Exploring 5-hydroxymethylfurfural hydrogenation pathways using NHC-stabilized water-soluble nanoparticles of various metals and alloys

Abstract

The catalytic valorization of 5-hydroxymethylfurfural (5-HMF), a key bio-based platform molecule, is central for sustainable chemical production. In this work, we report the aqueous-phase hydrogenation of 5-HMF using a family of water-soluble nanoparticles stabilized by an N-heterocyclic carbene (IMesPrSO₃) ligand. The nanoparticles were comprehensively characterized by BF-TEM, DLS, TGA, ICP-OES, and pair distribution function (PDF) analysis from wide-angle X-ray scattering (WAXS). By only varying the metal core (Ru, Pd, Ir, RuIr₂), and slightly modifying the reaction conditions, we accessed a diverse array of high-value products, including 2,5-bis(hydroxymethyl)furan (2,5-BHMF), 2,5-bis(hydroxymethyl)tetrahydrofuran (2,5-BHMTHF), oxidized cyclopentenones, and 1-hydroxyhexane-2,5-dione (HHD), all under environmentally friendly conditions (30–140 °C, 5 bar H₂, in water). Specifically, the Ru-based nanoparticles showed high selectivity to 2,5-BHMF (90%) at very mild conditions (30 °C), along with a promising recyclability. Reaction products were identified and quantified through extensive NMR spectroscopy, including ¹H, ¹³C, COSY, HSQC, and HMBC experiments. Our findings demonstrate that these truly colloidal catalytic systems can represent a tuneable and robust platform for green and sustainable biomass transformations.

---

Green foundation

1. This work advances green chemistry by using for the first time a family of water-soluble NHC-stabilized colloidal nanoparticles for 5-HMF hydrogenation in water, offering tuneable and, in some instances, unprecedented product selectivity under mild, sustainable conditions, without organic solvents.

2. The key green chemistry achievement lies in demonstrating high-yield and selective hydrogenation of 5-HMF (up to 90% yield to 2,5-BHMF at 30 °C, 5 bar H₂, in water) along with an efficient and recyclable catalyst. The process maximizes atom economy, avoids toxic solvents, and minimizes energy input, aligning with at least six Green Chemistry principles (1, 2, 5, 6, 7, 9).

3. A Greener development could be achieved by improving nanoparticle recovery to reduce catalyst loss, replacing noble metals with earth-abundant alternatives, and further minimizing side-product formation at higher temperatures via ligand tuning or metal composition optimization.

Introduction

The heavy reliance on non-renewable fossil feedstocks is recognized as a critical factor contributing to anthropogenic climate change. Fossil sources serve as a primary resource for a wide range of everyday products, including polymers, fuels, fine chemicals, and food additives. However, they are limited in supply and contribute significantly to greenhouse gas emissions.¹ Therefore, the catalytic conversion of alternative, clean, and renewable raw materials into fuels and chemicals is a scientific priority in dealing with global warming and contributing to the diversity of energy sources.^2,3

Among the various resources being explored as alternatives to petroleum and natural gas, biomass from waste or non-edible plant sources is a viable option for the sustainable manufacture of essential products for modern society.⁴ Notably, there is a growing consensus that catalytic strategies, both chemo- and bio-catalysis, should prioritize the selective conversion of agricultural and municipal waste into value-added chemicals and functional materials, rather than fuels or fuel additives, to maximize resource efficiency and economic viability.^5,6 In this context, the rational design of advanced catalytic systems capable of transforming bio-derived platform molecules under mild and sustainable conditions into a broad array of chemicals is of utmost importance.

Particularly, 5-(Hydroxymethyl)furfural (5-HMF) is one of the most valuable chemical compounds derived from holocellulose. It is produced through a stepwise dehydration process of various carbohydrates.^7–9 HMF has attracted significant attention due to its potential as a chemical platform for synthesizing polymers and environmentally friendly solvents, some of them nearing commercialization.^7–11 In fact, it was designated as one of the key platform chemicals by the US Department of Energy.¹²

More precisely, 5-HMF hydrogenation unlocks a diverse range of value-added products, with selectivity dictated by the specific activation of its functional groups (C=O, C=C and C–O–C bonds). This versatility enables the production of furan or tetrahydrofuran derivatives, linear oxygenated molecules through the combination of hydrogenation and (de)hydration processes, and cyclopentanones or cyclopentenones via hydrogenation and ring rearrangements (Scheme 1).^13–18 These bio-based molecules serve as essential building blocks for advanced materials, including thermosets and self-healing polymers (e.g. 2,5-bis(hydroxymethyl)furan: 2,5-BHMF),^19,20 surfactants, plasticizers and polymer precursors (e.g. 2,5-bis(hydroxymethyl)tetrahydrofuran: 2,5-BHMTHF).¹⁴ Additionally, they function as intermediates in pharmaceutical, food additive, and agrochemical molecules (e.g. 1-hydroxyhexane-2,5-dione: HHD),²¹ as well as for biologically active compounds and fragrances (cyclopentenones).^22–24 Therefore, harnessing selective hydrogenation strategies for 5-HMF transformation is a critical step toward establishing a sustainable bioeconomy.

Scheme 1 Overview of the various transformations in the catalytic hydrogenation of 5-HMF.

A significant concern in developing highly efficient and selective catalysts for 5-HMF hydrogenation lies in ensuring that the transformations occur under environmentally sustainable conditions. To address this issue, it is essential to employ mild reaction conditions and water as a solvent, owing to its non-toxic nature and cost-effectiveness.²⁵ Furthermore, biomass conversion typically results in the production of substantial quantities of water.

5-HMF hydrogenations have been frequently accomplished by using supported transition metal-based heterogeneous catalysts.²⁶ Nevertheless, the limited availability of active sites, in comparison to their homogeneous counterparts, necessitates the use of severe conditions regarding pressure (exceeding 10 bar of H₂) and temperature (≥150 °C).²⁷ These catalysts are also susceptible to water-mediated deactivation through active site blocking, sintering or leaching.^28,29 Furthermore, their selectivity is affected by the properties of the support, which can, in turn, be influenced by the presence of water.²⁹ Interestingly, we could not find any report on HMF hydrogenation catalyzed by colloidal metal nanoparticles (NPs) in an aqueous liquid phase. In that sense, our group has worked over the last decade to develop a series of water-soluble NPs stabilized by N-Heterocyclic Carbene (NHC) ligands.³⁰ NHC ligands are particularly effective in influencing the stability and reactivity of the NPs.^31–33 The high activity of these water-soluble nanoparticles in H/D exchange reactions, and their excellent resistance to leaching and sintering in aqueous media,^34–37 prompted us to utilize them for 5-HMF hydrogenation in water under mild reaction conditions (5 bar of H₂, ≤140 °C).

Thus, as a part of the present work, we successfully unlocked various hydrogenation pathways of 5-HMF by rationally selecting specific metals and reaction conditions (Scheme 2). We employed a range of catalysts, including Ru, Pd, Ir, and RuIr₂ water-soluble NPs stabilized by the same NHC ligand; namely 1-Mesityl-3-(3-sulfonatopropyl)-1,3-dihydro-2H-imidazol-2-ylidene (IMesPrSO₃). The selected ligand displays only very weak acidity (see ESI Table S7,† entries 1 and 2), allowing us to consider solely the metal effect. For partial and complete hydrogenation under mild conditions, Ru and Pd NPs were selected, respectively. Conversely, Ir NPs predominantly produced oxidized cyclopentanones, a rare selectivity in the catalytic hydrogenation of HMF. Furthermore, using RuIr₂ as a catalyst allowed the production of HHD, as long as silver tetrafluoroborate (AgBF₄) was introduced as a pH modifier. As the first example of colloidal water-soluble NPs for 5-HMF valorization, we restricted our focus to noble metals, thereby eliminating the necessity to stabilize the desired oxidation state in first-row transition metals, which adds complexity beyond the scope of this work. The benefits of the stabilization strategy have been demonstrated through analysis of the catalytic system before and after reaction, and the recyclability study of Ru@IMesPrSO₃.

Scheme 2 Overview of 5-HMF aqueous-phase hydrogenation (5 bar of H₂) strategies in this work with different M@IMesPrSO₃ NPs (M = Ru, Pd, Ir, RuIr₂). Maximum yields are provided in parentheses.

Experimental

Materials

The precursors [Ru(COD)(COT)] (COD = 1,5-cyclooctadiene, COT = 1,3,5-cyclooctatriene) and [Pd₂(DBA)₃] (DBA = dibenzylideneacetone) were purchased from Nanomeps Toulouse. [Ir(μ-OMe)(COD)]₂ (>99%), glacial acetic acid, aqueous formaldehyde (37 wt%), glyoxal (40 wt%), ammonium acetate (>98%), 2,4,6-trimethylaniline (98%), 1,3-propanesultone (98%), potassium tert-butoxide (98%), tetramethylammonium bromide (TMAB, 98%), 5-HMF (>99%), and 2,5-BHMF (>99%) were acquired in Sigma-Aldrich. NaHCO₃ (>99%) was purchased from Fluka. AgBF₄ (99%) was purchased from Alfa Aesar. All these reagents were used as received. H₂ was produced and directly used from a Hydrogen Gas Generator AVANTEC model 40H (water electrolysis). Organic solvents acetone, THF and pentane were purchased from VWR Prolabo. THF and pentane were also dried by passage through alumina in a solvent purification system and then further degassed by bubbling Ar for 30 min before use. Their use in the nanoparticle syntheses is due to the instability of organometallic precursors and the free carbene. The deionized water used in all catalytic tests was produced using a VEOLIA WATERS E300 22L equipment, also including a 1 μm pure polypropylene filter and a 5 μm activated carbon filter to remove organic contaminants.

