Moritz Otto
Haus
,
Yannik
Louven
and
Regina
Palkovits
*
Institut für Technische und Makromolekulare Chemie (ITMC), RWTH Aachen University, Worringerweg 2, Aachen, DE-52074, Germany. E-mail: palkovits@rwth-aachen.de
First published on 16th October 2019
The sustainable production of polymers from biogenic platform chemicals shows great promise to reduce the chemical industry's dependence on fossil resources. In this context, we propose a new two-step process leading from dicarboxylic acids, such as succinic and itaconic acid, to N-vinyl-2-pyrrolidone monomers. Firstly, the biogenic acid is reacted with ethanolamine and hydrogen using small amounts of water as solvent together with solid catalysts. For effective conversion, the optimal catalyst (carbon supported ruthenium) has to hold the ability of activating H2 as well as (imide) CO bonds. The obtained products, N-(2-hydroxyethyl)-2-pyrrolidones, are subsequently converted in a continuous gas phase dehydration over simple sodium-doped silica, with excellent selectivity of above 96 mol% and water as the sole by-product. With a final product yield of ≥72 mol% over two process steps and very little waste due to the use of heterogeneous catalysis, the proposed route appears promising – commercially as well as in terms of Green Chemistry.
The reduction of biogenic acids to valuable chemicals, such as diols, is well researched.3–9 However, selective transformations are often hard to achieve and mixtures of reduced products are obtained. Another approach that effectively utilizes the acid functionality presents the conversion into pyrrolidones by reaction with amines. In this context, the reductive amination of levulinic acid has found substantial attention.10–13 Here, the carbonyl functionality of the substrate readily forms an imine intermediate with the added amine. Consecutive reduction, even at comparably low hydrogen pressure and temperature, yields cyclic pyrrolidones, which may find applications as solvents or pharmaceutical intermediates. Touchy et al.10 tested the conversion of levulinic acid with n-octylamine and found promoted platinum catalysts holding superior performance in the synthesis of N-alkylpyrrolidones. Their finding that the acid properties of the promoting metal oxide play a prominent role in determining catalyst activity agrees well with the assumption that imine formation is the rate determining step. This hypothesis was later supported by research on TiO2-supported platinum catalysts, where decoration of the metal surface with TiOx proved to enhance catalyst acidity and activity.12 Using ethyl levulinate as substrate, a plethora of amines, including alkyl-, phenyl-, ether- and hydroxyl-functionalities, was shown to be suitable for reductive amination with good yields.
Dicarboxylic acids, such as succinic and itaconic acid, are not susceptible to reductive amination as they lack the necessary carbonyl group. Instead they form amides, diamides and imides, when reacted with amines at elevated temperatures. Amide reduction has long been considered a major challenge for heterogeneous catalysis.14 Consequently, there are few publications on the successful production of pyrrolidones from dicarboxylic acids. Budroni et al.15 have elaborated on the reduction of succinic anhydride to γ-butyrolactone, a main intermediate in today's pyrrolidone production, over Au/TiO2. The one pot conversion with phenylamine to phenylpyrrolidone was also demonstrated. However, the yield was limited to <60% due to the sequential formation of the respective pyrrolidine. Other studies concerned the production of N-methyl-2-pyrrolidone (NMP) from succinate containing fermentation broth.16 Here, N-methyl-succinimide was distilled from the broth to be reduced over a commercial rhodium catalyst, yielding up to 90% NMP at 220 °C and 100 bar H2. However, the substrate scope was not expanded to allow for the production of more valuable chemicals.
