Selectivity enhancement in the catalytic hydrogenation of propionitrile using ionic liquid multiphase reaction systems

Applied Catalysis A: General 356 (2009) 43–51

Contents lists available at ScienceDirect

Applied Catalysis A: General journal homepage: www.elsevier.com/locate/apcata

Selectivity enhancement in the catalytic hydrogenation of propionitrile using ionic liquid multiphase reaction systems Katharina Obert a, Daniel Roth a, Martin Ehrig a, Andreas Scho¨nweiz a, Daniel Assenbaum a, Harald Lange b, Peter Wasserscheid a,* a b

Lehrstuhl fu¨r Chemische Reaktionstechnik, Universita¨t Erlangen-Nu¨rnberg, Egerlandstr. 3, D-91058 Erlangen, Germany Lehrstuhl fu¨r Bioverfahrenstechnik, Universita¨t Erlangen-Nu¨rnberg, Paul-Gordan-Str. 3, D-91054 Erlangen, Germany

A R T I C L E I N F O

A B S T R A C T

Article history: Received 11 November 2008 Received in revised form 8 December 2008 Accepted 9 December 2008 Available online 24 December 2008

A new ionic liquid based, multiphase reaction system was investigated for the catalytic hydrogenation of propionitrile. The reaction system under investigation consisted of a solid heterogeneous Ru-catalyst (Ru on carbon), two liquid phases and the hydrogen gas-phase. By using two liquid phases in the reactor – of which one was an ionic liquid – it was possible to improve significantly the selectivity for the formation of propylamine by suppressing consecutive reactions to dipropylamine and tripropylamine. This enhanced selectivity resulted either from the protonation of the primary formed amine (in case of the Brønsted acidic ionic liquids dimethylcyclohexylammonium hydrogensulfate and 1-butylimidazolium hydrogensulfate) or from the extraction of the primary amine into the organic phase (in case of a 1-ethyl3-methylimidazolium ethylsulfate/1,2,4-trichlorobenzene biphasic system). ß 2008 Elsevier B.V. All rights reserved.

Keywords: Propylamine Hydrogenation Ionic liquid Biphasic

1. Introduction Amines constitute an important family of industrial chemicals and find use as e.g. building blocks for the synthesis of pharmaceuticals, agro-chemicals, polymers, detergents, disinfectants, stabilisers and corrosion inhibitors [1,2]. Commonly, amines are produced via two different chemical routes: Either by reductive amination of carbonyl compounds or by hydrogenation of nitriles. The hydrogenation of nitriles is usually carried out using heterogeneous catalysts containing transition metals such as palladium, platinum, rhodium, ruthenium, nickel, cobalt or iron. The reaction is technically performed in batch slurry systems or continuous fixed bed reactors using alcohols as organic solvents [1,2]. While the mechanism of nitrile hydrogenation is not understood in detail it is generally accepted that the catalytic hydrogenation follows the Langmuir–Hinshelwood-mechanism with a dissociative adsorption of hydrogen [3–5]. The first step of the reaction is the formation of an imine species, which is then further hydrogenated to the technically desired primary amine. Selectivity issues arise from the fact that the imine intermediate can also react with an already formed primary amine in a reaction that liberates ammonia and forms a secondary amine. Moreover, the secondary amine itself can react with the primary imine analogous to the above-mentioned mechanism giving tertiary amine as by-product. The effective

* Corresponding author. Tel.: +49 9131 85 27420; fax: +49 9131 85 27421. E-mail address: [email protected] (P. Wasserscheid). 0926-860X/$ – see front matter ß 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.apcata.2008.12.016

network of parallel and consecutive reactions in nitrile hydrogenation is depicted in Scheme 1 [6–10]. Several approaches have been reported in literature to enhance the selectivity of nitrile hydrogenation towards the formation of the primary amine. Already in 1928, a paper by Hartung proposed the addition of strong acids to the reaction mixture to protonate the initially formed primary amine thus preventing reaction with the imine [11]. More commonly, the selectivity to primary amines is enhanced by adding ammonia. This concept was first reported by Winans [12] and builds on the fact that free ammonia in the reaction system disfavours thermodynamically the undesired reaction of imine and primary amine. Today, this method is widely applied in industry although the added ammonia has to be separated after reaction from the product, a step that requires energy intensive condensation of ammonia at low temperatures. Nitrile hydrogenation has already been studied in multiphase reaction systems. Dow Chemical Co. USA patented in 1988 a rhodium-catalysed nitrile hydrogenation in which a biphasic system of water (pH 8) and organic solvents was applied. Selectivities to the primary amine of more than 80% were disclosed as the critical advantage of the biphasic reaction system [13]. Later, Hegedus and Mathe reported a Pd-catalysed nitrile hydrogenation in a biphasic system consisting of an acidic water and a dichloromethane phase. For this approach, selectivities to the primary amine of up to 95% have been reported [14]. Ionic liquids are salts with a melting point below 100 8C, which are characterised by very low vapour pressures [15]. In the last five

44

K. Obert et al. / Applied Catalysis A: General 356 (2009) 43–51

Scheme 1. Reaction pathway for the nitrile hydrogenation to primary (main reaction; MR) secondary (side reaction 1; SR1) and tertiary (side reaction 2; SR2) amines.