Water-soluble nanoparticle syntheses

Synthesis of 1-mesityl-1H-imidazole. In a two-necked round-bottom flask, glacial acetic acid (1.7 mL), formaldehyde (37 wt% in water, 0.52 mL, 6.45 mmol), and glyoxal (40 wt% in water, 0.84 mL, 7.18 mmol) were added and stirred at room temperature (r.t.). In a separate vial, ammonium acetate (0.31 g, 4.02 mmol) was mixed with glacial acetic acid (1.7 mL), water (0.15 mL) and 2,4,6-trimethylaniline (0.81 g, 0.84 mL, 6.00 mmol). This mixture was then added dropwise for 20 minutes into the two-necked flask, heated to 70 °C, and stirred for 20 hours. After cooling, the mixture was neutralized with saturated NaHCO₃ and stirred for 1 hour. The resulting brown precipitate was filtered with a Büchner funnel and washed twice with 10 mL of water. Finally, the intermediate product 1-mesityl-1H-imidazole was purified by sublimation of the brown solid at 80 °C overnight (reduced pressure of 0.1 bar), leading to a white solid (0.95 g, 85%).

Synthesis of 1-mesityl-3-(3-sulfonatopropyl)imidazolium (IMesPrSO₃H). IMesPrSO₃H, the NHC ligand precursor, was synthesized following the method of Shaughnessy and co-workers (ESI Scheme S1a and b†).³⁸

In a round-bottom flask, 1-mesityl-1H-imidazole (1.27 g, 6.82 mmol) was dissolved in 50 mL of acetone. Then, an excess of 1,3-propanesultone (2.09 g, 17.13 mmol) was added to the solution. Afterwards, the mixture was stirred for 5 days at r.t. The precipitated white solid was filtered off through a Büchner funnel, washed twice with acetone (3 × 20 mL), and dried under vacuum (1.4 g, 66%).

Subsequently, the unprotonated free carbene, IMesPrSO₃, was prepared as the NP stabilizer by adding, in a Schlenk tube, potassium tert-butoxide (0.0108 g, 1.2 mol eq.) to the imidazolium salt solution (0.0242 g, 1.0 mol eq.) in 15 mL of THF, and allowing it to react under magnetic stirring at r.t. for 7 hours and use it without further purification (ESI Scheme S1b†).

Metal nanoparticle syntheses. Finally, the NPs were synthesized using procedures previously reported in our group, involving an organometallic (bottom-up) approach (ESI Scheme S1c†).^34–37 Typically, IMesPrSO₃ was prepared before being used as a stabilizer, as was already mentioned, and added in a 0.2 eq. proportion corresponding to the metal. The solution resulting from adding potassium tert-butoxide to the imidazolium salt solution was transferred under an Ar atmosphere into a Fischer Porter (FP) bottle charged with a cold solution of [Ru(COD)(COT)], [Ir(μ-OMe)(COD)]₂ or a combination of both at −80 °C. In the case of Pd, [Pd₂(DBA)₃] precursor was added to a FP with THF under an inert atmosphere (glovebox), and the free carbene solution was added. In every case, 3 bars of H₂ (5 bars in the case of Ir) were added and allowed to react at room temperature overnight. After reaction, the volume of the final colloidal dispersion was reduced to 2–3 mL under vacuum. Finally, the NPs were precipitated with pentane (30 mL), washed twice with 30 mL of pentane and dried under vacuum.

Water-soluble nanoparticle characterization

The organic content of the NPs was measured using thermogravimetric analysis (TGA). Additionally, Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES) was used to determine the Ru : Ir ratio in the bimetallic case. The device used for ICP-OES is an ICAP 6300 Thermo RF Generator with a Polychromator – from 166 to 847 nm and a Charge Injection Device (CID) detector. The sample was analyzed after digestion in aqua regia at 80 °C for 12 hours. The TGA was conducted with a TGA/differential scanning calorimetry (DSC) 1 STAR System, equipped with an ultra-microbalance (UMX5), a gas switch, and sensors for differential thermal analysis (DTA) and DSC. The sample was oxidized under an air flow of 50 mL min⁻¹ until reaching 500 °C. A reduction followed this in a mixed gas flow of 4% H₂ and 96% Ar, also at 50 mL min⁻¹, up to a temperature of 700 °C.

The size and morphology of the as-prepared water-soluble NPs and after specific catalytic tests were assessed through bright-field transmission electron microscopy (BF-TEM) imaging, conducted using a 1400 JEOL microscope, operating at a voltage of 120 kV. The samples were analyzed by placing several drops of an aqueous suspension of the respective powders onto a carbon-coated copper grid. ImageJ software facilitated the acquisition of particle size distributions, which were subsequently analyzed to determine the mean size and standard deviation. The resulting data were fitted to the most appropriate statistical distribution.

Powder X-ray diffraction (PXRD) measurements for identifying crystalline phases were performed in reflection geometry using a PANalytical Empyrean diffractometer with Co-Kα radiation (λ = 0.1789 nm). The instrument operated at 35 kV and 45 mA. Data were collected over a 2θ range of 30–120° with a step size of 0.7° and a time per step of 500 s, resulting in a total acquisition time of ca. 46 minutes. The powdered samples were prepared and sealed with Kapton polyimide films in an Ar atmosphere to prevent oxidation.

Wide-angle X-ray scattering (WAXS) was used to determine the local structures and the morphologies acquired by the very small crystalline domains (<2 nm) present in our NPs. These measurements were performed on a Malvern Panalytical Empyrean III diffractometer in transmission geometry, equipped with a Mo anode (λ = 0.7107 nm) and a Galipix3D detector. The maximum reached scattering vector (q) value is q_max = 16.66 Å⁻¹, thus corresponding to a maximum angle of 2θ_max = 143°. Colloidal aqueous suspensions were prepared for each nanoparticle (5 mg in 1 mL) and placed in glass capillaries with a diameter of 1.0 mm and an internal diameter of 0.4 mm. For each specimen measurement, a capillary filled with water contribution was subtracted. Pair distribution functions (PDF) were extracted using PDFgetX3 software v2.1.2.³⁹ Least square refinements of structural models against experimental PDF were performed using WAXS toolbox software v1.0 suite,⁴⁰ based on the diffpy-cmi package.⁴¹ Crystalline structural models of plausible metal and metal oxide phases were generated using the ASE Python library,⁴² particularly the Icosahedron, Decahedron and Octahedron functions from ase.cluster to build corresponding particle morphologies. In the case of alloys (e.g. RuIr₂ particles), random atomic substitution was performed to reach the desired stoichiometry. The least square refinement algorithm was designed to refine the following quantities: i/scale factor, ii/average interatomic distance, iii/isotropic atomic displacement parameter, and iv/parameter to account for displacement correlation between neighboring atoms. All possible morphologies, including spherical particles, have been tested for each particle. The results presented here correspond to the best refinement results.

The nanoparticle dispersion in the reaction media was analyzed using dynamic light scattering (DLS) with a Nanotrack device, utilizing NP: solvent (H₂O) ratios consistent with those employed in the catalytic tests (i.e., 2 mg in 2.0 mL). Before analysis, the dispersion was subjected to sonication for 30 s.

Catalytic tests

Catalytic reactions were conducted in a batch-type glass Fischer–Porter (FP) reactor with a total volume of 80 mL. The reactor head was equipped with a pressure gauge and a sampling probe for effective monitoring. Additionally, a stirring bar was added to the FP. All reactions were carried out under a H₂ pressure of 5 bar, using a thermostatic oil bath set to various temperatures. The batch reactor was immersed in the bath until the oil level rose to ca. 2 cm above the surface of the liquid inside the glass reactor. The reaction mixture was magnetically stirred at 800 rpm.

Unless otherwise indicated, the reaction mixture comprised 0.500 mmol of the substrate (5-HMF), 5 mL of deionized water, and 5 mg of water-soluble NPs. Notably, due to the potential effect of DMSO on catalytic activity,⁴³ its content in commercial 5-HMF batches used during the present study is reported in the ESI (Table S2†). The NPs were stored within a glovebox and only exposed to the atmosphere when added to the FP reactor. Importantly, Ir-containing NPs must be passivated with water immediately after the vial is exposed to air to prevent self-ignition. Subsequently, the reactor was sealed and purged with cycles of 2 bar of H₂ and vacuum three times before establishing the target H₂ pressure of 5 bars.

For sampling, approximately 250 μL of the reaction mixture was collected. The kinetic experiments adhered to a sampling protocol designed to prevent the removal of more than 10% of the reaction mixture during the catalytic process. This approach ensures reproducibility and significance, as validated in the current (ESI Fig. S35†) and previous research studies.⁴⁴ When error bars are presented, they indicate the standard deviation calculated from at least three independent experiments where sample collection was conducted at the same data point. Trace by-products (<3% yield) are only indicated in the footnote of the corresponding graph.

