Multigram-scale flow synthesis of the chiral key intermediate of (–)-paroxetine enabled by solvent-free heterogeneous organocatalysis† †Electronic supplementary information (ESI) available: Synthetic procedures, additional batch and flow reaction data, characterization data, copies of NMR spectra and HPLC chromatograms. See DOI: 10.1039/c9sc04752b

The continuous flow synthesis of the chiral key intermediate of (–)-paroxetine is demonstrated via a solvent-free organocatalytic conjugate addition followed by a telescoped reductive amination–lactamization–amide/ester reduction sequence.


General information
All solvents and chemicals were obtained from commercial vendors (Sigma-Aldrich, TCI, Alfa Aesar, or VWR) and were used as received, without further purification. Continuous flow equipment was assembled from commercially available components as detailed in Sections 3-6.
Chromatographic purification was carried out by using a Biotage Isolera automated flash chromatography system with cartridges packed with KP-SIL, 60 Å (32-63 μm particle size). Analytical thin-layer chromatography (TLC) were performed on Merck silica gel 60 GF254 plates. Compounds were visualized by means of UV or KMnO4. 1 H-, 13 C-and 19 F-NMR spectra were recorded on a Bruker Avance III 300 MHz instrument at ambient temperature, in CDCl3 as solvent, at 300 MHz, 75 MHz and 282 MHz, respectively. Chemical shifts (δ) are reported in ppm using TMS as internal standard. Coupling constants are given in Hz units.
The ee of the compounds was determined by using a Shimadzu HPLC system (DGU-14A degasser, SCL-10A VP system controller, SPD-10 UV-VIS detector, LC-20AT pumps) and a Chiralpak ® AD-H chiral column with isocratic mixtures of hexane and iPrOH as eluent. Chromatographic conditions are listed in Section 7. A racemic reference sample of 2 was prepared by using a 1:1 mixture of (R)-and (S)-α,α-diphenyl-2-pyrrolidinemethanol trimethylsilyl ether as catalyst as follows. A mixture containing 1 equiv dimethyl malonate (cmalonate= 0.5 M), 2 equiv 4-fluorocinnamaldehyde, 0.3 equiv AcOH and 10 mol% catalyst was stirred for 24 h at RT in MeOH as solvent. The crude product was purified chromatographically using a mixture of ethyl acetate/40-60 petroleum ether as eluent. Racemic samples of 3 and 1 were prepared by means of standard flow procedures starting from (rac)-2.
Optical rotation was measured in CHCl3 (HPLC-grade) at 20 or 25 °C against the sodium D-line (λ= 589 nm) on a Perkin Elmer Polarimeter 341 using a 10-cm pathlength cell. IR spectra were recorded on a Bruker Tensor 27 / Diamond ATR FT-IR spectrometer.
Elemental analyses of catalyst 4 were performed on a LECO CHNS 932 micro-analyzer at the Universidad Complutense de Madrid, Spain.
Pt contents were determined by means of ICP-MS using an Agilent 7700x instrument. For the analysis, the samples were digested by using an MLS ultraCLAVE system (program: ramp in 30 min to 250 °C and then heating for 30 minutes at 250°C).
LC-MS analyses were performed on a Shimadzu HPLC system (DGU20A degasser, SIL-20A autosampler, CTO20A column oven, LC-20AD pumps) using a Macherey-Nagel Nucleodur C18 HTec column (150 mm × 4.6 mm, particle size 5 μm) at 37 °C with mobile phases A (H2O/acetonitrile 9:1 v/v + 0.1% TFA) and B (acetonitrile + 0.1% TFA) at a flow rate of 0.6 mLꞏmin -1 ). The detection of compounds was accomplished by a diode array detector (SPDM20A) prior electrospray ionization (ESI) using a Shimadzu LCMS-QP2020 instrument. The ESI-MS was operating either in positive or negative mode with in a scan range of 100-400 m/z or 350-750 m/z. The E-factor was calculated by dividing the mass of waste generated by the mass of product formed. The mass of the waste did not include the water.

