In-Situ Investigations of Polyoxometalate-Catalysed Biomass Oxidation to Formic Acid by Using Multinuclear High Resolution Flow NMR Spectroscopy

Graphical Abstract

This study explores the aqueous oxidation of Glycolaldehyde to Formic Acid catalysed by H₅PV₂Mo₁₀O₄₀ in aqueous and aqueous-methanolic solution gaining deep mechanistical insights using Multinuclear high-resolution Flow NMR spectroscopy.

Abstract

Biomass valorisation over polyoxometalate-based (POM) catalysts is a promising strategy for green and sustainable chemistry. In order to further improve such processes e. g. by using different additives, a better understanding of the chemical influence of various additives on the catalyst is paramount. The main objective of this study is to gain a deeper understanding of the effects caused by various additives on the catalytically active vanadium species in the selective aerobic oxidation of carbohydrates to formic acid (esters). After carefully choosing a suitable model system, the oxidation of glycolaldehyde using the heteropolyacid H₅PV₂Mo₁₀O₄₀ (HPA-2) as a catalyst was studied in situ by Flow Nuclear magnetic resonance (FlowNMR) spectroscopy. These measurements allowed to identify the catalytically active isomer of the HPA-2-catalyst and to observe the influence of additives on the reaction kinetics in situ. These findings provide a good basis for further improving POM-based catalytic aerobic oxidation of biomass.

Introduction

Due to the geopolitical and environmental issues attached to fossil resources (i. e. coal, oil, and natural gas), the demand for alternatives is growing. The most promising alternative for fossil resources is biomass, which is a renewable, carbon-neutral resource for the production of energy, materials, and chemicals.^{1, 2} Biomass includes a wide variety of materials such as agricultural crops, wood, algae, but also municipal waste of biological origin and from the food processing industry.³ A common system to categorize biomass is the division into different generations: the first generation consist of edible crops and its industrial use directly competes with food supply. The second generation contains agricultural waste and other lignocellulosic biomass which cannot be used as food. Algae make up the third generation of biomass, and the fourth generation of biomass is made up of modified organisms purposely created with valorisation in mind.^{4, 5} The valorisation of biomass can be performed by biochemical processes (e. g. fermentation),⁶ thermal processes (e. g. gasification or pyrolysis),⁷ catalytic processes (hydrolysis, oxidation, hydrogenation), or combinations of these.⁸ The products of these processes can be fuels or platform chemicals, which are used as feedstock for the production of various other higher value-added products.^9-11

The oxidation of carbohydrates or molecules derived therefrom is of growing interest in particular, as is evident by the rising number of publications on this topic.^12-21 One reason is the availability of the raw materials: lignocellulosic biomass is the most abundant type of biogenic material and typically contains more than 50 wt.% of carbohydrates.²² Typical reactions include hydrolysis of cellulose, hemicellulose, or other sugar-based polymers and subsequent oxidation or dehydration (e. g. to 5-hydroxymethylfurfural) followed by oxidation.^{10, 17, 18} The products of these processes are furan derivatives (e. g. 2,5-furandicarboxylic acid)¹⁷ or small carboxylic acids such as malonic acid, glycolic acid, acetic acid, and formic acid which can be used as intermediates.¹⁸

Promising catalysts for such selective biomass transformations are molecular metal-oxide clusters, so-called polyoxometalates (POMs).^23-25 POMs consist of metal-oxygen polyhedra, which are connected via their edges or corners by shared oxo ligands. Typical metals for such frameworks are molybdenum and tungsten, whereas some POM structures, such as the Keggin and Wells-Dawson structure types, also contain a central heteroatom such as phosphorus or silicon. This subclass of POMs is also known as heteropolyanions, or in their protonated form as heteropolyacids (HPAs).^26-28 The high Brönsted acidity of these compounds makes them excellent catalysts for the hydrolysis of carbohydrates. In order to also apply them as oxidation catalysts, the HPAs have to be modified: redox activity in POMs can be introduced by substituting the framework element with a redox-active transition element such as vanadium.^29-31 With this partial substitution of the framework metal different positional isomers become possible. In aqueous solution, partial dissociation of the HPA leads to isomerisation as well as to the formation of HPAs with higher or lower degree of substitution.^32-34 As a result, aqueous solutions of substituted HPAs typically contain a mixture of different species depending on the solution pH, which makes it challenging to identify structural motifs involved in catalytic activity.

