Continuous Flow Chiral Amine Racemization Applied to Continuously Recirculating Dynamic Diastereomeric Crystallizations

: A new, dynamic diastereomeric crystallization method has been developed, in which the mother liquors are continuously separated, racemized over a ﬁ xed-bed catalyst, and recirculated to the crystallizer in a resolution − racemization − recycle (R 3 ) process. Separating the racemization from crystallization overcomes problems of using catalysts in situ , that su ﬀ er con ﬂ icting sets of conditions, inhibition, and separation. Continuous racemization has been achieved through the covalent attachment of [IrCp * I 2 ] 2 SCRAM catalyst to Wang resin solid support to give a ﬁ xed-bed catalyst. One tertiary and a variety of secondary optically enriched amines have been racemized e ﬃ ciently, with residence times compatible with the crystallization (2.25 − 30 min). The catalyst demonstrates lower turnover (TOF) than the homogeneous analogue but with reuse shows a long lifetime (e.g., 40 recycles, 190 h) giving acceptable turnover number (TON) (up to 4907). The slow release of methylamine during racemization of N -methyl amines was found to inactivate the catalyst, which could be partially reactivated using hydroiodic acid. Dynamic crystallization is achieved in the R 3 process through the continual removal of the more soluble diastereomer and supply of the less soluble one. The solubility of the diastereomers was determined, and the di ﬀ erence correlates to the rate of resolution but is also a ﬀ ected by the rates of racemization, crystal growth, and dissolution. A variety of cyclic and acyclic amine salts were resolved using mandelic acid (MA) and ditoluoyl tartaric acid (DTTA) with higher resolvability ( S = yield × d.e.) than the simple diastereomeric crystallization alone. Comparing resolvabilities, resolutions were 1.6 − 44 times more e ﬀ ective with the R 3 process than batch, though one case was worse. Further investigation of this revealed an unusual thermodynamic switching behavior: rac - N -methylphenethyl-amine was initially resolved as an ( S , S )- bis -alkylammonium tartrate crystal but over time became the equivalent ( R , S ) salt. Thermal, mixing, concentration, stoichiometry, and seeding conditions were all found to a ﬀ ect the onset of the switching behavior which is only associated with difunctional resolving reagents. sample were dissolved in dichloromethane, dried in vacuo , and analyzed by 1 H NMR (CDCl 3 , 500 MHz). The pump was started, and samples were taken as described after each complete recirculation of the system volume, for the designated number of cycles. The activity of the catalyst was determined at the end of the reaction by pumping the optically active amine through the heated ﬁ xed bed for 3 − 4 reactor volumes (i.e., steady state) (Supporting Information section 9).


■ INTRODUCTION
When racemization is used in combination with chiral resolution, it is a powerful alternative to the asymmetric synthesis of chiral amines. These intermediates are used widely in the pharmaceutical and fine chemical industries but are often difficult to make in large quantities and as a result can be costly. 1−4 While catalytic asymmetric synthesis has largely superseded inefficient stoichiometric methods, both chemoand biocatalysts often perform poorly with highly functionalized compounds with issues of regio-and enantioselectivity, competing ligands, solubility, and product separation. 5 Racemic starting materials are commonly resolved using enantioselective enzymes 6 or by forming and breaking diastereomeric salts. 7 These processes are used frequently in industry, as they are simple and reliable, but they suffer low yields and high wastes. Improvements employ in situ racemization of the unwanted isomer, with the techniques known as dynamic kinetic resolution (DKR), crystallizationinduced asymmetric transformation (CIAT), and crystallization-induced diastereomeric transformation (CIDT), Figure  1. 8−16 DKR employs enzymes to react selectively with one enantiomer while racemizing the other in situ, Figure 1. The technique requires an enantioselective enzyme and substrate that racemizes under the same reaction conditions. The requirement for solvents, reactants, temperatures, and concentrations that are compatible with both catalysts compromises the optimal activity of each and reduces the overall DKR performance. CIAT works for compounds that form conglomerate crystals. The entropic barrier of going from a racemate to a single enantiomer can be overcome by adding seeds and mechanically grinding to make smaller particles, causing secondary nucleation, supersaturation, and regrowth. 17,18 Despite its elegance, the method is limited in scope because too few chiral compounds crystallize with the required space group and have easily racemizable chiral centers. 19−21 CIDT involves establishing a dynamic thermodynamic equilibrium between the salts epimerizing in solution and crystallization of the less soluble diastereomer. The equilibrium adjusts by dissolution of the more soluble diastereomer and results in a higher yield of the resolved solid, Figure 1. 22 Each of these techniques requires compounds that racemize easily under resolution conditions, relying on substrates with an acidic or enolizable chiral proton. However, racemization and resolution processes compete, with the former being faster at higher temperatures while the latter operated preferably under ambient conditions. This limitation can be overcome by continuously flowing the soluble reactants across a solid supported, heated, racemization catalyst and cooling back for the resolution. Furthermore, when kept separate, interference between the resolution and racemization is avoided, and the product is more easily separated from the catalyst.