Synthesis of catalyst 4
Catalyst 4 was immobilized on a cross-linked polystyrene resin (100-200 mesh) through a 1,2,3-triazole linker. The monomer synthesis and the azide-alkyne cycloaddition-based immobilization was carried out according to a recently published procedure. 1,2 The level of functionalization of the polystyrene-supported catalyst f (mmol of monomeric catalyst / gram of resin) was calculated based on the results of nitrogen elemental analysis by the following formula: 3 f (mmol g -1 ) = %N × 1000 × (number of N atoms) -1 × MW(N) -1 × 100 -1

Initial batch experiments
A typical procedure for the batch reactions is as follows: a 1-mL mixture containing 4-fluorocinnamaldehyde (1 equiv, 26.2 or 52.4 µL, 0.2 or 0.4 M), dimethyl malonate, a specified solvent and, occasionally, 0.6 equiv. of an additive was added into a glass vial. Catalyst 4 was next added (f= 0.464 mmol g -1 , 86 or 172 mg, 20 mol % loading), and the suspension was shaken for 24 h at 25, 50 or 75 °C. The mixture was filtered and the resin beads were washed with the same solvent used as reaction medium (5 × 1 mL). The solvent was concentrated under reduced pressure and the crude product was analyzed by means of 1 H-NMR and chiral HPLC.

Continuous flow experiments using CH2Cl2 as solvent
A typical procedure for the continuous flow experiments is as follows: the reaction mixture consisting of 4-fluorocinnamaldehyde (1 equiv.), dimethyl malonate and AcOH in CH2Cl2 was pumped by using a Syrris ® Asia syringe pump. 1 g of catalyst 4 (f= 0.464 mmol g -1 ) was encompassed in an adjustable Omnifit ® glass column (10 mm ID), which was heated by a Syrris ® column heater. The system was pressurized by applying a 10-bar BPR from IDEX. Prior to the reactions, the catalyst was swollen by pumping CH2Cl2 at 200 µL min -1 for 45 min. (The swollen bed was approximately 7 cm high). For the reactions, the flow rate was set to 100 µL min -1 , which corresponded to 35 min residence time on the catalyst bed. In each runs, 2 mL product solution was collected after reaching steady state, which was next concentrated under reduced pressure and analyzed by 1 H-NMR and chiral HPLC.

Comparison of catalyst swelling in different medium
Swelling properties of resin-supported catalyst 4 were compared in CH2Cl2 and in dimethyl malonate. To this end, 1 g of the material was loaded into an adjustable Omnifit ® glass column (10 mm ID) and was swollen by pumping the appropriate liquid at 200 µL min -1 for 45 min. In Figure S1, the difference in bed height represents the different swelling of the resin, which had significant effects on the residence times measured.

Continuous flow experiments under solvent-free conditions
A typical procedure for the continuous flow experiments is as follows: the reaction mixture consisting of 4-fluorocinnamaldehyde (1 equiv.), dimethyl malonate and AcOH was pumped by using a Syrris ® Asia syringe pump. 1 g of catalyst 4 (f= 0.464 mmol g -1 ) was filled into an adjustable Omnifit ® glass column (6.6 mm ID) which was heated by a Syrris ® column heater. The system was pressurized by applying a fixed-pressure BPR from IDEX. Prior to the reactions, the catalyst was swollen by pumping dimethyl malonate at 200 µL min -1 for 45 min. (The swollen bed was approximately 7 cm high). In each run, 2 mL product solution was collected after reaching steady state, which was next concentrated under reduced pressure and analyzed by 1 H-NMR and chiral HPLC.  Taking into account the ee, conversion and maximum attainable productivity, the conditions in Table S9, entry 5 were designated as optimum and applied in the subsequent preparative scale synthesis.

