Continuous Flow-Processing of Organometallic ... - ACS Publications

Continuous Flow-Processing of Organometallic Reagents Using an Advanced Peristaltic Pumping System and the Telescoped Flow Synthesis of (E/Z)-Tamoxife...
0 downloads 4 Views 2MB Size
Article pubs.acs.org/OPRD

Continuous Flow-Processing of Organometallic Reagents Using an Advanced Peristaltic Pumping System and the Telescoped Flow Synthesis of (E/Z)‑Tamoxifen Philip R. D. Murray,† Duncan L. Browne,† Julio C. Pastre,†,‡ Chris Butters,§ Duncan Guthrie,§ and Steven V. Ley*,† †

Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, United Kingdom Instituto de Química, University of Campinas - UNICAMP, CP 6154, 13083-970 Campinas, São Paulo, Brazil § Vapourtec Ltd., Park Farm Business Centre, Bury St. Edmunds IP28 6TS, United Kingdom ‡

S Supporting Information *

ABSTRACT: A new enabling technology for the pumping of organometallic reagents such as n-butyllithium, Grignard reagents, and DIBAL-H is reported, which utilises a newly developed, chemically resistant, peristaltic pumping system. Several representative examples of its use in common transformations using these reagents, including metal−halogen exchange, addition, addition−elimination, conjugate addition, and partial reduction, are reported along with examples of telescoping of the anionic reaction products. This platform allows for truly continuous pumping of these highly reactive substances (and examples are demonstrated over periods of several hours) to generate multigram quantities of products. This work culminates in an approach to the telescoped synthesis of (E/Z)-tamoxifen using continuous-flow organometallic reagent-mediated transformations.

1. INTRODUCTION Technological innovation in the development of flow chemistry platforms continues to grow1 as the organic synthesis community seeks ways to discover material- and energyefficient processes with improved safety profiles and the ability to automate routine tasks. Flow chemistry aims to complement existing batch manipulations by minimising labor-intensive and repetitive operations as well as allowing access to chemistry not easily achieved in standard glassware apparatus, such as the superheating of solvents2 and the ability to work with reactive gases.3 Quantitative inline reaction monitoring is also possible via infrared,4 ultraviolet,5 or mass spectrometry6 or Raman methods,7 and offline process monitoring via the use of cameraenabled technologies.8 The use of immobilised reagents,9 inline purification techniques, such as liquid/liquid extraction and separation devices,10 and polymer-supported scavengers11 all serve to increase efficiency through reaction telescoping of multiple, or otherwise incompatible, operations into a single, flowing reaction stream. Flow chemistry techniques and tools could offer potential utility to synthesis with organometallic reagents. The accurate temperature control of the potentially exothermic processes,12 the rapid and stoichiometric mixing of organometallic reagents and substrates,13 and the safe handling of the reactive components in a fully contained system are all examples of the potential benefits over traditional batch manipulations. By using the unique size constraints imposed by microcapillary reactors, Yoshida has also reported on a number of novel chemical transformations utilising organometallic reagents in flow, including the generation and quenching of an aryllithium species in the presence of pendant electrophilic functionality.14 Despite this potential, and the advances in flow tools in general, there exists an unmet need for the continuous-flow processing © XXXX American Chemical Society

of quantities of air-sensitive, sometimes pyrophoric, organometallic reagents on the mesoscale, defined loosely as between 1 g and 1 kg of material. Above this scale, synthesis with these reagents is achievable by siphoning from a raised reservoir of the reagent at controlled rates using a mass flow controller. The typical pumping hardware found in flow chemistry laboratories commonly consists of simple syringe pumps, which require considerable manual intervention and cannot operate under excessive pressures or for extended periods of time. Alternatively HPLC-type piston pumps, which although physically robust, accurate, and able to pressurise a sealed system, expose the wetted parts to ambient atmospheres, and hence, over a period of hours are prone to fouling and blocking via precipitation of hydrolytic materials. Peristaltic pumps, on the other hand, at no point expose the contained fluid to the moist external atmosphere. In a peristaltic pump, a central rotor compresses a flexible tube against a fixed housing, thus creating a pressure increase which progresses the contained fluid in a forward direction. As the tube relaxes back with forward motion of the rotor, a pressure decrease is observed which draws further fluid into the system; thus, the direction of rotation determines the direction of fluid flow. However, simple rotational motion of the rotor does not give a uniform flow rate, due to the compression/release cycles causing local pressure ‘waves’ in the fluid, and hence (idealised) sinusoidal behavior about an average flow rate (Figure 1a). The magnitude of these fluctuations is further exaggerated when a peristaltic pump is operated at the increased pressures more typical of flow chemistry applications (Figure 1b). Such a regimen causes problems when planning chemical processes Received: June 7, 2013

