Peter J. C.
Hausoul
*a,
Anna K.
Beine
a,
Leila
Neghadar
b and
Regina
Palkovits
*a
aInstitut für Technische und Makromolekulare Chemie, RWTH Aachen University, Worringerweg 2, 52074, Aachen, Germany. E-mail: hausoul@itmc.rwth-aachen.de; palkovits@itmc.rwth-aachen.de
bInorganic Chemistry and Catalysis, Utrecht University, Universiteitsweg 99, 3584 CG, Utrecht, The Netherlands
First published on 12th December 2016
The aqueous Ru/C-catalysed hydrogenolysis of xylitol and sorbitol was studied in a temperature range between 393–443 K under 6 MPa H2. For the three main reactions, stereoisomerisation, decarbonylation and deoxygenation, kinetic models were formulated and fitted to the experimental data. The obtained rate constants were used to determine apparent activation enthalpies via the Eyring method. The data reveals a clear dependence of the type and position of the reacting hydroxyl group as well as the length of the polyol on the activation energies. It is proposed that these differences are the result of increased stabilisation due to polydentate interactions with the metal surface.
The employed catalyst systems can roughly be divided into noble metal-based (e.g. Pt,4 Ru5,6) and non-noble metal-based (Ni,7 Cu8). In the presence of homogeneous acidic co-catalysts, noble metals are preferred due to increased stability. Ru-based catalysts are particularly attractive due to their good activity and comparatively low price. However, in addition to the desired hydrogenation reaction, Ru also catalyses side reactions leading to decreased yields of hexitols in the hydrolytic hydrogenation of cellulose.9 Recently, we have presented a detailed study on the product mixture and reaction network observed for the Ru/C-catalysed hydrogenolysis of biomass-based polyols under neutral and acidic conditions.10 The data revealed the formation of a complex product mixture due to the occurrence of (de)hydrogenation, stereoisomerisation, decarbonylation and deoxygenation reactions (Scheme 1).
This demonstrates that Ru exhibits a rich surface chemistry with polyols, which is still not adequately understood. Valuable insights into reaction pathways of volatile oxygenates on clean Ru surfaces (e.g. acids, alcohols, carbonyls) have been obtained through UHV techniques (e.g. HREELS) and TPD experiments.11 However, the multifunctional nature of higher polyols and the harsh reaction conditions (i.e. water, high temperature, high pressure) used in hydrogenolysis hinder the application of these methods. Therefore, the development of alternative concepts especially applicable to (hemi-)cellulose-based C5- and C6-polyols is essential. Fortunately, the complex substrate structure and broad product distributions of higher polyols can provide insight into the reactions occurring on the metal surface. Previously, we proposed simplified reaction mechanisms, though a detailed understanding of the interactions of polyols with the metal surface is still lacking. Here, we tackle this problem using a comprehensive kinetic analysis of the stereoisomerisation, decarbonylation and deoxygenation of sorbitol and xylitol. To this end, kinetic models were proposed and kinetic parameters were estimated. The obtained rate constants and equilibrium compositions were used to assess thermodynamic quantities such as apparent activation enthalpies and isomer energies. In addition, a description of the reactions of higher polyols with the Ru surface is proposed based on the obtained parameters.
Rate constant kxa corresponds to the epimerisation of xylitol at the 2- or 4-positions yielding ARA and kxr corresponds to epimerisation at the 3-position yielding RIB. For the hexitols (i.e., D-/L-sorbitol (SOR), D-/L-mannitol (MAN), allitol (ALL), galactitol (GAL), D-/L-iditol (IDI) and D-/L-talitol (TAL)), the same approach was applied (Scheme 3, see eqn (5)–(10) in the ESI†).
Here, ksm, ksa, ksg and ksi correspond to the rate constants for the epimerisation of SOR at the 2-position yielding MAN, 3-position yielding ALL, 4-position yielding GAL and 5-position yielding IDI, respectively.
From previous studies, it is known that Ru/C exhibits an induction period in these reactions.10 To avoid artefacts, time zero of the fitted curves was chosen 20–30 min after the reaction temperature had been reached. Fig. 1 shows the experimental data and curve fitting for the evolution of the various pentitols and hexitols relative to their total amount (see Fig. S1 and S2 in the ESI†).