Large-scale continuous flow synthesis of 2 under solvent-free conditions
The procedure for the large-scale asymmetric flow synthesis of 2 is as follows. The reaction mixture consisting of 4-fluorocinnamaldehyde (1 equiv.), dimethyl malonate (2 equiv.) and AcOH (0.6 equiv.) was pumped by using a Syrris ® Asia syringe pump. (In this reaction mixture, the concentration of 4-fluorocinnamaldehyde was 2.48 M as determined experimentally.) 1 g of catalyst 4 (f= 0.464 mmol g -1 ) was filled into an adjustable Omnifit ® glass column (6.6 mm ID) which was heated by a Syrris ® column heater at 60 °C. The system was pressurized by applying a 10bar BPR from IDEX. Prior to the reaction, the catalyst was swollen by pumping dimethyl malonate at 200 µL min -1 for 45 min. (The swollen bed was approximately 7 cm high). The flow rate was set to 70 µL min -1 (corresponded to 20 min residence time on the catalyst bed), and the product stream was collected continuously for 7 h after reaching steady state. During this period, samples were taken in every 15 minutes and conversion, chemoselectivity and ee were determined in all of them by means of 1 H-NMR and chiral HPLC, respectively.
According to the analysis of the samples collected, catalyst 4 proved highly robust during the large scale solventfree run. The selectivity of the catalyst remained unchanged: ee was constant in the range of 95-98%; side products were not detected, i.e. chemoselectivity was 100%. Only a small decrease in catalytic activity occurred during the experiment as indicated by a slight drop in conversion from 93 to 85%. Conversion and ee are represented as functions of time-on-stream in Figure S2. After the collection period, excess dimethyl malonate, unreacted aldehyde and residual AcOH could simply be removed in vacuo (55 °C, 10 -3 mbar) yielding 17.26 g analytically pure 2 without the need for chromatographic purification (84% isolated yield). The results of the experiment are summarized below in Table S10. Picture of the flow setup can be found as Figure S3. The same batch of catalyst was reused in two more preparative-scale runs to accumulate 2 for the optimization of the next step. Conversion and selectivity were not monitored regularly, but yield and ee were determined as follows: 75% yield and 96% ee in the first prep. run (7 h long), 64% yield and 95% ee in the second prep. run (5 h long).

Synthesis of lactam 3: tandem reductive amination-lactamization
Solutions of 2 and benzylamine were pumped as separate feeds (P1 and P2) by using a UNIQSIS Binary Pump Module and were combined in a Y-mixer. The module was equipped with two high-pressure HPLC pumps, two injection valves with sample loops and a pressure sensor to monitor system pressure. H2 gas was introduced into the system from a gas cylinder using a calibrated mass flow controller (MFC, Bronkhorst-EL). The inclusion of a check valve prevented any backflow of liquid towards the MFC. The gas flow rate was measured in units of mLn min −1 (n represents measurement under standard conditions: Tn= 0 °C, Pn = 1.01 bar). The liquid and gaseous streams were combined in a second Y-mixer at room temperature. A stainless steel column with internal dimensions of 4.6 × 100 mm was used as catalyst bed and was charged with a mixture of 200 mg of 5% Pt/C and 400 mg of activated charcoal. The packed column was sealed with compatible frits (0.5 µm pore size), and was placed into a Phoenix Flow Reactor™ (ThalesNano) for heating purposes. Prior to the catalytic reactions, dry THF was pumped through the packed bed at 200 µL min -1 for 45 min to remove water traces from the catalyst. The flow system was pressurized by applying a fixed-pressure BPR from IDEX.
A typical procedure for the continuous flow experiments is as follows. First, the carrier solvent flow was started. Then, when the pressure stabilized on the catalyst bed, the desired temperature was set on the Phoenix Flow Reactor™ and the gas flow was initiated by setting the desired flow rate on the MFC. Once a stable segmented flow regime was observed and the pressure and the temperature of the reactor were stabilized, the system was ready for feed injection. For small-scale reactions (parameter optimization), the starting material solutions (prepared under Ar atmosphere) were injected by using 2-mL sample loops (PFA tubing, 1/16" OD, 0.80 mm ID).
In each runs, the product stream was collected for 5 min after reaching steady state, which was next concentrated under reduced pressure and analyzed by 1 H-NMR and chiral HPLC.
CAUTION: H2 is extremely flammable, therefore extreme care must be taken when handling. All equipment must be set up in a well-ventilated fume hood. A thorough safety assessment should be made before conducting any experiments.   No dependence of conversion, chemo-, diastereo-or enantioselectivity was found utilizing different fluid flow rates in the range of 2 × 50-200 µL min -1 (P1 and P2) and different gas flow rates in the range of 5-25 mLn min -1 (c2= 0.2 M, cbenzylamine= 0.2 M, T= 100 °C, P= 3 bar). These results are therefore not represented in details.
In case of lower temperatures or an excess of 2 (Table S11, entries 1, 2, 7, 8 and Table S13, entry 2), side product 3a was formed via unwanted double alkyation as corroborated by 1 H-NMR and mass spectrometry ( Figure S4). Figure S4. 1 H-NMR and mass spectra of side product 3a.
In order to achieve lactam 3 in multigram scales, numerous preparative runs were carried out under optimum flow conditions (see : Table S13, entry 8). The system proved stable during the long runs, and resulted around 4 g h -1 of pure product with isolated yields in the range of 96-99% and ees of 96%. The product obtained was sufficiently pure without chromatographic purification, and was used in the next step directly after evaporation.
According to ICP MS measurements, practically no leaching occurred from the catalyst bed during the reactions (Pt contents detected in the crude product samples were in the range of 5-6 ppb). The complete optimization study and all the preparative runs were fulfilled by a single catalyst cartridge containing merely 10 mg of Pt (200 mg 5% Pt/C). In these experiments, approximately 75 mmol substrate was transformed resulting around 25 g lactam 3. These gave an effective catalyst loading of 0.07% and a turnover number of 1430.