Vanadium-substituted phosphomolybdates have been successfully applied for the catalysis of various oxidation reactions.^{23, 35, 36} It is well established that during the catalytic cycle V^V is reduced to V^IV while the organic substrate is being oxidised. For the reoxidation of the catalyst, either hydrogen peroxide or molecular oxygen can be used.^37-39 Specifically H₅PV₂Mo₁₀O₄₀ (HPA-2) has been successfully applied as an oxidation catalyst for multiple substrates.^40-43 The most prominent oxidation reaction catalysed by HPA-2 is the selective production of formic acid from cellulose, other carbohydrates, or even directly from complex biological materials such as algae or wood under comparatively mild conditions (90-120 °C, 20–30 bar O₂).^{36, 44-48} Although its higher-substituted derivative H₈PV₅Mo₇O₄₀ (HPA-5) has been shown to be a more efficient catalyst, HPA-2 remains in the focus of researchers as it follows the same reaction pathways but is easier to investigate.^{49, 50} The main reason for this is the lower number of positional isomers in the two-fold substituted POM (5 positional isomers) compared to the five-fold substituted POM (38 positional isomers).⁵¹

The chemical pathways of biomass oxidation have already been extensively studied. It is generally accepted that after initial hydrolysis, the substrate is broken down by oxidative C−C cleavage until formic acid and CO₂ are formed as the final products. It is important to note, however, that CO₂ is not formed by oxidation of formic acid, as it was shown that formic acid does not react further under the typical reaction conditions (90-120 °C, 20–30 bar O₂).^{42, 44} Typical intermediates on the pathway from biomass (carbohydrates) to formic acid/CO₂ include glyoxal and glycolaldehyde, which are then oxidatively cleaved into CO₂ and formic acid, presumably via the formation of glycolic acid or glyoxylic acid.⁵² It has been reported that the ratio between CO₂ and formic acid can be shifted towards the latter by addition of methanol. While some reports claim that methanol is unreactive in this system at temperatures of 90–120 °C, others have observed synergistic effects by simultaneous oxidation of methanol to formaldehyde at temperatures up to 190 °C.^53-55 A similar observation has been made by addition of isopropanol, which has been reported to act as a radical scavenger, thereby preventing a radical reaction pathway leading to the formation of CO₂.⁵⁶ In a different context, carboxylic acids have been reported to act as inhibitors for oxidation reactions, with the exception of oxalic acid which accelerates the reaction.⁵⁰ However, very little detail has been reported on the direct observation of the vanadium-substituted POM catalysts at work. DFT studies suggest a Mars-van Krevelen mechanism, but experimental studies on this topic mainly rely on observation of reaction rates without in situ catalyst analysis.^{21, 43} The main reason for this are the technical challenges associated with studying an aerobic oxidation catalyst during turnover.⁴¹ A potent technique for such observations is FlowNMR spectroscopy.⁵⁷ Real-time high resolution FlowNMR spectroscopy under reaction conditions is a powerful approach to gain insight into the speciation of a working catalyst,^{47, 58-62} especially when multinuclear in situ spectra can be correlated with the reaction progress.^{57, 63-66} Vanadium-substituted phosphomolybdates can be observed by NMR spectroscopy, either via their central ³¹P atom or the ⁵¹V atoms in diamagnetic oxidation states. The reduced catalyst containing V^IV is paramagnetic and as a result difficult to detect by NMR spectroscopy. ¹H NMR spectroscopy can help to monitor the reaction progress by observing changes in substrate and product peak intensity if the setup is suitably designed to withstand and maintain the reaction conditions.

In this study we tackled the challenges of applying in situ FlowNMR for the direct observation of HPA-2 during the aerobic oxidation of C2-model intermediates from carbohydrate oxidation to formic acid to reveal important details on how the catalyst works, which positional isomers are catalytically active, and how additives influence the catalytic performance.

Results and Discussion

Determination of a Suitable Reaction System for In-Situ Investigations

To facilitate the spectroscopic analysis, we chose to investigate the two-fold substituted HPA-2 as the catalyst due to its known solution speciation and well-known redox activity. HPA-2 is widely used to catalyse aerobic oxidation reactions like heteroaromatic sulfides,⁴⁰ aldehydes,⁶⁷ alkenes,⁶⁸ alcohols³⁷ and many more. Moreover, HPA-2 has also proven to be a suitable catalyst in combination with molecular oxygen for the extraction-coupled oxidative desulfurization of fuels.^{50, 69} Although slightly less active than HPA-5 in the aerobic oxidation of carbohydrates, HPA-2 has been shown to be an efficient catalytic system for carbohydrate-based biomass oxidation to formic acid.^{15, 24, 36, 70, 71}

In aqueous solution, five positional isomers of HPA-2 give rise to five partially overlapping ⁵¹V NMR signals in addition to some higher substituted species formed by dissociation equilibria as well as HPA-1 (H₄PVMo₁₁O₄₀) (Figure 1 and Table 1).^{23, 32}

Figure 1

⁵¹V NMR spectra of HPA-2 in H₂O with assignment of its positional isomers.³² The corresponding ³¹P NMR spectra are shown in SI (Figure S5).

Table 1. positional isomers of HPA-2 according to Petterson et al.;³² HPA-1=−533.5 ppm.