Qualitative and quantitative analyses of the reaction mixtures from different aliquots collected over various reaction times were performed by diluting the samples in ca. 400 μL of a solution of trimethylammonium bromide (TMAB) in D₂O (0.1 wt%). ¹H;¹³C{¹H}; ¹H–¹H COSY; ¹H–¹H TOCSY; ¹H–¹³C HSQC and, ¹H–¹³C HMBC NMR spectra in solution were recorded on a BrukerAvance 500 spectrometer equipped with a 5 mm triple resonance inverse Z-gradient probe (TBI ¹H/¹³C/BB). All the ¹H and ¹³C signals were assigned based on the chemical shifts, spin–spin coupling constants, splitting patterns and signal intensities at 298 K. Chemical shifts (δ) given in parts per million (ppm), are quoted relative to the internal standard (TMAB) for the ¹H and ¹³C{¹H} experiments. Coupling constants (ⁿJ_H–H) are given in Hertz. The reader is referred to the ESI† for full details on the NMR characterization of each compound discussed in the Results and discussion section (ESI sections 3.1, 3.2 and Fig. S12–S34†). After ¹H signal assignments, quantification required recording routine ¹H NMR spectra at 298 K. To obtain a proper quantification of the integrated signals, ¹H experiments were recorded with the suppression of the water signal (δ = 4.7 ppm) and with any default decoupling nuclei. The ¹H radio frequency pulse length (p1) was calibrated for each experiment (p1 ≈ 8.3 μs), and the spectral window (SW) was set from 15 to −5 ppm. Conversion rates, yields, and carbon balances were calculated using absolute mole values (ESI, section 3.3 and eqn (S1)–(S5)†) derived from comparing the most representative peak areas of the reagents, products, and the internal standard (TMAB).

The water-soluble NPs were sometimes recovered and utilized in consecutive catalytic cycles. The NP extraction involved cooling the reaction mixture to its freezing point, followed by gradual warming until the melting point was reached, after which centrifugation was employed (25 000 rpm, 10 °C, 15 min). Then, the NPs were washed with water and centrifuged once more before being applied in subsequent catalytic tests.

Results and discussion

Ru@IMesPrSO₃ for 2,5-BHMF production

Our group specifically reported several years ago the synthesis of Ru@IMesPrSO₃ NPs. These NPs demonstrated exceptional stability in aqueous media as no agglomeration was observed for at least six months.³⁶ We therefore looked at the possibility of using them for performing hydrogenation reactions in aqueous solutions.

Ru NPs are well known for their broad hydrogenation capabilities, including the potential reduction of the 5-HMF aromatic ring^45,46 or hydrodeoxygenation of hydroxymethylgroups.^47,48 However, in the present work, we deliberately constrained the mild reaction conditions to temperatures below 150 °C and a low hydrogen pressure of 5 bar, aiming to suppress such undesired side reactions. Indeed, evidence from the literature shows that Ru NPs onto carbon facilitate 5-HMF hydrogenation compared to 5-HMF hydrodeoxygenation and ring opening.⁴⁹ Furthermore, in aqueous media, Ru catalysts exhibit a strong preference for carbonyl group reduction.⁵⁰ Additionally, we hypothesized that the steric hindrance by NHC ligands may favour the adsorption of furan rings oriented tilted instead of parallel to the surface, favouring the activation of smaller groups like the C=O bond, likely in a η²(C,O)-aldehyde coordination mode, facilitating the formation of 2,5-BHMF.^51,52 In contrast, the furan ring might be hindered from effective coordination and hydrogenation. This rationale provided confidence that the Ru@IMesPrSO₃ system would selectively reduce the carbonyl group of 5-HMF to 2,5-BHMF, while operating under environmentally friendly conditions.

Therefore, the Ru@IMesPrSO₃ NPs were synthesized following our original procedure (Experimental section; Water-soluble nanoparticle syntheses).³⁶Fig. 1a illustrates a representative BF-TEM image and the corresponding particle size distribution, revealing an average size of 2.0 nm, indicative of a uniform and well-controlled synthesis process. Additional BF-TEM provided in the ESI (Fig. S2†) further confirms the homogeneity of the nanoparticle population. The DLS analysis shows that these NPs are moderately aggregated in aqueous media, with most of the aggregates displaying a size within the 10 to 20 nm (Fig. S6a†).

Fig. 1 (a) Particle size distribution of Ru@IMesPrSO₃ NPs and representative BF-TEM micrograph. (b) PDF obtained from WAXS measurement of Ru@IMesPrSO₃ NPs and comparison against an icosahedral model fit (five shells).

Moreover, in Fig. 1b, we present the structural analysis of these Ru@IMesPrSO₃ NPs using Pair Distribution Function (PDF) data from a WAXS measurement. The red line fitting the experimental data corresponds to a model based on a five-shell icosahedral structure with a diameter of an enveloping sphere of 20.8 Å (3D inset). The black line shows the residuals, indicating a minor difference between the experimental data and the model fit, suggesting a good agreement. The icosahedral structure, although scarcely documented,⁵³ can provide stability in small Ru NPs.⁵⁴

Using these nanoparticles, catalytic tests were performed at varying temperatures (30, 70, 100 and 120 °C), with aliquots collected at different time intervals to establish kinetic profiles (Fig. 2 and S36†). Interestingly, we found that the activity of the Ru@IMesPrSO₃ was sufficient at 30 °C to hydrogenate the aldehyde group, selectively producing 2,5-BHMF (90% yield after 16 h). Higher temperatures resulted in a selectivity loss, along with a loss of carbon balance. Probably, the tendency of 2,5-BHMF to undergo hydration-dehydration reactions increases with temperature.^55,56 These processes favour the formation of HHD, which can subsequently cyclize via intermolecular aldol condensation to 3-(Hydroxymethyl)cyclopent-2-en-1-one (HCPEN). Further hydrogenation of HCPEN results in the production of 3-(Hydroxymethyl)cyclopentan-1-one (HCPN). Moreover, it is well documented that furfural derivatives such as 5-HMF and 2,5-BHMF are prone to condensation reactions leading to insoluble humins (undetectable by NMR) at high temperatures.⁵⁷ In that sense, TGA of the solid residue after reaction at high temperatures (i.e., 120 °C) shows an increase of over 100% in organic content due to the presence of this carbonaceous matter on the NPs (Table S6†). Therefore, we could favour the desired reaction to 2,5-BHMF using lower temperatures.

Fig. 2 Temporal evolution of product distribution in the hydrogenation of 5-HMF using Ru@IMesPrSO₃ NPs as a catalyst. Reaction conditions: 63.1 mg (0.5 mmol) of 5-HMF, 5 mL of water, 5 mg of catalyst, at 30 °C, 5 bar of H₂, and 800 rpm. Notes: (i) Dashed lines are only visual guidelines. (ii) Error bars represent the standard deviation from at least three independent experiments at the same data point. (iii) Carbon balance (15 h): 97%.

Fig. 2 displays the kinetic profile for the hydrogenation of 5-HMF in the presence of Ru@IMesPrSO₃ under optimized conditions (5 bar of H₂ and 30 °C). The reaction approaches a quantitative yield of 2,5-BHMF after 15 hours, with the hydrogenation of the furan ring being the only minor side reaction (2,5-BHMTHF yield ≈5%). The observed decline in the 5-HMF consumption rate over time can partially be ascribed to apparent first-order kinetics (Fig. S38†). However, the decay is probably sharper than first-order kinetics would suggest. This is likely due to NP aggregation, as corroborated by DLS measurements after reaction (Fig. S48a†), which may lead to a reduced active surface area during the reaction. Despite this, BF-TEM analysis of the spent catalyst (Fig. S47b and c†) showed no significant nanoparticle growth, and ICP analysis confirmed negligible metal leaching (Table S5†), indicating the strong potential of Ru@IMesPrSO₃ for catalyst reusability.

The TOF of our catalyst (61 h⁻¹) was also calculated from the initial reaction rate (Fig. S37†), dividing it by the total number of Ru atoms in the media. The TON corresponding to this same time interval (i.e., 4 minutes) was 4. Notably, this catalytic system ranks among the most efficient heterogeneous catalysts operating under mild conditions (<100 °C, ≤10 bar) in aqueous media for the selective hydrogenation of 5-HMF to 2,5-BHMF (Table S3†). Specifically, the catalyst demonstrates exceptional competitiveness regarding the 2,5-BHMF produced per gram of catalyst and per unit time (specific productivity, Table S3†). We can partially ascribe this great catalytic behaviour to the fact that NHC ligands are strong σ-donors that increase the electron density on the Ru centre.^31,33 This enhanced electron density would make the metal more nucleophilic, promoting stronger interactions with the electrophilic C=O group of 5-HMF.

Pd@IMesPrSO₃ for 2,5-BHMTHF production

We believed a promising catalytic system to move one step forward in 5-HMF hydrogenation should be based on Pd. Pd species within the subnanometric to 10 nm size range exhibit notable catalytic activity in the total hydrogenation of the furan ring, with their reactivity influenced by particle size, surface structure, and reaction conditions.⁵⁸ Literature generally agrees that unsaturated Pd sites (corners, edges) are preferred for activating C=O in 5-HMF. In contrast, more saturated Pd sites at the NPs boost H₂ dissociation and are responsible for the strong interactions with the π electrons of the furan ring.^59–61 In that sense, small particles, such as those documented in our previous research on Pd@IMesPrSO₃ (ca. 2 nm), exhibit a distribution of surface sites with varying degrees of coordination. However, we previously proved by ATR-FTIR and Solid-State NMR with CO and ¹³CO, respectively, that, in these NPs, the NHC ligands coordinate mainly on the more unsaturated sites, leaving the flat surfaces free.³⁴ This effect could reduce competitive upright adsorption from C=O groups. Therefore, flat adsorption on exposed facets would dominate, favouring full furan ring hydrogenation. Furthermore, the electronic influence of the NHC ligand^31,33 is anticipated to enhance both C=O activation on the less active facets and H₂ dissociation, thereby potentially enhancing the efficiency of total hydrogenation.

