Ru/MgO-catalysed selective aerobic oxidation of 5-hydroxymethylfurfural to 2,5-furandicarboxylic acid

Material and reagents

All chemicals employed in this work were of AR grade and used as received. Magnesium oxide light, MgO (96%), was purchased from Loba Chemicals, India. Ruthenium(III) chloride hydrate (45–55 wt-% metal) was obtained from Sigma–Aldrich. Deionised water was used to prepare solutions. 5-Hydroxymethylfurfural (HMF, ≥99%), 5-hydroxymethyl-2-furancarboxylic acid (HMFCA), 5-formyl-2-furoic acid (FFCA), 2,5-furandicarboxylic acid (FDCA, 99%) and 2,5-furandicarboxaldehyde (DFF), were purchased from Sigma–Aldrich and Alfa Aesar. 4-Methylpentan-2-one (methyl isobutyl ketone, GR) and ethyl acetate (GR) were purchased from Loba Chemie, India. All reagents were used as received without any further purification.

Catalyst synthesis

Heterogeneous Ru/MgO catalysts were synthesised by the wet impregnation of RuCl₃ over different MgO supports. Supports were prepared by heating commercial MgO at 60°C for 16 h, and subsequently 150°C for 6 h under vaccum. The resulting white powders were then calcined in a muffle furnace at 300, 400, 500 or 600°C for 2 or 4 h, and stored in a vacuum desiccator before use. Appropriate concentrations of aqueous solutions of RuCl₃ in deionised water (0.098 mmol in 2 mL for 0.5 wt-% Ru) were added dropwise to calcined MgO supports, and the resulting slurries stirred for 16 h at 35°C. Water was then removed using a rotary evaporator, and the obtained solid dried at 60°C for 16 h, and subsequently at 150°C under vaccum for 6 h. Ru-impregnated MgO powders were then reduced under flowing H₂ (10 mL min⁻¹) at 150°C for 2 h. The light grey catalysts were stored in a desiccator to prevent water and CO₂ adsorption. Sample designations reflect the Ru metal loading and calcination temperature and time, e.g. 0.5 wt-% Ru/MgO, 300°C, 2 h.

HMF oxidation

All the reactions were carried out in a 100-mL Teflon-lined Parr steel autoclave. A known quantity of HMF, deionised water and catalyst were added to the reactor. The reactor was subsequently flushed multiple times with oxygen, filled with oxygen to the desired pressure, and heated to the desired temperature with stirring at 300 rpm to promote oxygen dissolution. Once the desired temperature was achieved, stirring was increased to 800 rpm to initiate reaction while the oxygen inlet was regulated to maintain a constant pressure. The solution pH was measured before and after reaction. The analytical protocol was as previously decsribed.³⁶ Reactions were typically performed at 160°C at which the vapour pressure of water is 0.7 MPa, far less than the 1.5 MPa of O₂ pressure typically maintained in reactions. Considering the polar nature of the reactants and products, we anticipate that all organic molecules remained in the aqueous phase under such conditions. Note that one experiment was performed at 160°C and 0.5 MPa of O₂, under which partitioning of HMF, HMFCA, DFF, FFCA and FDCA between liquid and vapour water phases is anticipated (this could account in part for the low reactivity discussed later).