Our group has recently shown the efficient conversion (90 mol% yield) of itaconic acid and ammonia to N-unsubstituted 3- and 4-methyl-2-pyrrolidones over carbon-supported ruthenium (conditions: 200 °C, 150 bar H2).17 The products of this conversion may subsequently be transformed into N-vinyl-2-pyrrolidone monomers by reaction with pressurized acetylene over a molecular potassium catalyst. The second reaction is known as part of the fossil-based value chain running from acetylene, over γ-butyrolactone and 2-pyrrolidone, to N-vinyl-2-pyrrolidone (NVP), pioneered by W. Reppe.18 NVP is polymerized to yield polyvinylpyrrolidone, a water-soluble, non-toxic polymer with growing application and demand, especially in the pharmaceutical industries.19
Despite its long-standing use, vinylation with pressurized acetylene raises concerns about its sustainability and safety. Specifically, the use of a hardly reusable catalyst and the necessity to remove traces of water to achieve effective conversion are undesirable process specifications.20,21 The most promising alternative to produce NVP is presented by the thermal reaction of ethanolamine and γ-butyrolactone. The product, N-(2-hydroxyethyl)-2-pyrrolidone (HEP) can be dehydrated in a gas-phase reaction leading to a simple, continuous production process, which is now in commercial application.22 Using conventional γ-butyrolactone as feedstock, this process remains fossil resource-based.
We herein present the reductive transformation of dicarboxylic acids with ethanolamine, leading to N-(2-hydroxyethyl)-2-pyrrolidones (reductive amidation), which are dehydrated to yield bio-based monomers analogous to NVP. Specifically, succinic acid is chosen as a model compound to study the underlying reaction network and kinetics of this transformation. Catalysts are screened for activity in the rate-determining step of the overall transformation and process conditions are optimized to show the potential for obtaining high yields of valuable products. Due to the spectrum of dicarboxylic acids obtainable from biomass, the substrate scope of the reductive transformation as well as of the continuous gas phase dehydration is explored. Overall, the envisioned two-step process (Fig. 1) yields NVP-like monomers on the basis of biogenic acids with water as sole by-product and, at the same time, solvent of the heterogeneously catalyzed reactions. Given the excellent yields obtained and the use of simple and/or commercially available catalysts, this seems to offer real potential for an applicable Green Chemistry value chain.
Fig. 1 Schematic of the green value chain proposed herein. Two steps lead from dicarboxylic acids over N-(2-hydroxyethyl)-2-pyrrolidones (HEP) to valuable N-vinyl-2-pyrrolidone monomers (NVP). (MP = methyl-2-pyrrolidone).36 |
Conversion and yield are calculated based on concentrations determined via HPLC through the assumption of mass conservation within the autoclave. For reasons that will shortly become evident, the following definitions are applied for the conversion X, yield Yi and selectivity Si:
(1) |
(2) |
(3) |
For recycling experiments the catalyst was retrieved by use of polycarbonate filter membranes after the reaction. Before re-use it was washed with 60 ml of deionized water and subsequently dried in a vacuum oven (60 °C, 14 h). These experiments were conducted with a slightly upscaled batch size (corresponding to 50 mg Ru/C) to allow for easier handling in the recycle process. Liquid samples of the reaction mixtures were diluted in 0.05 M HNO3 (1:1000) and analyzed by ICP-MS (Agilent 8800 ICP-MS Triple Quad) to quantify Ru leaching.
For experiments on imide reduction, the amount of substrate was adjusted to reflect the molar equivalent of 1.5 g of the respective acid. The amount of solvent was increased to ∼2 g, reflecting the water formed upon imide formation from the acid and ethanolamine.
(4) |
(5) |
Stability tests were performed with a reduced amount of catalyst (0.1 g), thus limiting conversion to intermediate levels.
Results for the reaction of succinic acid and ethanolamine (Fig. 2) underline the fundamental suitability of carbon-supported ruthenium for the reductive formation of N-(2-hydroxyethyl)-2-pyrrolidone (HEP). Moreover, the sequence of intermediates and products observed, allows for the construction of a consistent reaction network (Fig. 3): with increasing reaction time, the amount of acid in the solution rapidly declines, giving way to mono- and, to a lesser extent, diamides. Their subsequent cyclization to the imide limits the maximum concentration of amides observed during the course of reaction. N-(2-hydroxyethyl)succinimide shows good stability in aqueous solution and thus reaches a maximum intermediate yield close to 60% after 2 h. Its amount declines only as more and more pyrrolidone is formed by reduction.