A way to increase the scope of amine racemizations beyond those that rely on acidic, tautomerizable chiral centers is to employ hydrogen transfer catalysts which are able to dehydrogenate unactivated amines. Heterogenous amine racemization catalysts such as Pd/C or Pd/BaSO 4 lend themselves well to this but can suffer poor activity and selectivity. 23,24 A variety of homogeneous catalysts are known to borrow and replace the proton at the chiral center of secondary alcohols; however, few have been identified that perform a similar hydrogen transfer reaction with chiral amines. 25 Catalysts that have proven effective in this regard are the ruthenium based Shvöand analogues 26 and iodoiridium, SCRAM. 16 The former is better able to racemize primary amines, and the latter, secondary and tertiary chiral amines. This is because the iridium center coordinates and activates primary imines toward N-alkylation, resulting in dimerization. 27 Both of these homogeneous catalysts have been used in enzymatic DKR of amines but not in CIDT. 28−30 A silica-supported Shvöcatalyst has been reported recently for the transfer hydrogenation of levulinic acid 31 but is not reported for racemization. Similarly, a polymer supported chloro-iridium catalyst, tethered by a functionalized Cp* ligand, is reported for the reduction of benzaldehyde and repeated reuse. 32 The same supported metal, modified with chiral ligand, has shown a long lifetime in the continuous flow asymmetric transfer hydrogenation of ketones. 33 There are several ways in which a combined resolution and racemization can be configured. A DKR example reported by Falus et al. uses solid supported subtilisin enzyme to resolve an amine by esterifying selectively one enantiomer. 34 The mixture, containing the remaining alcohol enantiomer, is flowed across a fixed-bed racemization catalyst and the sequence repeated across 10 columns to give the amide in 98% e.e. and 79% yield with productivity of 8 g/L/h running over a week. In another example, de Miranda et al. packed into one column layers of solid-supported CalB lipase (to resolve phenethyl alcohol) and VOSO 4 (to racemize the unreacted alcohol enantiomer) using toluene at 70°C. 35 Four resolution and three racemizing layers were enough to generate the product in 90% e.e. and 82% yield. To avoid sequential columns, each separately supported The Journal of Organic Chemistry pubs.acs.org/joc Article catalyst can be intimately mixed; however, thermal control of each is then lost. Nevertheless, Farkas et al. achieved high e.e.'s and conversion in the DKR of benzylic primary amines using a supported CalB lipase and palladium-based hydrogen transfer catalyst. 36 Another operational mode is to continuously recirculate the solution between resolution and racemization, though this appears not to have been evaluated for dynamic crystallizations. This involves separating crystals from mother liquors and racemizing and recirculating them. The work described here starts by preparing an immobilized iodo-iridium racemization catalyst 1, exploring its use in continuous flow racemization, and then integrating this with diastereomeric crystallization to give the resolution−racemization−recycle R 3 process.

■ RESULTS AND DISCUSSION
Continuous Flow Racemization. The immobilized SCRAM catalyst, 1, was prepared as described previously, 32 by reacting the homogeneous chloro-iridium Cp*C 5 OH complex 3 first with excess sodium iodide to exchange the halide for an iodo-ligand, 4, and then activating the alcohol by triflation to make 5, which was reacted in situ with Wang resin, Scheme 1. After thorough washing of the beads, the amount of iridium supported on the resin was found by ICP to equate to 0.39 mmol of Ir/g.
To integrate the racemization and resolution processes, it is necessary to have a common solvent. It was found that the immobilized catalyst 1 is active in most solvents, except dipolar aprotics that may coordinate to the metal. This is convenient, because the selection of crystallization solvent is critical to the efficiency of the diastereomeric resolution. Solvents that are preferred for both steps maximize the difference in solubility between the diastereomeric salts and have a high enough boiling point to increase the racemization rate, but below 120°C , the point at which the polymer support breaks down. It was shown that alcoholic solvents, that act as hydrogen donors, are useful cosolvents, as they can increase the rate of racemization by enhancing imine reduction and reduce sidereactions. 37 Racemization in continuous flow was tested with optically pure amines 6−12, Figure 2.
A mass of immobilized catalyst 1, sufficient to give 5−10 mol % iridium loading, was mixed with washed sand, to manage swelling and compaction, and packed into an empty 4 mL, 23 cm HPLC column. The reactor's fluid volume was  measured by solvent mass difference as 1.5 mL. The column was placed in an electrically heated aluminum block, and amine solution was pumped at a temperature and flow rate enough to cause a significant amount of racemization in a short residence time. Table 1 shows the results of continuous flow racemization of optically pure amines 6−12 at steady state (Supporting Information section 7), measured as the fall in e.e. per minute and converted to turnover frequency (TOF) per hour. The primary amine (S)-6 is racemized slowly, entry 1, the reason being that side-reactions associated with the intermediate imine cause liberation of ammonia that deactivates the catalyst. An attempt to suppress the side reactions by adding the hindered hydrogen donor 2,4-dimethylpentanol 50% (v/v) was of limited success. 30 Benzylic amines (S)-7 to (S)-9 racemized with rates of 0.37−2.87% e.e./min at steady state. With flow rates adjusted to give residence times of 12−30 min, racemizations of 43, 14, and 11% ee for (S)-7, (S)-8, and (S)-9, respectively, were achieved in one pass, entries 2, 4, and 5. The TOFs of 1 are significantly less than the homogeneous catalyst [IrCp*I 2 ] 2 used with these substrates in batch (Supporting Information section 7); however, this is compensated using 5−10 mol % of 1. A 10-day continuous flow racemization was performed on (S)-7 to determine the impact of catalyst deactivation on extended use, Figure 3.