Catalyst characterisation

Bulk and surgface physicochemical properties of catalysts were investigated by X-ray diffraction (XRD), scanning electron microscopy (SEM), transmission electron microscopy (TEM), temperature-programmed desorption (TPD) and nitrogen porosimetry. The XRD patterns of MgO calcined at different temperatures (300, 400, 500 and 600°C) are shown in Fig. 1. The XRD patterns of MgO show main peaks corresponding to the MgO periclase phase (Joint Committee on Powder Diffraction Standards, JCPDS: 00-045-0946), and weak peaks corresponding to Mg(OH)₂ or the brucite phase (JCPDS: 84-2163) can also be seen. The formation of Mg(OH)₂ is due to the reaction of active MgO with water (moisture in the environment). The average crystallite size obtained from XRD is seen to increase with an increase in calcination temperature (Table 1). To understand the morphology and nature of the ruthenium loading, TEM was carried out. TEM images of the freshly prepared catalyst 0.5 wt-% Ru/MgO, 300°C, 2 h are provided in Supplementary Fig. S2. The presence of small black particles of ruthenium on the MgO with an average particle size of 20nm was confirmed using ImageJ software (ver. 1.54, W. S. Rasband, US National Institutes of Health, Bethesda, MD, USA, see https://imagej.nih.gov/ij/). N₂ physisorption was performed to examine the pore diameter and surface area of the catalysts (Table 1). The main goal of varying the calcination temperature was to study the redistribution of active catalytic sites and their effects on catalytic reactions. This is because redistribution can affect the accessibility of sites on the catalyst support. As seen, when MgO was calcined at 300°C, it showed a larger specific surface area of 72 m² g⁻¹ (determined by the Brunauer–Emmett–Teller method) than MgO calcined at higher temperatures. For structure-insensitive reactions, higher surface areas increase the concentration of active sites. Higher calcination temperatures can cause the fusion of support particles, thereby reducing the surface area of the support. With this, it was expected that Ru supported on these supports could form larger particles; pore volume and pore size data show collapse or restructuring of pores, diminishing the effective surface area available for catalysis with an increase in temperature. These results are in line with the XRD results, wherein with an increase in calcination temperature, average crystallite size increased.

Fig. 1.

X-Ray diffractograms of MgO support as a function of calcination temperature.

TPD analysis with CO₂ as a probe molecule indicated the total amount of basic sites present on the MgO support and catalyst (Table 2). The basicity of MgO primarily arises from cation vacancies (Lewis base sites) and hydroxyl groups (Brønsted base sites).^63–65 Extended calcination times and higher temperatures promote the formation of more stable (less defective) and dehydroxylated MgO surfaces, and hence typically confer lower basicity. Higher calcination temperatures also influence the morphology and structure of the MgO support, from N₂ sorption study. At elevated temperatures, transformations such as sintering may occur, affecting the distribution and strength of basic sites. Prolonged or high-temperature calcination can induce the formation of less basic phases on the MgO surface, such as periclase (MgO in a highly crystalline form), which may contribute to the observed decrease in basicity. Structural changes, such as particle growth or phase transitions, can decrease the number of active basic sites on the MgO surface. TPD measurements confirmed that higher calcination temperatures and longer calcination times decreased the total basic site density, consistent with an increase in MgO crystallite size and loss of active surface area, but had little effect on the strength (CO₂ desorption temperature) of basic sites.

Catalytic HMF oxidation

The effect of calcination temperature on 0.5 wt-% Ru/MgO catalysts was subsequently studied (Table 3). The catalytic results for reactions carried out at 140°C for 6.5 h suggest that elevating the calcination temperature translates into a reduction in yield of FDCA. When the results for entries 1 and 2 (samples calcined at 300°C for 2 or 4 h respectively) are compared, it is seen that the Ru/MgO, 300°C, 2 h catalyst formed more unidentified products. Note that based on the catalyst characterisation, with an increase in calcination temperature of MgO, decrease in overall basicity, surface area and pore diameter were observed, which play an important role in deciding the activity of the catalyst. Nevertheless, solutions of all catalysts exhibited a similar pH of ~9.5. The optimal calcination temperature for achieving maximum conversion of HMF (51%) and FDCA yield (23%) after 4 h was 300°C. As HMFCA and DFF were observed, it appears that both Pathways I and II operate (Scheme 2); the latter was not observed for Ru/C catalysts in our previous work.⁶⁶ From the results in Tables 2–3, FDCA production requires a support with good basicity and high surface area.