Fig. 2 Concentration-time-profile of the reductive amidation of succinic acid with ethanolamine. (Conditions: 150 °C, 150 bar H2, 750 rpm, 1.5 g acid, 1.5 g deionized water, 1 mol. equivalent of amine, 37.5 mg Ru/C) The corresponding reaction network is outlined in Fig. 3. *HEP = N-(2-hydroxyethyl)-2-pyrrolidone, **HEBA = N-(2-hydroxyethyl)-4-hydroxybutanamide. |
Fig. 3 Proposed reaction network for the reductive amidation of succinic acid with ethanolamine. Water and ethanolamine, which are consumed and/or formed, are implicit in the network. |
Given the fast initial rate of amide and imide formation, which is also observed in the absence of catalyst, and the comparably slow conversion to reduced products, the latter step emerges as rate-determining. N-(2-hydroxyethyl)succinimide, due to its abundance and cyclic structure, is identified as the most likely substrate to be hydrogenated on the ruthenium surface.16 Moreover, N-(2-hydroxyethyl)-4-hydroxybutanamide (HEBA) – the acyclic equivalent of HEP – is observed as a second product, which initially forms at a rate proportional to that of HEP. This is tentatively assigned to their formation from a joint substrate (N-(2-hydroxyethyl)succinimide) via two competing reactions on the ruthenium surface (hydrogenolysis of CO vs. C–N bond).14 On the same note, previous publications have reported that the catalytic hydrogenation of imides may lead to lactams or hydroxyl carboxamides depending on catalyst selectivity and process conditions.29
At any rate, HEBA formation from HEP via hydrolysis was excluded in separate experiments on product stability under reaction conditions. On the other hand, the apparent decline in N-(2-hydroxyethyl)-4-hydroxybutanamide formation in later reaction stages is rationalized by the conversion to HEP through intramolecular condensation. This reaction is known to proceed without catalyst at elevated temperatures,30 wherefore the HEBA formed here is not lost to the production of HEP and may be considered a valuable product.
Fig. 4 Screening of catalytic metals for N-(2-hydroxyethyl)succinimide hydrogenation. (Cond.: 150 °C, 150 bar H2, 6 h, 750 rpm, 37.5 mg catalyst). |
Carbonyl groups in amides and, by analogy, also in imides are known to be extremely stable against reduction due to their low electrophilicity.14,31 Consequently, any metal with the ability to activate the imide CO functionality would be superior for the tested reaction. While platinum for example has an outstanding ability to activate hydrogen, ruthenium combines the potential for hydrogen activation with that for CO-bond activation, as has recently been underlined by experimental32 and theoretical35 considerations. This combination then proves beneficial for the reduction of imides, as is underlined by the presented data.
Ruthenium as the most promising metal may then be supported on different carrier materials, possibly influencing the catalytic performance (Table 1).26 It is observed that zirconia and alumina lead to an increase in the overall yield of valuable products (HEP and HEBA), whereas yields stay the same over Ru/TiO2 and decrease for Ru/SiO2. Since differences such as the above often originate from metal dispersion and/or support acidity, these catalyst properties were subjected to further study (see ESI,† p. 5). While slight variations of both properties were evident in our set of oxide-supported catalysts, a simple correlation with catalyst activity was not found. Further studies to elucidate these aspects are ongoing. In the meantime, commercial Ru/C already offers promising catalytic activity and is thus used for the further development of the targeted value chain. The obtained trends are expected to be transferable to a range of support materials, excluding specific cases of strong metal-support-interaction and bifunctionality due to a direct participation of the support in the ongoing reaction.31
Furthermore, a positive and linear correlation is found between hydrogen pressure (50–150 bar) and the desired product yield (Fig. 5b). While this effect might be attributed to gas-liquid mass transfer limitations, results obtained for the reduction of itaconic acid in the same setup show that hydrogen mass transfer is orders of magnitude faster than the reductive amidation studied herein.17 The results from experiments conducted at different stirring intensities and catalyst loadings further support this (see Fig. S2†). It is therefore possible that p(H2) exerts its influence on the reaction by inducing a higher coverage of the catalysts metal surface with activated hydrogen (θH), which benefits the ongoing reduction. A slight change in product selectivity hints towards the different degree of hydrogen influence in desired and undesired reaction pathways.