years catalysis in ionic liquids has become a field of great research activity, a fact that has also created an extensive reviewing practice [15]. Obviously, there are many good reasons to study ionic liquids in catalytic reactions and liquid–liquid biphasic catalysis with homogeneously dissolved transition metal complexes. In this context the possibility of adjusting solubility properties of ionic liquids by different cation/anion combinations is an important feature. Another interesting feature arises from the fact that functional and reactive groups can be easily incorporated into ionic liquids (those ionic liquids are known as TSILs -task-specific ionic liquids [16,17]). Among the numerous reported examples, Brønsted acidic ionic liquids are of special importance [18]. The latter have even found industrial use in the BASILTM process of BASF SE [19]. Recently, a first example of the application of protic ionic liquids for the protection of primary amines has been reported by Sunitha et al. [20]. These authors treated various amines with a tert-butyl carbonyl group (Boc) in 1-methylimidazolium tetrafluoroborate which led in most cases to good yields and selectivities to the corresponding Boc-protected amines. In contrast to the large number of examples using homogeneously dissolved metal complexes in ionic liquids for catalysis, literature reports applying heterogeneous catalysts suspended in ionic liquids are rather rare. Nevertheless, the hydrogenation of some systems in ionic liquids has been reported, e.g. unsaturated aldehydes have been hydrogenated using a supported palladium catalyst [21]. Another example has been published by Claus et al. dealing with the hydrogenation of citral applying Rh-Sn/SiO2 suspended in 1-butyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide ([BMIM][NTf2]) [22]. A special form of heterogeneous hydrogenation catalysis in ionic liquids is constituted by nanoparticle catalysed reactions in which the particles are generated and stabilised by the ionic liquids. This approach has been successfully realised for the hydrogenation of arenes [23], dienes [24], aldehydes [25] and allylalcohols [26]. Recently, examples have been reported where heterogeneous hydrogenation catalysts were coated with a thin film of ionic liquid [27]. In this way the selectivity in the consecutive hydrogenation of cyclooctadiene could be manipulated. In this paper we present a new approach to enhance the selectivity in the heterogeneously catalysed nitrile hydrogenation namely the application of ionic liquid based biphasic systems to prevent the reaction of the desired primary amine with the imine intermediate. The investigated reaction systems comprise four phases: The heterogeneous ruthenium catalyst constitutes the solid phase; hydrogen as reactant represents the gaseous phase. Furthermore, an ionic liquid and an organic solvent forming a miscibility gap constitute two liquid phases. Depending on the chemical reactivity of the ionic liquid, two different approaches to improve the selectivity for primary amine formation have been studied.

Approach 1 Hydrogenation of propionitrile in the presence of a Brønstedacidic ionic liquid acting as the protonating medium for the formed primary amine. Approach 2 Hydrogenation of propionitrile in presence of a neutral ionic liquid acting as the reaction medium from which the primary amine is extracted by means of an immiscible, organic extraction phase. In Approach 1 the nitrile hydrogenation was intended to take place in the organic phase. The Brønsted-acidic phase should constitute a reactive extraction phase protonating the primary amine. Afterwards, the corresponding ammonium salt should be extracted into the ionic liquid phase. Of course, to realise this concept, it would be necessary to suspend the heterogeneous catalyst in the organic phase. This was attempted by combining an organic phase with high density and an IL phase with a relatively low density. In such a system the heterogeneous catalyst was expected to follow gravity and therefore to be suspended in the heavy organic phase (see Fig. 1). Certainly, this approach requires an efficient work-up of the ionic liquid phase (containing the propylamine product in form of its ammonium salt) after the hydrogenation reaction. Besides product isolation this work-up procedure has to regenerate the acidic ionic liquid phase for its reuse in the next batch or for its regeneration in a continuous catalytic process. In Approach 2 the hydrogenation takes place in a neutral ionic liquid that contains the suspended heterogeneous Ru-catalyst (see Fig. 2). In this approach the formed propylamine is not protected from a consecutive protonation step but simply extracted into an organic solvent showing preferential solubility for the amine compared to the nitrile. In this way it is intended to keep the concentration of primary amine in the catalytic ionic liquid phase low even at high nitrile conversion. The effective removal of primary amine from the catalyst should result in a significant suppression of the consecutive reaction forming secondary and tertiary amines. This second approach is somewhat related to a patent by BASF SE that has been published in 2005 describing in very general terms the concept to perform nitrile hydrogenation reactions in ionic liquid based slurry systems [28]. However, no detailed catalytic results have been given in the BASF patent. Thus, from the patent a proper evaluation of the ‘‘ionic liquid effect’’ in the heterogeneously catalysed nitrile hydrogenation is not possible. The results of both approaches have been compared with the ionic liquid free standard system. In order to perform this

K. Obert et al. / Applied Catalysis A: General 356 (2009) 43–51

45

2.4. Reactor set-up

Fig. 1. Intended strategy to extract primary amines from a dense organic phase with suspended Ru-catalyst by Brønsted-acidic ionic liquids.

The measurements both for the IL-free reference system and for the IL-containing system were carried out in a stirred tank autoclave (PARR Instrument GmbH, Germany) made of stainless steel and equipped with a gas entrainment stirrer, a cooling coil, and a temperature and pressure control. All reactions were carried out in semi-batch mode with hydrogen kept at a constant pressure during the experiment. The experiments for Approach 1 were carried out in a reactor vessel with 600 ml volume (PARR Instrument GmbH, Germany). The data for Approach 2 were obtained using a 300 ml vessel (PARR Instrument GmbH, Germany). 2.5. Catalyst pre-treatment

Fig. 2. Second approach to improve the selectivity in the hydrogenation of propionitrile – extraction of primary amine by an organic phase from the aprotic ionic liquid containing the suspended Ru-catalyst.

comparison in a meaningful manner baseline experiments in the ionic liquid free system have been carried out first using the same batch of Ru-catalyst, the same reaction conditions and the same autoclave equipment as later applied for the ionic liquid multiphase experiments. 2. Experimental 2.1. Chemicals Propionitrile was obtained from Acros Organics, Belgium (99%). 1,2,4-trichlorobenzene (99%), benzotrichloride (98%) and N,Ndimethylcyclohexylamine (98%+) were purchased from Alfa Aesar GmbH and Co. KG, Germany. Sulphuric acid (for analysis, > 98%) and di-n-butyl ether (for synthesis) were both obtained from Merck KGaA, Germany. All these chemicals were applied without further purification. As heterogeneous catalyst for our experiments 5 wt.% ruthenium on carbon (Heraeus GmbH, Germany) was used. The catalyst contained 50% of water. All experiments were carried out with the same batch of Ru-catalyst obtained from Heraeus GmbH, Germany. 1-Ethyl-2-methylimidazolium ethylsulfate ([EMIM][EtOSO3]) purchased from Solvent Innovation GmbH, Germany (ECOENG 212, 99%) was applied without further purification.