On day 1, the e.e. reached steady state after two to three vials, falling from 98% (stock solution) to 12.5% ee. Thereafter, the activity of the catalyst fell almost linearly at 2.8% e.e./h, until after 10 days the amine was only racemized by 13% e.e, with a TON of 30 compared with 187 at the start. Nevertheless, the cumulative TON was 4907, showing an effective catalyst use of 0.02 mol %. The cause of deactivation has been determined as methylamine coordination to the iridium which arises from hydrolysis of the imine intermediate. 38 Reducing the moisture content or adding isopropanol hydrogen donor reduces but does not eliminate the catalyst deactivation. The addition of potassium iodide at day 9 had little effect on reversing the inhibition; however, it was found that treating the spent catalyst with dilute hydroiodic acid (0.1 M) for 2 h recovered about 25% of the activity. More concentrated acid or extended exposure led to cleavage of the catalyst from the resin. Unfortunately, this reactivation procedure is tedious and requires extensive washing to remove the acid. As it was only partially successful, emptying and repacking the fixed bed using fresh catalyst was adopted instead.
Salsolidine (S)-10 was racemized at 37.8% e.e./min, so a residence time (t res ) of 2.25 min gave a fall of 85% e.e., entry 6. The same packed column of catalyst was reused in different experiments 130 times, giving a total TON of 1105. The mother liquor solution derived from the diastereomeric crystallization of 10 with (S)-mandelic acid ((S)-MA) gave a reduced TOF of 25.8 h −1 ; nevertheless, in a single pass, or 1 reactor volume, the catalyst was able to epimerize the diastereomer by 26% d.e., indicating that continuous racemization of crystallization mother liquor (ML) would be feasible, entry 7. The possibility that the epimerization was caused by racemization of (S)-MA, or the mandelate anion, rather than the amine, was tested by stirring (S)-MA or its mandelate salt with catalyst 1 in batch under the same conditions as entry 7. Neither of them showed any racemization. However, after 12 h, 3.5% phenylglyoxalate was formed from the mandelate salt by dehydrogenation of the alcohol group. Since epimerization of the diastereomeric salt involves only the amine, the process is hereinafter described as racemization. Surprisingly, the racemization of heterocyclic amine (S)-11 failed when methyl tert-butyl ether (MTBE) and iso-propyl alcohol were used, entry 8. However, iso-propyl acetate solvent worked well, giving a racemization rate of 13.2% e.e./min with a catalyst TOF of 158 h −1 , entry 9. The continuous flow epimerization of sertraline (1S,4S)-12 in toluene at 80°C was successful with a catalyst TOF of 252 h −1 , though over 9 reaction volumes the activity fell by half. As with amine 7, a linear deactivation of the catalyst was found with 12, at a rate of 14% d.e./h which is also caused by methylamine. 38 Development of a Resolution−Racemization−Recycle Process. With a continuous flow racemization method in hand, we sought to integrate this with the diastereomeric The Journal of Organic Chemistry pubs.acs.org/joc Article crystallization. Direct addition of the homogeneous SCRAM catalyst to the crystallization was unsuccessful because the racemization rates were too low at ambient temperature. Furthermore, the homogeneous catalyst interfered with the crystallization (the crystals formed were dark rather than white solids) and made purification of the product more difficult. This provided the rationale for the recycle system using immobilized catalyst 1. An important consideration was the acidity of the system, as carboxylic acids are known to retard hydrogen transfer by reacting with the iridium hydride intermediate to form dihydrogen (hence the use of triethylammonium formate as hydrogen donor rather than formic acid 39 ). It was already shown that 1 was active in racemizing the diastereomeric salt, Table 1, entry 7, and that (S)-MA salt was insensitive to racemization. To ensure a neutral or basic solution, a slight under-charge of the resolving acid was used at the start of the work; however, it was later shown as unnecessary and a 1:1 stoichiometry was satisfactory.
The equipment used to integrate the resolution, racemization, and recycle steps is shown in Figure 4.
A temperature-controlled crystallizer with gas-tight PTFE lid, drilled with holes to support the in-and outflow tubes, is gently stirred with a magnetic stirrer bar, and the system is a closed loop. Solutions of the acid and amine were introduced in a solvent at a concentration enough to cause precipitation of solid. The solubility of each diastereomer was measured to facilitate selection of the optimal concentration. The mother liquors, that are enriched with the more soluble diastereomer, were separated from the solid using a sintered glass in-line filter that pumped the saturated solution into the electrically heated racemization column. The hot outflow is now enriched in the less soluble diastereomer and subsaturated in the more soluble one. As the solution is returned to the crystallizer, it becomes supersaturated in the less soluble diastereomer which further crystallizes. This happens as the more soluble diastereomer dissolves, thus providing a dynamic equilibrium, Scheme 2.
The scope of the system was tested with (S)-MA salts of amines 7−11, Table 2 The solvent and substrate concentrations were selected from those that gave a substantial amount of solid in batch at ambient temperature. The yield of solid produced during the reaction was inferred by subtracting the concentration of soluble amine (and derived impurities) from the initial amount using calibrated GC. The d.e. was measured from a sample of the solid, dried, and dissolved in solvent for analysis by chiral HPLC, 1 H NMR, or GC to determine the e.e. of the free amine generated from the salt. The solubility of each pure diastereomer was measured by analyzing its saturated solution in the corresponding solvent. It is useful to compare the performance of the batch and R 3 processes; however, yield and d.e. can be competing functions and determining optimum conditions can be time-consuming. Resolvability, S, is the product of both variables and can be useful to compare outcomes. In a perfect batch resolution, S max = 1.0, i.e., 2 × 50% yield × 100% d.e., and in a dynamic system, S max = 2.0. 40 The resolvability of 7-(S)-MA was found to be the same in batch diastereomeric crystallization as the R 3 process, S = 0.36, Table 2, entries 1 and 2; however, the solid yield is 30% with 60% d.e. in the R 3 process but 49% yield with 37% d.e. in batch. Acetophenone, formed by hydrolysis of the intermediate imine, was seen to accumulate slowly over 190 h to about 10% of the reaction mass. Since methylamine is a potent catalyst inhibitor, catalyst activity was assayed at the end of the experiment and was 42% less active than at the start. Nevertheless, this does not explain the low solid yield in the R 3 process; rather, this indicates a change in crystallinity, perhaps because of the impurities or crystal growth inhibition. A different resolving reagent gave better results, entries 11 and 12, but also exposed some unusual behaviors that are discussed later.