Therefore, we reproduced the synthesis of Pd@IMesPrSO₃ nanoparticles following our original procedure as detailed in the Experimental section; Water-soluble nanoparticle syntheses.³⁴Fig. 3a illustrates a representative BF-TEM image and the corresponding particle size distribution, revealing an average size of 2.1 nm with a narrow distribution, indicative again of a uniform and well-controlled synthesis process. More BF-TEM images provided in the ESI (Fig. S3†) further confirm the homogeneity of the nanoparticle population. As in the previous case, DLS analysis confirms a stable colloidal suspension formation, with aggregate sizes predominantly in the 20–50 nm range (Fig. S6b†).

Fig. 3 (a) Particle size distribution of Pd@IMesPrSO₃ NPs and representative BF-TEM micrograph. (b) PDF obtained from WAXS measurement of Pd@IMesPrSO₃ NPs and comparison against a pentatwin decahedral (fcc) model fit (6-1-0).

Furthermore, in Fig. 3b, we present the WAXS structural analysis of the Pd@IMesPrSO₃ nanoparticles for the first time, utilizing the derived PDF. The red line depicts the best-fitted model based on a pentatwin geometry, resulting in a decahedral shape with a diameter of an enveloping sphere of 23.3 Å, in agreement with BF-TEM observed sizes. Less satisfactory fitting attempts can be found in the ESI (Fig. S9†). The 3D inset illustrates the three-dimensional representation of the decahedral structure utilized in the modelling.

With these NPs, several catalytic tests were conducted at different temperatures (25, 40, 70, 100, and 120 °C) to determine the most suitable reaction conditions for the total hydrogenation of 5-HMF (Fig. 4 and S39†). Surprisingly, the selectivity toward full hydrogenation to 2,5-BHMTHF was more pronounced at lower temperatures (25 and 40 °C; ≈55% yield). Indeed, a decline in the reaction rate is observed at all temperatures, but it occurs more significantly at high temperatures, with 2,5-BHMTHF being stabilized at lower yields (Fig. 4 and S39†). This observation correlates with greater growth in nanoparticle size in the spent catalyst at higher temperatures (Fig. S46 and S47e, f†). This suggests that faster catalyst deactivation at elevated temperatures reduces the availability of active sites necessary for furan ring hydrogenation. The reaction profile at 40 °C in Fig. 4 shows the rapid and complete consumption of 5-HMF within approximately 2.5 hours. 2,5-BHMF appears rapidly at the early reaction stage, reaching a maximum yield of around 10% before declining slightly. The rapid increase in 2,5-BHMTHF appears to correlate with two consecutive first-order reactions, where the rate of the second step (2,5-BHMF to 2,5-BHMTHF) is significantly higher than the hydrogenation of the carbonyl group. Additionally, the by-product poly-5-(hydroxymethyl)tetrahydrofuran-2-carbaldehyde (PHMTHF), formed through polymerization of the full hydrogenated furan ring was identified by NMR (Fig. S25–S28†). The carbon balance remained above 88%, confirming that most products and intermediates were accounted for.

Fig. 4 Temporal evolution of product distribution in the hydrogenation of 5-HMF using Pd@IMesPrSO₃ NPs as a catalyst. Reaction conditions: 63.1 mg (0.5 mmol) of 5-HMF, 5 mL of water, 5 mg of catalyst, at 40 °C, 5 bar of H₂, and 800 rpm. Notes: (i) Dashed lines are only visual guidelines. (ii) Error bars represent the standard deviation from at least three independent experiments at the same data point. (iii) Carbon balance (15 h): 88%.

Remarkably, it is difficult to find studies for Pd-catalyzed complete hydrogenation of 5-HMF to 2,5-BHMTHF, or the conversion of furfural into tetrahydrofurfuryl alcohol, in aqueous media under mild conditions (<150 °C, ≤10 bar H₂).^62–65 In that sense, a comparative Table can be found in the ESI (Table S4†). As hypothesized before, the behaviour of our pentatwinned Pd 6-1-0 decahedral NPs can be tentatively attributed to a unique combination of ligand effects and selective site exposure, as supported by the existing literature. First, the NHC ligands selectively coordinate to step or edge sites,³⁴ preventing the usual edge-dominated C=O hydrogenation pathway and shifting both C=O and furan ring hydrogenation toward the twin-modified (111) facets, where strong π-interactions with the furan ring are promoted.⁶¹ Second, it is reasonable that the electronic contribution from NHC ligands^31,33 might promote C=O activation and H₂ dissociation.

Despite its promising catalytic properties, the Pd@IMesPrSO₃ system exhibits progressive nanoparticle growth under reaction conditions, as evidenced by DLS (Fig. S48b†) and BF-TEM analysis of the spent catalyst (Fig. S43†). While leaching was not observed (ESI Table S5†), NP aggregation and growth ultimately led to catalyst deactivation, reducing overall yield to the desired 2,5-BHMTHF. Future optimization may focus on designing bimetallic systems, or ligands with similar coordination behaviour and a higher stabilization effect to achieve the full potential of this Pd-based system. Indeed, only increasing the IMesPrSO₃ equivalents in the synthesis already offered a glimpse of the potential improvement that could be achieved with the aforementioned ligand optimization strategies (Table 1).

Table 1 Effect of IMesPrSO₃ on total 5-HMF hydrogenation with Pd-based NPs

Equivalents of IMesPrSO3	5-HMF Conversion/%	2,5-BHMTHF Yield/%
Reaction conditions: 63.1 mg (0.5 mmol) of 5-HMF, 5 mL of water, 0.02 mmol of Pd, at 40 °C, 5 bar of H2, and 800 rpm, for 15 h.
0.2	100	55
0.5	99	60

---

Ir@IMesPrSO₃ for oxidated cyclopentenone production

In the next phase of our study, we selected Ir as the metal centre for our NHC-stabilized NPs to achieve enhanced thermal stability compared to previously explored Pd and Ru systems. While IMesPrSO₃-stabilized Pd and Ru have shown interesting hydrogenation capabilities, they also suffer at high temperature from particle agglomeration (Fig. S48†), growth (Fig. S47†), and fouling (Table S6†). All this leads to deactivation under high-temperature reaction conditions (Fig. S36 and S39†). Iridium, with its superior thermal stability, was chosen to enable access to high-temperature 5-HMF-derived products, particularly those involving hydration/dehydration reactions and ring rearrangements, such as HHD or 1,2,6-hexanetriol (HTO). At the same time, we acknowledge that Ir generally exhibits lower hydrogenation activity than Pd and Ru, necessitating harsher reaction conditions for H₂ activation.⁶⁶ Nevertheless, superior thermal stability of Ir may prevent deactivation,⁶⁷ ensuring prolonged structural integrity and sustained reactivity under more demanding conditions.

Based on that, we reproduced the synthesis of Ir@IMesPrSO₃ NPs following the original procedure, as described in the Experimental section; Water-soluble nanoparticle syntheses.³⁵Fig. 5a illustrates the particle size distribution obtained for this system alongside a representative BF-TEM image of the NPs. The distribution reveals an average particle size of 2.0 nm. The BF-TEM images (Fig. 5a and S4†) show a uniform morphology across the sample. Moreover, DLS measurements were performed to characterize the aggregation state of the Ir@IMesPrSO₃ NPs in aqueous media. The results in Fig. S6c† indicate an aggregate size distribution predominantly in the 1 to 10 nm range. Although DLS has reduced accuracy for particles below 10 nm due to hydrodynamic effects, the absence of larger aggregates suggests that the nanoparticles remain well-dispersed in water, with aggregate sizes likely below 10 nm.

Fig. 5 (a) Particle size distribution of Ir@IMesPrSO₃ NPs and representative BF-TEM micrograph. (b) PDF obtained from WAXS measurement of Ir@IMesPrSO₃ NPs and comparison against a cuboctahedron (fcc) model fit (9-4).

Structural characterization using WAXS (Fig. 5b) provided further insight into the atomic arrangement of Ir@IMesPrSO₃ NPs. The PDF analysis (red curve) yielded a best-fit model corresponding to a cuboctahedral geometry with a diameter of an enveloping sphere of 21.6 Å, in agreement with BF-TEM observed sizes. The model agrees with the experimental data indicated by the low residual value. Attempts to fit the data using alternative models yielded less satisfactory results, as shown in the ESI (Fig. S10†). A 3D representation of the cuboctahedral structure used in the modelling is provided as an inset, offering a visual perspective of the nanoparticle's atomic arrangement.

To investigate 5-HMF hydrogenation using these Ir@IMesPrSO₃ nanoparticles, experiments were conducted at 5 H₂ bar and various temperatures (80, 120, 140, and 160 °C; Fig. 6 and S40†). The goal of these experiments was to explore reaction pathways involving ring opening and/or rearrangement. Excitingly, among the tested conditions, 140 °C gave a surprisingly high yield (>40%) of oxidized cyclopentenones at 4 h, specifically 4-hydroxy-4-(hydroxymethyl)cyclopent-2-en-1-one (4-HHCPEN; 38%), and a low amount of 4-hydroxy-3-(hydroxymethyl)cyclopent-2-en-1-one (4-H-3-HCPEN; 4%). The reaction mechanism appears to involve a first hydrogenation of 5-HMF to 2,5-BHMF, the key intermediate, followed by a ring rearrangement step, resembling a Piancatelli-type reaction, to form the cyclopentenones.⁶⁸ At 140 °C, the temporal evolution shows steady 5-HMF consumption, with 2,5-BHMF peaking early and then converting into cyclopentenones. Lower temperatures (80–120 °C) limit the rearrangement process, resulting in incomplete 5-HMF conversion and 2,5-BHMF as the dominant product. At 160 °C, faster 5-HMF consumption led to an earlier peak of 2,5-BHMF, but the yield of oxidized cyclopentenones decreases slightly due to competing side reactions such as 2,5-BHMF hydration and reduction to HHD. These results highlight 140 °C as the ideal temperature, balancing 5-HMF hydrogenation and rearrangement to maximize the formation of the cyclopentenones (Fig. 6).