Polyhedral structure
Designation (enantiomer)	α1,4	α1,2	α1,5; (α-1,8)	α1,6; (α-1,7)	α1,11
Statistical abundance	18.2 %	18.2 %	18.2 %	36.4 %	9.1 %
^51V NMR chemical shift [ppm]	-526.0	-535.0	-535.8	-536.0	-538.1

Due to their relevance in various biomass oxidation processes we chose to investigate the aerobic oxidation of a range of C₂ intermediates with different oxygen-containing functionalities (acid, aldehyde, alcohol) found in real biomass and their oxidative cleavage reactions. Preliminary experiments under batch conditions (setup see Figure S1) were used to identify reaction conditions suitable for the in situ FlowNMR investigations. Using mild conditions of 70 °C and 5 bar O₂ ethylene glycol was barely oxidized to CO₂ over 4 h (X=3.0 %), but glycolic acid (X=13.2 %), glycol aldehyde (X=33.0 %), glyoxal (X=100 %) and glyoxylic acid (X=100 %) gave reasonable conversion. The results are shown in detail in the supporting information (Table S2) and are in agreement with previous studies from different groups.^{23, 56, 70} For further investigations, we chose glycol aldehyde (GA) as a substrate as it showed reasonable conversion under the conditions applied with room for investigating accelerating and decelerating effects. Control reactions without catalyst showed no activity. Catalytic conversion (X) and yields (Y) with HPA-2 were largely concentration-independent to range between X=52–59 % and Y_FA=47–52 % (Table S3). Reaction conditions were set to 80 °C with a pressure of 5 bar oxygen, 100 mmol/L substrate and 10 mmol/L catalyst, to combine optimal observability in the FlowNMR setup with the conditions close to best reaction conditions reported previously.

Selection of Suitable Additives

Influence of Selectivity Enhancing Additives on the Isomer Distribution of HPA-2

The addition of 2 vol.% methanol (MeOH) almost entirely suppressed the formation of CO₂ as undesired side product compared to 40 % CO₂ yield in pure water, but increasing the MeOH content to 10 vol.% significantly decreased substrate conversion. The same was true for the addition of ethanol (EtOH), but iso-propanol (iPrOH) lead to a significantly lower conversion even at 2 vol.%. However, the differences between the addition of 2 vol.% and 10 vol.% of iso-propanol are negligible, contradicting the results recently published by He et al..⁵⁶ Full results are shown Table S4.

To investigate the interaction of the catalyst with the solvent additives ⁵¹V NMR spectra of these mixtures were measured at different pH values (Figure 2). While the pH value has a significant influence on the chemical shifts of the different positional isomers, the signal for HPA-1 can be used as a reference since it doesn't change with varying pH values according to Petterson et al.³² On the other hand, the pH-value has a huge influence on the line shape and chemical shift of the other signals. The α1,4 positional isomer (−-532 ppm at pH 2.0) in particular shifts significantly with changes in the solution pH. Furthermore, the change in pH induced by the formation of formic acid over the observed reaction time is insignificant as shown by control experiments (Table S7). Therefore, any shift observed can be attributed to chemical interactions/changes of the catalyst.

Figure 2

⁵¹V NMR spectra of HPA-2 dissolved in different solvent mixtures at different pH values. c(HPA-2)=10 mmol/L, pH=1, 1.5 and 2 in H₂O/D₂O, T=25 °C, addition of 10 vol.% methanol or ethanol.

Moreover, with the introduction of solvent additives the resulting spectra differ significantly from the aqueous system regarding line shape and position. For MeOH, the α1,4 positional isomer is shifted towards high field and has another broader signal underlying. Also, the signal of the positional isomers α1,2, α1,5 and α1,6 have an increased width at half height showing a change in electronic environment of the corresponding vanadium atoms, which is observable within the ⁵¹V NMR spectra. The broader signal might also be attributed to the α1,2 positional isomer, showing the chemical similarity to the α1,5 and α1,6 positional isomers due to the addition of MeOH. For the addition of EtOH, the α1,4 positional isomer signal is shifted even more high-field compared to MeOH. As the signal is overlapping with the HPA-1 reference signal, the observation of EtOH solvent effects on the catalyst and referencing might not be possible. Therefore, it is suggested to investigate the effect of MeOH in situ as the most promising additive for the study.

Influence of Rate-Enhancing Additives on the Isomer Distribution of HPA-2

In a previous study by Poller et al. it was shown that oxalic acid, acetic acid and formic acid have activating or deactivating effects on polyoxometalate-catalysed three-phase liquid-liquid-gas reactions.⁵⁰ It was shown that the addition of oxalic acid promoted the formation of the active V^IV species increasing the rate of the reaction. Acetic acid and formic acid, however, formed stable complexes with the vanadium centres of the HPA-2 catalyst inhibiting its catalytic activity in oxidative desulfurization.

This knowledge is now transferred to the oxidation of biomass-derived substrates. In order to investigate the effect of the mentioned acids on the catalyst and the biomass oxidation reaction, oxalic acid and acetic acid were tested as additives. The use of formic acid was omitted, as it is the main product of the biomass oxidation reaction. Here, acetic and oxalic acid were tested as equimolar catalyst additives under optimized conditions, each alone and in combination with MeOH as selectivity enhancing additive (the full results are shown in Table S5).