There is a large difference in the solubilities of the 8-(S)-MA diastereomers, which provides a good driving force for the dynamic resolution, Table 2, entries 3 and 4. The R 3 process was run at a flow rate of 0.1 mL/min, giving a recycle time of 4.75 h. The d.e. steadily increased to 89% after 30 recycles with an isolated solid yield of 62%, Figure 5.
The concentration and e.e. of amine 8 in the mother liquors (MLs) fell gradually over the course of the experiment. However, the rate of fall does not correlate with the solid, indicating that the rate-limiting step may be the growth or dissolution of the solid, not racemization. It would be interesting to determine these parameters, as the effect of constant supersaturation of one diastereomer and subsaturation of the other may be different from those in the batch diastereomeric crystallization. The resolvability of amine 8 was improved to S = 1.1, compared with the batch diastereomeric crystallization S = 0.67. Based on the measured solubilities, the

Scheme 2. Proposed Mechanism for the Formation of the Predominant Diastereomer
The Journal of Organic Chemistry pubs.acs.org/joc Article predicted yield is 80%, while the isolated yield was 62%; the difference is due mainly to formation of acetophenone (16.7%). Resolution of amine 9 using (S)-MA failed, as the diastereomeric salts failed to crystallize in any of the solvents tested; in iso-propyl acetate, the solubility at ambient temperature exceeded 43 mg/mL, entries 5 and 6. Instead, the diacid resolving reagent (S,S)-DTTA was tried with results discussed later, entries 13 and 14.
Racemic 1-methyl tetrahydro-iso-quinoline 10 forms diastereomeric salts easily with (S)-MA, and a solvent composition of ethyl acetate and iso-propanol (7:1 v/v) was found to give a high d.e. solid at ambient temperature, Table 2, entries 7 and 8. The batch diastereomeric crystallization gave 80% yield and 7% d.e., S = 0.11; however, the R 3 process gave a resolvability of S = 1.25. Figure 6 shows the reaction profile with change in solid yield, d.e., and impurities over 30 recycles; each recycle was 48 min, at a flow rate of 0.5 mL/min.
After one recycle, 83% of the amine forms a solid and this remains relatively constant; however, the initial 44% d.e. drops to 17% after one racemization recycle and then slowly rises over the next 30 recycles (24 h) to finish at 96% d.e. The improved performance of R 3 was encouraging, but a limitation identified was that the yield of the solid is determined by the solubility of the least soluble diastereomer. The solubility of both diastereomers was measured as 3.0 and 0.7 mg/mL. This 4.3-fold difference explains the driving force on the dynamic equilibrium but also that 9% of the mass remains in solution, giving a theory maximum yield of 91%, versus the 86% yield observed. The remaining mass is the side-reaction impurities dihydro-and iso-quinoline resulting from transfer dehydrogenation. However, they remained at low levels using iso-propanol as a hydrogen donor which serves to reduce them back into 10.
When (S)-MA was added to racemic amine 11 in batch, 53% solid of 20% d.e. formed immediately, Table 2, entries 9 and 10. Applying these conditions to the R 3 process, a flow rate of 0.25 mL/min gave a recycle time of 114 min and racemization residence time of 6 min at 80°C. The resolvability in the R 3 process was S = 0.93 compared to S = 0.11 in batch, with the solid d.e. improved from 20 to 89% but little change in yield. The solubility of both diastereomers was determined as 7.5 and 2.4 mg/mL, giving a theoretical yield of 61% solid versus 52% isolated. A yellow colored impurity formed during racemization was tentatively assigned by MS as a trimeric species. 41 A similar trend was observed with d.e. increasing with time, reaching 89% after 16 recycles (30 h), Figure 7.
The e.e. and concentration of the amine in the mother liquors fell gradually, though at a higher rate than the solid, perhaps indicating crystal growth or dissolution as rate limiting. The activity of the catalyst was determined at the end of the experiment and had the same TOF as at the start,    The Journal of Organic Chemistry pubs.acs.org/joc Article though after two further runs of similar duration it was found to have lost half of its activity.
To improve the resolution of amine 7, a half mole equivalent of chiral acid, (S,S)-DTTA, was tested, Table 2, entries 11 and 12. Using ethyl acetate with 20% (v/v) iso-propanol, an insoluble precipitate was isolated after 6 h in 84% yield. Analysis indicated a 2:1 amine:acid salt of only 2% d.e. The diastereomeric crystallization was carried out in the presence of homogeneous SCRAM catalyst in a one-pot recycle system. The concentration of amine 7 in solution and d.e. of the solid remained at 66 and 75%, respectively, for about 25 h. Surprisingly, a sample taken after 30 h revealed a significant drop of concentration of 7 in solution to <40% and the d.e. of the solid decreased to about 6%. 1 H NMR analysis of the DTTA showed no epimerization, and the 2:1 stoichiometry of the salt was the same. The same system was tested in batch without racemization and was sampled with time. It was found that the initial d.e. of the solid formed was 70% but fell to 6% over 6−12 h, Figure 8.
