Why Ru‐MACHO‐BH Is Poor in Coupling Two Ethanol to N‐Bu

Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

1 2 3 The Guerbet Reaction Network – a Ball-in-a-Maze-Game or: 4 5 Why Ru-MACHO-BH is Poor in Coupling two Ethanol to 6 n-Butanol 7 8 Andreas Ohligschläger, Nils van Staalduinen, Carsten Cormann, Jan Mühlhans, Jan Wurm, 9 and Marcel A. Liauw*[a] 10 11 12 The Guerbet reaction from two alcohols to a long-chain alcohol subsystems of the network. In-situ-infrared spectroscopy is 13 and water requires a redox catalyst and a strong base in applied to determine time-resolved concentration profiles. 14 homogeneous liquid systems. Especially, the reaction from Adapted kinetic models of the single steps are integrated into a 15 ethanol to n-butanol is a challenging example of the reaction microkinetic model of the whole network. The simulation of the 16 that suffers from low yields and selectivities in comparison with reaction network reveals dependencies between temperature, 17 reactions of higher alcohols. The most important side reactions hydrogen pressure, initial concentrations and the yield and 18 are the polymerization of acetaldehyde to C -components and selectivity of n-butanol. Finally, it is shown that Ru-MACHO 19 6+ the saponification of ethyl acetate under consumption of the does not lead to high yields in the reaction, because the 20 base. This work pursues the systematic kinetic investigation of dehydrogenation to ethyl acetate exhibits a too low activation 21 the Guerbet reaction network by experiments with isolated barrier. 22 23 24 1. Introduction 25 26 The Guerbet reaction[1] from ethanol to n-butanol (Scheme 1, Scheme 1. Overall description of the Guerbet reaction from ethanol to n- 27 below referred to as butanol) is appealing, since butanol has butanol. 28 advantageous properties in fuel applications and ethanol can 29 be obtained sustainably from fermentation of sugars.[2–4] In 30 detail, butanol has a higher energy density and is less miscible 1-ethoxyethanol and the saponification of ethyl acetate. In 31 with water. Unfortunately, only a few successful examples of contrast, the production of acetate is often described as 32 the reaction from ethanol to butanol are published[4–13] and Cannizzarro reaction between two acetaldehyde molecules.[4,8,9] 33 ethanol has been referred to as a “sluggish substrate”.[5] The choice of the route via the hemiacetal will be explained in 34 It is generally accepted that the Guerbet reaction proceeds the results. All these reactions span a reaction network that has 35 in three steps:[14,15] the dehydrogenation from ethanol to one desired destination and two undesired outcomes 36 acetaldehyde, the aldol condensation from two acetaldehyde (Scheme 2). 37 molecules to crotonaldehyde and water and the hydrogenation It is worth to mentioning that the acetate and polymer 38 of crotonaldehyde to butanol. The most important side production is often neglected in the quantitative analysis of the 39 reactions are the polymerization to C -compounds and the product mixture since both side products are not detectable in 40 6+ production of acetate salts. Both reactions are irreversible and gas chromatography (GC). Positive counterexamples are the 41 decrease yield and selectivity. The production of acetate salts is publications of Jones et al.[7] and Cavani et al.[4] which quantify a 42 here assumed to proceed via the dehydrogenation of the carbon loss by calibration with an internal standard. 43 hemiacetal Established redox catalysts are based on bidentate or 44 tridentate ligands and iridium, ruthenium or manganese as 45 central atoms. Sodium or potassium alkoxides are commonly 46 [a] A. Ohligschläger, N. van Staalduinen, C. Cormann, J. Mühlhans, J. Wurm, applied as strong bases. An outstanding example is the work of 47 Prof. M. A. Liauw Cavani et al.[4] in which a ruthenium catalyst is applied which 48 Institut für Technische und Makromolekulare Chemie accomplishes to get 25% yield with 50% selectivity in the 49 RWTH Aachen University Worringerweg 1 presence of 5 vol-% water. 50 52074 Aachen (Germany) While many publications deal with the optimization of the 51 E-mail: [email protected] applied redox catalyst, the aim of this work is to gain insight 52 Supporting information for this article is available on the WWW under into the kinetic characteristics of the Guerbet reaction network. 53 https://doi.org/10.1002/cmtd.202000056 © 2021 The Authors. Published by Wiley-VCH GmbH. This is an open access Since base catalyzed steps as well as redox catalyzed steps 54 article under the terms of the Creative Commons Attribution Non-Com- need to gear into each other to produce butanol, the network 55 mercial NoDerivs License, which permits use and distribution in any med- is first divided into subsystems that can be observed in isolated 56 ium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. experiments: 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 181/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