Acetic acid yielded almost the same amounts of FA and CO₂ as the reaction without additive.^{50, 72} In combination with 10 vol.% MeOH even the best benchmark result was slightly improved. Oxalic acid as reactive additive improved the conversion of GA to 100 % but increased the CO₂ yield as well (partly due to its own decomposition to CO₂ under the applied reaction conditions). MeOH, acetic acid, and the combination of both seemed to have no impact on the conversion of GA. Using oxalic acid on the other hand greatly increased the reaction rate but lowers selectivity to formic acid.

For further investigations, the resulting ⁵¹V NMR spectra after the reaction were compared to HPA-2 dissolved in H₂O (pH 2.0) as shown in Figure 3. The catalyst before and after the reaction in aqueous solution showed the same peaks, showing successful re-oxidation. This was also true for the reactions using oxalic acid and acetic acid additives. Here, no further complexation and interaction was visible after reaction. However, looking at the reactions with MeOH and other additives there were visible influences. When no reactive additive was used, the α1,4 positional isomer shifted high-field compared to the HPA-1 reference signal (6 ppm shift). The other positional isomers were mildly affected by the addition of MeOH and only shifted marginally (1 ppm) high-field. Using acetic acid as reactive additive lead to a further high-field shift of the α1,4 positional isomer (1 ppm) compared the HPA-1 signal, but the difference in signal shape was significant. Here, a sawtooth-shape to the left side is visible. Using oxalic acid, a sawtooth-shape to the right is visible. Also, no significant broadening of the signals is visible, indicating that the influence of oxalic acid on the reaction is negligible.

Figure 3

Comparison of ⁵¹V NMR spectra before (bottom, pH 2.0) and after HPA-2 catalyzed oxidation reactions of GA. (GA (100 mmol/L), HPA-2 (10 mmol/L), *acetic acid (10 mmol/L), **oxalic acid (10 mmol/L), ⁺10 vol.% MeOH as solvent additive, 14 h.)

In Situ Investigations Using Multinuclear High Resolution FlowNMR Spectroscopy

In order to follow the reaction progress as well as any observable changes in the POM catalyst we investigated the reaction with interleaved ¹H, ³¹P{¹H} and ⁵¹V FlowNMR experiments from a pressure reactor under 7 bar O₂ circulated through an InsightMR flow tube lined with 1/16” PEEK tubing (780 μm ID) at 4 mL/min using a piston pump (setup see Figure S2). Each measurement cycle took about 7 min, and the reactor as well as all sample lines were heated to 80 °C.

In-Situ Investigations of Glycolaldehyde Oxidation in Pure Water

First, we investigated the oxidation of GA (100 mmol/L) using HPA-2 as a catalyst (10 mmol/L) in aqueous solution (30 mL DI-water). The reactor was pressurized with oxygen after the first complete measurement cycle. The corresponding ¹H FlowNMR spectra (Figure S8) quantified against tert-butanol as internal standard (shown to be inert, see Table S6) saw a steady decrease in substrate concentration with increasing amounts of formic acid formed as the main product. As expected, the substrate GA is consumed linearly over the whole reaction time as published in previous batch experiments^{51, 70} and formic acid yield increases linearly over time with a maximum yield of around 60 % after 15 h reaction time.

With respect to the HPA-2 catalyst, some minor changes where observed in the ⁵¹V FlowNMR spectra during the first 100 minutes (Figure 4) with no significant changes detected afterwards. The corresponding ³¹P{¹H} FlowNMR spectra exhibit only minor changes throughout the reaction (see Figure S9). From the ⁵¹V FlowNMR spectra it appears that the catalyst reaches a steady state after the first hour of the reaction (~10 measurement cycles). As before, using the ⁵¹V NMR signal of HPA-1 as a reference, changes of the signals of the isomers α1,2 and α1,4 may be analysed quantitatively.

Figure 4

⁵¹V FlowNMR spectra of the HPA-2 catalyst during the aqueous oxidation of GA under 7 bar O₂ at 80 °C (time interval=7 min). Signal C shows HPA-1, C1 is the α1,4 isomer, C2 shows other α isomers, C3 is the α1,11 isomer.

As previously shown by Weinstock et al.,⁷³ VO₂⁺ (pervanadyl) ions were released from HPA-2 by decreasing the pH-value adding sulfuric acid to their reaction solution. The authors referred that by ⁵¹V NMR spectroscopy due to an increasing and broader signal at −540 ppm. Kim et al.⁷⁴ describe a line width decrease with increasing temperature for pervanadyl ions due to the formation of VO²⁺ (vanadyl) ions and their paramagnetic effect on the NMR spectra. As we didn't observe any significant line broadening or decrease of V^V POM species during the reaction we can rule out the in-situ formation of pervanadyl and vanadyl ions under the conditions applied.