A

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

Figure 1. Comparison in flow rate and pressure behavior of a traditional, uncorrected peristaltic pump at (a) 0 bar system pressure and (b) 8 bar system pressure, showing oscillation about the set-point rate and pressure, and (c) behavior of the newly developed Vapourtec V-3 pump.

due to issues of achieving constant reactant stoichiometry, especially at elevated flow rates and when multiple pumps are operating at different rates. As a consequence, these types of pumps have been largely neglected for flow chemistry applications by the synthetic chemistry community. Recently, a solution to these problems of operating peristaltic pumps at elevated pressures and flow rates, whilst achieving smooth flow, has been developed in the form of the new V-3 pump in the Vapourtec E-series flow reactor suite (Figure 2). A nonlinear rotor system, which adjusts its rotation rate as it compresses the tubing, has been developed to maintain a constant fluid flow rate (Figure 1c). In parallel with this innovation, two fluoropolymers have been identified that exhibit extraordinary chemical compatibility with a broad range of substances found in typical synthetic environments. The two types of tubing, conveniently color-coded by red and blue crimped end connectors, complement each other in their compatibilities to cover the majority of commonly encountered reagents and solvents including n-butyllithium, Grignard reagents, DIBAL-H, strong organic and aqueous acids and bases, and hydrocarbon, alcohol, polar aprotic, ethereal, ketonic, and halogenated solvents (Table 1). The robustness of any pumping system must be determined by the lifetime of any consumable parts employed in the system. The blue and red fluoropolymer peristaltic flow tubes have a finite lifetime, determined by the material compatibility of the polymer with the fluid being pumped, and the delivery pressure of the pumps. The tube selection guide provided by

Figure 2. Vapourtec E-series flow platform shown in typical tworeactor experimental setup (e.g., Schemes 1, 2, 3, and 6), equipped with three V-3 chemically resistant peristaltic pumps.

the manufacturer gives the compatibility of a tube to a particular chemical (Table 1), expressed as either ‘compatible’, ‘not compatible’, or ‘reduced-life’. The reduced-life designation B

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

Table 1. Compatibility of the two fluoropolymer peristaltic tubes toward common synthetic solvents and reagents (list is nonexhaustive)

*

Reduced tube life.

indicates that the tube material is affected by the chemical, normally exhibited as slight swelling and associated small reduction in mechanical properties but is still capable of safe pumping at consistent delivery flow rates. A conservative expectation for tube lifetime when pumping fluids listed as ‘compatible’ is over 30 L when operating at 8 bar (high pressure) and over 40 L when operating at 3 bar (moderate pressure). When pumping fluids listed as ‘reduced-life’, the expected lifetime is 10 L at 8 bar, and 15 L at 3 bar. When a tube begins to fail, a slowly increasing overdelivery of fluid from the calibrated set-point is observed. The system permits a correction to account for this, achieved by calibration of the pumps. The correction can be increased, but with further use of the tube, the skew beyond the set-point calibration then becomes more rapid. For this reason, we recommend replacing a tube when the correction reaches −10% from the set-point. At no point have we observed a tube to burst, leak, or rupture in any way, as long as chemical compatibility was checked first. It should be noted that contact of the blue tube with n-butyllithium solutions, concentrated aqueous nitric acid solution, or liquid bromine does lead to rapid failure and leakage, and so this must be avoided.

When moving to continuous-flow processing of chemical substances, there is often a lack of clarity in the term ‘continuous’, leading to wide interpretation, depending upon its specific application. This is commonly the case when sample loading loops are required to introduce reagents to the system. These either require a sophisticated dual-loop setup to recharge one whilst the other is being emptied,15 or manual refilling of the loops in a segmented-flow fashion. This new flow platform, by virtue of the peristaltic design and chemical resistance of the components, is able to surmount these limitations by pumping directly from a reagent reservoir and therefore allowing for truly continuous pumping over a working day. Herein we report a number of applications of the use of this new flow platform to reactions with commercially available organometallic reagents, where the material is pumped directly from the suppliers bottle, through the pump to prepare multigram quantities of product in a continuous fashion, over periods of several hours without any manual intervention throughout the product collection. Several different types of reaction are described, including lithium−halogen exchange, Grignard reagent addition, magnesium−halogen exchange and addition−elimination, transmetalation of Grignard reagents C

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

Scheme 1. Continuous-flow synthesis of aryl boronic pinacol ester 1 via three-stream lithiation/electrophilic quench methodology