Further experiments focused on the dependence of reductive amidation on substrate concentration cS, which was modified through the amount of water added to the reaction mixture. Following common adsorption theory, lower amounts of water (higher cS) increase the substrate surface coverage θS of the catalyst and thus influence the rate of surface reactions. Assuming the latter to be rate-determining, product yields after a specified time vary accordingly. However, no significant change in product yields is observed in the respective data (Fig. 5c). Notwithstanding future in depth analysis, this may be associated with a decoupling of cS and θS, i.e. saturation of the substrate adsorption isotherm. In the case at hand, this means that the chosen solvent (water) does not compete effectively with the organic substrate for the adsorption sites on Ru/C, wherefore the surface is likely covered with organic moieties. This, in turn, agrees well with the behavior of carbon-supported Ru in the aqueous phase hydrogenation of acetic acid with molecular hydrogen.32
To further elucidate the role of substrate adsorption on Ru/C, the imide structure to be reduced in our experiments was systematically varied (Table 2). It is interesting to see that methyl substituents attached to the carbon atoms of the succinimide cycle hardly have any influence on the yield of pyrrolidone (compare entries 1 and 2 or 3 and 4). However, adding an ethyl-substituent to the imide N-atom (entries 4 and 5), which greatly increases the thermodynamic stability of the succinimide in an aqueous environment, doubles the pyrrolidone yield obtained under otherwise equivalent conditions. This may further underline the importance of the respective imides in the rate determining step of reductive amidation. Finally, the fastest conversion rates are obtained for substrates containing an additional hydroxyl group derived from ethanolamine. Since the hydroxyl group, due to its position, is unlikely to have a direct electronic effect on the carbonyl to be reduced, the effect is tentatively assigned to a change in substrate adsorption.
Dicarboxylic acid substrate | Conversion [mol mol−1] | Y (Pyrrolidone)a [mol mol−1] | Y (Butanamide)b [mol mol−1] |
---|---|---|---|
a Equivalent of HEP in the case of succinic acid. b Equivalent of HEBA in the case of succinic acid. | |||
Succinic acid (150 °C) | 0.75 | 0.40 | 0.17 |
Succinic acid (200 °C) | 0.98 | 0.74 | 0.01 |
Methylsuccinic acid (150 °C) | 0.81 | 0.50 | 0.10 |
Methylsuccinic acid (200 °C) | 0.88 | 0.74 | 0.00 |
Itaconic acid (150 °C) | 0.73 | 0.39 | 0.06 |
Itaconic acid (200 °C) | 0.96 | 0.39 | 0.02 |
The conversion of succinic acid is especially successful, yielding more than 60 mol% of valuable products at 150 °C. Utilizing the improved rates of imide hydrogenation and HEBA condensation at 200 °C this value is increased to 75 mol%, while maximizing the final HEP content at the same time. Further enhancements are then limited by the chemoselectivity of the catalyst, which still allows for the formation of N-ethyl-substituted by-products and substrate hydrogenation without nitrogen incorporation (e.g. formation of γ-butyrolactone and butanols).
These results are well transferable to methylsuccinic acid, reaching up to 74 mol% yield of N-(2-hydroxyethyl)-methylpyrrolidones. For itaconic acid, however, yields are consistently lower than for the other two acids, despite comparable conversion levels. This is rationalized by the possibility of additional reactions taking place on the methylene functionality prior to its reduction. More specifically, ethanolamine is known to react with itaconic acid in a thermally-activated aza-Michael addition yielding N-(2-hydroxyethyl)-2-pyrrolidone-4-carboxylic acid – another monomer from biomass.33 The resulting undesired competition between addition and reductive amidation may, however, be circumvented by the selective reduction of itaconic acid to methylsuccinic acid at mild conditions previous to the introduction of ethanolamine. For example, the electrochemical production of methylsuccinic acid from itaconic acid containing fermentation broth has been reported.34
Finally, the viability of reductive amidation as a production tool depends on the stability of the noble metal catalyst. In this context, no signs of catalyst deactivation are observed during five consecutive batch reactions (Fig. 6). After a slight increase in the yield of desired products from the first to the second use cycle of a catalyst batch, further recycling experiments show constant catalytic performance. ICP-MS analysis of the filtered product solution indicates a small and decreasing fraction of ruthenium leaching from the catalyst over sequential use cycles. However, the observed stability of catalyst performance would indicate a negligible contribution of the leached species to the ongoing reduction. Finally, it was tested whether the removal of Ru/C from an ongoing reaction precludes further substrate conversion (Fig. S3†). Since hydrogenation reactions are completely suppressed after the filtration step, the heterogeneous nature of the applied catalyst is verified.