the influence of solvent effects that play a dominant role in 1 aldol condensation and saponification (see below). 2 All experiments of this work employ potassium-tert-butano- 3 late as a base and Ru-MACHO (in form of Ru-MACHO-BH 4 precursor) as the redox catalyst (Scheme 3). Ru-MACHO is a 5 popular catalyst for the hydrogenation of esters[24] and the 6 acceptorless dehydrogenation of alcohols.[25] Thus, it appears as 7 a suitable catalyst for the Guerbet reaction at first sight and was 8 already applied for Guerbet reactions starting from 9 methanol.[26,27] However, Wass et al.[8] reported that Ru-MACHO 10 is an inferior catalyst for the Guerbet reaction, but they could 11 not find an explanation based on the structure of the catalyst. A 12 recent study shows that Ru-MACHO degrades to a mixture of 13 several catalytically active complexes in alkaline media.[28] In our 14 work, we could only reproduce that Ru-MACHO is not a suitable 15 catalyst for the Guerbet reaction from ethanol to butanol. The 16 Scheme 2. The Guerbet reaction network that is assumed in this work. Redox additional benefit of the microkinetic simulation consists of 17 reactions are depicted in vertical direction, while base dependent reactions are depicted in horizontal direction. For kinetic experiments, subsystems are understanding why the catalyst fails in the reaction. This 18 defined that can be observed isolated from each other (transparent knowledge helps to identify critical properties of a successful 19 rectangles). redox catalyst for the Guerbet reaction. 20 21 22 1. The redox system of C -substances (ethanol, acetaldehyde/ 2. Results and Discussion 23 2 1-ethoxyethanol and ethyl acetate), 24 2. the redox system of C -substances (crotonaldehyde, croty- This section is ordered by results of the kinetic experiments of 25 4 lalcohol, butanal, butanol), the four subsystems, experiments of the whole Guerbet 26 3. the aldol condensation of acetaldehyde and reaction, assembly of the microkinetic model, simulation of the 27 4. the saponification of ethyl acetate. Guerbet reaction network and lastly a simulation of a modified 28 All subsystems are investigated with in-situ-IR spectroscopy. catalyst. Though it may be appealing to read only the results of 29 Thus, time-resolved concentration profiles are accessible after the simulation, it is recommended to take a look at the 30 the application of chemometric models. That allows a descrip- experimental results as they convey a feeling for the behavior 31 tion of the kinetic behavior in detail with a comparably low of the reaction steps. 32 experimental effort.[16–19] While the aldol condensation and the 33 saponification can be described with an accurate kinetic model 34 without further ado, the redox subsystems themselves consist 2.1. Aldol Condensation of Acetaldehyde 35 of several reaction steps, in particular the hydrogenation of the 36 catalyst. Those single steps cannot be resolved by IR-spectro- The aldol condensation from acetaldehyde to crotonaldehyde is 37 scopy independently, because the reduced and oxidized a key step in the Guerbet reaction network, since this step leads 38 catalyst species are not detectable under process conditions. to the formation of the new CÀ C-bond. Reported kinetics of this 39 Thus, the kinetic model of the redox steps is created from step refer to water as the reaction medium.[29–32] It is assumed 40 activation barriers of DFT calculations found in literature,[20–22] that the Guerbet reaction takes place with alkaline ethanol as 41 but the barrier heights are adapted to reproduce the the major reaction medium, but water and potassium acetate 42 experimental concentration profiles. The kinetic models of all are formed with ongoing conversion. Thus, the kinetic experi- 43 subsystems are finally assembled to a microkinetic model of the ments of the aldol condensation of acetaldehyde with 44 Guerbet reaction network in MatLab. The assumed “reactor” for potassium hydroxide as a base are split in four series: 45 the simulations is an isothermal and isobaric batch reactor with 46 homogeneous concentration distribution at all times. Simula- 47 tions are performed with variations of temperature, hydrogen 48 pressure and starting concentrations of potassium acetate, 49 water and ethyl acetate. In a final step, the activation barriers of 50 the redox steps are varied to simulate modifications of the 51 catalyst. The only published example of the simulation of 52 homogeneously catalyzed Guerbet reaction is the work of 53 Pathak et al.[23] In contrast to this work, they based their 54 Scheme 3. For this work, potassium tert-butanolate is used as a base (left) simulation solely on the activation barriers of DFT calculations 55 and Ru-MACHO is used as redox catalyst (right). Ru-MACHO is formed in this without experimental basis. Thus, their simulation might miss study from the precursor Ru-MACHO-BH by the separation of a BH -group 56 3 under alkaline conditions (not shown here). 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 182/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

1. In ethanol alone (V1), 1 2. in ethanol with water (V2), 2 3. in ethanol with potassium acetate (V3), 3 4. in ethanol with water and potassium acetate (V4). 4 As the reaction is known to proceed at room temperature, 5 the temperature range for the experiments is limited between 6 25 °C and 60 °C to gain suitable concentration-time-profiles for 7 further investigation. That is a disadvantage, since the Guerbet 8 reaction is known to proceed at temperatures above 120 °C and 9 the rate coefficients have to be extrapolated over a big range. 10 At higher temperatures, the reaction is expected to proceed so 11 fast that too few spectra may be acquired, mixing issues may 12 influence the results and the high vapor pressure of 13 acetaldehyde complicates proper handling of the reaction 14 mixture. A possible solution for that would be the use of a 15 microreactor with fast mixing, a defined residence time and in- 16 line spectroscopy at the end of the reactor or even in the small 17 reactor capillaries.[33–36] Since there was no access to such a 18 reactor setup, the results of this work are limited to the IR- 19 spectra measured in a covered glass beaker. The use of IR- 20 spectroscopy was in particular useful for this reaction, since the 21 presence of the hemiacetal species 1-ethoxyethanol 22 (acetaldehyde+ethanol) or 1,1-ethanediol (acetaldehyde+ 23 water) was observed that does not appear in the aprotic 24 medium cyclohexane (Figure 1). This observation is in agree- 25 ment with the NMR results from Scheithauer et al who observed 26 similar species in aqueous solution (Figure 2).[30] 27 Through a mass balance (ESI chapters 2 and 3), both species 28 could be quantified independently in the calibration of the IR- 29 spectra. Since no different reactivity could be identified for both 30 species, it is assumed that the interconverion of acetaldehyde 31 Figure 1. IR spectra of acetaldehyde mixed with cyclohexane (top) or ethanol and the hemiacetal proceeds much faster than the aldol 32 (bottom). Mixtures are shown with a blue line and pure solvents are shown condensation. Thus, the sum of both concentrations was used with a red line. While the mixture spectrum on the left contains only bands 33 as the substrate concentration. A linearization in the manner of that can be ascribed to the “monomer” acetaldehyde, the mixture spectrum 34 on the right shows additionally an absorption band (black circle) that can be a second order reaction (Equation (1), Figure 3) makes the 35 assigned to a CÀ O single bond of 1-ethoxyethanol. apparent rate coefficient accessible. The results of all exper- 36 imental series are given in Table 1. 37 38 1 1 39 ¼ 2k � c À � t þ c app;AC OH c (1) 40 AcH AcH;0 41 Figure 2. Possible C2- and C4-species to appear in the aldol condensation of 42 acetaldehyde in ethanol/water media. c concentration 43 i k apparent rate coefficient of aldol condensation 44 app;AC t time 45 Table 1. Apparent rate coefficients of the aldol condensation in dependence from temperature and experimental series. At 45 °C and 60 °C, the 46 Obviously, the presence of salt increases the reaction rate potassium hydroxide concentration was decreased to slow down the 47 reaction to a measurable time scale. This leads to lower apparent rate and that effect is reinforced in the presence of water. For a 48 coefficients than one might expect. kinetic model, this is interpreted as a salting-out-effect[37] that 49 kapp in 25 °C 35 °C 45 °C 60 °C correlates the logarithmic apparent rate coefficient with the 2 À 2 À 1 50 L mol min dissociated salt concentration ac . Fit parameters are the 51 salt V1 (–) 2.07 18.4 18.2 163 Setchenow-constant KS and the logarithmic rate coefficient in 52 V2 (H2O) 2.64 16.3 13.9 137 salt-free conditions lnk0;AC (Equation (2)). V3 (KOAc) 4.09 29.0 21.8 465 53 + The emphasis is given on dissociated, since water facilitates V4 (H2O KOAc) 7.83 50.0 48.2 685 54 the dissociation of salts in comparison with ethanol and water 55 reinforces the salt effect as well. The sum of potassium 56 hydroxide and potassium acetate concentration is considered 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 183/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