The catalyst was subject to a pre-treatment procedure prior to the catalytic reactions. 250 ml – resp. 125 ml for Approach 2 – of organic solvent and the Ru-catalyst were filled into the reactor and the remaining air was purged with inert gas (helium). The reactor was pressurised with 100 bar of hydrogen and heated to 50 8C. During pre-treatment, the catalyst was intensively contacted (stirrer speed: 1200 rpm) over 16 h with the hydrogen atmosphere. During this procedure the catalyst was suspended in the same organic solvent that was later used in the catalytic experiments (the nitrile substrate was absent during the preformation time). After the pre-treatment procedure, the reactor was cooled down to room temperature, depressurised, purged with He and opened. 2.6. Baseline experiments in the IL-free system For the determination of reference data in the IL-free reaction system (one liquid reaction phase only), 10 g of Ru/C was pretreated as described above. We used 1,2,4-trichlorobenzene as standard organic solvent; baseline experiments were also carried out with methanol and cyclohexane. Afterwards, 100 g of propionitrile were filled into the reactor, which was then closed, purged with inert gas, pressurised with hydrogen and heated to the desired reaction temperature. The temperature variation was performed between 80 and 120 8C at 100 bar hydrogen pressure. The corresponding pressure variation was carried out at 100 8C with pressures between 40 and 100 bar being applied. As soon as the reaction temperature was reached the stirrer was switched on and this time was set as t = 0 for the catalytic experiment. It was assumed that prior to this time almost no reaction took place as stirring is necessary to contact propionitrile, hydrogen and catalyst in the autoclave. Liquid phase samples were taken regularly and analysed via gas chromatography (see below) to determine the concentrations of all reactants in the organic liquid phase.

2.2. Syntheses of the ionic liquids 2.7. Brønsted acidic IL-containing systems – Approach 1 The ionic liquids 1-butylimidazolium hydrogensulfate ([BIM][HSO4]), dimethylcyclohexylammonium hydrogensulfate, trioctylammonium hydrogensulfate, 1-octylimidazolium hydrogensulfate were synthesised and characterised according to [29]. 2.3. Miscibility tests For the miscibility tests, 50 ml of the respective ionic liquid (see Table 3) and 100 ml of the organic solvents 1,2,4-trichlorobenzene, resp. benzotrichloride were mixed in a flask and heated to the desired temperature of 100 8C. The biphasic system was stirred for several hours.

For the measurement in the IL-containing four-phase-system, 4 g of the catalyst was pre-treated in 1,2,4-trichlorobenzene as a first step. 100 ml of the ionic liquid were added and afterwards 20 g of propionitrile was introduced into the reactor. The reactor was closed, purged, pressurised and heated. The experiments for Approach 1 were carried out at 100 bar hydrogen pressure and 100 8C. As soon as the reaction temperature was reached, the stirrer was switched on (1200 rpm) and this moment was set as t = 0 of the reaction. Organic phase samples were taken regularly and analysed via gas chromatography. Approximately 1 min before the liquid samples were taken the stirrer was switched off to allow

46

K. Obert et al. / Applied Catalysis A: General 356 (2009) 43–51

phase separation. Immediately after sampling, the stirrer was switched on again to allow the reaction to proceed. 2.8. Product analysis – Approach 1 After the experiment was completed, the reactor was cooled to room temperature, purged with inert gas, opened and the biphasic content was poured into a phase separator. Approximately 10 ml of the upper ionic liquid phase was separated for the product analysis. To these 10 ml of product – ionic liquid mixture, 30 ml of saturated aqueous sodium hydroxide solution were added and the mixture was carefully mixed to deprotonate all ammonium salts. The resulting aqueous phase was extracted with a double excess of di-n-butyl ether, the ether phase was filtrated and was then analysed by gas chromatography. 2.9. Attempts for product isolation by vacuum distillation – Approach 1 After the experiment was completed, the organic phase was removed from the reactor by bottom discharge. When only the ionic liquid phase was left in the reactor the bottom sampling valve was closed, the reactor was set to the distillation temperature and depressurised. Then vacuum of 40 mbar was applied directly to the reaction vessel and all volatile components were collected in a cooling trap. The content of the cooling trap was than analysed by GC. 2.10. Aprotic IL-containing systems – Approach 2 5 g of the ruthenium catalyst was pre-treated in 1,2,4trichlorobenzene, 55 g of the ionic liquid and finally 20 g of propionitrile was added. The reactor was closed, purged, pressurised with 100 bar hydrogen and heated to 100 8C. As soon as the desired reaction temperature was reached the stirrer was switched on at 1200 rpm, the time was set as t = 0 of the reaction. During reaction, liquid samples were taken from the organic phase. For this purpose, the stirrer was switched off approximately 1 min before the sample was taken to allow phase separation. The samples were analysed by gas chromatography.