Figure 3. (a) Three-input, low-temperature reactor assembly, equipped with a 10 mL perfluoroalkoxy (PFA) reactor coil, shown mounted (left) on the E-series machine. (b) Schematic flow diagram to represent this reactor assembly.

decomposition temperature. The scaling up of these reactions is therefore correspondingly difficult and potentially unpredictable. For these many reasons, we sought to investigate this process under continuous-flow conditions where the large surface-area to volume ratio of a flow tube reactor is ideally suited to providing the rapid heat transfer required to contain exothermic events. (Hetero)aryl pinacol boronic esters, (Het)ArBPins, as typically shelf-stable transmetallating reagents in Suzuki coupling reactions, are useful building blocks for many applications which rely upon efficient carbon−carbon bondforming processes to deliver new functional compounds. We have previously demonstrated their preparation under segmented-flow conditions, and here begin this study by defining a continuous-flow process.12,17 To demonstrate the capability of the new technology we chose to prepare 4fluorophenyl pinacol boronic ester (1), as the parent ring system is an important pharmacophore, appearing in 8 of the top 100 pharmaceutical drugs.18 In order to bring about the lithium−halogen exchange of 1bromo-4-fluorobenzene, n-butyllithium solution was pumped

with copper catalysts and subsequent cuprate addition and conjugate addition, and partial reduction of an ester with DIBAL-H. Examples of enhanced selectivities and yields over batch procedures are described, with examples of in-line reaction telescoping. This work culminates in the continuousflow synthesis of the important breast cancer drug, tamoxifen.

2. RESULTS AND DISCUSSION 2.1. Continuous-Flow Lithium−Halogen Exchange Reaction of an Aryl Bromide and Conversion to an Aryl Boronic Ester. The lithium−halogen exchange reaction of (hetero)aryl halides, commonly bromides or iodides, to generate a nucleophilic (hetero)aryllithium species, before quenching with a suitable electrophile is a common synthetic transformation. These reactions in batch are usually carried out at −78 °C, to avoid isomerisation of the aryl−lithium bond,16 premature quenching or other undesirable decomposition. The metal−halogen exchange and electrophilic quench processes are normally exothermic, such that efficient mixing and heat transfer are essential to contain the exotherm below D

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

Scheme 2. Continuous-flow preparation of allyl acetate 2 via the addition of vinylmagnesium bromide to acetophenone, and subsequent in-line quenching of the alkoxide product with acetic anhydride

mechanically strong material. Introduction of precooled isopropoxy pinacolborolane reagent (iPrOBPin) as a quenching stream at this temperature, before passing the flow to a short reactor coil maintained at ambient temperature, resulted in formation of the lithium arylboronate as expected. An offline workup involving an aqueous ammonium chloride wash gave the desired fluorophenyl pinacol boronate ester (1) in near quantitative yield (99%), in >99% purity directly after evaporation. The only detected very minor side product was butyl pinacol boronic ester owning to excesses of reagents. This is easily removed via evaporation under high vacuum. Whilst reagent concentrations above 0.4 M lead to viscous reaction solutions, the flow rates are high enough and residence times short enough to process substantial quantities of material in short periods of time. Once the flow equipment and reaction parameters were optimised with respect to temperature, residence times, and stoichiometry of reagents, the preparation was run continuously over 4 h to generate 27.5 g of the product, consuming ∼90 mL of a 1.54 M n-butyllithium solution in the process. Whilst the flow chemistry preparation of boronic ester 1 described above may initially appear convoluted when compared with conventional batch methods, its value becomes apparent when one considers the scaling up issues of working with quantities of air sensitive reagents, which use labor intensive and repetitive practices, and are becoming increasingly unacceptable. This flow system provides a working solution to these problems and gives reproducible, walk-up access to multigram quantities of desired product, whilst being modular enough in design to rapidly reconfigure the system for a different chemical process. 2.2. Continuous-Flow Grignard Reagent Addition and Exchange Chemistry. 2.2.1. Grignard Reagent Addition to an Enolizable Ketone and In-line Telescoping of the Alkoxide Product. Grignard reagents ((hetero)aryl, vinyl, allyl and alkyl magnesium halides) feature in many carbon−carbon bond forming processes, many of which are commercially available in ethereal solvents. The addition of a Grignard reagent to a ketone or aldehyde is a reliable method for the formation of alcohols, whereas metalation of an amine, followed by addition to an ester is a method for direct amidation. We and others have demonstrated such transformations in segmented flow