Crystals taken at the start and end of the experiment were analyzed by powder and single crystal X-ray diffraction, Figure  8. It was found that a switch in amine enantiomers within the solid phase had occurred, going from the [(S)-7] 2 -(S,S)-DTTA diastereomer to a, presumably more stable, meso-like form [(S)-7][(R)-7]-(S,S)-DTTA containing a racemate of both amine enantiomers, hence the observed fall in d.e. The experiment was repeated twice with similar results. Ferreira et al. have observed this behavior previously with rac-6-(S,S)-DTTA, and they developed a model to explain the different solid forms that were related to solubility, acid:amine ratio, and pH. 42 The unusual switch in solid form was investigated further, Figure 9. Figure 9a shows that the higher the temperature, the more rapid the switch from [(S)-7] 2 -(S,S)-DTTA to [(S)-7][(R)-7]-(S,S)-DTTA, which may be related to increased mass transfer and nucleation rate. Figure 9b shows that a 2:1 stoichiometry of the amine and diacid is required, and with only a quarter equivalent of (S,S)-DTTA to (S)-7, the [(S)-7] 2 -(S,S)-DTTA diastereomeric salt remains stable over 36 h. Figure 9c illustrates that higher concentrations of (S)-7 cause a faster rate of switching at the same 2:1 amine/acid stoichiometry, presumably because of an increased nucleation rate. Similarly, higher rates of mixing cause more crystal attrition and are likely  Table 2, entries 11 and 12. A resolvability of 1.34 was achieved, a 44-fold improvement compared to diastereomeric resolution alone. It is interesting to speculate whether other diastereomeric crystallizations that employ diacids might be susceptible to this effect: initially, the diastereomer is formed under kinetic control, giving a metastable product, and with time, a more stable crystal morphology is formed. When resolving reagents and conditions are screened, samples are rarely taken over time, so a low d.e. will be dismissed as a failure. Though of little practical use, as no resolution is achieved, the thermodynamic phenomenon appears to be the opposite of CIAT that is under kinetic control. Unlike amine 7, the N,N-dimethyl analogue 9 forms a 1:1 rather than 2:1 salt with (S,S)-DTTA, using either 0.5 or 1 equiv of the diacid. A solvent screen was carried out and ethyl acetate−ethanol (7:3) gave the best yield (42%) and d.e. (59%) with S = 0.49, Table 2, entries 13 and 14. Since the racemization of 9 was found to be slow, a residence time of 2.5 h was selected as well as a whole process recirculation time of 14.3 h. At the start, a gradual increase in d.e. of the solid was  The Journal of Organic Chemistry pubs.acs.org/joc Article observed; however, its yield decreased as the solution concentration of 9 increased, concurrent with its racemization. The reaction was run over 242 h, with a final yield of 17% solid with 77% d.e., S = 0.26, worse than that in batch. The apparent change in diastereomer solubility was shown not to be epimerization of (S,S)-DTTA, so a hypothesis that the less soluble (S)-9-(S,S)-DTTA crystallizes more slowly than dissolution of the more soluble (R)-9-(S,S)-DTTA diastereomer was tested. Both diastereomers are less soluble in isopropanol:iso-propyl acetate (3:7); however, running the R 3 process showed no improvement in either the yield or d.e. of (S)-9-(S,S)-DTTA; instead, all of the crystals dissolved after 15 recycles. The cause of this is not clear, but an explanation might be the iminium−enamine intermediates, that occur during racemization, inhibit crystal nucleation/growth.

■ CONCLUSIONS
An immobilized iodo-iridium racemization catalyst has been prepared, characterized, and used in fixed-bed continuous flow. Though its activity is around 10 times lower than its homogeneous counterpart. When used in the fixed bed at  38 Reducing the moisture content of the system and using isopropanol as a cosolvent and hydrogen donor suppressed this side-reaction. A partial catalyst reactivation was identified using hydroiodic acid, but a lengthy washing procedure precluded its further use. The continuous racemization was applied in a new resolution−racemization−recycle (R 3 ) process in which tubing The Journal of Organic Chemistry pubs.acs.org/joc Article connected the racemization catalyst to the crystallizer, allowing continuous recirculation of the mother liquors. This allowed both reactions to be operated separately at different temperatures. The resolvability of amines 8−11 (S)-MA salts was improved using the R 3 process by 1.6−11 times over batch.
The exception was amine 7, but its resolution was more successful using the diacid (S,S)-DTTA. In this case, a surprising effect was observed, when, instead of harvesting the solid soon after it was formed, it was left until the next day, the d.e. dropped from >70% to almost zero. Investigation of the crystal structures showed a change in molecular structure, producing a more stable meso-like or "neutral" form in which the chiralities of both N-methylphenethylammonium molecules were (R) and (S) rather than being both (S) as they were at the start. 42 The racemization of DTTA was eliminated. The switch in structure appeared to be stochastic and affected by temperature, mixing, concentration, stoichiometry, and seeding. These parameters are those that affect secondary nucleation rates, like those seen in CIAT and in polymorph switches. By slow addition of the diacid with (fast) racemization of (R)-7, the switch to the meso-like form can be avoided, giving a 44-fold improvement in resolvability over the control diastereomeric crystallization, achieving 78% yield and 86% d.e. The difficulty this exposes is that a good or bad resolution with a diacid depends on when the sample is taken; a low d.e. might be dismissed as a failed set of conditions, whereas a sample taken with high d.e. might cause problems later on when the crystallization is held. A different explanation is needed for the failure to resolve N,N-dimethylphenethylamine 9 which forms a 1:1 salt with (S,S)-DTTA. In this case, racemization and recirculation cause dissolution of the solid which might be accumulation of impurities that slow crystal nucleation and growth but not dissolution. Further work is required to confirm this. As with other CIDT processes, the increased optical purity is likely to occur through dissolution of the more soluble diastereomer, and crystallization of the less soluble one, and is driven by the difference in concentration between the raffinate and racemized feed. As material of low d.e. is returned to the crystallizer, it is enriched in the less soluble diastereomer and causes its supersaturation, while the dynamic equilibrium is maintained by dissolution of the more soluble one. Since the dynamic system in the R 3 process in most cases allows formation of the most thermodynamically stable diastereomer, the yield and d.e. are both determined by the diastereomer solubilities and the resolution rates by the racemization, crystal growth, and dissolution. More work is required to quantify and model these. Prior determination of these parameters in different solvents might overcome the usual need for careful solvent, temperature, and chiral acid screening to find the optimal resolution conditions. The R 3 process provides a simple, reproducible, low cost, and low waste method of producing chiral amines. If the catalyst activity could be increased further, the process might be amenable for use in chiral amine manufacture and studies on this are in progress. Continuous flow racemization might also be applied to CIAT, enzyme DKR, and chiral chromatographic resolution processes. Potential applications are in the end-of-pipe resolution processes to recover waste amine, for example, in chromatographic, enzymatic, or crystal resolution processes.

