Heterogeneously Catalyzed Continuous-Flow Hydrogenation Using Segmented Flow in Capillary Columns

This paper explores how the visible features of segmented-flow, under reaction conditions, can be used in lab-scale multiphase heterogeneous catalysis. Continuous-flow microreactors are now routinely used in bench-scale synthesis [1] and optimization [2] applications, owing to their small reactant inventory, negligible heat effects at small scales, and fast mixing. [3] How-ever, miniaturizing multiphase heterogeneous catalysis on chips is considerably more difficult than homogeneous liquid phase chemistry. Several applications of gas–liquid [2b,4] and gas–liquid–solid [5] reactions have been reported. In these con-tinuous-flow devices, the catalyst was immobilized on the wall of the channel [5a,6] or incorporated as powder. [5e,7] A powder packed-bed gas–liquid microreactor may appear ideal for off-the-shelf catalysts, but in practice such reactors are cumbersome: critical packing parameters vary from one instance to the next and channeling and flow hysteresis abound, as we have recently visualized. [8] For immobilized catalysts, Kobayashi et al. have advocated creating a thin film of liquid on the walls, sheared along by a fast-flowing gas stream. [5a] A drawback of this system is that it is hard to control or visualize how long the reactants are in contact with the catalyst, because both phases each move at their own velocity. Especially for more complex pathways, the spread in residence time reduces yields. heterogeneous catalysis and synthesis routes, such as solvent effects, competitive adsorption and irreversible poisoning.


2
A solution of Palladium (II) acetate (Pd(OAc) 2 , Alfa Aesar, 45.9 wt% Pd) in toluene (J.T. Baker) was prepared and stirred for 60 min. The capillaries were filled with the Pd(OAc) 2 solution, capped on both sides with GC septa and ion-exchange occurred for 24 h. Capillaries were impregnated with solutions of different Pd(OAc) 2 amounts aiming for Pd catalysts between 0.003 and 7 wt% Pd on γ-Al 2 O 3 .
In a typical experiment, the amount of Pd in the solution was aimed at 7 wt% per unit γ-Al 2 O 3 weight.
After the impregnation the capillary was emptied and washed with n-hexane followed by flowing N 2 for 20 min at room temperature. After flushing, they were dried and calcined with a flow of air for 1 h at 120°C, followed by 2 h at 250°C, and then cooled to room temperature, with a heating rate up and down of 2 K min -1 . Reduction of the Pd nanoparticles took place at room temperature for 12 h prior to a hydrogenation experiment. The H 2 flow and pressure during reduction was 1 ml(STP) min -1 and 1.1 bar.
The amount of Pd in a capillary was determined by atomic adsorption spectroscopy (AAS) on a Perkin-Elmer 4100ZL. Transmission electron microscopy (TEM) was performed using a Philips CM30T electron microscope with a LaB6 filament as a source of the electrons equipped with an energy dispersive X-ray Spectrometer. Scanning electron microscopy (SEM) was performed using a Philips XL20 electron microscope to investigate the thickness and homogeneity of the γ-Al 2 O 3 layer. The γ-Al 2 O 3 layer thickness was determined at several points in the capillary and was 60.5 µm. N 2 physisorption, performed on a Quantachrome Autosorb-6B, was used to determine the specific surface area S BET , the average pore diameter, and the pore volume of the γ-Al 2 O 3 coating. The S BET of the γ-Al 2 O 3 layer was 195 m 2 g -1 , the average pore diameter was 10 nm, the pore volume was 0.47 cc g -1 , and the porosity was 0.63.
AAS revealed that 80% of the aimed amount of Pd was deposited in the capillary resulting in 5.7 wt% Pd loading and all Pd was deposited with aimed amounts of <3 wt%. Pd nanoparticles (d = 5 nm, determined with TEM and CO chemisorption) were homogenously distributed over the entire length of a capillary of 0.5 m. Capillaries with lower wt% Pd resulted in smaller Pd nanoparticles (determined with TEM): for 1.1 and 2.7 wt% Pd capillaries the particles size of Pd was 2.4 and 3.3 nm, respectively.