Synthesis of phenylpiperidin 1: BH3-mediated amide/ester reduction
The starting material solutions were pumped as separate feeds (P1 and P2) by using a UNIQSIS Binary Pump Module equipped with two high-pressure HPLC pumps, two injection valves with sample loops and a pressure sensor to monitor system pressure. The liquid streams were combined in a Y-mixer at room temperature and the resulting solution was directed through a 12-mL reaction coil (PFA tubing, 1/8" OD, 1.58 mm ID) which was heated in an oil bath. The flow system was pressurized by applying a 10-bar BPR from IDEX. In each runs, the product stream was collected for 3-5 min after reaching steady state. In order to safely decompose the unreacted reducing agent, the stream exiting the reactor was collected into a flask containing a well-stirred 1:1 mixture of 3 M HCl and 2-MeTHF. After the collection period, the solution was refluxed for 30 min in order to remove BH3 adducts and was next treated with NaOH solution (2 M) until pH 10. The resultant mixture was extracted three times with EtOAc. The combined organic layers were washed with brine and dried over Na2SO4. The filtrate was concentrated under reduced pressure and analyzed by 1 H-NMR and chiral HPLC.
CAUTION: Borane reagents are extremely dangerous. They decompose thermally or in the presence of atmospheric moisture, water and acids resulting flammable gases (B2H6 and H2) and boric acid (possible blockage in reactor channels). Extreme care must therefore be taken when handling. Dry conditions must be ensured during experimentation and all equipment must be set up in a well-ventilated fume hood. A thorough safety assessment should be made before conducting any experiments.

S14
At temperatures around 100-120 °C, occasional gas formation and precipitation occurred in the reaction coil. To ensure stable and safe operation, reaction temperature was maximized in 90 °C during parameter optimization.  Under optimum flow conditions (see : Table S16, entry 6), a preparative-scale run was carried out by using neat BH3ꞏDMS as reducing agent. For this, the product stream was collected for 30 min after reaching steady state. After extractive work-up, 2.28 g crude product was obtained. According to 1 H-NMR measurements, the material was acceptably pure without chromatographic purification. However, in order to remove dimethyl sulfide traces (smelly in very low concentrations), column chromatographic purification was carried out using a mixture of ethyl acetate/40-60 petroleum ether as eluent in the presence 1% trimethylamine as additive. After purification, 1.97 g of 1 was isolated (84% yield), and ee was 96%. The process ensured an outstanding productivity of 3.94 g h -1 of pure product.