A frequency factor of aldol condensation 1 AC E activation energy of aldol condensation 2 A,AC R gas constant 3 T temperature 4 K dissociation constant 5 D DG dissociation enthalpy 6 diss x molar fractions 7 i 8 9 10 Figure 3. Reciprocal acetaldehyde concentration versus time as linearization 2.2. Saponification of Ethyl Acetate 11 of a second-order reaction with potassium hydroxide at 25 °C without water 12 and without salt. The slope of the linear fit delivers the rate coefficient of the aldol condensation. The saponification of ethyl acetate is a well-studied reaction 13 that serves as a model reaction for second-order-reactions. 14 Nonetheless, the determination of the kinetics is performed 15 as a (constant) initial salt concentration. To gain the dissociated here since the reaction medium of the Guerbet reaction 16 salt concentrations with respect to temperature and water somewhat special for the saponification. The experimental 17 content, conductivity values of potassium acetate in water- series are equivalent to the study of the aldol condensation: 18 ethanol mixtures[38] were converted to the dissociation constant 1. In ethanol alone (V5), 19 with Ostwald’s law of dilution[39] (Equation (4)). Furthermore, the 2. in ethanol with water (V6), 20 dissociation enthalpy and entropy are separated by the 3. in ethanol with potassium acetate (V7), 21 assumption that all binary salts have the constant dissociation 4. in ethanol with water and potassium acetate (V8). 22 entropy DS of À 133.89 JKÀ 1 molÀ 1.[40] A linear correlation of Potassium hydroxide and ethyl acetate are used as sub- 23 diss the molar fraction of water with the dissociation enthalpies strates. Again, the reaction temperature is limited to the range 24 leads to separated dissociation enthalpies in water DH from 25 °C to 60 °C and the reaction is performed in a covered 25 diss;H2O (À 33.2 kJmolÀ 1) and in ethanol DH (À 22.9 kJmolÀ 1) (Equa- glass beaker. This time, the concentration-time-profiles are 26 diss;EtOH tion (6)). That allows the prediction of the dissociated salt converted to reaction rates by numerical derivation of the ethyl 27 concentration at any given temperature and solvent composi- acetate concentration with respect to time. A quadratic fit of 28 tion. The mean value of the Setchenow constants derived at the reaction rate vs. the ethyl acetate concentration (Equa- 29 each temperature K is 68.3 LmolÀ 1 that is assumed to be tion (7), Figure 4) leads to the apparent rate coefficient, because 30 S independent of temperature. From the rate coefficients under both substrates are consumed in equimolar amounts. The 31 salt free conditions, the activation energy E (87 kJmolÀ 1) as results of all experiments are shown in Table 2. 32 A;AC well as the frequency factor A (5.13 1013 L2 molÀ 2 sÀ 1) can be 33 AC extracted in an Arrhenius plot (Equation (3)). The activation r ¼ kapp;ScEtOAccOHÀ 34 [23,41] (7) energy is higher than reported values, because a salt free ¼ 2 þ � � 35 kapp;ScEtOAc kapp;S Z cEtOAc reaction medium is not accomplishable in an experiment with a 36 Brønsted base. Concluding, the rate coefficient of the aldol 37 condensation can be described by following Equations (2) to 38 (6). The same rate coefficient will be assumed for further aldol 39 condensation steps to C - and C -substances, as no kinetic 40 6 8+ experiments were performed for the aldol condensation 41 between crotonaldehyde and acetaldehyde. This will over- 42 estimate the reaction rate of the side reaction and under- 43 estimate the butanol-yield, because crotonaldehyde is known 44 to be less reactive in the aldol condensation.[23] 45 46 lnk ¼ lnk þ K ac 47 app;AC 0;AC S salt (2) 48 lnk0;AC ¼ lnAAC À EA;AC=RT (3) 49 50 a2 K ¼ � c (4) 51 D 1 À a salt 52

53 DGdiss ¼ À RT � lnKD (5) 54 D ¼ ðD � þ D � Þ À D 55 Gdiss Hdiss;EtOH xEtOH Hdiss;H2O xH2 O T Sdiss (6) Figure 4. Reaction rate of saponification versus concentration of ethyl acetate with potassium hydroxide at 35 °C without salt and without water. 56 The fit parameters of the quadratic fit deliver the apparent rate coefficient. 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 184/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