Hydrogenation experiments
The fused silica capillaries with Pd nanoparticles deposited on the γ-Al 2 O 3 layer were connected to a gas mass flow controller for the H 2 supply (Bronkhorst HI-TEC model F-200CV-FAC-33-V with a maximum of 1 H 2 ml(STP) min -1 ) and to a syringe pump (Harvard, PHD2000 programmable) containing a 50 ml syringe (SGE Europe Ltd.) filled with a liquid reactant/solvent mixture. Simple quartz Y-mixer (Restek) and stainless steel T-and Y-mixers (Valco and Swagelok) were used as gas-liquid distributors which were attached upstream of a capillary.
Continuous flow experiments were performed in the segmented flow regime. The analysis of the reaction mixture effluent was only considered after a stable flow pattern of segmented flow had established inside the transparent capillary. Samples were only collected and analysed by HPLC and/or GC analysis when a stable segmented flow was observed.
Initially, the capillaries were reduced for 12 h at room temperature in a H 2 flow of 1 ml(STP) min -1 .
Experimental results obtained in the non-steady state induction period were discarded. No improvement was found in the reaction rate and stability in comparison with capillaries that were not reduced in advance, and the time-consuming procedure was abandoned and capillaries were used as prepared. Before every hydrogenation experiment, the H 2 flow was switched on first and after 10 min the liquid flow was switched on. This short procedure resulted in capillaries that were fully active at the start of experiments.
The capillaries were submerged in a temperature controlled stirred water bath (IKA RCT basic and IKA ETS-D4 Fuzzy). The progress of reactions was measured using off-line GC analysis (also HPLC for the hydrogenation of 3-phenyl-propyl-azide) and also by visual observation of the H 2 bubble shrinkage during the hydrogenation of 3-methyl-1-pentyn-3-ol and cyclohexene.
A capillary of 0.5 m was cut into several pieces of equal length and these pieces were tested separately for cyclohexene hydrogenation activity. This gave hydrogenation conversions within 3% confirming that the Pd was distributed homogenously in the axial direction, as was also shown by TEM analysis of the separate pieces. No (detectible) activity was observed for capillaries without Pd when performing the hydrogenation of cyclohexene, 3-methyl-1-pentyn-3-ol and 3-phenyl-propyl-azide.
The Pd content on the capillary was measured with AAS before and after all hydrogenation reactions and no decrease of the Pd content was observed except for the hydrogenation of 3-phenyl-propyl-azide in NMP. Note that the other solvents used in the hydrogenation 3-phenyl-propyl-azide (ethanol/water and toluene) showed no Pd content decrease. After reaction, several product mixtures were analysed by inductively coupled plasma optical emission spectroscopy (ICP-OES) and no Pd was detected in these mixtures (at least no Pd dissolution above the detection limit). ICP-OES was carried out with a Perkin-Elmer Optima 5300. Solutions of the products of the hydrogenations in the appropriate solvent were used as matrix in the calibration samples. TEM analysis of the post-mortem catalyst at several points did not show a particle size decrease or an accumulation of Pd (readsorption of leached Pd) at the end part of the capillary for the hydrogenation of 3-methyl-1-pentyn-3-ol, cyclohexene and 3-phenyl-propyl-azide in ethanol/water and toluene. Continuous experiments with increasing time on stream for the hydrogenation of cyclohexene and 3-methyl-1-pentyn-3-ol showed no deactivation under kinetically controlled conditions and capillaries could be used for several weeks without deactivation. The hydrogenation of 3phenyl-propyl-azide showed strong deactivation which was mainly caused by poisoning (reversible and irreversible adsorption) except for the hydrogenation of 3-phenyl-propyl-azide in NMP, where the reaction mixture had a yellowish colour and the capillary became completely yellow again pointing towards full Pd dissolution without readsorption of leached Pd downstream or at the end of the capillary.
AAS analysis of the used capillaries in the hydrogenation of 3-phenyl-propyl-azide in NMP showed that most Pd leached into the reaction mixture (less than 0.2 wt% left from the 5.7 wt% Pd originally on the capillary) caused by complex formation with components in the highly polar reaction mixture. Moreover, N 2 physisorption revealed that the S BET of the γ-Al 2 O 3 support decreased significantly from 195 to 100 m 2 g -1 indicating γ-Al 2 O 3 sintering and/or dissolution because of the high pH (~10) of the reaction mixture.