Reaction conditions: 4 mmol HMF, 1.5 MPa of O₂, 0.5 wt-% catalyst (substrate:metal molar ratio = 120), 30 mL of deionised water, 140°C, 6.5 h.

The most active and selective 0.5 wt-% Ru/MgO, 300°C, 4 h catalyst was subsequently chosen to optimise reaction conditions. The duration of the reaction was tested over a range of 4–10 h and a favourable yield of FDCA (68%) was achieved in 6.5 h (Fig. 2). It was observed that the rate-determining step in the reaction is the conversion of FFCA to FDCA. Therefore, prolonging the reaction time can lead to a higher yield of FDCA with a decrease in the FFCA yield. The time study also indicated that both mass balance and product distribution are dependent on reaction time. As the reaction time increased the total mass balance also fell associated with the formation of undetected products (likely HMF oligomers or humins). It is also interesting to note that though time was increased to 10 h, DFF and HMFCA still did not yield FFCA as their yields remained almost constant. This may be due to greater interaction between FFCA and the catalyst that did not allow DFF and HMFCA to interact with the catalyst surface. Alternatively, this may also suggest that the catalyst is no longer active for oxidising the alcohol group (in HMFCA) to an aldehyde group. However, at the same time, the aldehyde group in FFCA could be oxidised to a carboxylic acid group. From this trend, subsequent reactions were conducted for 6.5 h, identified as the optimum point because the highest mass balance of detected products was seen at this reaction time.

Fig. 2.

HMF conversion and product selectivity as a function of reaction time. Reaction conditions: 4 mmol HMF, 1.5 MPa of O₂, 0.5 wt-% Ru/MgO, 300°C, 4 h (substrate:metal molar ratio = 120), 30 mL of deionised water, 160°C.

Fig. 3 shows the results on the HMF conversion and product yields of a change in reaction temperature from 120 to 160°C. At 140°C, HMF conversion reached 51% with the predominant formation of FFCA (13%) and FDCA (23%), and formation of DFF (5%) with HMFCA (5%) was also seen. Reactions performed at higher temperatures of 160 and 180°C increased HMF conversion, but the maximum FDCA yield of 68% was obtained at 160°C with 96% HMF conversion. As the temperature increased to 180°C, the yield of FDCA decreased, mostly owing to HMF degradation being a temperature-sensitive reaction and thus this reaction taking precedence over the oxidation reaction, as evidenced by an increase in unidentified products. The same observation was also made when HMF was subjected to isolation from an organic solvent in the laboratory (Fig. S3–S6) and with an increase in rotary evaporator temperature, HMF degradation increased.

Fig. 3.

HMF conversion and product selectivity as a function of reaction temperature. Reaction conditions: 4 mmol HMF, 1.5 MPa of O₂, 0.5 wt-% Ru/MgO, 300°C, 4 h (substrate:metal molar ratio = 120), 30 mL of deionised water, 6.5 h.

The effect of oxygen pressure on FDCA production was also investigated (Fig. 4), with 1.5 MPa of O₂ the optimum. If the pressure is higher, FFCA gets converted to FDCA because higher pressure can enhance reaction rates, especially in gas–liquid reactions. It is well known that elevated pressures increase the concentration of oxygen molecules in the reaction system (water), promoting faster reaction rates. HMF conversion was 100% in the O₂ pressure range of 1.5–2.0 MPa. Increasing the O₂ pressure from 1.0 to 1.5 MPa, the FDCA yield increased from 45 to 68%, and unidentified products decreased from 17 to 10%; this then remained the same up to 2.0 MPa. This is because with lower availability of oxygen for the reaction (at lower pressures, owing to lower solubility), HMF instead of undergoing oxidation reactions forms degradation products, which is likely under elevated temperatures, as seen from the temperature study. Under pressure conditions between 1.0 and 2.0 MPa, the reaction may favour the conversion of FFCA to FDCA. When higher pressures were employed, complete conversion of intermediates such as DFF, HMFCA and FFCA to FDCA was still not possible. This may reflect reversible site-blocking due to adsorption of products or catalyst deactivation. Higher oxygen pressures can increase the oxidation of FFCA to FDCA up to certain limit because the thermodynamics of the oxidation reaction are influenced by the partial pressure of oxygen. Thus, understanding and manipulating thermodynamic aspects are crucial for achieving optimal FDCA yields. At elevated oyxgen, pressures are anticipated to increase oxygen solubility and hence circumvent mass-transport limitations that will otherwise lower the measured catalyst activity. Careful control of oxygen pressure is necessary to prevent undesired side reactions, mitigate the risk of explosive conditions and ensure the overall safety of the oxidation process.