1 Table 2. Apparent rate coefficients of the saponification in dependence from temperature and experimental series.

2 kapp in 25 °C 35 °C 45 °C 60 °C À 1 À 1 3 L mol min 4 V5 (–) 0.075 0.126 0.297 0.870 5 V6 (H2O) 0.335 0.677 1.664 5.785 V7 (KOAc) 0.080 0.120 0.267 0.825 6 V8 (H2O+KOAc) 0.310 0.763 1.886 4.736 7 8 9 r 2.3. Redox System of C -Substances 10 2 reaction rate 11 The hydrogenation of ethyl acetate spans a reaction equilibrium 12 k with the dehydrogenation of ethanol. The position of the 13 app;S apparent rate coefficient of saponification equilibrium is determined by the temperature as well as the 14 hydrogen pressure. Thus, kinetic experiments of the hydro- 15 Z genation of ethyl acetate were performed in an autoclave[19] 16 difference between base and ester concentration under isobaric conditions (in constant connection with the 17 This time, the rate coefficients increase with the water hydrogen gas cylinder) with varied hydrogen pressures from 18 content, but the salt addition has a negligible influence. The one experiment to another. In all experiments, the catalyst Ru- 19 Arrhenius plots of all four series show that the activation energy MACHO-BH was first activated in a mixture of ethanol and 20 is nearly the same, but the frequency factor increases with the potassium-tert-butanolate before adding the substrate ethyl 21 water content (Figure 5). The activation energy is in the range acetate to prevent saponification reactions between ethyl 22 of reported activation energies. This behavior can be modeled acetate, the base and residual water in ethanol. A reverse 23 with the assumption that the saponification proceeds via two addition did not lead to the activation of the catalyst. Figure 6 24 pathways, one with and one without water. This assumption is shows a representative concentration profile of the hydro- 25 covered by DFT calculations of saponification reactions[42] that genation of ethyl acetate to ethanol. In comparison of all 26 also assume two parallel reaction pathways. Concluding, the experiments, it could be shown that measurable amounts of 27 saponification can be described by Equations (8) and (9). ethyl acetate are left when a hydrogen partial pressure of 28 10 bar or below is applied. At higher temperatures, the residual 29 À � r ¼ k þ k � c � c c À concentration of ethyl acetate increases or the dehydrogen- 30 1;S 2;S H2O EtOAc OH (8) À � ation is preferred, respectively. 31 þ � ¼ ð þ � Þ � À = k1;S k2;S cH2O A1;S A2;S cH2O exp EA;S RT (9) For a kinetic model, the microkinetic MatLab-simulation (see 32 below) was performed under the assumption that no base is 33 present and aldol condensation as well as saponification can be 34 k rate coefficient of saponification path without water neglected. As starting values for the activation barriers, the 35 1;S k rate coefficient of saponification path with water results from DFT calculations of Ru-MACHO or at least 36 2;S E activation energy of saponification ruthenium-pincer-complex catalyzed reactions were taken from 37 A;S A frequency factor of saponification path without water literature.[20–22] Those barriers were adapted to reproduce the 38 1;S A frequency factor of saponification path with water 39 2;S 40 41 42 43 44 45 46 47 48 49 50 51 52 Figure 5. Arrhenius plots of the four experimental series of the saponifica- 53 tion. The upper two plots contain water, the lower two plots contain no 54 water. The red dots were excluded from the fit since water from the Figure 6. Concentration time profiles of ethyl acetate (orange) and ethanol 55 environment was absorbed from the environmental atmosphere while (blue) from the hydrogenation employing Ru-MACHO-BH at 120 °C and 6 bar potassium hydroxide was dissolved. Thus, those two experiments contain constant external hydrogen pressure. Under these conditions, a high residual 56 overestimated reaction rates. of ethyl acetate persists. 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 185/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

experimental concentration profiles (ESI chapter 5 and 6). A 1 change of the activation enthalpies seems to be allowed, since 2 all DFT calculations either neglect the solvent influence or do 3 not take the actual solvent composition into account. The 4 biggest change was made in the activation entropy of the 5 dehydrogenation of acetaldehyde to ethyl acetate. This needed 6 to be done, because the DFT calculation assumed a termolecu- 7 lar mechanism starting from acetaldehyde, ethanol and oxi- 8 dized catalyst (Scheme 4, top). Such a termolecular reaction 9 leads to a high activation entropy which makes the reaction 10 nearly impossible at elevated temperatures. In contrast, this 11 work assumes the dehydrogenation of acetaldehyde to proceed 12 via the hemiacetal 1-ethoxyethanol which leads to a bimolecu- 13 lar mechanism with an accessible activation entropy (Scheme 4, 14 bottom). As the equilibrium between the hemiacetal and 15 acetaldehyde is not determined, the sum of both concentra- 16 tions is used in the kinetic differential equations for reactions 17 that require one of the substances as a substrate. To the best of 18 our knowledge, the dehydrogenation of the hemiacetal has not 19 been calculated with DFT methods yet. 20 21 22 2.4. Redox System of C -Substances 23 4 24 For the investigation of the C -substances, crotyl alcohol was 25 4 used as a model substrate in the transfer hydrogenation with 26 ethanol, because crotonaldehyde is known to be toxic and the 27 direct hydrogenation of crotyl alcohol to butanol is inhibited by 28 a high activation barrier.[20] Thus, crotyl alcohol is first converted 29 to crotonaldehyde and butanal before butanol is formed and all Figure 7. Time dependent IR spectra of transfer hydrogenations from crotyl 30 alcohol (970 cmÀ 1 and 1000 cmÀ 1, falling) to butanol (1080 cmÀ 1, rising) C -species are present in solution. Additional hydrogen pressure 31 4 employing Ru-MACHO-BH. Top: ethanol is used as a hydrogen source (80 °C). was not needed as the solvent ethanol is a sufficient hydrogen Bottom: methanol is used as a hydrogen source (120 °C). 32 source for the reaction. The experiments show that crotyl 33 alcohol is converted to butanol even at moderate 60 °C temper- 34 ature and most of the substrate is consumed while heating Again, the kinetic model was assembled by activation 35 when higher temperatures are applied (Figure 7). This observa- barriers from DFT-calculations. This time, no further adaptations 36 tion matches the results of De Vries et al.[20] After consumption needed to be made to match the experimental concentration 37 of the crotyl alcohol, the IR-probe is covered with polymeric profiles. 38 structures of aldol condensation products. To eliminate the 39 conclusion that all crotyl alcohol is transformed to the polymer, 40 the same reaction was performed in methanol which does not 2.5. Experiments of the Whole Guerbet Reaction 41 form polymeric structures based on its aldehyde (Figure 7, 42 right). In this reaction, a clear build-up of butanol can be seen In experiments of the whole Guerbet reaction, no butanol 43 that is not followed by a polymerization. formation could be detected in the IR spectra. Nonetheless, 44 these experiments are helpful to confirm the assumed Guerbet 45 reaction network. Figure 8 shows a representative profile of the 46 time-dependent spectra. From the very beginning (or while 47 heating), ethyl acetate and water are formed. Since no water 48 was added in the starting mixture, the water formation 49 corresponds to the formation of aldol condensation polymers. 50 The polymer could not be detected, as it did not sediment on 51 the IR-probe in this experiment. Next, the saponification in 52 presence of water, ethyl acetate and the base leads to acetate 53 formation. As the acetate concentration reaches a plateau value, 54 Scheme 4. Dehydrogenation of acetaldehyde via a termolecular pathway more water is formed and the intermediate build-up of a 55 À 1 (top) and via the hemiacetal 1-ethoxyethanol in bimolecular reactions (mid species with 1500 cm absorbance frequency can be observed. 56 and bottom). This species is guessed to be the adduct of hydroxide with 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 186/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