Interestingly, after addition of 7 bar oxygen (Figure 5) to the reaction vessel the observed signals shifted. The signal of the α1,4 isomer showed a low-field shift (0.25 ppm) while the other signals, especially the α1,2 isomer, shifted to lower field by only 0.05 ppm. As the formation of formic acid is not significant at that point in time (ref. Figure S8), the shift can't be explained by a change of pH (1.49 at Y(FA)=0 % compared to 1.44 at Y(FA)=100 %). Therefore, it must be caused by the interaction of the active adjacent vanadium centres inside the HPA-2 structure. This leads us to the assumption, that the α1,4 isomer plays the most significant role in catalysing this reaction (as it has the highest low-field shift) with prolonging reaction. In agreement with the published catalytic cycle by Albert et al., molecular oxygen is needed to re-oxidize the catalyst to its original state.²⁴ This confirms the hypothesis that two adjacent V-oxygen octahedra are required for the catalytic oxidation with HPA-2.⁵⁰ As the α1,4 and the α1,2 isomer provide this specific structure and shift the most in the oxidation of GA to formic acid it leads to the assumption, that the hypotheses drawn beforehand is also true for the biomass oxidation reaction catalysed by vanadium-containing POM-catalysts. This is also well in line with the conclusions drawn by Wu et al.⁵⁷ that HPA-2 is following a Mars-van-Krevelen-type mechanism.

Figure 5

⁵¹V FlowNMR spectra of HPA-2. Spectrum 1 shows the first measurement (ambient pressure) to which 7 bar of O₂ were added from the second spectrum onwards. Time scale is as shown in Figure 4.

When comparing the shifts for the overall reaction progress, shifts of α1,4 are 2 ppm, whereas the other isomers only change by 0.5 ppm. In the spectrum recorded after 11 h of reaction time another small shift was observed for the signal of α1,4 (shown by red arrow in Figure 5). Corresponding to this is also a minor high field shift of the α1,2 signal into the α1,5/α1,6 signal. This further shift can be explained with an increasing content of peroxo species, which are formed during the re-oxidation step of the catalyst.⁵⁰ The formation of peroxo species in non-tetrahedral coordination leads to a high field shift in ⁵¹V-NMR as shown by Rehder.⁷⁵ A tetrahedral peroxo species would even yield a decrease in shielding and a low field shift.⁷⁶ As the peroxo species, which is formed during the re-oxidation of the catalyst, are discussed to be the reason for an increased overoxidation of the substrate to CO₂,⁵⁶ could not be observed here, we speculate that paramagnetic V^IV species formed during substrate oxidation also play a role here.

Aside from the changes in chemical shift, changes in signal intensity were observed. Due to its very fast relaxation (longest T1 time is 31.5 ms for HPA-1, ESI Figures S6 and S7, Table S1), all measured ⁵¹V-spectra were quantitative, and the signal areas can be compared within the measurement cycle.

Figure 6 shows the signal intensity for each positional isomer relative to HPA-1. At the start (reaction time zero) the smallest integral values for all isomers can be found. Within a short reaction time, and after pressurising the reaction vessel with oxygen, a rapid increase in the integral value until 100 min reaction time could be observed. Here, a steady state is reached. This can be explained by the nature of the V^V/V^IV redox cycle, as the V^IV is paramagnetic and therefore NMR silent.⁷⁷ At the start of the measurement cycle where no oxygen for re-oxidation is present, the catalyst is mostly reduced due to the reaction progress while heating the reactor and the NMR probe. After the addition of oxygen, the catalyst is increasingly re-oxidised and after 100 min reaction time an equilibrium is reached.

Figure 6

Signal-intensity plot of different positional isomers referring to the benchmark HPA-1 signal which is set constant.

As a general conclusion of the first experiment in the Flow NMR setup, the data confirm the suitability of the reaction conditions that were chosen in the initial investigation, and shows that analysis of the HPA-2 catalyst by in-situ NMR is possible for the homogenous catalysed oxidation of GA to formic acid. In general, the performed experiment confirms the previous ex-situ analysis results. Also, it was shown, that two adjacent vanadium-containing octahedra in a single catalyst are needed for the ability to oxidize the substrate confirming the previous hypothesis.

In-Situ Investigations of Glycolaldehyde Oxidation Using the Selectivity Enhancing Additive Methanol

Subsequently, the effect of MeOH addition on the catalyst and the product selectivity were investigated by in situ FlowNMR. In prior publications, a vanadate-methanolate complex [VO(OMe)₃]_n was suggested to be the active species in water-methanolic solution under oxidative conditions, identified by a low field ⁵¹V NMR resonance at −486 ppm.⁵³ So far, spectroscopic investigations have only been carried out for high methanol contents >30 vol.%. Therefore, we wanted to shed further light on the effect of MeOH addition in small amounts between 2–10 vol.%, which have also shown a drastic selectivity shift towards formic acid in glucose oxidation.⁵⁶

The FlowNMR experiment was carried out using the same parameters and procedures as in pure aqueous solution (Figure 4) but this time with 10 vol % MeOH added. Based on the ¹H NMR spectra, we found, that mainly formic acid and methyl formate are formed (Figure S11). The CO₂-suppression is not directly observable, as measurements of gas-phase could not be performed in this setup. However, based on our preliminary experiments (Table S4) it can be inferred that CO₂ formation was efficiently supressed by adding MeOH as previously reported.⁵³ Similar to the purely aqueous GA oxidation shown above, ¹H-NMR signals for the products (formic acid 8.25 ppm, methyl formate 8.15 ppm) increase linearly, whereas the signals attributed to GA (9.55 ppm) decreases linearly (Figure S11).