Fig. 4.

HMF conversion and product selectivity as a function of oxygen pressure. Reaction conditions: 4 mmol HMF, 0.5 wt-% Ru/MgO, 300°C, 4 h (substrate:metal molar ratio = 120), 30 mL of deionised water, 160°C, 6.5 h.

From an industrial perspective, reducing the catalyst loading, and particularly precious metal usage (termed ‘thrifting’) is critical to demonstrate technoeconomic feasibility. The molar ratio of substrate:Ru was therefore varied between 240 and 80 (Fig. 5). Although increasing the mass of catalyst from 0.1 g (substrate:Ru = 240) to 0.3 g (substrate:Ru = 80) doubled the yield of FDCA to ~68%, further increases offered minimal benefit, with 0.2 g (substrate:Ru = 120) optimal and delivering 96% HMF conversion and 68% FDCA yield. High catalyst loadings also increased the proportion of undesired side products and hence compromised the carbon mass balance.

Fig. 5.

HMF conversion and product selectivity as a function of HMF:Ru molar ratio. Reaction condition: 4 mmol HMF, 1.5 MPa of O₂, 0.5 wt-% Ru/MgO, 300°C, 4 h, 30 mL of deionised water, 160°C, 6.5 h.

Catalyst stability is an important aspect of their potential for commercialisation, and hence reuse of the 0.5 wt-% Ru/MgO, 300°C, 4 h catalyst was examined (Fig. 6). Post-reaction catalysts were recovered from the reaction mixture by centrifugation, then washed three times with deionised water to remove any soluble organic compounds and returned to a freshly prepared reaction mixture without drying or adjusting the catalyst mass. HMF conversion remained >95% over four reaction cycles, evidencing that the catalyst retained significant activity post reaction without the need for an aggressive regeneration protocol (such as calcination to remove heavy organic residues). Deactivation of base sites would be expected to suppress HMF conversion, which is generally accepted to react through the gem-diol precursor to form HMFCA. However, these recycle measurements do not permit quantification of the degree of deactivation, for which such studies must be performed at low HMF conversion where reactions are free from potential mass-transfer limitations.⁶⁷ The FDCA yield progressively decreased with each reaction cycle (from 68 to 35% over four cycles), accompanied by an increase in unidentified degradation products, suggesting Ru site blocking by water-insoluble organic residues, probably HMF oligomers. Quantitative studies of catalyst lifetime and reactivation will be explored in future studies.

Fig. 6.

HMF conversion and product selectivity over 0.5 wt-% Ru/MgO, 300°C, 4 h as a function of re-use. Reaction condition: 4 mmol HMF, 1.5 MPa of O₂, 30 mL of deionised water, 160°C, 6.5 h.