Again, the moment when the stirrer was switched on (1200 rpm) was considered to be the starting point of the reaction. Liquid samples for GC analyses were taken regularly during reaction after phase separation. 2.13. Analytical techniques GC measurements were carried out using a Varian Type 3900 machine equipped with a flame ionisation detector. For the GC–MS measurements a Varian Saturn 2100T MS was applied. Both devices operated with helium as carrier gas and were equipped with an amine column (Optima-5-Amin-1.5 mm, length 30 m, inner diameter 0.32 mm from Macherey-Nagel). The temperature profile started at a temperature of 60 8C and was heated to 280 8C with a rate of 8 8C per minute. All measurements were carried out at a column flow of 2 ml/ min, applying an injector temperature of 250 8C and a FID temperature of 280 8C. MALDI-ToF mass spectra were acquired by an AXIMA CFR curved field reflectron time-of-flight (ToF) mass spectrometer (AXIMA CFRplus, Kratos Analytical, UK) in linear ion mode using positive ion mode. In order to irradiate the sample for desorption and ionisation a nitrogen laser (337 nm, 3 ns pulse, maximum pulse rate 10 Hz, 21.5 mJ) was used. A constant pressure of 104 Pa was prevalent over the ion source during the measurements. The acceleration voltage was set to 20 kV. Each spectrum constituted an average of 200 profiles composed of the accumulation of two single laser shots. The external calibration was carried out using 2  a-cyano-4-hydroxycinamic acid (m/z 379,34 bradykinin fragment (m/z 757.3997) and angiotensin 2 human (m/z 1,046.5423). For optimal mass accuracy delay time was set to 87.6 ns. Mass error was calculated to 200 ppm. Reported m/z values showed nominal masses (i.e., values after 1st decimal place have been truncated). Acquisition and data processing were controlled by Launchpad Software (Kratos Analytical, UK), version 2.7. For sample depositing a 384 position stainless-steel sample plate was used. 1 mL of the sample was pipetted to the target and allowed to air-dry. 3. Results and discussion

2.11. Mass balance for the aprotic IL-containing systems – Approach 2

3.1. Baseline experiments in the IL-free system

For performing a mass balance the IL (12 g resp. 8 g of [EMIM][EtOSO3]) was added to the pre-treated catalyst and the TCB. Then the reactor was closed and stirred at 5 bar helium for 2 min in order to assure that the rather small amount of IL could contact the catalyst. Then the reactor was depressurised, opened and 20 g of propionitrile was added. The reactor was closed afterwards, purged with helium, pressurised with 100 bar and heated to 100 8C. The moment when the stirrer was switched on (1200 rpm) was considered to be the starting point of the reaction. Liquid samples for GC analysis were taken from the organic phase (after phase separation).

In order to allow a proper evaluation of the effect of added ionic liquid in the hydrogenation systems under investigation, a number of baseline experiments have been carried out first. For this (and for all other experiments) the ruthenium catalyst underwent a pretreatment with hydrogen over night. This procedure was necessary as previous experiments had shown that the commercial Rucatalyst was not active when applied without the H2 treatment. As ruthenium tends to form ruthenium oxide when contacted with air the pre-treatment with hydrogen allowed reducing the Ru to its metallic form. These experiments focused mainly on the influence of the nature of the organic solvent in which the reaction was carried out (Table 1). For the envisaged Approach 1 (using a Brønsted acidic ionic liquid extraction phase) the experiment in the dense organic solvent 1,2,4-trichlorobenzene (density = 1.45 g/ml) was of special interest to see whether the chlorinated solvent would have a detrimental influence on the heterogeneous Ru-catalyst. Interestingly, the catalytic reaction in 1,2,4-trichlorobenzene resulted in the highest yield to the desired primary amine. Consequently, this solvent was chosen for more detailed kinetic investigations in the IL-free system. Fig. 3 shows how the composition of the reaction mixture developed for a typical experiment over time.

2.12. Ionic liquid/catalyst recycling experiments in Approach 2 After 9 h reaction time, the reactor was cooled down and left under 100 bar hydrogen pressure over night. The organic phase was collected from the autoclave by bottom discharge using the sampling device. When all the organic phase was removed from the autoclave the reactor was depressurised and purged with He. The reactor vessel was opened and the initial amount of solvent and reactant i.e. 125 ml TCB and 20 g propionitrile were reloaded to the reactor vessel. Finally, the reactor was closed, purged with helium, pressurised with 100 bar hydrogen and heated to 100 8C.

K. Obert et al. / Applied Catalysis A: General 356 (2009) 43–51

47

Table 1 Amount of propylamine obtained for different organic solvents in the ionic liquid free baseline system. Solvent

t (min) (99%cv.)

S(PA)a

S(DPA)a

S(TPA)a

Cyclohexane Methanol 1,2,4-Trichlorobenzene

180 100 160

34.8 26.8 48.3

46.6 45.7 26.9

4.1 8.7 0.3

Conditions: T = 100 8C, pH2 ¼ 100 bar, 10 g Ru/C, 125 ml propionitrile, 250 ml of organic solvent t(99% cv): time to reach 99% conversion; PA = propylamine, DPA=dipropylamine, TPA=tripropylamine. a Selectivitiy in mol% based on GC analysis.

Fig. 4. Time-dependent nitrile conversion and amine product yield (T = 100 8C, pH2 ¼ 100 bar). Table 2 Kinetic parameters of the Ru/C catalysed hydrogenation of propionitrile; data were obtained in TCB as organic solvent.

Activation energy of propylamine Reaction order of propionitrile (n)

Fig. 3. Hydrogenation of propionitrile with Ru/C in 1,2,4-trichlorobenzene – composition of the reaction mixture over time (T = 100 8C, pH2 ¼ 100 bar).