assembly of the microkinetic model, all reaction steps are 1 numbered as shown in Scheme 5. 2 Redox steps with the oxidized catalyst have the suffix “_1”, 3 while redox steps with the reduced catalyst have the suffix “_2”. 4 The rate terms of all single steps (ESI Table S8) are then 5 combined to the component balances of all species (ESI 6 Table S9). Those rate laws are a set of ordinary differential 7 equations that can be solved numerically with the ODE solver- 8 functions implemented in MatLab. As a result, concentration- 9 time-profiles are received. For this work, the solver ode23 t was 10 used, because it is suitable for so called “stiff” problems that 11 include fast processes parallel to slow processes. In addition to 12 the ODE solver, the microkinetic simulation includes also: 13 1. for-loops for the automated simulation of different temper- 14 Figure 8. Time-dependent IR-spectra of an unsuccessful Guerbet-reaction at atures, hydrogen partial pressures and starting concentra- 15 125 °C starting from ethanol, Ru-MACHO-BH and potassium tert-butoxide. At tions, 16 the beginning, ethyl acetate (1750 cmÀ 1) and water (1650 cmÀ 1) can be À 1 2. The determination of the dissociated salt concentration, 17 detected. The evolution of acetate (1580 cm ) and presumably hydroxide (1500 cmÀ 1) follow. 3. An estimation of the equilibrium between potassium 18 ethoxide, potassium hydroxide, water and ethanol, 19 4. An estimation of the hydrogen solubility with respect to 20 temperature and salt concentration. 21 The latter two estimations are included to get more realistic 22 results with respect to changes in the reaction mixture 23 Figure 9. Depiction of one possible ethanol-hydroxide adduct formed by composition. In conclusion, the microkinetic model evolved 24 À 1 hydrogen bonding. The vibration band at 1500 cm may be attributed to here integrates all relevant reaction steps on an experimental 25 the HÀ OÀ H-bending vibration. basis. Furthermore, solvent and salt effects are included as well 26 as the influence of acid/base equilibrium and hydrogen 27 solubility. This allows getting profound insights to the parame- 28 ethanol (Figure 9), since its decrease is connected with the ter correlations of the Guerbet reaction Network that exceed 29 build-up of water and no further absorption band in the the findings of Pathak et al.[23] Nonetheless, room for improve- 30 spectrum is connected with it. At higher water contents, ment would be the separation of the aldol condensation 31 hydroxide is solvated with water and does not appear with a kinetics with respect to the substrates and kinetic experiments 32 unique absorption band. This guess would be supported by the of saponification and aldol condensation at elevated temper- 33 calculation of the vibrational spectrum of hydroxide-ethanol atures. Furthermore, the integration of more reactor models 34 adduct(s). would improve the results (see below). 35 In total, the sequence of appearing side products confirms 36 the assumed reaction network, because all assumed reaction 37 steps are essential to describe the formation of the side 38 products. On the other hand, no further reaction steps are 39 needed to describe the formation of all substances which 40 makes the assumed reaction network as simple as possible. 41 42 43 2.6. Assembly of the Microkinetic Model 44 45 A central assumption underlies the microkinetic simulation. The 46 reactor is assumed to be an ideal, isobaric batch reactor with no 47 mixing issues. Thus, for all simulations a hydrogen pressure 48 above zero needs to be assumed, because at 0 bar hydrogen 49 pressure the reactor would work under the conditions of the 50 acceptorless dehydrogenation of ethanol under vacuum and all 51 ethanol would be instantly dehydrogenated. Besides hydrogen, 52 all substances are assumed to have no own partial pressure. In 53 contrast to the experiments, potassium ethanolate is used as a 54 Scheme 5. The Guerbet reaction network with the numbering that is also base in the simulation to simplify the reaction mixture. 55 employed in the MatLab simulation script. As the hemiacetal 1,1-eth- Now, all rate coefficients of the reaction steps can be oxyethanol and acetaldehyde are only described by one concentration 56 described with a kinetic model on an experimental basis. For an parameter in the simulation, the hemiacetal was excluded from the scheme. 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 187/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

2.7. Screening of Reaction Parameters in Simulation enough base is present. Meanwhile, the equilibrium between 1 ethanol and ethyl acetate is established. A closer look reveals 2 The behavior of the Guerbet reaction network is investigated that the (side) products are produced in the order ethyl acetate, 3 under the following aspects: butanol and oligomer and lastly acetate. This observation 4 1. A grid of hydrogen pressures and reaction temperatures is coincides with the experimental results. The only difference is 5 varied in the window from 8 bar and 100 °C to 116 bar and that no water build-up can be observed in the simulation, 6 160 °C in equidistant steps. All points of the grid are also because the saponification reaction rate may be overestimated 7 applied in combination with the other variations. at elevated temperatures. Thus, an additional acceleration of 8 2. The starting concentration of the base potassium ethanolate the saponification or aldol condensation steps due to the 9 is set to 0.8 molLÀ 1, 1.6 molLÀ 1, 2.4 molLÀ 1, 3.2 molLÀ 1 and presence of water does not arise. An interesting aspect is that 10 5.0 molLÀ 1. butanol is formed with higher selectivity at later reaction time, 11 3. Water (1.5 molLÀ 1), potassium acetate (0.6 molLÀ 1) or a because the ratio of redox catalyst to base continuously rises 12 combination of both is added. and thus the redox steps are favored over the base catalyzed 13 4. Ethyl acetate (1/5 molar amount of ethanol) is added as a steps. 14 hydrogen acceptor. For a comparison of all simulated experiments of the 15 The results of a single simulation run are the concentration temperature-pressure-grid, the butanol concentration after 5 h 16 profiles of all present species under constant temperature and reaction time is extracted. The Figures 11 and 12 show that 17 hydrogen pressure. An example of a simulated reaction three main domains are existent: 18 progress starting from ethanol, the catalyst and the base alone A) At high hydrogen pressures or low temperatures the 19 is shown in Figure 10. butanol yield is near zero, since the initial dehydrogenation 20 Ethanol is converted to the target product butanol and the of ethanol is prohibited. 21 side products acetate and acetaldehyde oligomer, as long as 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 Figure 11. Division of the temperature/hydrogen pressure-parameter space 37 for the Guerbet reaction. In area A, no butanol is formed since the 38 dehydrogenation of ethanol is suppressed. In area B, the butanol yield is 39 maximized as the kinetics of all single steps gear into each other. In area C, no butanol is formed as the reaction rate of the aldol condensation 40 outperforms the hydrogenation of crotonaldehyde. 41 42 43 44 45 46 47 48 49 50 51 52 53 54 Figure 10. Simulated concentration profiles of the Guerbet reaction starting 55 from ethanol, Ru-MACHO-BH and 1.6 molLÀ 1 potassium ethoxide (160 °C, 104 bar hydrogen pressure). The upper plot includes all organic species, the Figure 12. Temperature-pressure-screening of the Guerbet reaction with 56 À lower plot shows all organic species excluding the substrate ethanol. respect to butanol yield at 1.6 molL 1 potassium ethoxide concentration. 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 188/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