■ EXPERIMENTAL SECTION
General Information. Unless otherwise stated, all chemicals were obtained from Sigma-Aldrich, Fisher Scientific, Alfa Aesar, or Fluorochem and were used without further purification. All solvents used were HPLC grade. The syringe pumps used were Harvard Apparatus Model 11 from Scientific Support. Column chromatographic purifications were performed using Biotage (Isolera Spektra One, silica column = RediSepR f ). Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker 500 UltraShield Spectrometer operating at 500 MHz ( 1 H NMR) and 126 MHz ( 13 C NMR). Liquid chromatography−mass spectrometry (LCMS) was performed on an Agilent Technologies 1200 Series LC system with a C18 column (Phenomenex Luna 5u C18(2), 50 mm × 2 mm × 5 μm) eluted with an acetonitrile/water gradient (15−95% acetonitrile, 0.1% formic acid over 3 min) and a Bruker Daltonics HCTUltra system equipped with an ion trap MS detector, or a Thermo Ultimate 3000 UHPLC system with a C18 column (Phenomenex Kinetex, 50 mm × 2.1 mm × 2.6 μm) eluted with an acetonitrile/water gradient (2−95% acetonitrile, 0.1% formic acid over 1 min) and a Bruker Amazon Speed ion trap mass spectrometer. High resolution mass spectrometry (HRMS) was performed on a Bruker Maxis Impact Electron Spray Ionization (ESI) spectrometer with a qTOF mass analyzer. For clarity in the assignment of MS peaks of the amine salts, the cationic and anionic fragments were assigned as X and Y, respectively. Melting points were recorded using a Stuart Scientific SMP3 melting point apparatus. FT-IR was recorded using a Bruker Platinum-ATR spectrometer. Optical rotations ([α] D ) were measured using a Polartronic H532 at 589 nm. Achiral and chiral gas chromatography (GC) were used to determine the conversion and enantiomeric excess (ee) of the substrates (7, 8, 10, 11, and 12), respectively. They were performed using a HP6890 series GC system, Agilent Technologies 7683 series injector, and HP7683 series autosampler. The e.e. of 10 and diastereomeric excess (d.e.) of mandelate salts were analyzed by chiral high performance liquid chromatography (HPLC) using an Agilent Technologies 1290 Infinity system, while the e.e. of (S)-MA was analyzed by chiral HPLC using an Agilent 1100 Series system. The calculations of resolvability, yield, residence time, and total process time are given in Supporting Information section 1. The supported SCRAM catalyst 1 was prepared as previously reported. 32 Optical rotations ([α] D ) were measured using a Polartronic H532 at 589 nm and were calculated as where T = temperature; c = concentration of sample in g/100 mL; and l = path length, which is the length of the cell in dm (the cell used was 1 dm). Achiral and chiral gas chromatography (GC) were used to determine the conversion and enantiomeric excess (ee) of the substrates (7, 8, and 11), respectively. They were performed using a HP6890 series GC system, Agilent Technologies 7683 series injector, and HP7683 series autosampler. The ee of 10 and diastereomeric excess (de) of the MA salts of 10 were analyzed by chiral high performance liquid chromatography (HPLC) using an Agilent Technologies 1290 Infinity system, while the ee of MA was analyzed by chiral HPLC using an Agilent 1100 Series system.
In all resolution and R 3 processes, the resolvability of the system S was calculated as where RV is the volume of the reactor.
The total process time of the R 3 process was calculated as the product of residence time τ r and number of cycles (eq S1.5). General Procedure for Continuous Flow Racemization. All of the racemic and optically active amines used in the study are known, and methods for their synthesis, along with characterization data, are described below. Fresh catalyst 1 (ICP 6.3%, 0.32 g, 0.11 mmol of Ir) was used to racemize each optically active amine using the equipment shown in Supporting Information section 8. The catalyst was mixed with sand and loaded into an empty stainless HPLC column (4 mL). The column was placed inside an aluminum block which was connected to an electrical heater and equilibrated with the selected solvent. The liquid volume was determined by mass difference. A solution of amine (2.1 mmol) in solvent (20 mL) at room temperature was pumped from a syringe through the heated column at a specific flow rate. The eluent was collected from the second reactor volume onward. When all of the reaction mixture was pumped, the column was washed with the reaction solvent or isopropyl alcohol while cooling to room temperature. Each collected fraction was sampled and analyzed to determine its e.e. Detail procedures of sampling and analysis were shown in Supporting Information section 7. The e.e. at steady state was calculated as the average from reactor volumes 2−5.