![]() | ||
Fig. 1 Ru/C-catalysed stereoisomerisation of xylitol (left) and sorbitol (right) (conditions: 2 g substrate, 20 ml H2O, 0.4 g Ru/C (5 wt%), 433 K, 6 MPa H2). |
XYL isomerises to ARA and RIB, with ARA as the major product. At temperatures above 423 K, the composition equilibrates within the monitored timeframe. In all cases, a final composition with on average 51.5% ARA, 30.5% XYL and 18.0% RIB is observed (Table 1). This is well in line with isomer distributions observed previously for the Ru/C-catalysed hydrogenation and isomerisation of xylose syrup by Beck et al.14 From the isomer distribution, it follows that XYL and RIB are respectively 1.9 and 3.8 kJ mol−1 higher in Gibbs energy than ARA.
Isomer | Equilibrium composition (%) | ΔG423K (kJ mol−1) | |
---|---|---|---|
Pentitols | Lit.14 | This work | |
Arabitol | 50 | 51.5 | 0 |
Xylitol | 29 | 30.5 | 1.9 |
Ribitol | 21 | 18.0 | 3.8 |
Hexitols | Lit.15 | ||
---|---|---|---|
Sorbitol | 41.4 | 31.4 | 0 |
Talitol | — | 18.4 | 1.9 |
Iditol | 26.5 | 16.3 | 2.4 |
Mannitol | 31.5 | 14.5 | 2.8 |
Galactitol | — | 13.7 | 3.0 |
Allitol | — | 5.8 | 6.1 |
A similar behaviour was observed for the isomerisation of SOR. SOR converts to all other isomers and at 433 K an apparent equilibrium is reached after 3 h. The final composition contains 31.4% SOR, 18.4% TAL, 16.3% IDI, 14.5% MAN, 13.7% GAL and 5.8% ALL. It follows that ALL is 6.1 kJ mol−1 higher in energy than SOR, whereas TAL, IDI, MAN and GAL are between 1.9–3.0 kJ mol−1 higher. To the best of our knowledge, data concerning the equilibrium composition of hexitols from Ru/C-catalysed isomerisation has not been reported before. A report by Wright et al. describes the isomerisation of hexitols catalysed by Ni/kieselguhr under alkaline conditions.15 The authors proposed that the isomerisation proceeds via aldose formation and enediol intermediates.
This proposal is supported by the absence of TAL, GAL and ALL, which are only available by isomerisation at the 3-/4-positions. Our data differs significantly and shows that the isomerisation occurs on all positions of the polyol.
The rate constants obtained from curve fitting at different reaction temperatures were used to calculate apparent activation energies via the Eyring method (Table 2). The Eyring plots all show a clear linear trend within the temperature range of 403–443 K. The data for XYL reveals that the apparent activation enthalpy of kxa (2-/4-pos., 82.5 kJ mol−1) is 32.3 kJ mol−1 lower than that of kxr (3-pos., 114.8 kJ mol−1). Thus, isomerisation at the 3-position is significantly more difficult than that at the 2-/4-positions. A similar behaviour is observed for SOR where the activation enthalpies of ksm (2-pos., 56.7) and ksi (5-pos., 62.7) are roughly 25 kJ mol−1 lower than those of ksa (3-pos., 81.2) and ksg (4-pos., 88.1). These data suggest that large differences exist between the reactivity of the different hydroxyl positions on the Ru surface. This is rather puzzling since these hydroxyl groups are chemically and sterically equivalent.
To further verify the validity of the model, additional reactions were performed with MAN and GAL (Fig. 2). As expected, MAN converts to the other isomers with SOR as the main product. The composition after 160 min resembles the equilibrium composition observed for SOR. Interestingly, for GAL, this is not the case. Here, the main product of the reaction is TAL. This again shows that isomerisation at the 2-/5-position (GAL to TAL) is significantly faster than that at the 3-/4-position (GAL to SOR). Leaving the reaction for a longer period was not possible due to conversion of hexitols via decarbonylation and deoxygenation reactions. Even though the equilibrium was not reached in this case, these data confirm the general observations for the XYL and SOR reactions.