To investigate the effect of MeOH addition on the HPA-2 catalyst, ⁵¹V FlowNMR spectra were also measured and the results are shown in Figure 7.

Figure 7

⁵¹V FlowNMR spectra of HPA-2 during methanolic oxidation of GA to formic acid and methyl formate under 7 bar O₂ at 80 °C (time interval=7 min).

Compared with the results from the same experiments without MeOH (Figure 4), the α1,4 isomer of HPA-2 was shifted more high-field (6 ppm) relative to the HPA-1 signal. This could be caused by an increased shielding due to the increased electronegativity introduced by the exchange of an oxygen atom with an methoxy-ligand.⁷⁵ The pH-dependency can be neglected, as Petterson et al. showed a high field shift of 12 ppm by changing the pH by one. Therefore, the shift of the α1,4 isomer is most probably related to the addition of MeOH and its previously suggested complexation to the two adjacent V-atoms of HPA-2 (compare Figure 2).

In the first spectrum shown in Figure 7 (bottom) where no oxygen was present, the catalyst would have been partially reduced to NMR-silent V^IV species as in the aqueous system (Figure 4), and after the addition of 7 bar O₂ the intensity of the diamagnetic V^V species increased over time to reach a steady-state after 28 mins (third spectrum). The reactive α1,4 isomer is shifted low field by 0.25 ppm by the addition of oxygen pressure and forms an isolated peak at around −534 ppm. After 35 min (=5 spectra) the signal is shifted slightly more high-field compared to the starting point. Also, the signal doesn't shift any further after that, the catalyst has reached a steady state. This is also true for the α1,2 isomer, however its peaks are overlapping with the other isomers α1,5 and α1,6.

An enlargement of the spectra showing the signals of HPA-1 and the α1,4 isomer is shown in Figure 8 together with a chemical shift-evolution graph, which is following the signal shift with ongoing reaction time. The small shifts after 10 h can be correlated to the slightly acidified reaction solution by the formation of formic acid as the main product. The addition of MeOH stabilizes the isomerization behaviour of the HPA-2 catalyst and even buffers pH-effects.

Figure 8

Zoom-in in the ⁵¹V-NMR spectra with signals of HPA-1 (left) and methanolic α1,4 isomer with visible high field shift. Right picture shows chemical shift-evolution, visibly catalyst doesn't shift after 60 min.

As there is no significant high field shift visible with prolonging reaction progress in the methanolic system compared to the pure aqueous system, we suppose that the formation of peroxo species does not take place in the methanolic system. Also, MeOH doesn't block the substrate by forming hydrogen bonds as it supresses only the formation of CO₂, as it would be the case with using DMSO as additive.⁷⁸ As MeOH has little radical scavenging abilities it must be a different process than the one described by He et al. for isopropanol.⁵⁶

Therefore, we suggest that methanol is incorporated into the structure as a bridging alkoxide ligand connecting two adjacent vanadium octahedra. This was already shown for polyoxovanadium clusters by Hartl et al..⁷⁹ This methoxy ligand changes the electronic structure of the vanadium complex and followingly the energetic positions of the transition states.

As a general conclusion of the second experiment, the positive influence of the MeOH added to the catalyst was shown. It was observable, that the catalyst was modified and therefore stable for the overall experiment. No significant changes were induced by the raising formic acid yield. Also, MeOH is influencing only the two adjacent vanadium-containing octahedra (α1,4 and α1,2 isomers) in a single catalyst, which are also the ones needed for the catalytic activity as shown before.

In-Situ Investigations of Glycolaldehyde Oxidation Using the Rate-Enhancing Additive Oxalic Acid

Finally, oxalic acid was tested with regard to the acceleration of the reaction and its influence on the HPA-2 catalyst. Based on observations by Bertleff et al.⁸⁰ for the oxidative desulfurization of fuels catalysed by vanadium substituted phosphomolybdates and on computational studies, Poller et al. suggested, that the initial oxidation (Figure 9, step 1) of the substrate by the catalyst is the rate-determining step in the catalytic cycle.⁵⁰

Figure 9

Schematic catalytic cycle of substrate oxidation catalysed by vanadium-substituted phosphomolybdates according to Poller et al.⁵⁰

By adding oxalic acid as a substrate this step can be accelerated and the oxidation of the actual substrate can proceed faster in step 3. Although this phenomenon has been conclusively confirmed for the oxidative desulfurization of fuels, it has not yet been tested for oxidation of water-soluble bio-based substrates. We therefore tested the effect of oxalic acid as a rate enhancing additive in the oxidation of GA, observing the formation of formic acid by ¹H NMR and the catalyst by ⁵¹V NMR spectroscopy. As oxalic acid is easily decomposed to CO₂ by the catalyst in its oxidized state, the catalyst was added to the reaction solution after the temperature and the FlowNMR measurements had been set. This was accompanied by a colour change to dark brown/blue, indicating immediate reduction of the catalyst (Figure 9, Step 1). Then, after the first measurement cycle the reaction vessel was pressurized with 7 bar oxygen to initiate catalyst re-oxidation.