Telescoped flow synthesis
Aldehyde 2 was obtained in continuous flow organocatalytic conjugate addition between 4-fluorocinnamaldehyde and dimethyl malonate under solvent-free conditions as described in section 3.5 and was used directly after removal of unreacted reaction components by evaporation.
The optimum conditions (Table S13, entry 8 and Table S16, entry 6) determined during the step-by-step experiments were taken into account when designing the telescoped sequence. In order to match flow rates, the 12-mL reaction coil used for the amide/ester reduction was exchanged to a 9.3-mL one.
For the tandem reductive amination-lactamization, 2.0 M solutions of 2 and benzylamine were prepared in dry 2-MeTHF (under Ar atmosphere) and were pumped as separate feeds (P1 and P2) at 100 µL min -1 from 15-mL sample loops with dry 2-MeTHF as carrier solvent by using a UNIQSIS Binary Pump Module. (6-port injection valves were integrated into the pump module.) The liquid streams were combined in a Y-mixer at room temperature. H2 gas was introduced into the system from a gas cylinder using a calibrated MFC (Bronkhorst-EL). The inclusion of a check valve prevented any backflow of liquid towards the MFC. The liquid and gaseous streams were combined in a second Y-mixer at room temperature. The resulting gas-liquid feed entered a 4.6 × 100 mm stainless steel column packed with a mixture of 200 mg of 5% Pt/C and 400 mg of activated charcoal. (Prior to the experiment, dry THF was pumped through the packed bed at 200 µL min -1 for 45 min to remove water traces from the catalyst.) The Pt/C column was heated at 100 °C by using a Phoenix Flow Reactor™ (ThalesNano); 3 bar was maintained by a fixed-pressure BPR from IDEX. During reductive amination, one equivalent of water is released which must be removed in order to prevent decomposition of BH3ꞏDMS downstream. The gas-liquid mixture exiting the Pt/C column was therefore passed through a 10 × 100 mm stainless steel column packed with 5 g of freshly activated 4 Å MS. H2 gas was separated from the liquid mixture through a buffer flask. The dried and degassed stream containing approximately 1 M 2-MeTHF solution of lactam 3 was re-incorporated for the subsequent amide/ester reduction through a 3-port valve by using a Knauer Azura P 4.1S HPLC pump (P3) at a flow rate of 200 µL min -1 . Neat BH3ꞏDMS was streamed from a 20-mL sample loop with dry 2-MeTHF as carrier solvent by using a Knauer WellChrom K-120 HPLC pump (P4) and a manual 6-port injection valve from IDEX. Both liquid lines were pressurized by 3-bar BPRs from IDEX and were combined in a Y-mixer at room temperature. The resulting solution was directed through a 9.3-mL reaction coil (PFA tubing, 1/8" OD, 1.58 mm ID) which was heated at 90 °C in an oil bath. The system was pressurized applying a 10-bar BPR from IDEX. Pictures of the telescoped setup can be found in Figures S5 and S6.
A typical procedure for the experiment is as follows: first, the carrier solvent flows were started (P1, P2, P3 and P4). When the pressure stabilized on the Pt/C column, the desired temperature was set on the Phoenix Flow Reactor™ and the gas flow was initiated by setting the desired flow rate on the MFC. At the same time, the coil reactor was also heated up. Once a stable segmented flow regime was observed after the gas-liquid mixer and the pressure and the temperature of the reactors were stabilized, the system was ready for feed injection. Solutions of 2 and benzylamine were injected first. After 15 min, the reductive amination-lactamization stream exiting the 4 Å MS column reached steady state. Then the output was placed into the gas separator, and after 5 more min, P3 was switched from 2-MeTHF to the degassed stream of 3. Simultaneously, the neat BH3ꞏDMS feed was initiated by injection. The product stream exiting the heated reaction coil was collected continuously for 100 min after reaching steady state (1 h after injection of 2 and benzylamine). In order to safely decompose the unreacted reducing agent, the outcome from the reactor was directed into a flask containing a well-stirred 1:1 mixture of 3 M HCl and 2-MeTHF. After the collection period, the solution was refluxed for 30 min in order to remove BH3 adducts and was next treated with NaOH solution (2 M) until pH 10. The resulting mixture was extracted three times with EtOAc. The combined organic layers were washed with brine, dried over Na2SO4 and concentrated under reduced pressure. Column chromatographic purification was carried out using a mixture of ethyl acetate/40-60 petroleum ether as eluent in the presence 1% trimethylamine as additive. The results of the telescoped experiment are summarized below in Table S17.  S17 Figure S6. Details of the telescoped system.