continuously through the pump, equipped with the compatible ‘red’ fluoropolymer peristaltic tube, directly from the commercial suppliers bottle by inserting a needle and argon line through the septum seal (Scheme 1). No predilution or use of loading loops was required. The continuous-flow preparation of this boronic ester was then carried out using a three-stream reactor assembly (Figure 3) which features short precoils on each of the three input streams, and a post-reactor quench-coil. The reactor housing is insulated, and the temperature is controllable continuously from ambient down to −70 °C, using a cooled nitrogen or dry-air stream. The purpose of the precoils are to cool the substrate, the organometallic reagent, and the quench streams to a preset reaction temperature before mixing, increasing control in cryogenic reactions which show a ceiling temperature. The quench coil provides a temperaturecontrolled zone to dissipate any exotherm on quenching of the reactive species. The assembly also serves to cool the Tpieces at the junctions between the initial two substrates, and the reacting solution and quench solution, the key points at which the reactions begin. The ability to continuously vary the temperature range and accurately fixate on and maintain a chosen temperature are further benefits of the new system over traditional cooling-bath methods, especially over prolonged reaction times. With the ability to pump these reagents over long periods of time without concern for the pumps seizing up, we could rapidly optimise the desired reaction by making a change in the reaction conditions, giving sufficient residence time for the desired transformation to occur, then monitoring the output by TLC, GC, or NMR analysis. This provides a noticeable increase in productivity and greater confidence that, when the process is run for longer reaction times, the same result will be achieved. In moving from segmented-flow to continuous-flow synthesis one often finds that this is not the case, as dispersion effects on the reagent plugs may affect reaction stoichiometries and effective concentrations. We found that the lithium−halogen exchange proceeds rapidly to completion at −50 °C without any observed side reaction or decomposition of the aryllithium species (Scheme 1). All flow tubes, with the exception of the peristaltic components were constructed of 1 mm i.d. perfluoroalkoxy (PFA) polymer, an inexpensive yet chemically resistant and E

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

Scheme 3. Continuous-flow preparation of diarylmethanol 3 via the Grignard exchange reaction of m-iodotoluene and addition of this metalated arene to p-chlorobenzaldehyde

fashion,4a,19 and we now wish to utilise the new machinery to examine the addition of commercially available vinyl magnesium bromide to acetophenone under continuous-f low conditions, followed by a telescoped acylation of the alkoxide product in line with acetic anhydride (Scheme 2). In this experiment, a THF solution of vinylmagnesium bromide from a commercial source, and THF solutions of acetophenone and acetic anhydride were introduced to the flow system via three peristaltic pumps equipped with blue tubes suitable for ethereal solutions and Grignard reagents (Scheme 2). The organometallic reagent was pumped directly from the suppliers bottle via the septum seal, with no requirement for predilution or use of loading loops. We found that the Grignard addition to acetophenone proceeded rapidly to completion at room temperature. In batch, these reactions are commonly initiated at 0 °C by dropwise addition of one reagent to the other, prior to warming to room temperature for a period of time to complete the process.20 In flow, this approach was not necessary, and direct stoichiometric mixing at room temperature was achievable without an uncontrollable exotherm, leading to much reduced reaction times and increased material throughput. The alkoxide output was immediately mixed in-line with the acetic anhydride solution, and the resultant passed to a 10 mL reactor held at 50 °C, with a residence time of 3.0 min at these flow rates. This afforded, in excellent yield the allyl acetate product (2). It was noted that these telescoped reactions using an acyl chloride (rather than the anhydride), lead to precipitation of the magnesium halide salts in the reactor, whereas the mixed bromide/acetate salts were soluble in THF solution. Over 3 h of continuous collection, 31.8 g of allyl acetate 2 resulted in a 92% yield of the theoretical amount. This telescoped approach nicely demonstrates that intermediate alkoxides can be utilised in further reactions, representing an improvement in process economy. This reaction was once again developed and optimised in response to GC and TLC conversions of collected aliquots whilst running continuously, by adjusting parameters such as flow rates and residence times, temperatures, and concentrations accordingly. The reaction was first optimised with respect to conversion to the alcohol, before introducing the acylating reagent in-line, and adjusting residence time and temperature of the second reactor to give full conversion from the alcohol.