The fact that the selectivities of propylamine (PA), dipropylamine (DPA) and tripropylamine (TPA) at full conversions did not sum up to the amount of one (see Table 1) was found to originate from a remarkable side reaction that led to the formation of oligomeric products. In fact, less than 80 wt.% of the propionitrile (PN) feedstock was converted into PA, DPA and TPA. However, no other products could be detected in the GC analysis of the reaction mixture. By plotting the conversion of propionitrile versus the sum of all GC detected products as a function of time it became quite obvious that the formation of products that are not detectable by the GC analysis become more important at high conversions of propionitrile (Fig. 4). In order to confirm the oligomeric nature of the products, MALDI-ToF mass spectrometry (matrix assisted laser desorption ionisation – time of flight) of the product mixture has been carried out. The MALDI-ToF analyses indeed showed significant amounts of oligomers with an average weight of approximately 1180 g/mol. In order to find out whether the observed oligomerisation reaction in this hydrogenation system is linked to the special choice of our solvent 1,2,4-trichlorobenzene we also analysed the reactions in methanol and carried out additional measurements in cyclohexane. These reactions showed the same behaviour with approximately 20% of the propionitrile being transformed at full conversion into oligomers which were not GC-detectable. To the best of our knowledge this side reaction in the hydrogenation of propionitrile has not been described in detail in the open literature yet. For the determination of kinetic parameters, the formation of all side products, i.e. the formation of secondary and tertiary amines as well as the formation of oligomeric products, were taken into account. Measurements were carried out for pressures from 40 to

Graphic determination

Iterative determination

Data from literature

50.6 kJ/mol

48.4 kJ/mol

50.0 kJ/mol

0.48

0.51

0.5–1

100 bar and for temperatures from 80 to 120 8C. The kinetic parameters were calculated for low nitrile conversions according to Eq. (1) and are summarised in Table 2. The iterative determination was based on a least-square-fit using MATLAB1. All kinetic parameters obtained in 1,2,4-trichlorobenzene correspond well to literature values obtained for the same reaction in other solvents [11,23], which shows that the reaction system is not affected by 1,2,4-trichlorobenzene. Furthermore, as the kinetic data were determined from experiments far away from full conversion, also the formation of oligomeric products at high conversion did not influence significantly the determined kinetic data. m n r ¼ kcH c 2 nitrile

(1)

In the next set of experiments the propylamine selectivity was determined as a function of reaction temperature and of hydrogen pressure at full conversion of propionitrile. Of course, the formation of the oligomeric products was taken into account in these analyses. As depicted in Figs. 5 and 6, the reaction system showed a significant dependency of temperature and pressure on the amine selectivity. Improved propylamine selectivities were found for high reaction temperatures and low hydrogen pressure. As shown in Figs. 5 and 6, the maximum selectivities in propylamine were limited at full nitrile conversion to approximately 60% which indeed leaves room for improvement by applying multiphase approaches. 3.2. Propionitrile hydrogenation in systems containing Brønsted acidic ionic liquids - Approach 1 In our first approach to use ionic liquids in the hydrogenation of propionitrile we intended to apply a biphasic system containing an organic reaction phase and a Brønsted-acidic ionic liquid extraction phase. In this approach the acidic protons of the ionic liquid

48

K. Obert et al. / Applied Catalysis A: General 356 (2009) 43–51

Fig. 5. Selectivity of propylamine formation as a function of reaction temperature ( pH2 ¼ 100 bar, X > 99%).

Fig. 6. Selectivity of propylamine formation as a function of hydrogen pressure (T = 100 8C, X > 99%).

were expected to convert the primary amine instantaneously into the corresponding propylammonium hydrogensulfate salt. In this way, any possibility for the primary amine to react with imine to the undesired higher order amines should be avoided. We assumed the suspended Ru/C-catalyst to remain in the phase of higher density due to gravity. Therefore we expected to need an organic reaction phase with higher density compared to the ionic liquid extraction phase to realise this concept. Consequently, we first selected a combination of an organic solvent and a Brønsted acidic ionic liquid exhibiting a miscibility gap in which the ionic liquid was the phase of lower density.

As the dense organic phase, chlorinated aromatic solvents were chosen as they combine acceptable prices, high densities, chemical stabilities and suitable boiling points. Since we were looking for an industrially attractive system we applied in our ionic liquid screening only ionic liquids that are liquid at room temperature and can be obtained by direct protonation of an amine with sulphuric acid. Table 3 shows selected results from our solvent selection experiments. Only with the organic solvent 1,2,4-trichlorobenzene the desired phase behaviour for our catalytic approach could be realised. Indeed, dimethylcyclohexylammonium hydrogensulphate, 1-butylimidazolium hydrogensulfate and 1-octylimidazolium hydrogensulfate (see Fig. 7 for chemical structures) showed in combination with the solvent 1,2,4-trichlorobenzene the desired phase behaviour at the reaction temperature of 100 8C. Under these conditions the ionic liquid constituted the upper phase of the liquid–liquid biphasic system. In order to assure that the observed biphasic systems were also existent in presence of the feedstock propionitrile, the same miscibility experiments were repeated after addition of propionitrile in the typical amount of the catalytic experiment. With all three ionic liquids shown in Fig. 7 no change in the phase behaviour was found at 100 8C. Surprisingly, when the heterogeneous catalyst Ru on C was added to the biphasic mixture, the catalyst distribution among the two phases showed a very different behaviour from our expectations. As shown in Fig. 8 for the example of the dimethylcyclohexylammonium hydrogensulfate/1,2,4-trichlorobenzene system the catalyst was uniquely found in the lighter ionic liquid phase after the reactor was opened (the black ‘‘dots’’ at the bottom of the beaker stem from some ionic liquid droplets adsorbed to the glass surface). This unexpected result can be rationalised by taking into account the obviously much better wettability of this catalyst by the ionic liquid compared to the organic solvent. In contrast to our originally intended concept, this surprising catalyst distribution behaviour resulted in the fact that both, hydrogenation and product protonation, happened effectively in the ionic liquid phase. Thus the organic phase played only the role of being a reservoir of dissolved hydrogen and dissolved propionitrile for the reaction. Moreover the organic phase allowed taking samples for GC analyses to monitor the conversion of propionitrile during the reaction. For the calculation of propionitrile conversion from the composition of the organic phase it was assumed that the reaction rate was slow compared to the mass transfer between the two liquid phases. Furthermore, we assumed that the partition coefficient of propionitrile between the two phases did not change significantly during the course of the reaction. With these two reasonable assumptions the changes in composition of the organic phase could be used as direct measure for the reaction progress. In Fig. 9, the change of composition in the organic phase with reaction time is depicted for a typical propionitrile hydrogenation experiment. The concentration of propionitrile decreased constantly over time whereas less than 2% of amines were detected in the 1,2,4-trichlorobenzene phase (at a conversion of about 50%). This result gives already a very strong indication that the amines