Fig. 6 Temporal evolution of product distribution in the hydrogenation of 5-HMF using Ir@IMesPrSO₃ NPs as a catalyst. Reaction conditions: 63.1 mg (0.5 mmol) of 5-HMF, 5 mL of water, 5 mg of catalyst, at 140 °C, 5 bar of H₂, and 800 rpm. Notes: (i) Dashed lines are only visual guidelines. (ii) Error bars represent the standard deviation from at least three independent experiments at the same data point. (iii) Carbon balance (4 h): 82%.

Kinetic observations suggest a reaction mechanism (Scheme 3) wherein 5-HMF is hydrogenated to 2,5-BHMF, a key intermediate, which then undergoes a Piancatelli-type rearrangement (Scheme S2, ESI†) to form cyclopentenones. An alternative pathway might involve the intramolecular aldol condensation of 1-hydroxyhex-3-ene-2,5-dione (HHED) derived from 2,5-BHMF. Indeed, experiments conducted under Ar instead of H₂ (with RuIr₂-based catalyst, vide infra) resulted in the accumulation of HHED (Fig. S29 and S30†). However, as no HHED was detected in these experiments with H₂ and the Ir-based catalyst, this accounts for a short-lived intermediate that is more compatible with a faster hydrogenation of the C=C to yield HHD. Additionally, a control experiment showed that 2,5-BHMF, but not 5-HMF, could easily convert into cyclopentenones without a catalyst at 160 °C (56% yield, 3 h), reinforcing the notion that Ir is not essential for the Piancatelli rearrangement but plays a crucial role in selective 5-HMF hydrogenation under elevated temperatures. Other studies have previously established the purely thermal feasibility of the Piancatelli rearrangement from 2,5-BHMF.⁶⁹ Remarkably, efforts to employ Ru@IMesPrSO₃ at temperatures exceeding 100 °C were unsuccessful in achieving similar selectivity (Fig. S36b and c†). This outcome may be attributed to Ru over-reactivity (Scheme 3), resulting in competitive pathways such as the hydrogenation of the oxidized cyclopentenones to HCPEN and HCPN, or the formation and hydrogenation of HHED, which results in HHD.

Scheme 3 Main reaction network proposed in 5-HMF hydrogenation with 5 H₂ bar and Ru@IMesPrSO₃, Pd@IMesPrSO₃ or Ir@IMesPrSO₃ as catalysts. [a]: Piancatelli rearrangement. M: Ru, Pd, Ir.

A catalyst-free conversion of 2,5-BHMF to oxidized cyclopentenones in water was previously reported in the literature. While achieving a 54% yield, this strategy requires prolonged reaction times (up to 24 h) and a previous (separate) conversion of 5-HMF into 2,5-BHMF.⁶⁹ Additionally, microwave heating studies catalyzed Piancatelli rearrangements of 2,5-BHMF derivatives, achieving higher yields to 4-HHCPEN (∼57%) but again requiring pre-synthesized dicarbinol intermediates.⁷⁰ While our Ir-mediated approach operates at 140 °C, meaning it does not improve upon previous studies in terms of temperature, its ability to start directly from 5-HMF in a single reaction system is a major advantage. This highlights the versatility of the Ir catalyst, which enables both hydrogenation and subsequent thermal rearrangement within a streamlined catalytic approach, offering a more direct and efficient route for cyclopentenone synthesis from this biomass-derived feedstock. Moreover, despite some aggregation detected by DLS (Fig. S48c†), the catalyst does not experience NP growth (Fig. S44†) after the reaction under the optimal conditions (140 °C and 5 H₂ bar).

RuIr₂@IMesPrSO₃ for HHD production

The previous studies employing Ir@IMesPrSO₃ demonstrated selectivity toward Piancatelli-type rearrangements, converting 5-HMF into oxygenated cyclopentenones such as 4-HHCPEN and 4-H-3-HCPEN. In contrast, Ru@IMesPrSO₃ at temperatures ≥100 °C, directed the reaction toward less oxygenated cyclopentanones (HCPEN and HCPN; Fig. S36b and c†), either by promoting full hydrogenation of the Piancatelli-derived cyclopentenones or by facilitating the hydrogenation of HHED to HHD, which can undergo intramolecular aldol condensation to form HCPEN and HCPN (Scheme 3). However, while Ru@IMesPrSO₃ exhibits a stronger hydrogenation capability than Ir@IMesPrSO₃, it suffers from deactivation (>100 °C; Fig. S36b and c†). Therefore, we hypothesized introducing Ir into a Ru-based system (i.e., RuIr₂@IMesPrSO₃) might unlock new selectivity trends in 5-HMF hydrogenation at higher temperatures, combining Ru's enhanced hydrogenation activity with Ir's superior thermal stability.

Consequently, our previously reported RuIr₂@IMesPrSO₃ NPs were prepared according to our established procedure.³⁵ Particle size characterization by BF-TEM (Fig. 7a and S5†) revealed narrowly dispersed NPs with an average diameter of 1.7 ± 0.3 nm. The DLS results shown in Fig. S6d† reveal a predominant peak in the 1–2 nm range. As discussed earlier, DLS is less accurate for aggregates smaller than 10 nm due to reduced scattering intensity and the inclusion of solvation shells and surface ligands in the measured size. Nevertheless, this measurement confirms that the RuIr₂@IMesPrSO₃ NPs exhibit a small size distribution with no significant large aggregates, highlighting their effective stabilization in aqueous media.

Fig. 7 (a) Particle size distribution of RuIr₂@IMesPrSO₃ NPs and representative BF-TEM micrograph. (b) PDF obtained from WAXS measurement of RuIr₂@IMesPrSO₃ NPs and comparison against a decahedron (fcc) model fit (5-1-0).

Fig. 7b presents the results of the WAXS structural analysis of the RuIr₂@IMesPrSO₃ NPs. For the first time, a PDF derived from WAXS data has been employed to model the structure of these NPs. The red line corresponds to the best-fit structural model, which assumes a decahedral geometry with 5 (1) atoms on the 100 facets perpendicular (parallel) to the five-fold axis, resulting in a diameter of an enveloping sphere of 18.7 Å. The low residual value (R_w = 0.22) indicates a high level of agreement between the model and the experimental data. Models that do not demonstrate the same goodness are outlined in the ESI (Fig. S11†). The inset presents a 3D representation of the decahedral nanoparticle structure utilized in the modelling, illustrating a plausible atomic arrangement within the RuIr₂ alloy.

The reaction profiles for RuIr₂@IMesPrSO₃ at different temperatures (100, 120, and 140 °C; Fig. S41†) under 5 bar H₂ reveal significant variations in product distribution. In all cases, 5-HMF undergoes rapid initial conversion, with 2,5-BHMF as a major early intermediate. At 100 °C, 2,5-BHMF accumulates significantly, while the formation of HHD, HHCPEN, and HCPEN, remains relatively low, suggesting further reactions are slow. At 120 °C, 2,5-BHMF reaches a maximum concentration within the first one and a half hours and then decreases, while HHD and HCPEN gradually increase, indicating that higher temperatures facilitate (de)hydration, Piancatelli rearrangement and aldol condensations (Scheme 3). At 140 °C, 2,5-BHMF accumulation is less pronounced, and HCPEN and HCPN become more prominent, suggesting that higher temperatures favour deeper hydrogenation and conversion into stable cyclopentanone derivatives. Notably, at this temperature, HHD concentration is stabilized, suggesting that its conversion to cyclopentanones (HCPEN and HCPN) via aldol condensation is no longer the main route. Instead, product evolution supports a Piancatelli-mediated pathway, where 5-HMF → 2,5-BHMF → 4-HHCPEN → HCPEN → HCPN emerges as the predominant sequence at elevated temperatures.

In summary, at 140 °C, the reaction exhibits a combination of selectivities, with products arising from both Piancatelli rearrangement (leading to HCPEN/HCPN) and the hydration-dehydration-hydrogenation of 2,5-BHMF (leading to HHD), indicating that both pathways are active under these conditions. From the reaction profile, the final yields show that HCPEN and HCPN together account for approximately 35%, while HHD stabilizes at around 20%. Notably, across all temperatures, the carbon balance decreases progressively over time, indicating the formation of undetected byproducts. At 140 °C, the carbon balance drops the fastest, in good agreement with a polymerization (humin formation) favoured at elevated temperatures.⁵⁷

Given that Brønsted acids could promote the hydrolysis of the 2,5-BHMF to give HHD,⁵⁵ potentially enhancing the yield of this compound, a slight pH modifier (i.e., AgBF₄) was selected to fine-tune selectivity towards the hydrolysis pathway. Although AgBF₄ is commonly used as a Lewis acid in organic media, upon exposure to water, it potentially produces hydrogen fluoride.⁷¹ This effect has been validated by evaluating its impact on the pH of the reaction mixture, leading to a significant decrease from 4.38 to 1.43 (Table S7†).