Increasing the Ru loading from 0.1 to 5 wt-% offered only a small increase (10%) in HMF conversion (Supplementary Table S2); complete conversion was achieved for loadings ≥1 wt-% under our reaction conditions (160°C, substrate:metal molar ratio = 120, 1.5 MPa of O₂ and 6.5 h). However, product selectivity was far more sensitive to Ru loading, with the FDCA yield rising from only 44 to 88% across the Ru loading range (Supplementary Table S2), accompanied by a corresponding fall in the yields of all reactive intermediates (HMFCA, FFCA and DFF). Higher loadings also improved the carbon mass balance, which reached 95% for Ru loadings ≥1 wt-% wherein HMF oxidation was heavily favoured over, for example, competing oligomerisation pathways. The increase in FDCA selectivity with Ru loading is analogous to that observed for Au/MgAl-hydrotalcite catalysts,⁴³ wherein strong HMF adsorption over metal nanoparticles site-blocked oxidation of its reactively formed products and, hence, HMFCA oxidation to FDCA (by FFCA) was only efficient at >90% HMF conversion. Control reactions in the absence of catalyst or oxygen (Supplementary Table S3) evidenced modest HMF conversion (≤25%) and minimal FDCA formation (≤12%); the poorer mass balances indicate that slow thermal oxidation competes against significant oligomerisation. It is notable that no DFF was observed without catalyst and oxygen, suggesting that HMF oxidation is favoured through an HMFCA intermediate (Path B in Fig. 2). After reaction, the catalyst was filtered from the reaction mixture and the filtrate analysed by inductively coupled plasma optical emission spectroscopy (ICP-OES) to check for metal leaching. Ruthenium leaching was not observed, but low concentrations of Mg were detected, indicative of support instability and consistent with the observed increase in solution pH from hydroxide formation (Table 4).

Reaction pathways for HMF oxidation to FDCA

The selective oxidation of HMF to FDCA may occur through two pathways: initial oxidation of the hydroxyl group of HMF to form DFF, or of the aldehyde group to form HMFCA, and their subsequent oxidation to FFCA and then FDCA.^68–72 In almost all reactions, HMFCA was formed in larger amounts than DFF after 6 h. Reaction profiles also evidence that oxidation of these intermediates to FFCA was rapid, whereas the oxidation of FFCA to FDCA was slower and hence rate-determining. Considering various literature reports for supported Ru catalysts, it is likely that HMFCA is the reactive intermediate with DFF a spectator, as for 5 wt-% Ru over carbon and mixed metal oxide supports³⁶ (albeit in the presence of a homogeneous base such as sodium carbonate). Zheng et al. reached the same conclusion by studying the individual reaction steps in HMF oxidation to FDCA over Ru/C.³¹ Elementary steps for the oxidation of HMF to DFF over Ru/C in aqueous solution were proposed by Xie et al,³² HMF and O₂ being dissociatively chemisorbed over Ru to form an alcoholate (R–CH₂O*; R hereafter denotes the furfural CHO–C₄H₂O‐–part of HMF) species and atomic H* and O* (Scheme 3). Subsequent β‐H elimination of R–CH₂O* by reaction with O* is irreversible and forms DFF (R–CHO*) and hydroxyl (OH*) species, with the reactively formed R–CHO* desorbing as DFF. Note that OH* can also be formed by the reaction of O* and H* adatoms. Where the transformation of HMF to DFF requires elimination of two hydrogen atoms, oxidation of the formyl group of DFF to FFCA and FDCA requires the insertion of one and two oxygen atoms respectively, from either O₂ or H₂O. By contrast, kinetic studies of HMF and HMFCA oxidation over Au/MgAl-hydrotalcite catalysts evidence similar kinetics with HMFCA as the key reactive intermediate, and DFF never observed.⁴³ Davis and co-workers also propose an HMFCA intermediate to FDCA over Au/TiO₂ and Pt/C catalysts, with oxygen atoms originating from water,^73,74 in accordance with Nie et al.,³² who showed that HMF oxidation in toluene over Ru/C, Ru/Al₂O₃ and Ru/ZrO₂ produced mainly DFF (>96% selectivity).

Scheme 3.

Proposed mechanisms for HMF oxidation to FDCA through HMFCA or DFF intermediates.