Acetophenone (10 g, 83.2 mmol) and degassed ethanol (40 mL) were added to a round-bottom flask containing molecular sieves (3 Å, 16 g) which was purged with nitrogen, followed by the addition of methylamine solution in ethanol (33% wt, 22 mL). The mixture was stirred at room temperature for 48 h. The reaction was then filtered through Celite and was washed with dichloromethane. The solvent was removed in vacuo to give a yellow oil (12.5 g) which contained about 10% acetophenone (by 1 H NMR). The E/Z isomeric ratio of crude imine was not determined and was used directly in the next step without further purification. 1  Acetone (250 mL) was added to racemic 1-phenylethylamine (10.0 g, 82.5 mmol), and the mixture was heated to reflux using a heating mantle. After 24 h, the reaction was cooled to room temperature and the solvent was removed in vacuo to give a yellow oil. Methanol (200 mL) was added, and the solution was cooled on ice. Sodium borohydride (5.6 g, 148.0 mmol) was added, and the mixture was stirred on ice for 1.5 h and for a further 2.5 h at room temperature. The mixture was acidified by aqueous hydrochloric acid (2 M), and the solvent was removed in vacuo to give a white solid. Water and dichloromethane were added, and the mixture was basified with aqueous sodium hydroxide (1 M). The layers were separated. The organic layer was washed with brine and dried with anhydrous MgSO 4 , and the solvent was removed in vacuo to give a yellow oil ( The Journal of Organic Chemistry pubs.acs.org/joc Article (R)-N-Isopropyl-α-methylbenzylamine, (R)-8.
[α] D 23 = +56.1°(c 0.52, chloroform); the 1 H and 13 C NMR, IR, and MS data correspond to those reported for the racemate. H and 13 C NMR, IR, and MS data correspond to those reported for the racemate. N-(3,4-Dimethoxyphenethyl)acetamide. 45 3,4-Dimethoxyphenethylamine (97%, 9.9 g, 53.1 mmol) was charged to a mixture of dichloromethane (50 mL) and triethylamine (12 mL, 86.3 mmol), and the mixture was cooled to 0°C. With stirring, acetyl chloride (5 mL, 70.3 mmol) was charged dropwise to the reaction mixture over 0.5 h. The pH was checked to ensure the reaction mixture stayed basic. The reaction mixture was warmed to room temperature and stirred overnight. The content was then poured into ice-cold water and was acidified by aqueous hydrochloric acid (1 M). The organic and aqueous layers were separated, and the aqueous layer was further extracted with dichloromethane (3 × 50 mL). The combined organic layer was washed with water (100 mL), dried by anhydrous MgSO 4 , and filtered. The solvent was removed in vacuo to yield a pale yellow solid (12.1 g). The product was recrystallized in hexane/ethyl acetate 1:1 to yield the acetamide product as off-white crystals (9.2 g, 78% The acetamide from stage 1 (6.0 g, 27.0 mmol) was added to toluene (30 mL, dried over 4 Å molecular sieve, 1−2 mm beads) in a roundbottom flask fitted with an air condenser and calcium chloride drying tube. The mixture was warmed to 40°C using a heating mantle. Phosphoryl chloride (6 mL, 64.4 mmol) was charged dropwise into the mixture over 20 min. When the addition was complete, the mixture was heated to reflux for 3 h using a heating mantle. The reaction mixture was then cooled to 0°C for 2 h. The mixture was decanted, and water was charged to the remaining solid at 0°C. The content was warmed to room temperature and stirred overnight. The reaction mixture was basified by aqueous sodium hydroxide (5 M) and extracted with dichloromethane (4 × 50 mL). The combined organic layer was dried over anhydrous MgSO 4 and filtered, and the solvent was removed in vacuo to yield 6,7-dimethoxy-1-methyl-3,4dihydroisoquinoline as an orange solid (5.4 g, 97%). The product was used in the next step without purification. Racemic 6,7-Dimethoxy-1-methyl-1,2,3,4-tetrahydroisoquinoline, 10. 46 6,7-Dimethoxy-1-methyl-3,4-dihydroisoquinoline (4.3 g, 21.0 mmol) was dissolved into methanol (70 mL) and was cooled to 0°C. Sodium borohydride (4.0 g, 105.5 mmol) was added slowly in portions over 20 min. The reaction mixture was stirred at room temperature overnight. The reaction mixture was acidified by aqueous hydrochloric acid (1 M) and then concentrated in vacuo. The white solid obtained was basified with aqueous sodium hydroxide (1 M) and extracted with dichloromethane (3 × 100 mL). The combined organic layer was washed with brine (1 × 50 mL), dried over anhydrous MgSO 4 , and filtered, and the solvent was removed in vacuo to yield a light brown oil. The crude product was purified by column chromatography using Biotage (ethyl acetate/triethylamine gradient, 95−96.5% ethyl acetate); TLC: silica plate, ethyl acetate/triethylamine 95:5, visualization by UV (254 nm; R f of product = 0.15) to yield rac-10 as a pale yellow oil. Off-white solid was formed overnight Formic acid (7.2 mL) was added to acetonitrile (20 mL) under nitrogen and stirred at 0°C. Triethylamine (10.6 mL) was added slowly to maintain the temperature at below 10°C. After the addition was complete, the reaction mixture was warmed to room temperature. Ru(p-cymene)(R,R-TsDPEN) (0.2 g, 0.3 mmol) was dissolved in acetonitrile (25 mL) and added to the formic acid−triethylamine mixture. The content was stirred for 15 min. A solution of 6,7dimethoxy-1-methyl-3,4-dihydroisoquinoline (6.2 g, 30.3 mmol) in acetonitrile (35 mL) was charged to the reaction mixture and was warmed to 30°C using a heating mantle. The reaction was stirred overnight. After the reaction was complete, the reaction mixture was cooled to room temperature and concentrated in vacuo. The residue was acidified with aqueous hydrochloric acid (2 M) and washed with toluene (2 × 20 mL). The aqueous layer was basified with aqueous sodium hydroxide (5 M) and was extracted with dichloromethane (3  48 [α] D 20 = −41.3°(c 1.10, ethanol)}. The spectral and MS data correspond to those reported for rac-10 and in the literature. 48,50 (R)-6,7-Dimethoxy-1-methyl-1,2,3,4-tetrahydroisoquinoline, (R)-10.