![]() | ||
Fig. 2 Ru/C-catalysed stereoisomerisation of mannitol (left) and galactitol (right) (conditions: 2 g substrate, 20 ml H2O, 0.4 g Ru/C (5 wt%), 433 K, 6 MPa H2). |
![]() | ||
Scheme 4 Kinetic models for simultaneous decarbonylation and deoxygenation (left: only decarbonylation and deoxygenation (model 1); right: with an additional reaction path (model 2)). |
Here, decarbonylation and deoxygenation are treated as irreversible reactions. Kinetic modelling of XYL at 423 K (Fig. 3, left) using model 1 shows a good fit for decarbonylation, whereas for deoxygenation, the fit shows a systematic error indicating a problem with the model. When a third reaction path (kwp) is included (model 2), the fit for deoxygenation improves significantly (Fig. 3, right). This suggests that another decomposition reaction may be involved in the hydrogenolysis reaction. However, GC and LC analysis of the reaction mixture did not reveal any other products. A possible explanation for this is that adsorbed species can react multiple times without intermediate dissociation, i.e. the first model assumes that the fragments formed from cleavage reactions are completely hydrogenated and desorbed before further reaction. However, if adsorption of the polyol is strong, a second cleavage reaction may ensue without desorption. Given the improved fits, the latter model was used for the Eyring analysis of the rate constants kpt and k54 (Table 3, see eqn (11)–(13) and Fig. S3 in the ESI†). The data shows that the apparent activation enthalpy for deoxygenation (83.1 kJ mol−1) is significantly lower than that for decarbonylation (124.3 kJ mol−1). Interestingly, deoxygenation is considerably slower than decarbonylation. This suggests that there are fewer sites catalysing deoxygenation.
The second approach to model the hydrogenolysis kinetics is to consider consecutive decarbonylation or consecutive deoxygenation. Unfortunately, consecutive deoxygenation could not be determined with sufficient accuracy (i.e. low amounts, overlapping peaks in GC). Scheme 5 shows the reaction sequence for the consecutive decarbonylation of hexitols down to ethylene glycol.
Here, khp represents the rate constant for the decarbonylation of hexitols to pentitols and kwh indicates the rate constant corresponding to the loss of hexitols to side reactions. All stereoisomers of a polyol were treated as equally reactive and grouped. As with stereoisomerisation, first-order kinetics was assumed for the formulation of the rate equations (see eqn (14)–(18) and Fig. S4 and S5 in the ESI†). Fig. 4 shows the experimental data and curve fitting of sorbitol and xylitol at 433 K. As expected, the model exhibits excellent agreement with the experimental data. The maximum accumulation of pentitols, tetritols and glycerol decreases with decreasing chain length. In view of accuracy, only the first few follow-up reactions were considered for Eyring analysis. Table 4 shows the Eyring plots of the rate constants khp, kpt and ktg obtained for the SOR and XYL reactions in the temperature range of 403–443 K.
![]() | ||
Fig. 4 Ru/C-catalysed decarbonylation of pentitols (left) and hexitols (right) (conditions: 2 g substrate, 20 ml H2O, 0.4 g Ru/C (5 wt%), 433 K, 6 MPa H2). |
The plots show a clear linear trend and the calculated apparent activation enthalpies are clustered between 94–122 kJ mol−1. Comparison of the same transformations e.g. decarbonylation of pentitols (kpt) and tetritols (ktg) observed for the SOR and XYL reactions shows that they are comparable and have differences between 6.1–10.2 kJ mol−1. Interestingly in both cases, the activation energy slightly decreases with each subsequent step. This suggests that the length of the polyol has an influence on the activation energy. The activation enthalpy of kpt obtained from the XYL dataset using the simultaneous deoxygenation and decarbonylation model (Scheme 4, Table 3, 124.3 kJ mol−1) is approximately 13 kJ mol−1 higher than that for the consecutive decarbonylation model (Scheme 5, Table 4, 111.7 kJ mol−1). This shows that both models provide comparable results. However, further studies are required to determine these activation energies with greater accuracy.
When taking this situation as a reference state, polyol adsorption cannot proceed without prior desorption of adsorbates (e.g. water). A recent study by Murzin et al. revealed that the adsorption energy of the primary hydroxyl group of Cn (n = 1–4) polyols on clean Ru surfaces increases from −0.65 (n = 1) to −0.93 eV (n = 3, 4).17 Compared to the adsorption energy of water (−0.47 to −0.54 eV), this confirms that for higher polyols there is still a significant net adsorption energy. The results of Miller et al. on the adsorption of glycerol on Ru-sponge also suggest stronger adsorption for higher polyols.18 In addition, polydentate polyol adsorption will be entropically favoured due to the successive release of adsorbates (Scheme 7). For this adsorption process, many permutations exist which differ in the order of hydroxyl attachment and polyol geometries on the surface. Even if full polydentate adsorption is not possible (e.g. due to steric constraints), it is expected that hydrogen-bonding interactions will lead to additional stabilisation of the polyol on the surface.