In order to assess the influence of the additive on the reaction rate, we have compared the relative changes of the product signals over time in each reaction (8.25 ppm for FA, 9.65 ppm for GA). Thereby, we determined the following relative reaction rates: addition of oxalic acid>purely aqueous reaction>addition of methanol (Figure 10).

Figure 10

Increase of product signals over time (normalized to internal standard ^tBuOH) for the oxidation of GA in aqueous solution, with addition of MeOH, and with addition of oxalic acid.

In the ⁵¹V NMR spectra of this reaction, all of the isomers observed in the purely aqueous reaction can be found as well (ref. Figure 4). In addition, an additional broad signal left to the α1,4 isomer in the initial spectra (Figure 11) can be observed. We attribute this signal to the peroxo-species of the POM (between Step 2 and Step 3 in Figure 9). Which further confirms the conclusion, that the α1,4 isomer is dominating the catalytic activity. This species could not be detected by Poller et al. because its lifetime is too short for ex-situ analysis. It vanishes within the time of one measurement cycle, by completing the catalytic cycle according to Figure 9 (Step 3). Since oxalic acid is a sacrificial additive, the peroxo species cannot be formed in larger quantities again, once the oxalic acid is used up.

Figure 11

⁵¹V FlowNMR spectra of the GA oxidation with HPA-2 as catalyst and oxalic acid as additive.

During the course of the reaction, the α1,4 isomer shifts significantly to low field, while the other isomers shift low field to a lesser degree (Figure 11 and Figure 12).

Figure 12

Signal intensity for oxalic acid promoted HPA-2 catalyzed oxidation of GA to formic acid (left); chemical shift evolution graph (right).

The chemical shift evolution (Figure 12, right) indicates, that the low field shift (1.2 ppm) of the α1,4 isomer is finished after 45 min reaction time, the corresponding shift of the α1,2 isomer is smaller (0.6 ppm) and finished already after 30 min. This shows again that the α1,4 isomer is the most involved species in this catalytic mechanism. A look at the signal intensity reveals, that the signals of all HPA-2 isomers decrease relative to the more stable HPA-1. This indicates initial reduction of the catalyst (Step 1 in Figure 9) until a steady-state is reached after approximately 50 min. This is contrary to the purely aqueous system and the experiment with addition of MeOH. The reason for this is the reduction of the peroxo-complex (Step 3 and Step 1 in Figure 9), which was accumulated after the rapid initial reaction with oxalic acid.

Correlating the interpretation of the ¹H and ⁵¹V spectra, we can relate the higher activity at the start of the reaction due to the pre-reduced catalyst, which leads to a faster substrate oxidation in step 3 of the catalytic cycle. Our in-situ investigation thereby further confirms the hypothesis of Poller et al.⁵⁰ and proves that the same mechanism is applicable for the aqueous oxidation of bio-based substrates catalysed by vanadium substituted phosphomolybdates.

Conclusions

In summary, we have successfully employed ¹H and ⁵¹V FlowNMR spectroscopy to observe the aqueous oxidation of GA catalysed by HPA-2 and gained valuable mechanistic insights. Specifically, we have identified the catalytically active isomer (designated α1,4), which features two adjacent V atoms. By using methanol as a selectivity enhancing additive we were able to show the influence of methanol enacted on the catalyst, even with at lower methanol content than used in previous studies. Based on the spectra we hypothesize that methanol directly interacts with the catalyst as opposed to forming a separate vanadium methanolate complex. The addition of oxalic acid as an additive showed the applicability of this rate enhancement for aqueous oxidation of bio-based substrates, and confirmed an earlier hypothesis regarding the catalytic mechanism by observing the postulated peroxo-complex for the first time. Overall, this study greatly enhances the understanding of the catalytic mechanism of vanadium-substituted phosphomolybdates and provides valuable insights for future catalyst development and catalytic applications.