Upon this addition, the temporal evolution of product distribution in Fig. 8 shows a significant accumulation of HHD, reaching over 70% yield. At the same time, the formation of HCPEN and HCPN is strongly suppressed, indicating that Piancatelli rearrangement is minimized. The overall lower yield of HCPEN/HCPN (< 5%) compared to the reaction without AgBF₄ suggests that, in the presence of Brønsted acid activation, 4-HHCPEN from 2,5-BHMF is not efficiently formed because of the competition of HHD formation, preventing its subsequent hydrogenation into HCPEN and HCPN. Additionally, the carbon balance reaches 85%, suggesting that side reactions leading to undetected (insoluble) byproducts are less pronounced compared to the reaction without AgBF₄. These results demonstrate that Brønsted acid conditions favour hydrolysis-driven pathways over rearrangement, thereby providing an effective means of switching product selectivity.

Fig. 8 Temporal evolution of product distribution in the hydrogenation of 5-HMF using RuIr₂@IMesPrSO₃ NPs as a catalyst. Reaction conditions: 63.1 mg (0.5 mmol) of 5-HMF, 5 mL of water, 5 mg of catalyst, 5 mg AgBF₄, at 140 °C, 5 bar of H₂, and 800 rpm. Notes: (i) Dashed lines are only visual guidelines. (ii) Error bars represent the standard deviation from at least three independent experiments at the same data point. (iii) Carbon balance (8 h): 85%. Note: The control experiment using AgBF₄ without RuIr₂@IMesPrSO₃ NPs provided 4% yield of HHD.

To the best of our knowledge, only two systems have so far been described producing yields above 20% to HHD from 5-HMF, without requiring temperatures above 150 °C, H₂ pressures higher than 10 bar, or organic solvents. First, the group of van Bekkum was able to produce up to 28% yield of HHD in water using 10% Pd/C and HCl under mild reaction conditions (1 bar of H₂, 60 °C, 6 h).⁷² Later, Y. Cao and co-workers developed a catalyst comprising Pd single atoms dispersed on gold nanoparticles, which are in turn supported on TiO₂, for the synthesis di- and tri-ketones from biomass-derived furans. The catalyst was found to be highly efficient and afforded HHD from 5-HMF in an excellent yield (ca. 90%) under moderate reaction conditions (10 bar of H₂, 120 °C, 3 h).⁷³ This lack of solid catalysts for the conversion of 5-HMF into HHD under mild aqueous conditions, coupled with the effective performance of our system that utilizes RuIr₂@IMesPrSO₃ nanoparticles and AgBF₄, signifies a significant advancement in the valorization of 5-HMF.

Comparative catalyst performance and Ru@IMesPrSO₃ reusability

To sum up, Table 2 provides a comprehensive comparison of the catalytic performances across all nanoparticle systems explored in this work under their respective optimized conditions. This table highlights the distinct selectivity profiles enabled by the rational choice of metal, with each system steering the hydrogenation of 5-HMF toward specific high-value-added products.

Table 2 Overview of the different catalytic performance at two conversion levels

Catalyst	Temperature/°C	5-HMF Conversion/%	Selectivity to Main Product(s)^b/%
Reaction conditions: 63.1 mg (0.5 mmol) of 5-HMF, 5 mL of water, 5 mg of catalyst, 5 bar of H2, and 800 rpm.a 5 mg of AgBF4 was used as an additive.b Only products with selectivity exceeding 15% are included.
Ru@IMesPrSO3	30	45	2,5-BHMF: 80
		95	2,5-BHMF: 89
Pd@IMesPrSO3	40	54	2,5-BHMTHF: 40; 2,5-BHMF: 16
		99	2,5-BHMTHF: 55
Ir@IMesPrSO3	140	39	2,5-BHMF: 42; HHCPENs: 21
		87	HHCPENs: 48; HHD: 20
RuIr2@IMesPrSO3 ^a	140	40	HHD: 53; 2,5-BHMF: 17
		91	HHD: 75

---

Among all catalysts, Ru@IMesPrSO₃ emerges as the most selective and efficient system, reaching ≈90% selectivity toward 2,5-BHMF under exceptionally mild conditions (30 °C, 5 bar H₂). This system also ranks among the best reported in the literature for aqueous-phase hydrogenation of 5-HMF under similarly mild conditions (see Table S3†), as discussed throughout the manuscript. Importantly, the Ru catalyst avoids over-hydrogenation and undesirable side reactions, achieving near-complete carbon balance (i.e. 97%) at full conversion. Furthermore, unlike Pd@IMesPrSO₃, which suffers from temperature-induced nanoparticle growth and progressive deactivation, Ru@IMesPrSO₃ maintains its nanostructure post-reaction, with no observable growth by TEM (Fig. S42†). These findings support the superior thermal and colloidal stability of the Ru-based system in aqueous media. To further assess the catalyst's viability for sustainable processing, its recyclability was explored. Although recovering the water-soluble NPs was challenging, we could devise a method to successfully recover the NPs by cooling the reaction mixture before centrifugation (Experimental section; Catalytic tests). The recovered NPs were then washed with water and reused in consecutive catalytic cycles.

First, results showing a decline in 5-HMF conversion from 92% to 68% over three uses (Fig. 9a) were achieved, which also translated into a decline in the yield of 2,5-BHMF from 85% to 60% (Fig. 9b), despite the preservation of the selectivity. Remarkably, we noted that, during the work-up to recover the catalyst, much of the initial catalyst mass was lost, which would explain the partial loss of the catalytic activity. In order to minimize the loss percentage, the process was successfully scaled up by a factor of five (Experimental section; Catalytic tests), resulting in yields exceeding 90% of 2,5-BHMF after a reaction time of 2.5 h. Again, the first three tests in the red bars of Fig. 9a described a decline in the catalytic activity over time, but the drop was less pronounced (from 100 to 83%) this time, and the selectivity to 2,5-BHMF was well preserved. Indeed, TEM analysis after 3 uses confirmed preservation of nanoparticle size and morphology (Fig. 9c), while WAXS-derived PDF analysis revealed no Ru–O bond features (Fig. 9d), ruling out oxidation or structural reorganization. Unfortunately, we found that approximately 25% of the initial catalyst was still lost during the recovery process. To ascertain whether this loss accounts for the observed reduction in catalytic activity, subsequent experiments were conducted in which fresh catalyst was added to the recovered NPs until the initial amount of catalyst was restored. In this case, the catalytic results in Fig. 9 (diagonal stripes) point to a total restoration of the catalytic activity. These results, together with the catalyst characterization after three uses, indicate that the decline in catalytic activity was likely due to load loss, not NP growth or oxidation.

Fig. 9 Reusability study of Ru@IMesPrSO₃ NPs in the hydrogenation of 5-HMF. (a) 5-HMF conversion over three consecutive uses. (b) 2,5-BHMF yield over three consecutive uses. Reaction conditions: 63.1 mg (0.5 mmol) of 5-HMF, 5 mL of water, 5 mg of catalyst, at 30 °C, 5 bar of H₂, and 800 rpm, during 2.5 h. (c) Particle size distribution comparison between fresh and spent Ru@IMesPrSO₃ NPs. (d) PDF obtained from WAXS measurements of fresh and spent Ru@IMesPrSO₃ NPs.

Conclusions

This study reports the selective hydrogenation of 5-HMF in aqueous media using water-soluble Ru, Pd, Ir, and RuIr₂ nanoparticles stabilized by a water-soluble NHC ligand (IMesPrSO₃). XRD, BF-TEM, and DLS were employed to verify the success of the syntheses. Moreover, we provide here the PDF resulting from WAXS analyses of these NPs for the first time.

A key outcome of this work is demonstrating that distinct hydrogenation pathways and product selectivities can be achieved upon varying the metal composition and reaction conditions. This result highlights the versatility of this colloidal platform, allowing targeted synthesis of high-value compounds from 5-HMF, ranging from 2,5-BHMF and 2,5-BHMTHF, to oxidized cyclopentenones (4-H-3-HCPEN and 4-HHCPEN) and HHD, all without changing the stabilizing ligand or introducing solid supports. Specifically, Ru@IMesPrSO₃ stands out as a rare example of a potentially reusable catalyst capable of achieving partial 5-HMF hydrogenation (2,5-BHMF) under exceptionally mild conditions (30 °C, 5 bar H₂) in water, showcasing the potential of NHC-stabilized NPs in the selective hydrogenation of biomass derivatives under environmentally friendly conditions. Ir@IMesPrSO₃ unlocked a selectivity toward oxidized cyclopentenones only previously reported if starting from 2,5-BHMF, while RuIr₂@IMesPrSO₃, aided by AgBF₄, directed the reaction toward HHD formation.

We believe this work opens an avenue to exploit the potential of NHC-stabilized water-soluble nanoparticles in the valorization of specific biomass derivatives. Future optimizations could involve modifying the stabilizing ligand, adjusting its molar ratio, or exploring different organometallic precursors to further control nanoparticle shape and reactivity, thermal stability, and acid–base properties. Such refinements could improve catalyst reusability and extend their applicability to more complex reactions. Moreover, our on-going efforts are currently focused on extending this strategy to first-row transition metals such as Ni and Cu, with the aim of developing more sustainable and cost-effective catalytic systems based on earth-abundant elements.