Similar procedures as (S)-10 were employed in the synthesis of (R)-10 with Ru(p-cymene)(S,S-TsDPEN). Toluene was used instead of dichloromethane in the extraction of product from the aqueous layer during workup. A pale brown oil (5.7 g, 94%) was obtained with an ee of 96% (by chiral HPLC). (S)-MA (0.11 g, 0.74 mmol) was dissolved into ethyl isobutyrate (2 mL), and the solution was added to (R)-7 (0.10 g, 0.74 mmol). The slurry was heated to reflux until all solid dissolved. The mixture was cooled to room temperature with reduced stirring and was filtered to give a white solid which was dried in air overnight (0.17 g, 82%) with d.e. = 100% (by 1   (S)-MA (0.11 g, 0.74 mmol) was dissolved into ethyl isobutyrate (1 mL), and the solution was added to (S)-7 (0.10 g, 0.74 mmol). The mixture was stirred overnight at room temperature for solid to precipitate, which was then heated to reflux. After all solid dissolved, the mixture was cooled to room temperature with reduced stirring overnight. The slurry was filtered, and the solid collected was dried in air overnight (0.14 g, 67%) with d.e. = 100% (by 1 1 g, 0.62 mmol) were dissolved in isopropyl acetate (1.5 and 0.5 mL, respectively). The two solutions were mixed, and more isopropyl acetate (8 mL) was added. The mixture was heated to reflux using a heating mantle. Isopropanol was added until all solid dissolved. The mixture was then cooled to room temperature with reduced stirring and was filtered to give a white fluffy solid (0.14 g, 74%) with 100% d.e. (by 1   were dissolved into isopropyl acetate (3 and 2 mL, respectively). The two solutions were mixed, stirred, and heated to reflux using a heating mantle. More isopropyl acetate (9 mL) was added until all solid dissolved. Stirring was stopped. The reaction mixture was cooled to room temperature and filtered to give a white crystalline solid (0.24 g, 67%). A small amount of crystals was dissolved into aqueous sodium carbonate solution (10% w/v, 1 mL) and extracted with ethyl acetate (1 mL). The organic layer was dried with anhydrous MgSO 4 and filtered through a plug of Celite. A solution of (S)-MA in methanol (21 mg mL −1 , 100 μL) was added, and the mixture was sonicated and then dried in vacuo. The residue was analyzed by 1 2984, 2731, 1719, 1680, 1610, 1456, 1406, 1388, 1338, 1265 A similar procedure to that above was used to synthesize (S)-9-(S,S)-DTTA. After more isopropyl acetate was added to the hot solution of (S)-9/(S,S)-DTTA, isopropyl alcohol (8 mL A solution of (S)-MA (0.3 g, 1.9 mmol) in ethyl acetate/methanol 7:1 (2 mL) was added to a solution of (R)-10 (0.4 g, 1.9 mmol) in ethyl acetate/methanol 7:1 (2 mL). The mixture was heated to reflux using a heating mantle. More ethyl acetate/methanol 7:1 was added until all of the solids dissolved. The stirring was stopped and the heat switched off, allowing the mixture to cool to room temperature. The slurry was filtered to give an off-white crystalline solid (0.4 g, 47%) with 100% d.e. (by 1 H NMR in CDCl 3 , 500 MHz). The resolution results are shown in Supporting Information  Figures S8.1 and S8.2. The 50 mL magnetically stirred crystallizer, used at ambient temperature, was fitted with an in-line sintered glass filter, suitable to retain the crystals and allow the separation from the mother liquors. A PTFE lid with inlet and outlet tubes was tightly fitted to prevent solvent evaporation. The mother liquors were pumped through the heated racemization column using a Jasco pump and recirculated back to the crystallizer. The parameters used for each of the R 3 processes are shown in Supporting Information Tables S8.1− S8.12. The diastereomeric salts were made by dissolving the racemic amine and resolving acid in the chosen solvent at a supersaturating concentration to cause precipitation. Otherwise, reactions were seeded with the pure diastereomeric salt (5% w/w). An initial sample was taken (150 μL) and prepared for analysis by filtering the mother liquor and diluting in ethyl acetate (2850 μL) and basified by aqueous Na 2 CO 3 (10% w/w, 3 mL); the organic layer was dried with MgSO 4 and filtered through a plug of Celite. The crystals collected from each The Journal of Organic Chemistry pubs.acs.org/joc Article sample were dissolved in dichloromethane, dried in vacuo, and analyzed by 1 H NMR (CDCl 3 , 500 MHz). The pump was started, and samples were taken as described after each complete recirculation of the system volume, for the designated number of cycles. The activity of the catalyst was determined at the end of the reaction by pumping the optically active amine through the heated fixed bed for 3−4 reactor volumes (i.e., steady state) (Supporting Information section 9).
■ ASSOCIATED CONTENT