B) Here, the butanol yield has a maximum along a pressure- between conversion of ethanol, conversion of base, butanol 1 temperature-curve, but no sharp maximum can be identi- yield, butanol selectivity and the initial concentrations are 2 fied. shown in Table 3. As ethanol and ethyl acetate span a reaction 3 C) At low hydrogen pressures or high temperatures the equilibrium independently from the Guerbet reaction, both 4 butanol yield again approaches zero, since the reaction rate substances are summed up as residual substrate for the 5 of the aldol condensation outperforms the hydrogenation prediction of the selectivity. 6 of crotonaldehyde. An increase of the base concentration leads to an increase 7 In domains B and C, a majority of the base is converted to of conversion of ethanol, as the irreversible reaction steps are 8 acetate. As no accumulation of water can be observed, the accelerated and more water can be captured, until the base is 9 conversion of ethanol is connected to the initial base concen- consumed. As both side reactions are accelerated, the selectiv- 10 tration. There is no parameter combination that suppresses the ity of butanol drops. At very high base concentrations, even the 11 saponification, so a useful parameter to describe the selectivity yield of butanol drops. Furthermore, the base itself also acts as 12 is the weighted ratio of butanol and acetaldehyde oligomer a salt and less hydrogen can be dissolved in the liquid phase. 13 concentration 2c =4c . Figure 13 shows the plot of This leads to a shift of the domain B to higher hydrogen 14 BuOH C8À Oligomer the selectivity ratio versus the temperature and hydrogen pressures and lower temperatures. 15 pressure. The maximum is located at high hydrogen pressures An addition of potassium acetate leads to similar effects as 16 and low temperatures, but there is no parameter combination the increase of the base concentration. The higher salt 17 in domain B that favors butanol formation over oligomer concentration decreases the hydrogen solubility and shifts 18 formation. Again, the selectivity ratio shows no global max- domain B to higher hydrogen pressures and lower temper- 19 imum along the maximum butanol yield curve. Thus, no atures. Additionally, more salt accelerates the aldol condensa- 20 parameter set of the pressure-temperature-grid can be declared tion which diminishes the butanol yield and selectivity. 21 as optimal combination. Water is disadvantageous in the Guerbet reaction, as it 22 The further parameter variations (base-, salt-, water concen- increases the saponification reaction rate and decreases butanol 23 tration, addition of ethyl acetate) did not lead to a sufficient yield and selectivity. The position of domain B is not changed 24 increase of the butanol yield that would make Ru-MACHO a by the presence of water. 25 useful catalyst for the Guerbet reaction of ethanol. Plots In the limits of the simulation, ethyl acetate has no big 26 equivalent to Figure 12 all show a similar behavior with the influence on the Guerbet reaction. Ethyl acetate and ethanol 27 domains A, B and C. Position and height of domain B are quickly approach the hydrogenation equilibrium due to the 28 sensitive to the initial concentrations. Qualitative correlations infinite hydrogen reservoir, whereby the reaction cannot be 29 observed under hydrogen deficient conditions. For this pur- 30 pose, another reactor model with a closed mass balance of gas 31 and liquid phase needs to be included in the simulation. In the 32 very beginning, the high ethyl acetate concentration leads to 33 an increase of saponification reaction rate which in turn leads 34 to a small drop of butanol yield and selectivity. 35 36 37 2.8. Simulation of a Modified Catalyst 38 39 The preceding results show clearly that no parameter variation 40 leads to high butanol yields with the use of Ru-MACHO as a 41 redox catalyst. A main disadvantage is that the saponification 42 reaction proceeds as soon as water is present. The publications 43 of Jones et al.[7] and Cavani et al.[4] show that Guerbet reactions 44 with a “water tolerant” catalyst are accessible. The most likely 45 Figure 13. Relative selectivities of butanol and acetaldehyde-polymer in feature of these catalysts is a higher activation barrier of the 46 dependence of hydrogen pressure and temperature in area B. A maximum can be seen in direction of high pressure and low temperature, but the ratio dehydrogenation of 1-ethoxyethanol (acetaldehyde hemiacetal) 47 is always below 1. in comparison to the activation barrier of Ru-MACHO for this 48 step. The raise of this activation barrier would lead to a 49 depletion of the ethyl acetate concentration and so to a 50 Table 3. Qualitative influences of the starting concentrations of the side suppression of the saponification. To test this hypothesis, 51 products on conversion, yield, selectivity and position of area B. another simulation series was performed with an incremental 52 Increase Influence on c X Y S p ðBÞ TðBÞ increase of the activation enthalpy for step 3_1 (ESI Table S12). 53 A EtOH BuOH BuOH H2 An increase of the activation enthalpy up to 10 kJmolÀ 1 leads to 54 KOEt " " # " # KOAc � # # " # no significant improvement, but more oligomer is formed. An 55 À 1 H O � # # � � increase of the activation enthalpy of 15À 20 kJ mol leads to a 56 2 EtOAc � # # � � significant increase of the butanol yield, but the selectivity 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 189/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