Table 3 Results of miscibility experiments – phase behaviour of selected chlorinated aromatic solvents (1,2,4–trichlorobenzene and benzotrichloride) with ammonium and imidazolium hydrogensulfate ionic liquids. Ionic liquid

1,2,4-Trichlorobenzene (density = 1.45 g/ml)

Benzotrichloride (density ionic liquid multiphase reaction systems = 1.37 g/ml)

Trioctylammonium hydrogensulfate Dimethylcyclohexylammonium hydrogensulfate 1-Butylimidazolium hydrogensulfate 1-Octylimidazolium hydrogensulfate

Completely miscible Biphasic (100 8C), IL upper phase Biphasic (100 8C), IL upper phase Biphasic (100 8C), IL upper phase

Completely Completely Completely Completely

miscible miscible miscible miscible

K. Obert et al. / Applied Catalysis A: General 356 (2009) 43–51

49

Fig. 7. Chemical structures of the Brønsted acidic ionic liquids applied for Approach 1 of our study (R=butyl, octyl).

Fig. 9. Change in composition of the dense organic phase over time in a propionitrile hydrogenation experiment using Ru on carbon in the system 1,2,4trichlorobenzene/dimethylcyclohexylammonium hydrogensulphate (T = 100 8C, pH2 ¼ 100 bar).

Fig. 8. View of the solvent system 1,2,4-trichlorobenzene (bottom phase)/ dimethylcyclohexylammonium hydrogensulphate (top phase) with a suspended Ru on carbon catalyst uniquely dispersed in the lighter ionic liquid phase.

formed in the ionic liquid-catalyst phase were indeed readily protonated. Therefore hardly any free amine was detectable in the organic phase. The analysis of the ionic liquid phase (after aqueous work-up under basic conditions, for details see Section 2) yielded for the same conversion of 50% a selectivity of the primary amine (PA) of 82.7% in the GC-detectable products. 6.6% of dipropylamine (DPA), 0.5% of tripropylamine (TPA) and 10.2% of propylimine were found in the volatile products. These selectivities compare to the values obtained for the ionic liquid free system at 50% conversion where 61.3% of propylamine, 23.7% of dipropylamine and 0.4% of tripropylamine were detected within the volatile amines (oligomeric structures taken into account). More remarkably, the analysis of all GC-detectable products resulted in the ionic liquid case in a closed mass balance indicating that the catalytic reaction in the IL-based system suppressed effectively the formation of oligomeric products. In order to understand the influence of the cation on the selectivity we also investigated the same approach using 1butylimidazolium hydrogensulfate ([H-BIM][HSO4]) as the acidic ionic liquid phase. In this case the work-up of the IL-phase after 60% conversion yielded up to 87% selectivity to propylamine vs. 56% in the case of the ionic liquid free system for the same conversion. Based on these promising results we studied the conversion dependency of isolated propylamine yields in more detail for the [H-BIM][HSO4] system. The results are shown in Fig. 10. Remarkably, the selectivity and yield of propylamine exceeds 85% even for complete nitrile conversion after 8 h reaction time compared to SPA,IL-free = 48% at full conversion. In addition, MALDI-ToF analyses of the organic phase were carried out both for the ionic liquid based system and for the IL-free reference system. No oligomeric products were identified after the reaction in the ionic liquid whereas the IL-free reference system showed significant amounts of oligomers (spectra are available in

the Supplementary Material). Consequently, a good part of the improved yield to the primary amine in the case of the Brønsted acidic IL approach is due to the very effective suppression of oligomer formation. Encouraged by these results we attempted to optimise our experimental protocol to allow a reuse of the Ru on C/Brønsted acidic ionic liquid phase. Of course, a direct recycling is not possible after basic work-up as the latter also leads to a complete deprotonation of the ionic liquid cation to the corresponding amine. Therefore we tried a direct recycling method by product distillation from the Ru-catalyst/ionic liquid phase. We expected the low boiling products (e.g. the boiling point of propylamine is only 49 8C under atmospheric pressure) to evaporate from the ionic liquids formed by protonation of high boiling amines (boiling point of dimethylcyclohexylamine is 155–165 8C at atmospheric pressure, boiling point of 1-butylimidazole is 116 8C at 1.6  102 bar) under suitable temperature and pressure conditions. Unfortunately, all attempts of this kind failed. Under mild conditions (at room temperature or 95 8C) the amount of isolated amines in the cooling trap of the vacuum line (40 mbar) was very low. Under

Fig. 10. Conversion and total propylamine selectivity in the IL-phase for the [HBIM][HSO4] system (T = 100 8C, pH2 ¼ 100 bar).