Experimental

Materials and Catalyst Synthesis

All chemicals were obtained commercially and used as received without further purification. The model substrate glycol aldehyde was supplied in its dimer form by Sigma Aldrich; glycolic acid (99 %) was supplied by Acros organics; Oxalic acid dihydrate and glyoxal (40 % solution in water) were supplied by Merck KGaA; Glyoxylic acid (50 % w/w) was supplied by Alfa Aesar; Acetic acid (glacial) was supplied by VWR chemicals and tert-butanol (99 %) as internal standard for NMR measurement was supplied by Grüssing GmbH. The vanadium-substituted catalyst HPA-2 (H₅PV₂Mo₁₀O₄₀) was synthesized according to the literature.^{24, 29, 81} The characterization of the catalyst has been carried out using a Fa. Spectro Arcos ICP-OES device resulting in a Mo/P/V ratio of 10/1.2/2. A corresponding IR-spectrum can be found in Figure S3. Oxygen (5.0) was bought from Linde AG. Demineralized water (DI), methanol (99.8 %, VWR BDH Chemicals), ethanol (absolute, VWR BDH Chemicals) and iso-propanol (99.95 %, Roth) were used as additives.

Experimental Procedure for Selection of a Suitable Reaction System

All experiments for the first section were performed in a 10-fold high-pressure screening plant (see Figure S1) in a batch mode setup. Here, up to ten 20 mL autoclaves made of Hastelloy C276 can be used. All pipes, valves and fittings were made of stainless steel 1.4571. The gaskets used were made of Teflon and a high-performance grease for oxygen fittings by Fuchs Lubritech was used. The autoclaves were connected in parallel to a single oxygen supply line via individual couplings and placed inside a heating plate in order to adjust the required temperature. The heating plate was equipped with magnetic stirrer bars. Additionally, each reactor was connected to a rupture disk with a burst pressure maximum of 90 bar.

Usually, a liquid volume of 10 mL of solvent with various compositions (as noted) was used for the catalytic oxidation reactions. Each reactor was filled with different amounts (as noted) of substrate and catalyst. The reactors were sealed and purged with 5 bar pure oxygen to remove any residual air out of the reactors. Afterwards, the reactors were pressurized to 5 bar, the magnetic stirrer bars were set to 300 rpm and the heating was switched on.

Experimental Procedure for In-Situ Investigations in the Flow-NMR Setup

All experiments using the multinuclear high resolution FlowNMR setup were per-formed at the DReaM facility at the University Bath. The general setup was shown in previous publications (see Figure S2).^{60, 63, 65, 66, 82} For the performed experiments the flow tube and flow path were cleaned and thoroughly rinsed with DI-water for a minimum of 20 min at 4 mL/min. As reactor, a 100 mL Büchi “miniclave inert” was used which was charged with 25 mL of chosen solvent composition, glycol aldehyde (100 mmol/L) and catalyst (HPA-2, 10 mmol/L). The flow path was connected via polyether ether ketone (PEEK) tubing with the miniclave. Subsequently, the NMR-probe and the miniclave were heated to reaction temperature while flowing the reaction solution trough the flow path. After reaching the reaction temperature (80 °C) the NMR acquisition cycle was started and the reaction vessel was pressurized with oxygen. For the reaction with oxalic acid, only oxalic acid (10 mmol/L) and glycol aldehyde (100 mmol/L) were charged into the miniclave. The catalyst was added after reaching the reaction temperature and the NMR acquisition cycle was started.

Determination of Quantitative Reaction Parameters

After the experiments all products were quantitatively determined by HPLC- (High Performance Liquid Chromatography), NMR- (Nuclear Magnetic Resonance) and GC- (Gas Chromatography) analysis. Liquid phase analysis was carried out using HPLC and ¹H-NMR spectroscopy. The conversion of substrates and yields of all liquid products were determined by means of HPLC measurements using a Nexera-40 HPLC from Shimadzu equipped with a 300 mm×7.8 mm Aminex 87-H column from Biorad and a refractive index detector. 5 mmol of an aqueous sulfuric acid solution was used as eluent and the samples were filtrated before analysis through a syringe filter (45 μm).

The yields of formic acid and the corresponding esters were quantified by ¹H-NMR using a Bruker Avance III HD 600 MHz spectrometer of the central analytics department of the University of Hamburg. Both acid and ester yields were calculated as n(product)/n(C-atoms substrate). The determination of the gaseous by-products CO₂ and CO was done by means of GC analysis using a Varian GC 450 equipped with a 2 m×0.75 mm ID ShinCarbon ST column and calculated as n(CO2 resp. CO)/n(C-atoms substrate). No other gaseous products could be detected by the used GC.

Acknowledgments

The authors thank the Royal Society (UF160458 fellowship to UH) and the EPSRC Dynamic Reaction Monitoring Facility at the University of Bath (EP/P001475/1). Funded by the European Union (ERC, BioValCat, Project 101086573). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. J.-D. H. Krueger and M. J. Poller thank the International Office of Universität Hamburg for covering their travel expenses to the DReaM facility at the University of Bath. We thank our research interns Thu Hanh Pham and Emma Petersen for their contributions to the experimental work. We also thank the division for NMR spectroscopy and the division for central element analytics of the Department of Chemistry conducting NMR experiments and ICP measurements. Open Access funding enabled and organized by Projekt DEAL.

Conflict of Interests

There are no conflicts to declare.

Open Research

Data Availability Statement

Data will be made available on request.