drops dramatically. The simulated butanol yield and selectivity In reverse, this feature may be used for the design of new 1 do not reach reported values of successful catalysts, since the Guerbet redox catalysts. 2 rate coefficients of aldol condensations leading to C -oligom- 3 6 + ers are overestimated in the model. Figure 14 shows that under 4 these conditions water accumulates in the reaction mixture – 3. Conclusions 5 the Guerbet reaction “tolerates” water. Further increase of the 6 activation enthalpy does not lead to further changes of butanol As a metaphor, the Guerbet reaction network can be described 7 yield and selectivity, as the ethyl acetate formation and the as a 2D-ball-in-a-maze with two axes of rotation (Figure 15). In 8 saponification are shut down completely. Under these con- this maze, paths along the vertical direction are the redox steps 9 ditions, the “uncatalyzed” Cannizzarro reaction of acetaldehyde while paths along the horizontal direction depend on the base. 10 should be included into the model to depict the behavior in Between two substances, a hill has to be overcome that 11 reality. corresponds to the activation barrier. The relative height of the 12 These results show that a high activation barrier for the substances illustrates the reaction enthalpy. 13 dehydrogenation of the acetaldehyde hemiacetal 1-eth- This ball-in-a-maze-game allows playing the Guerbet reac- 14 oxyethanol is an important descriptor of a good Guerbet-redox- tion. The ball is set on the ethanol spot in the beginning and 15 catalyst. Another proof would be the comparison of calculated the axes are moved. The product of the reaction is determined 16 activation barriers for this step and the dehydrogenation of by the spot where the ball stops. Conversion as well as product 17 ethanol for a series of several (un-)successful Guerbet catalysts. distribution are determined if one plays the game over and 18 over and counts the events of finishing spots. 19 With the figure of the maze, some relations of the Guerbet 20 reaction network can be described. The axis can be independ- 21 ently moved more easily, the more redox catalyst or base are 22 induced. A higher temperature allows rotating the axis to 23 higher degrees. The hydrogen pressure leads to a tilt of the 24 mean position in direction of saturated alcohols or ethyl 25 acetate. The water concentration leads to a tilt of the mean 26 position in direction of acetate or acetaldehyde oligomer. At 27 low temperatures or high hydrogen pressures, the ball is fixed 28 to the ethanol spot. In contrast, no butanol is formed either at 29 high temperatures and low hydrogen pressures, because the 30 ball rolls through to the oligomer after one aldol condensation. 31 For the formation of butanol, the ball has to follow the ideal 32 path (dehydrogenation, aldol condensation, and hydrogena- 33 tion). A wrong turn would lead to the ball falling into a “hole” 34 of acetate or acetaldehyde oligomer. Thus, a balance of all 35 Figure 14. Simulated concentration profiles of the Guerbet reaction starting from ethanol, a modified catalyst with 15 kJmolÀ 1 higher activation barrier reaction parameters is important. A high activation barrier of 36 À 1 for step 3_1 and 2.4 molL potassium ethoxide (134 °C, 116 bar hydrogen the dehydrogenation of the acetaldehyde hemiacetal 1-eth- 37 pressure). In contrast to simulations with an unmodified catalyst, water is oxyethanol helps to stay on the ideal path. 38 accumulated with ongoing reaction. The labelled concentrations belong to water (1.17 molLÀ 1) and butanol (0.48 molLÀ 1) at the end of the simulation. 39 40 Experimental Section 41 42 All IR-spectra were recorded using a Bruker Matrix-MF spectrometer 43 with a diamond ATR-IR-probe exhibiting two internal reflection elements (infrared fiber sensors Aachen). The apparent measuring 44 parameters are listed in ESI Table S1. Due to the fast reaction rate in 45 saponification and aldol condensation, the number of scans was 46 reduced here to 10, while for all other experiments 100 scans were 47 averaged in one measurement. The sources and purities of the 48 used chemicals are listed in ESI Table S2. 49 To enable quantification of IR spectra, samples with known 50 concentration are prepared at room temperature and measured 51 with the number of scans employed in the kinetic experiments. As 52 chemometric models either peak integration (PI) or indirect hard 53 modelling (IHM) were applied using PEAXACT 4 (S-PACT, ESI chapter 3). It was taken care that the systematic errors or biases 54 were close to zero so that the predicted concentrations are not 55 over- or underestimated on average. The ratio between root mean Figure 15. Representation of the Guerbet reaction network as a ball-in-a- 56 maze game. square errors of cross-validation (RMSECV) and applied concen- 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 190/191] 1 Full Papers Chemistry—Methods doi.org/10.1002/cmtd.202000056