Fig. 7 Simplified schematic of the gas phase dehydration of N-(2-hydroxyethyl)pyrrolidones in lab-scale. Vinylpyrrolidone monomers are obtained after condensing the product stream. |
Firstly, the functionality of a simple Na2O/SiO2 catalyst (1:20 molar ratio)21 was verified using commercially available N-(2-hydroxyethyl)-2-pyrrolidone as test substrate (Table 4). Here, 2-pyrrolidone appears as main side-product, due to the undesired cleavage of the C–N bond. However, 2-pyrrolidone formation can be mitigated by the choice of reaction conditions. In detail, the right balance between residence time and reaction temperature has to be achieved, so that the gas mixture leaves the catalyst bed immediately after full conversion is reached. A positive example of this is obtained at WHSV = 2.4 gsubstrate gcat−1 h−1 and T = 350 °C, where the reaction yields 95 mol% N-vinyl-2-pyrrolidone. The stability of the catalytic material at reaction conditions was further assessed by running the setup for 8 h at constant operating conditions (Fig. S6†). No obvious deactivation is observed in line with a very limited amount of catalyst coking evidenced by thermogravimetric analysis.
Temperature [°C] | WHSVa [h−1] | Conversion [mol mol−1] | Y (NVP) [mol mol−1] | Y (2-pyrrolidone) [mol mol−1] |
---|---|---|---|---|
a Weight hourly space velocity on a gsubstrate gcatalyst−1 basis. | ||||
350 | 1.2 | 1.00 | 0.83 | 0.16 |
350 | 2.4 | 0.99 | 0.95 | 0.03 |
350 | 3.6 | 0.89 | 0.86 | 0.01 |
400 | 1.2 | 1.00 | 0.78 | 0.15 |
400 | 2.4 | 1.00 | 0.93 | 0.05 |
400 | 3.6 | 1.00 | 0.96 | 0.02 |
Further tests concern the applicability of the same catalyst and process conditions to the dehydration of methyl-N-(2-hydroxyethyl)-2-pyrrolidones, which are obtained by the conversion of itaconic and methylsuccinic acids. To satisfy the need for relatively large substrate quantities in order to operate the continuous setup, 3-methyl-N-(2-hydroxyethyl)-2-pyrrolidone was synthesized by thermally reacting 3-methyl-γ-buytrolactone and ethanolamine (see ESI†). The desired product (≥95% NMR, see Fig. S7†) is obtained after vacuum distillation. Subsequent gas phase dehydration on the continuously operated setup yields 75 mol% of 3-methyl-N-vinyl-2-pyrrolidone at previously optimized conditions (WHSV = 2.4 gsubstrate gcat−1 h−1 and T = 350 °C). Since the main by-product is 3-methyl-2-pyrrolidone, a reduction in residence time through an increase in substrate flow rate (WHSV = 3.6 gsubstrate gcat−1 h−1) increases yields to 90 mol% of vinylpyrrolidone. Overall, the setup was operated continuously for 7 h converting 20 gsubstrate gcatalyst−1 (see Fig. S10†). The collected product mixture was vacuum distilled leading to the facile separation of 3-methyl-N-vinyl-2-pyrrolidone (≥95% NMR, see Fig. S8†), which may be used for polymer synthesis. It thus appears that an industrial process could make use of rather simple and efficient separation techniques.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c9gc01488h |
This journal is © The Royal Society of Chemistry 2019 |