Author contributions

Oscar Suárez-Riaño: formal analysis (equal); investigation (lead); validation (equal); visualization (equal); writing – original draft preparation (supporting); writing – review and editing(equal). Jaime Mazarío: conceptualization (lead); formal analysis (equal); funding acquisition (equal); investigation (equal); methodology (lead); validation (equal); visualization (equal); writing – original draft preparation (lead); writing – review and editing (equal). Gabriel Mencia: conceptualization (equal); formal analysis (lead); investigation (supporting); validation (equal); writing – review and editing (supporting). Víctor Varela-Izquierdo: conceptualization (supporting); formal analysis (equal); investigation (supporting); methodology (supporting); validation (supporting); visualization (equal); writing – review and editing (supporting). Nicolas Ratel-Ramond: formal analysis (equal); software (lead); writing – review and editing (supporting). Antonio Martín-Pinillos: investigation (supporting). Edwin A. Baquero: supervision (supporting); writing – review and editing (equal). Luis M. Martínez-Prieto: funding acquisition (supporting); supervision (supporting); writing – review and editing (supporting). Simon Tricard: supervision (supporting); writing – review and editing (supporting). Bruno Chaudret: conceptualization (equal); funding acquisition (lead); project administration (lead); resources (lead); supervision (equal); writing – original draft preparation (supporting); writing – review and editing (supporting).

Conflicts of interest

There are no conflicts to declare.

Data availability

All data supporting this article have been included as part of the ESI.† Additionally, unformatted raw data is available from the authors upon request.

Acknowledgements

J. M. acknowledges his MSCA individual postdoctoral fellowship from the Horizon Europe research and innovation program (Project 101109254 – BIOCATMAG). V. V. I. and B. C. are grateful to the French State aid managed by the Agence Nationale de la Recherche under France 2030 plan, bearing the reference code ANR-22-PESP-0010: Projet ciblé “POWERCO2” within the PEPR project SPLEEN. E. A. B. is grateful to the Vicerrectoría de Investigación of Universidad Nacional de Colombia for their financial support through project with HERMES code 59659. L. M. M. P. acknowledges Junta de Andalucía (ProyExcel_00706) for financial support, as well as Grants PID2021-126080OA-I00, TED2021-132087A-I00 and CNS2023-145078 funded by MICIU/AEI/10.13039/501100011033, “ERDF/EU” and “European Union NextGenerationEU/PRTR”. The authors greatly acknowledge Guy Lippens for his support during NMR analyses.