tration in calibration was always less than 2%. Even in the adverse [7] N. V. Kulkarni, W. W. Brennessel, W. D. Jones, ACS Catal. 2018, 8, 997– 1 case that this ratio rises to 10% under experimental conditions, it 1002. 2 can be assumed that derived quantities such as rate coefficients are [8] H. Aitchison, R. L. Wingad, D. F. Wass, ACS Catal. 2016, 6, 7125–7132. 3 specified with a relative error of 15–20%.[17] [9] Y. Xie, Y. Ben-David, L. J. W. Shimon, D. Milstein, J. Am. Chem. Soc. 2016, 138, 9077–9080. 4 The experiments of aldol condensation and saponification were [10] S. Fu, Z. Shao, Y. Wang, Q. Liu, J. Am. Chem. Soc. 2017, 139, 11941– 5 performed in a well-stirred 100 mL glass beaker. The IR-probe was 11948. [11] K.-N. T. Tseng, S. Lin, J. W. Kampf, N. K. Szymczak, Chem. Commun. 2016, 6 guided through an enclosure made from aluminium foil and 52, 2901–2904. 7 immersed into the reaction mixture. After heating to reaction [12] G. R. M. Dowson, M. F. Haddow, J. Lee, R. L. Wingad, D. F. Wass, Angew. 8 temperature and dissolution of potassium hydroxide and eventually Chem. Int. Ed. 2013, 52, 9005–9008; Angew. Chem. 2013, 125, 9175– 9 potassium acetate, acetaldehyde or ethyl acetate were injected and 9178. the continuous IR measurement was started. The saponification [13] G. R. M. Dowson, M. F. Haddow, J. Lee, R. L. Wingad, D. F. Wass, Angew. 10 reactions were stopped as soon as the IR spectra did not change in Chem. 2013, 125, 9175–9178; Angew. Chem. Int. Ed. 2013, 52, 9005– 11 structure for several minutes. The aldol condensation reactions 9008. 12 were stopped as soon as the reaction mixture turned yellow and [14] S. Veibel, J. I. Nielsen, Tetrahedron 1967, 23, 1723–1733. [15] H. Machemer, Angew. Chem. 1952, 64, 213–220. 13 obviously polymeric products are formed. [16] T. Eifert, M. A. Liauw, React. Chem. Eng. 2016, 1, 521–532. 14 All experiments with the redox catalyst Ru-MACHO(-BH) were [17] A. Ohligschläger, C. Gertig, D. Coenen, S. Brosch, D. Firaha, K. Leonhard, M. A. Liauw, Chem. Eng. J. 2019, 368, DOI 10.1016/j.cej.2019.02.195. 15 conducted under inert (transfer hydrogenation) or hydrogen [18] M. Rößler, P. U. Huth, M. A. Liauw, React. Chem. Eng. 2020, 5, 1992–2002. 16 atmosphere (hydrogenation of ethyl acetate, experiments of the [19] S. Hardy, I. M. de Wispelaere, W. Leitner, M. A. Liauw, Analyst 2013, 138, 17 whole network) in a 20 mL window autoclave that was designed for 819–824. [19] 18 multiphase-systems in combination with in-situ-spectroscopy. [20] R. A. Farrar-Tobar, Z. Wei, H. Jiao, S. Hinze, J. G. de Vries, Chem. A Eur. J. Small amounts of potassium-tert-butanolate were premixed with 2018, 24, 2725–2734. 19 ethanol or ethanol/ethyl acetate mixtures while great amounts [21] X. Chen, Y. Jing, X. Yang, Chem. A Eur. J. 2016, 22, 1950–1957. 20 were put in the autoclave before assembly. [22] J. Neumann, C. Bornschein, H. Jiao, K. Junge, M. Beller, Eur. J. Org. Chem. 21 2015, 5944–5948. Further details and applied amounts of chemicals are listed in ESI [23] K. S. Rawat, S. C. Mandal, P. Bhauriyal, P. Garg, B. Pathak, Catal. Sci. 22 chapter 4. Technol. 2019, 9, 2794–2805. 23 [24] W. Kuriyama, T. Matsumoto, O. Ogata, Y. Ino, K. Aoki, S. Tanaka, K. 24 Ishida, T. Kobayashi, N. Sayo, T. Saito, Org. Process Res. Dev. 2012, 16, 166–171. 25 Acknowledgements [25] E. Alberico, A. J. J. Lennox, L. K. Vogt, H. Jiao, W. Baumann, H.-J. Drexler, 26 M. Nielsen, A. Spannenberg, M. P. Checinski, H. Junge, M. Beller, J. Am. 27 Chem. Soc. 2016, 138, 14890–14904. We thank Niklas Westhues and Jürgen Klankermayer for the [26] A. Kaithal, M. Schmitz, M. Hölscher, W. Leitner, ChemCatChem 2019, 11, 28 initiation of this research. We gratefully appreciate the support 5287–5291. 29 [27] A. Kaithal, M. Schmitz, M. Hölscher, W. Leitner, ChemCatChem 2020, 12, received from the Federal Ministry of Education and Research 30 781–787. (Bundesministerium für Bildung und Forschung, BMBF, [28] A. Anaby, M. Schelwies, J. Schwaben, F. Rominger, A. S. K. Hashmi, T. 31 03EK3041D, Verbundvorhaben Carbon2Chem-L4: SynAlk – Teilpro- Schaub, Organometallics 2018, 37, 2193–2201. 32 [29] J. P. Guthrie, Can. J. Chem. 1974, 52, 2037–2040. jekt ‘Herstellung von C2+-Alkoholen auf Basis von H2, CO und 33 [30] A. Scheithauer, T. Grützner, C. Rijksen, D. Zollinger, E. von Harbou, W. R. CO2 aus Kuppelgasen’). Thiel, H. Hasse, Ind. Eng. Chem. Res. 2014, 53, 8395–8403. 34 [31] J. B. Anderson, M. S. Peters, J. Chem. Eng. Data 1960, 5, 359–364. 35 [32] A. Scheithauer, T. Grützner, D. Zollinger, E. von Harbou, H. Hasse, Chem. 36 Eng. Process. Process Intensif. 2016, 101, 101–111. Conflict of Interest [33] C. Zhang, J. Zhang, G. Luo, J. Flow Chem. 2020, 10, 219–226. 37 [34] J. Keybl, K. F. Jensen, Ind. Eng. Chem. Res. 2011, 50, 11013–11022. 38 [35] K. C. Aroh, K. F. Jensen, React. Chem. Eng. 2018, 3, 94–101. The authors declare no conflict of interest. 39 [36] S. Marre, A. Adamo, S. Basak, C. Aymonier, K. F. Jensen, Ind. Eng. Chem. Res. 2010, 49, 11310–11320. 40 [37] J. Setschenow, Zeitschrift für Phys. Chemie 1889, 4 U, DOI 10.1515/zpch– 41 Keywords: homogeneous catalysis · solvent effects · 1889–0409. 42 sustainable chemistry · reaction kinetics · time-resolved [38] K. Mishima, M. Sakemi, M. Nagatani, S. Yonezawa, Y. Arai, Kagaku Kogaku Ronbunshu 1992, 18, 732–739. 43 spectroscopy [39] P. W. Atkins, Physikalische Chemie, Wiley-VCH, Weinheim, 2006. 44 [40] J. T. Denison, J. B. Ramsey, J. Am. Chem. Soc. 1955, 77, 2615–2621. 45 [41] J. E. Rekoske, M. A. Barteau, Ind. Eng. Chem. Res. 2011, 50, 41–51. [1] M. Guerbet, Comptes rendus l’Académie des Sci. 1899, 128, 511–513. [42] R. Gómez-Bombarelli, E. Calle, J. Casado, J. Org. Chem. 2013, 78, 6868– 46 [2] A. Elfasakhany, A.-F. Mahrous, Alexandria Eng. J. 2016, 55, 3015–3024. 6879. 47 [3] R. Sharma, D. Kumar, M. Chhabra, G. Dwivedi, 2019, pp. 605–612. 48 [4] R. Mazzoni, C. Cesari, V. Zanotti, C. Lucarelli, T. Tabanelli, F. Puzzo, F. Passarini, E. Neri, G. Marani, R. Prati, F. Viganò, A. Conversano, F. Cavani, 49 ACS Sustainable Chem. Eng. 2019, 7, 224–237. 50 [5] K. Koda, T. Matsuura, Y. Obora, Y. Ishii, Chem. Lett. 2009, 38, 838–839. 51 [6] S. Chakraborty, P. E. Piszel, C. E. Hayes, R. T. Baker, W. D. Jones, J. Am. Manuscript received: October 28, 2020 Chem. Soc. 2015, 137, 14264–14267. Version of record online: February 24, 2021 52 53 54 55 56 57

Wiley VCH Freitag, 26.03.2021 2104 / 195694 [S. 191/191] 1