50

K. Obert et al. / Applied Catalysis A: General 356 (2009) 43–51

more severe conditions (160 8C, 40 mbar), however, an undesired reaction took place in the bottom of the distillation column (note that the Ru-catalyst was still present in the ionic liquid product mixture) forming a black and highly viscous residue. Apparently, the deprotonation equilibrium of propylammonium hydrogensulfate is shifted so far to the protonated side that there is hardly any free propylamine existing under realistic distillation conditions. It should be noted, however, that exactly this low concentration of free propylamine in the reaction mixture is the reason for the very high propylamine selectivity during the reaction, as all side reactions involving free amines are effectively suppressed in the Brønsted-acidic ionic liquid system. 3.3. Propionitrile hydrogenation in aprotic IL-containing systems – Approach 2 An alternative concept to recycle the ionic liquid phase in a direct manner would be a reaction system in which the product propylamine is not protected by protonation. In Approach 2 the consecutive reaction was avoided by extracting the propylamine into another liquid phase (see Fig. 2). This approach has the intrinsic advantage that the extraction from the reaction phase represents at the same time the product isolation step. Consequently, the ionic liquid catalyst phase can be directly recycled without additional steps for product separation. In order to allow a comparison between the Approaches 1 and 2 we intended to realise Approach 2 with a system of great similarity to the 1,2,4-trichlorobenzene/[H-BIM][HSO4] system applied in Approach 1. Therefore, 1-ethyl-3-methylimidazolium ethylsulfat ([EMIM][EtOSO3]) was chosen as the aprotic ionic liquid for all further experiments and 1,2,4-trichlorobenzene remained the applied organic phase. The latter constituted the denser phase of the liquid–liquid biphasic system under investigation. Once more, the Ru/C-catalyst was uniquely found in the upper ionic reaction phase. Fig. 11 shows the composition of the organic phase for a typical propionitrile hydrogenation experiment according to Approach 2. After reaction, the whole reaction mixture, i.e. the ionic liquid/Ru on carbon phase and the organic phase were stored under hydrogen in the autoclave over night. The next day, the organic phase was drained from the reactor and the ionic liquid/catalyst phase was re-used in a second experiment without addition of fresh catalyst or ionic liquid. After loading the reactor with identical amounts of 1,2,4-trichlorobenzene and propionitrile as for the original run, the reaction was started as described in Section 2. Our results showed that the aprotic ionic liquid catalyst phase can be recycled, although a certain loss in reaction rate was observed (as can be seen from the reduced negative slope of the propionitrile consumption curve in the recycling experiment). The latter may be related to the known deactivation behaviour of heterogeneous hydrogenation catalysts [30–32]. Interestingly, the results of the recycling run gave a clear indication that the majority of amine products is indeed extracted into the 1,2,4-trichlorobenzene phase. This can be concluded from the first data points of the recycling run which in fact constitutes some kind of product extraction from the ionic liquid phase into the fresh 1,2,4-trichlorobenzene phase. It is also noteworthy that in these experiments propylimine was detected as the hydrogenation intermediate. In the early stages of the hydrogenation this intermediate even exceeded the amount of the DPA side product. As expected from the hydrogenation mechanism, the imine is converted during the reaction mainly to propylamine and dipropylamine so that its amount decreases at longer reaction times.

Fig. 11. Hydrogenation of propionitrile with Ru/C in liquid–liquid biphasic system 1,2,4-trichlorobenzene/[EMIM][EtOSO3] – composition of the organic phase over time for the original hydrogenation experiment and one catalyst phase reuse starting at t = 23 h (T = 100 8C, pH2 ¼ 100 bar).

From both runs presented in Fig. 11 we derived the total propylamine selectivities as a function of propionitrile conversion. As depicted in Fig. 12 both runs gave virtually identical selectivities. Remarkably, both runs showed rising propylamine selectivities with enhanced conversion. Total selectivities reached 66% of PA at full conversion for the original run and again 66% for the recycle run at 90% conversion which represents a clear selectivity advantage over the IL-free system in which under identical reaction conditions a total propylamine selectivity of only 48% was obtained. Most likely the enhanced PA selectivities were again due to the suppression of oligomeric structures. The easiest way to validate this assumption is to perform a mass balance. For this purpose the data depicted in Fig. 11 were not applicable as the rather large amount of ionic liquid led to significant cross-solubility effects. As a consequence the distribution coefficients of all components under reaction condition would be needed to calculate the mass balance in a proper way. Instead, we decided to carry out two reactions with a strongly reduced amount of the [EMIM][EtOSO3] ionic liquid, namely 12 and 8 g (see Section 2).

Fig. 12. Total selectivity of propylamine as a function of conversion for the [EMIM][EtOSO3] – 1,2,4-trichlorobenzene system – comparison between original run and recycling run (100 bar, 100 8C).

K. Obert et al. / Applied Catalysis A: General 356 (2009) 43–51

The results of these experiments showed a PA selectivity of 70  2% at conversions >90% which was in good accordance to the values from Fig. 12. More importantly, the mass balances could be successfully closed for both cases. When 12 g IL was applied the deviation between the initially applied amount of nitrile and the resulting products was 3.4 mol%. As expected, the second experiment (8 g IL) showed an even smaller deviation, namely 1.1 mol% (note that in the IL-free reference system 20% of the nitrile was converted into oligomers that are not detectable in GC). These mass balances proved that the application of the neutral ionic liquid [EMIM][EtOSO3] improved the propylamine selectivity mainly due to the greatly reduced oligomer formation. 4. Conclusions In this paper two different, multiphasic approaches using ionic liquids were applied to improve the selectivity in catalytic propionitrile hydrogenation. All experiments used Ru on carbon as the heterogeneous catalyst and avoided the addition of ammonia, which is applied in the actual industrial processes to improve the selectivity for the primary amine. First, baseline experiments in different IL-free systems were carried out and yielded total selectivities to propylamine of less than 50% at full conversion. The rather low amine selectivity in the conventional system could be attributed to the formation of dipropylamine but also to a significant formation of oligomeric, non-GC detectable products (approx. 20% yield at full propionitrile conversion). The latter have been identified by MALDI-ToF spectrometry. The application of Brønsted acidic ionic liquids (Approach 1) resulted in a significant increase in total primary amine selectivity at full conversion with maximum values of 85%. The enhanced selectivity is caused by an efficient scavenging of the propylamine product by the acidic ionic liquids and by the suppression of oligomer formation in the ionic liquid system. However, this approach required an aqueous work-up under basic conditions for product isolation making several product isolation steps necessary and continuous processing difficult. In a second approach (Approach 2), the aprotic ionic liquid [EMIM][EtOSO3] was applied. In this case the Ru on carbon catalyst was completely dispersed in the ionic liquid phase. Approach 2 led to PA selectivities of up to 70% at full conversion. By performing a mass balance it could be demonstrated that the addition of [EMIM][EtOSO3] reduced the formation of oligomeric sideproducts in the reactor effectively. Moreover, the recyclability of the ionic liquid/catalyst phase by simple decantation could be demonstrated. Recycling was achieved by simple phase separation. The improved selectivity and the demonstrated recyclability of the neutral ionic liquid system may open an attractive way for technical nitrile hydrogenations under ammonia-free hydrogenation conditions. Acknowledgements The authors like to thank Dr. Klemens Massonne (BASF SE) for fruitful discussions and many helpful comments. BASF SE is