References

1. P. G. Levi and J. M. Cullen, Environ. Sci. Technol., 2018, 52, 1725–1734,  DOI:10.1021/acs.est.7b04573.
2. A. Abbas, M. Cross, X. Duan, S. Jeschke, M. Konarova, G. W. Huber, A. F. Lee, E. C. Lovell, J. Y. C. Lim, A. Polyzos, R. Richards and K. Wilson, One Earth, 2024, 7(5), 738–741,  DOI:10.1016/j.oneear.2024.04.018.
3. P. T. Anastas, M. M. Kirchhoff and T. C. Williamson, Appl. Catal., A, 2001, 221(1–2), 3–13,  DOI:10.1016/S0926-860X(01)00793-1.
4. J. R. Banu, S. Kavitha, V. K. Tyagi, M. Gunasekaran, O. P. Karthikeyan and G. Kumar, Fuel, 2021, 302, 121086,  DOI:10.1016/j.fuel.2021.121086.
5. V. G. Yadav, G. D. Yadav and S. C. Patankar, Clean Technol. Environ. Policy, 2020, 22, 1757–1774,  DOI:10.1007/S10098-020-01945-5.
6. E. T. C. Vogt and B. M. Weckhuysen, Nature, 2024, 629(8011), 295–306,  DOI:10.1038/s41586-024-07322-2.
7. Q. Hou, X. Qi, M. Zhen, H. Qian, Y. Nie, C. Bai, S. Zhang, X. Bai and M. Ju, Green Chem., 2021, 23(1), 119–231,  10.1039/D0GC02770G.
8. R. J. van Putten, J. C. Van Der Waal, E. D. De Jong, C. B. Rasrendra, H. J. Heeres and J. G. de Vries, Chem. Rev., 2013, 113(3), 1499–1597,  DOI:10.1021/cr300182k.
9. A. A. Rosatella, S. P. Simeonov, R. F. M. Frade and C. A. M. Afonso, Green Chem., 2011, 13(4), 754–793,  10.1039/C0GC00401D.
10. E. de Jong, H. R. A. Visser, A. S. Dias, C. Harvey and G. J. M. Gruter, Polymers, 2022, 14(5), 943,  DOI:10.3390/polym14050943.
11. Y. Román-Leshkov, C. J. Barrett, Z. Y. Liu and J. A. Dumesic, Nature, 2007, 447(7147), 982–985,  DOI:10.1038/nature05923.
12. J. J. Bozell and G. R. Petersen, Green Chem., 2010, 12(4), 539–554,  10.1039/B922014C.
13. Y. Nakagawa, M. Tamura and K. Tomishige, ACS Catal., 2013, 3, 2655–2668,  DOI:10.1021/cs400616p.
14. C. Chien Truong, D. Kumar Mishra, S. Hyeok Ko, Y. Jin Kim and Y.-W. Suh, ChemSusChem, 2022, 15(13), e202200178,  DOI:10.1002/cssc.202200178.
15. L. Hu, J. Xu, S. Zhou, A. He, X. Tang, L. Lin, J. Xu and Y. Zhao, ACS Catal., 2018, 8(4), 2959–2980,  DOI:10.1021/acscatal.7b03530.
16. S. Dutta and N. S. Bhat, ACS Omega, 2021, 6(51), 35145–35172,  DOI:10.1021/acsomega.1c05861.
17. C. Castillo, M. H. Brijaldo, L. P. C. Silva and F. B. Passos, ChemCatChem, 2023, 15(15), e202300480,  DOI:10.1002/cctc.202300480.
18. X. Tang, J. Wei, N. Ding, Y. Sun, X. Zeng, L. Hu, S. Liu, T. Lei and L. Lin, Renewable Sustainable Energy Rev., 2017, 77, 287–296,  DOI:10.1016/j.rser.2017.04.013.
19. J. Zhang, T. Wang, X. Tang, L. Peng, J. Wei and L. Lin, BioResources, 2018, 13(3), 7137–7154,  DOI:10.15376/biores.13.3.Zhang.
20. E. A. Baroncini, S. Kumar Yadav, G. R. Palmese and J. F. Stanzione III, J. Appl. Polym. Sci., 2016, 133(45), 44103,  DOI:10.1002/app.44103.
21. B. Wozniak, S. Tin and J. G. de Vries, Chem. Sci., 2019, 10, 6024–6034,  10.1039/C9SC01309A.
22. S. P. Roche and D. J. Aitken, Eur. J. Org. Chem., 2010, 5339–5358,  DOI:10.1002/ejoc.201000704.
23. S. H. Burstein, Prostaglandins Other Lipid Mediators, 2020, 148, 106408,  DOI:10.1016/j.prostaglandins.2020.106408.
24. D. Belsito, D. Bickers, M. Bruze, P. Calow, M. L. Dagli, W. Dekant, A. D. Fryer, H. Greim, Y. Miyachi, J. H. Saurat and I. G. Sipes, Food Chem. Toxicol., 2012, 50, S517–S556,  DOI:10.1016/j.fct.2012.04.019.
25. P. Anastas and N. Eghbali, Chem. Soc. Rev., 2009, 39(1), 301–312,  10.1039/B918763B.
26. K. Gupta, R. K. Rai and S. K. Singh, ChemCatChem, 2018, 10(11), 2326–2349,  DOI:10.1002/cctc.201701754.
27. A. Messori, A. Fasolini and R. Mazzoni, ChemSusChem, 2022, 15(13), e202200228,  DOI:10.1002/cssc.202200228.
28. C. Mondelli, G. Gözaydın, N. Yan and J. Pérez-Ramírez, Chem. Soc. Rev., 2020, 49(12), 3764–3782,  10.1039/D0CS00130A.
29. J. Huo, J.-P. Tessonnier and B. H. Shanks, ACS Catal., 2021, 11(9), 5248–5270,  DOI:10.1021/acscatal.1c00197.
30. O. Suárez-Riaño and E. A. Baquero, in Surface Functionalized Metal Catalysts, ed. L. M. Martínez-Prieto, Topics in Organometallic Chemistry, Springer Cham., 2024, ch. 5, vol. 75, pp. 199–242.  DOI:10.1007/3418_2024_118.
31. M. Koy, P. Bellotti, M. Das and F. Glorius, Nat. Catal., 2021, 4(5), 352–363,  DOI:10.1038/s41929-021-00607-z.
32. C. A. Smith, M. R. Narouz, P. A. Lummis, I. Singh, A. Nazemi, C.-H. Li and C. M. Crudden, Chem. Rev., 2019, 119(8), 4986–5056,  DOI:10.1021/acs.chemrev.8b00514.
33. M. N. Hopkinson, C. Richter, M. Schedler and F. Glorius, Nature, 2014, 510(7506), 485–496,  DOI:10.1038/nature13384.
34. O. Suárez-Riaño, G. Mencia, S. Tricard, J. Esvan, P. F. Fazzini, B. Chaudret and E. A. Baquero, Chem. Commun., 2023, 59(8), 1062–1065,  10.1039/D2CC06019A.
35. A. Zuluaga-Villamil, G. Mencia, J. M. Asensio, P. F. Fazzini, E. A. Baquero and B. Chaudret, Organometallics, 2022, 41(22), 3313–3319,  DOI:10.1021/acs.organomet.2c00288.
36. L. M. Martínez-Prieto, E. A. Baquero, G. Pieters, J. C. Flores, E. de Jesus, C. Nayral, F. Delpech, P. W. N. M. van Leeuwen, G. Lippens and B. Chaudret, Chem. Commun., 2017, 53(43), 5850–5853,  10.1039/C7CC02445B.
37. D. Bouzouita, G. Lippens, E. A. Baquero, P. F. Fazzini, G. Pieters, Y. Coppel, P. Lecante, S. Tricard, L. M. Martínez-Prieto and B. Chaudret, Nanoscale, 2019, 11, 16544–16552,  10.1039/C9NR04149D.
38. L. R. Moore, S. M. Cooks, M. S. Anderson, H.-J. Schanz, S. T. Griffin, R. D. Rogers, M. C. Kirk and K. H. Shaughnessy, Organometallics, 2006, 25(21), 5151–5158,  DOI:10.1021/om060552b.
39. P. Juhás, T. Davis, C. L. Farrow and S. J. L. Billinge, J. Appl. Crystallogr., 2013, 46, 560–566,  DOI:10.1107/S0021889813005190.
40. N. Ratel-Ramond, WAXS_toolbox: PDF-for-nano (v1.0.0-alpha), 2024.  DOI:10.5281/zenodo.10694185.
41. P. Juhás, C. Farrow, X. Yang, K. Knox and S. Billinge, Acta Crystallogr., Sect. A:Found. Adv., 2015, 71, 562–568,  DOI:10.1107/S2053273315014473.
42. A. H. Larsen, J. J. Mortensen, J. Blomqvist, I. E. Castelli, R. Christensen, M. Dułak, J. Friis, M. N. Groves, B. Hammer, C. Hargus, E. D. Hermes, P. C. Jennings, P. B. Jensen, J. Kermode, J. R. Kitchin, E. L. Kolsbjerg, J. Kubal, K. Kaasbjerg, S. Lysgaard, J. B. Maronsson, T. Maxson, T. Olsen, L. Pastewka, A. Peterson, C. Rostgaard, J. Schiøtz, O. Schütt, M. Strange, K. S. Thygesen, T. Vegge, L. Vilhelmsen, M. Walter, Z. Zeng and K. W. Jacobsen, J. Phys.: Condens. Matter, 2017, 29, 273002,  DOI:10.1088/1361-648X/aa680e.
43. A. Turkin, S. Eyley, G. Preegel, W. Thielemans, E. Makshina and B. F. Sels, ACS Catal., 2021, 11(15), 9204–9209,  DOI:10.1002/cssc.202200412.
44. J. Mazarío, Z. Raad, P. Concepción, C. Cerda-Moreno and M. E. Domine, Catal. Sci. Technol., 2020, 10(23), 8049–8063,  10.1039/D0CY01423K.
45. D. K. Mishra, H. J. Lee, C. C. Truong, J. Kim, Y.-W. Suh, J. Baek and Y. J. Kim, Mol. Catal., 2020, 484, 110722,  DOI:10.1016/j.mcat.2019.110722.
46. S. Fulignati, C. Antonetti, E. Wilbers, D. Licursi, H. J. Heeres and A. M. R. Galletti, J. Ind. Eng. Chem., 2021, 100, 390.e1–390.e9,  DOI:10.1016/j.jiec.2021.04.057.
47. Y. Yang, Q. Liu, D. Li, J. Tan, Q. I. Zhang, C. Wang and L. Ma, RSC Adv., 2017, 7(27), 16311–16318,  10.1039/C7RA00605E.
48. P. Priecel, N. A. Endot, P. D. Carà and J. A. López-Sánchez, Ind. Eng. Chem. Res., 2018, 57(6), 1991–2002,  DOI:10.1021/acs.iecr.7b04715.
49. S. Fulignati, C. Antonetti, D. Licursi, M. Pieraccioni, E. Wilbers, H. J. Heeres and A. M. R. Galletti, Appl. Catal., A, 2019, 578, 122–133,  DOI:10.1016/j.apcata.2019.04.007.
50. A. B. Jain and P. D. Vaidya, Int. J. Chem. Kinet., 2016, 48(6), 318–328,  DOI:10.1002/kin.20992.
51. M. Mavrikakis and M. A. Barteau, J. Mol. Catal. A:Chem., 1998, 131(1–3), 135–147,  DOI:10.1016/S1381-1169(97)00261-6.
52. F. Q. Yan, M. H. Qiao, X. M. Wei, Q. P. Liu, J. F. Deng and G. Q. Xu, J. Chem. Phys., 1999, 111(17), 8068–8076,  DOI:10.1063/1.480140.
53. R. K. Ramamoorthy, K. Soulantica, I. Del Rosal, R. Arenal, P. Decorse, J. Y. Piquemal, B. Chaudret, R. Poteau and G. Viau, Chem. Mat., 2022, 34(7), 2931–2944,  DOI:10.1021/acs.chemmater.1c03452.
54. Y. Nanba, T. Ishimoto and M. Koyama, J. Phys. Chem. C, 2017, 121(49), 27445–27452,  DOI:10.1021/acs.jpcc.7b08672.
55. W. P. Wu, Y. J. Xu, R. Zhu, M. S. Cui, X. L. Li, J. Deng and Y. Fu, ChemSusChem, 2016, 9(10), 1209–1215,  DOI:10.1002/cssc.201501625.
56. Y. Yang, D. Yang, C. Zhang, M. Zheng and Y. Duan, Molecules, 2020, 25(11), 2475,  DOI:10.3390/molecules25112475.
57. G. Tsilomelekis, M. J. Orella, Z. Lin, Z. Cheng, W. Zheng, V. Nikolakis and D. G. Vlachos, Green Chem., 2018, 18(7), 1983–1993,  10.1039/C7GC03054A.
58. S. Chen, R. Wojcieszak, F. Dumeignil, E. Marceau and S. Royer, Chem. Rev., 2018, 118, 11023–11117,  DOI:10.1021/acs.chemrev.8b00134.
59. X. Jia, K. Yuan, G. Wang, S. Zhu, M. Dong, J. Wang, X. Li and W. Fan, Chem. Eng. J., 2024, 1(487), 150482,  DOI:10.1016/j.cej.2024.150482.
60. S. M. Rogers, C. R. A. Catlow, C. E. Chan-Thaw, A. Chutia, N. Jian, R. E. Palmer, M. Perdjon, A. Thetford, N. Dimitratos, A. Villa and P. P. Wells, ACS Catal., 2017, 7(4), 2266–2274,  DOI:10.1021/acscatal.6b03190.
61. S. H. Pang, C. A. Schoenbaum, D. K. Schwartz and J. W. Medlin, Nat. Commun., 2013, 4(1), 2448,  DOI:10.1038/ncomms3448.
62. C. Wang, A. Wang, Z. Yu, Y. Wang, Z. Sun, V. M. Kogan and Y. Y. Liu, Catal. Commun., 2021, 148, 106178,  DOI:10.1016/j.catcom.2020.106178.
63. Y. Wu, Y. Tong, H. Liang, J. Peng, X. K. Gu and M. Ding, Chem. Eng. J., 2023, 460, 141779,  DOI:10.1016/j.cej.2023.141779.
64. J. Chen, R. Liu, Y. Guo, L. Chen and H. Gao, ACS Catal., 2015, 5(2), 722–733,  DOI:10.1021/cs5012926.
65. W. W. Lestari, R. S. R. Suharbiansah, L. Larasati, F. Rahmawati, U. S. F. Arrozi, S. Durini, F. Rohman, R. Iskandar and E. Hey-Hawkins, Chem. Pap., 2022, 76(8), 4719–4731,  DOI:10.1007/s11696-022-02193-1.
66. L. M. Martínez-Prieto, I. Cano and P. W. van Leeuwen, in Iridium Catalysts for Organic Reactions, ed. L. A. Oro and C. Claver, Springer, Cham, 2021, ch.10, pp. 397–454.
67. K. E. Ohriner, Platinum Met. Rev., 2008, 52(3), 186–197,  DOI:10.1595/147106708X333827.
68. G. Piancatelli, A. Scettri and S. Barbadoro, Tetrahedron Lett., 1976, 17, 3555,  DOI:10.1016/S0040-4039(00)71357-8.
69. H. Nakatsuji, T. Kamishima, T. Nonaka, Y. Koseki and H. Kasai, Chem. Lett., 2022, 51(5), 497–499,  DOI:10.1246/cl.220027.
70. F. Cacheux, G. Le Goff, J. Ouazzani, J. Bignon, P. Retailleau, A. Marinetti, A. Voituriez and J.-F. Betzer, Org. Chem. Front., 2021, 8(11), 2449–2455,  10.1039/D1QO00268F.
71. R. E. Mesmer, K. M. Palen and C. F. Baes Jr, Inorg. Chem., 1973, 12(1), 89–95,  DOI:10.1021/ja01183a101.
72. G. C. A. Luijkx, N. P. M. Huck, F. van Rantwijk, L. Maat and H. van Bekkum, Heterocycles, 2009, 77(2), 1037,  DOI:10.3987/COM-08-S(F)81.
73. M. M. Zhu, X. L. Du, Y. Zhao, B. B. Mei, Q. Zhang, F. F. Sun, Z. Jiang, Y.-M. Liu, H.-Y. He and Y. Cao, ACS Catal., 2019, 9(7), 6212–6222,  DOI:10.1021/acscatal.9b00489.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5gc01961c

‡ These authors contributed equally to this work.

---