![]() | ||
Scheme 8 Schematic representation of Ru-catalysed epimerisation reaction (hydrogen, water and other adsorbates not shown for clarity). |
The H/D-exchange data also revealed that exchange occurs readily with retention of configuration at temperatures around 373 K.10 This indicates that the formation of species A*, A** and C** is reversible and fast and that conversion of C** to C* or C is rate determining for epimerisation. Based on this, we propose that the activation enthalpies of stereoisomerisation (Table 2) are related to desorption of the ketone. As a reference, we have calculated the reaction energy between species A and C + H2 for the dehydrogenation of xylitol at the 1-/5-, 2-/4- and 3-positions in aqueous solvent. The results are summarized in Table 5.
ΔH298K (kJ mol−1) | ΔG298K (kJ mol−1) | ΔS (J mol−1 K−1) | |
---|---|---|---|
1-/5-position | 105.3 | 138.6 | −111.8 |
2-/4-position | 91.7 | 123.2 | −105.8 |
3-position | 85.4 | 117.4 | −107.1 |
The enthalpy for aldose formation (105.3 kJ mol−1) is 13–20 kJ mol−1 higher than those for ketose formation (85.4–91.7 kJ mol−1). Interestingly, the enthalpy for 3-ketose formation is 6.3 kJ mol−1 lower than that for 2-,4-ketose formation. In contrast, the apparent activation enthalpies of the epimerisation at the different hydroxyl groups of SOR and XYL (Table 2) revealed that the enthalpies for the innermost positions (3-,4- or 3-) are approximately 25–30 kJ mol−1 higher than those for the outer positions (2-,5- or 2,4-). Structurally, these positions differ only in the nature of neighbouring carbon atoms (i.e. 1° or 2°) and the number of neighbouring hydroxyl groups. The increases in activation energy can originate either from increased stabilisation of the reactant state or from destabilisation of the transition state. Although subtle differences in stability are to be expected for positional isomers and stereoisomers, on average their stabilisation at the surface is expectedly comparable. Also, the data for the 3- and 4-positions of SOR show that the stereochemical configuration of the neighboring positions has a relatively small influence on the activation energy. It appears therefore that the increased activation energy results from destabilisation of the transition state. A possible way to account for this is to consider desorption of the carbonyl, which is necessary for stereoinversion. Scheme 9 shows a schematic representation of the proposed desorption of 2-/4- and 3-ketose intermediates.
In the initial state (1), the hydroxyl groups and the carbonyl are color coded red to signify interaction with the surface (e.g. as adsorbed or via hydrogen bonding). To invert the stereotopic plane of the carbonyl on the metal, desorption of the carbonyl and neighbouring hydroxyl groups is necessary (2). For the 2-ketose, this entails desorption of at least the 1-OH, 2-CO and most likely also 3-OH groups, to allow for free rotation of the carbonyl. For the 3-ketose, the required desorption is even more extensive (i.e. 1-OH, 2-OH, 3-CO and 4-OH). Thus, it is proposed that the higher apparent activation enthalpy for epimerisation of the inner positions is due to more difficult desorption of the carbonyl groups as a result of polydentate interactions of the neighbouring hydroxyl groups.
![]() | ||
Scheme 10 Schematic representation of Ru-catalysed decarbonylation reaction (hydrogen and other adsorbates not shown for clarity). |
Previous H/D exchange experiments on SOR showed that the C–H groups at the 1-/6-positions are rapidly and completely exchanged at 373 K without the occurrence of decarbonylation.10 This indicates that the reaction path between A and C** is fast and reversible. It is proposed that the observed activation energy is related to the formation of either Ac* or CO* + R*. The data in Table 4 show that the apparent activation enthalpy decreases with decreasing polyol chain length. Also here, the chemical nature of the reacting group is identical and the change in activation energy may be attributed to polydentate interactions of neighbouring hydroxyl groups. However, contrary to epimerisation where the reaction can be traced to desorption and rotation of a single carbonyl fragment, it is not obvious how the increasing polydentate binding leads to higher reaction barriers in this case. One possible way to account for this is to consider that the reacting group must go through several geometrical changes and bonding modes. Since the rest of the polyol chain is also attached to the reacting group, it must also accommodate these changes by reorganisation on the surface. This reorganisation again depends on desorption of the polyol, which will be increasingly more difficult with increasing chain length.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6cy02104b |
This journal is © The Royal Society of Chemistry 2017 |