51

gratefully acknowledged for financial support. This work was further supported by the Excellence Cluster ‘‘Engineering of Advanced Materials’’ granted to the University of ErlangenNuremberg. Appendix A Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.apcata.2008.12.016. References [1] S. Gomez, J.A. Peters, T. Maschmeyer, Adv. Synth. Catal. 344 (10) (2002) 1037– 1057. [2] Ullmann’s Encyclopedia of Industrial Chemistry, 7th ed., Wiley-VCH, Weinheim, 2006. [3] C. Mathieu, E. Dietrich, H. Delmas, J. Jenck, Chem. Eng. Sci. 47 (1992) 2289– 2294. [4] B.W. Hoffer, P.H.J. Schoenmakers, P.R.M. Mooijman, G.M. Hamminga, R.J. Berger, A.D. van Langeveld, J.A. Moulijn, Chem. Eng. Sci. 59 (2004) 259–269. [5] C. Joly-Vuillemin, D. Gavroy, G. Cordier, C. de Bellefon, H. Delmas, Chem. Eng. Sci. 49 (1994) 4839–4849. [6] A. Chojecki, Dissertation, TU Mu¨nchen, 2004. [7] P.N. Rylander, Catalytic Hydrogenation in Organic Syntheses, Acad. Press, New York, 1979. [8] P.N. Rylander, Hydrogenation Methods, Acad. Press, London, 1985. [9] J.L. Dallons, A. Van Gysel, G. Jannes, Catal. Org. React. 47 (1992) 93–104. [10] S.N. Thomas-Pryor, T.A. Manz, Z. Liu, T.A. Koch, S.K. Sengupta, W.N. Delgass, Catal. Org. React. 75 (1998) 195–206. [11] W.H. Hartung, J. Am. Chem. Soc. 50 (1928) 3370–3374. [12] C.F. Winans, J. Am. Chem. Soc. 61 (1939) 3566–3567. [13] M.F. Zuckerman, patent application US 4739120 (to Dow Chemical Co., USA), 1988. [14] L. Hegedus, T. Mathe, Appl. Catal. A: General 296 (2005) 209–215. [15] (a) V. Parvulescu, C. Hardacre, Chem. Rev. 107 (6) (2007) 2615; (b) P. Wasserscheid, P.S. Schulz, in: P. Wasserscheid, T. Welton (Eds.), Ionic Liquids in Synthesis, second ed., Wiley-VCH, Weinheim, 2007, pp. 369–463; (c) T. Welton, Coord. Chem. Rev. 248 (2004) 2459; (d) P. Wasserscheid, W. Keim, Angew. Chem., Int. Ed. Engl. 39 (2000) 3772. [16] A. Visser, R. Swatloski, M. Reichert, R. Mayton, J. Sheff, A. Wierzbicki, J. Davis, R. Rogers, Chem. Comm. (2001) 135–136. [17] P. Dyson, T. Geldbach, Electrochem. Society Interface 16 (1) (2007) 50–53. [18] T.L. Greaves, C.J. Drummond, Chem. Rev. 108 (1) (2008) 206–237. [19] M. Maase, K. Massonne, K. Halbritter, R. Noe, M. Bartsch, W. Siegel, V. Stegmann, M. Flores, O. Huttenloch, M. Becker, patent application WO 2003062171 (to BASF AG), 2003. [20] S. Sunitha, S. Kanjilal, S. Reddy, R. Prasad, Tetrahedron Lett. 49 (2008) 2527–2532. [21] K. Anderson, P. Goodrich, C. Hardacre, D. Rooney, Green Chem. 5 (2003) 448–453. [22] M. Steffan, M. Lucas, A. Brandner, M. Wollny, N. Oldenburg, P. Claus, Chem. Eng. Technol. 30 (4) (2007) 481–486. [23] B. Le´ger, A. Denicourt-Nowicki, A. Roucoux, H. Olivier-Bourbigou, Adv. Synth. Catal. 350 (2008) 153–159. [24] G.S. Fonseca, J.B. Domingos, F. Nome, J. Faruk, Dupont, J. Mol. Catal. A: Chem. 248 (2006) 10–16. [25] K. Anderson, S. Ferna´ndez, C. Hardacre, P. Marr, Inorg. Chem. Comm. 7 (2004) 73– 76. [26] P. Dash, N. Dehm, R. Scott, J. Mol. Catal. A: Chem. 286 (2008) 114–119. [27] U. Kernchen, B. Etzold, W. Korth, A. Jess, Chem. Eng. Technol. 30 (8) (2007) 985– 995. [28] V. Weiskopf, T. Gerlach, K. Wenz, Kirsten, patent application WO 2005061429, (to BASF AG), 2005. [29] M. Ehrig, Dissertation Universita¨t Erlangen-Nu¨rnberg, 2007. (http://www.opus.ub.uni-erlangen.de/opus/volltexte/2008/900/pdf/MartinEhrigDissertation.pdf, 24.6.2008). [30] J. Patzlaff, J. Gaube, Chem. Ing. Technol. 69 (1997) 1462–1466. [31] Z. Belohlav, P. Kluson, L. Cerveny, Res. Chem. Intermediates 23 (2) (1997) 161– 168. [32] N. Thakar, T. Schildhauer, W. Buijs, F. Kapteijn, J. Moulijn, J. Catal. 248 (2007) 249– 257.