mathrm{l}}}\times 100$$

(1)

$$\mathrm{T}\mathrm{O}\mathrm{F}\left({\mathrm{h}}^{-1}\right)={\displaystyle \frac{\mathrm{m}\mathrm{o}\mathrm{l}\mathrm{e}\mathrm{s} \mathrm{o}\mathrm{f} \mathrm{d}\mathrm{i}\mathrm{f}\mathrm{u}\mathrm{r}\mathrm{f}\mathrm{u}\mathrm{r}\mathrm{y}\mathrm{l}\mathrm{e}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{r}}{\mathrm{m}\mathrm{o}\mathrm{l}\mathrm{e}\mathrm{s} \mathrm{o}\mathrm{f} \mathrm{R}\mathrm{u} \mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e} \mathrm{s}\mathrm{i}\mathrm{t}\mathrm{e}\mathrm{s}\times \mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}}$$

(2)

Turnover frequencies (TOFs) were calculated at low conversions on the basis of the total number of Ru atoms (from ICP-AES results). Due to the sub-nanometer scale of Ru clusters, there are no bulk Ru atoms, as shown in Figure 4.

4. Conclusions

Single-atom and sub-nano cluster heterogeneous catalysts are a paradigm shift in catalysis, offering control over physicochemical features and catalytic performances at the atomic scale. In this paper, we report an unfailing reproducible sonication-assisted synthesis route for the successful deposition of uniformly and highly dispersed ruthenium sub-nano clusters on the surface of silica–alumina mixed oxides, at ruthenium loadings up to 2wt%. Raman spectroscopy mapping carried out on micro-scale areas highlighted a wide coverage of the catalyst’s surface by ruthenium, which is only possible if the ruthenium is highly dispersed on the surface of the catalysts. HAADF-STEM revealed that 100% of the ruthenium was at the sub-nano scale, ranging from single atoms (24 to 74% of total Ru) up to eight Ru atoms per cluster, corresponding to an average size from 0.3 to 0.7 nm. These materials showed highly selective size-dependent catalytic performances in transportation biofuel additive production from biomass-derived furfural, with turnover frequencies up to 1148 h⁻¹. An exceptional resistance to sintering and agglomeration, with a Ru particle size variation of ±0.1 nm and no leaching, was evidenced. The unique structural and electronic properties at the sub-nano scale, the high surface/bulk ratio and the consequent abundance of active sites open the doors to versatile catalytic applications in various types of chemical reactions.