In addition, regeneration attempts by H 2 and N 2 flow at elevated temperatures did not result in any activity increase, indicating Pd loss for the hydrogenation in NMP.

Cyclohexene hydrogenation: determination of kinetic parameters
Cyclohexene hydrogenation experiments were analyzed using Varian CP-3380 GC with a Chrompack Cyclohexene (Aldrich, used as received without purification) hydrogenation was performed in a 0.2 m capillary with a Pd loading of 2.7 wt%. The cyclohexene feed concentration was varied from 30 to 600 mol m -3 in dry ethanol (Aldrich) and n-decane (Merck), which allowed us to determine the first-order part and the zero-order part of the Langmuir-Hinshelwood kinetic expression (Figure 4). The adsorption constant of cyclohexene was found to be K ≈ 0.01 m 3 mol -1 . The kinetic parameters were obtained at low cyclohexene conversion (<20%), at intermediate conversion levels (20% to 70%) or at high conversion levels (>70%). In all cases, the same values were obtained for the kinetic constants, indicating that axial dispersion or back-mixing had no negative impact on the measurement of kinetics.
Reactions were performed with temperatures up to 90°C. It was found that the observed activation energy decreased at temperatures above 55°C to a new value of 14 kJ mol -1 , indicating that at higher temperatures diffusion inside the γ-Al 2 O 3 layer was starting to limit the reaction rate.
After several hydrogenation runs at least two new peaks (retention times: 4.1 min and 4.9 min) appeared in the HPLC diagrams. The molar weight of a peak with a retention time of 4.9 min (determined by LC-DAD-MS) was 253 g mol -1 corresponding with the secondary amine di-(3-phenyl-propyl)-amine 5b, which was formed at high reactant concentrations, high temperature, and longer time on stream. In samples taken from hydrogenation runs in ethanol a third peak was detected, which has a molecular weight of 163 g mol -1 . It was identified by using LC-DAD-MS as a secondary amine 5c formed from condensation of 5 with the solvent ethanol. A fourth by-product (retention time = 4.1 min) was detected with a very small peak in the HPLC chromatograms. Most likely, this is a tertiary amine, as it was also found in batch experiments with long reaction times (20 h). Moreover, the conversion of 4 was followed by GC analysis. This analysis was performed by using a HP6890 with a CP-SIL 5CB-MS column.
4 was synthesised by azidation of 3, derived from 1, or by azidation of 3-phenyl-propyl-chloride 2. 3 was synthesized by reacting 1 with methane sulphonyl-chloride (Fluka) in CH 2 Cl 2 (Aldrich). Aqueous NaN 3 (Janssen Chimica) reacted with 3 in NMP. NMP is a very polar solvent that dissolves NaN 3 and that provides some homogeneity. The azidation was performed in a batch reactor at 55°C and in a continuous-flow stainless-steel tubular microreactor at 110°C, and both gave >95% yield of 4 (Table 2, used as reactant in entry 1-2).
A solution of 4 in toluene was prepared by contacting 3 in toluene and a mixture of ammonium chloride in propanol/water with aqueous NaN 3 at 82°C. Subsequently, adding toluene resulted in phase separation, the toluene was washed with water and azeotropic distillation resulted in a solution of 4 in toluene (Table 2, entry 6). 2 (Aldrich) was used without further purifications and dissolved in ethanol.
Azidation was performed by reacting 2 with aqueous NaN 3 in a continuous-flow steel microreactor at temperatures up to 200°C to 100% yield of 4. The liquid flow rate was varied from 100 to 300 l min -1 .
However, the catalyst activity decreased to 2% after a WHSV*TOS of 2. The chloride-derived 3-phenylpropyl-azide 4 shows an increased catalyst lifetime and still retained 12% activity after WHSV*TOS of 9.
However, a drop in selectivity to 3-phenyl-propyl-amine 5 is from 95 to75% with increasing TOS. The activity could be increased to 30% (without any further loss of activity with TOS) after regeneration of the Pd catalyst in flowing N 2 (g) at 200°C for 2h (not shown).
The dashed lines are guidance for the eyes.