Also of interest is that batch mode addition of Grignard reagents to ketones with enolizable protons often results in selfcondensation of the carbonyl compound leading to wasted material. Indeed, product yields for the addition of vinylmagnesium bromide to acetophenone using standard batch procedures are reported between 30 and 60%, the rest being made up of the difficult to separate self-aldol product of acetophenone.20 This can be solved in batch by addition of catalytic quantities of cerium salts to the reaction mixture,21 but under continuous-flow conditions without this addition, this wasteful side reaction was largely suppressed, and less than 5% was detected in the crude product stream. This is an important observation, which we postulate may be an effect of mixing the carbonyl compound and Grignard reagent directly in a matched stoichiometry at a temperature where reaction is very rapid. 2.2.2. Grignard Metalation of an Aryl Iodide and Addition to a Benzaldehyde: Bioactive Diarylmethanol Preparation. Many noncommercially available Grignard reagents can be accessed through magnesiation of arenes or aryl halides using a range of commercially available reagents developed by Knochel.22 Typically employed in magnesium−halogen exchange reactions is either isopropylmagnesium chloride or its lithium chloride complex (Turbo Grignard), which shows enhanced reactivity over the uncomplexed variant in terms of functional group tolerance and reactivity. We have previously demonstrated segmented flow preparation of arylmaganesium halides from (hetero)aryl iodides and bromides via such metalation, with in-line quenching against benzaldehydes,4a and it is now attractive to import this procedure to the new peristaltic pumping system for continuous processing (Scheme 3). Diaryl methanols display interesting and varied biological activities and form the core of several major medications, such as Cetirizine (antihistamine), and thus are targets worthy of attention.23 By using parameters such as concentrations, residence times, and reaction temperatures developed in the segmented flow preparations of these products as a starting point, and optimising rapidly in response to GC conversions using micro-workup approaches without full isolation, we quickly found the ideal set of conditions for both the metalation and quench reactions. The Grignard exchange of m-iodotoluene, followed by quenching with p-chlorobenzaldehyde was not investigated in our work on segmented flow preparations, but F

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

Scheme 4. Continuous-flow copper-catalyzed conjugate addition reaction of phenylmagnesium bromide to cyclohexenone displays very high regioselectivity for the 1,4-product (4), with reproducible selectivity and yield between different machine operators

carbonyl directly (1,2-).24 This alternative reaction pathway increases the utility of Grignard reagents, and conjugate addition is widely used in natural product synthesis,25 with asymmetric variants utilising a variety of chiral ligands having been developed.26 We therefore decided that transmetalation of Grignard reagents to organocopper species in situ under continuous-flow conditions was an avenue worthy of investigation. Cuprate chemistry also has a reputation of being capricious, so the ability to use a machine-assisted approach to achieve reproducible cuprate chemistry is highly attractive. Once again, this is a class of reactions where, in batch, the Grignard reagent is often kept as the limiting reagent and is added dropwise to the substrate and copper catalyst to limit the direct addition.26a The meeting of equal stiochiometric quantities of Grignard reagent, enone, and only catalytic copper as would be the case in flow, one could expect a reduced selectivity for the conjugate position, weighted for the relative rates of the transmetalation process and conjugate addition, versus the direct addition. We chose to investigate the transmetalation of phenylmagnesium bromide and its conjugate addition to cyclohexenone (Scheme 4), as it is reported that this Grignard reagent in particular is slower to transmetallate than alkyl Grignard reagents, and in a batch procedure, gave a 1:1 regioselectivity for the conjugate product (4) in the addition to cyclohexenone.26a We found that the direct mixing of a cyclohexenone solution containing catalytic quantities of copper bromide dimethyl sulfide complex solubilised with triphenylphosphine, through one pump line, and phenylmagnesium bromide in a second stream, in a matched stoichiometry at −10 °C in the cooled reactor assembly (Figure 2) gave the conjugate product 4 in >99:1 regioselectivity, with a 95% yield of the product (Scheme 4). Over 8 h of continuous collection, we obtained 16.2 g of ketone 4. We were able to pump the substrate and Grignard reagent solutions with the ‘blue’ fluoropolymer peristaltic tube, as this material was found to be compatible with the catalytic

with this platform we were quickly able to optimise to these specific substrates from the general starting point known from our previous work. Both the metalation and quench are carried out at room temperature, removing the need for energyintensive cryogenic conditions sometimes employed for these reactions in batch.22a The residence time for the metalation is rapid compared to the time requirement in batch, although longer than for previous processes investigated in this work. Despite this, over 6.3 h of continuous collection, 19.3 g of the diarylmethanol product (3) was collected, representing a 97% yield. This example demonstrates how we were able to take an existing preparation and rapidly develop it into a reproducible, continuous-flow synthesis, which delivers gram-scale quantities of the product. The original publication we adapted in this example was a flow procedure, but we have also been able to adapt batch procedures into continuous-flow operations, using the batch-optimised conditions as a starting point, then adjusting in particular reagent concentrations and stoichiometries to avoid precipitation in the flow tubes, and thereafter residence times and temperatures to make the process go to completion. One could also employ this system to rapidly develop reaction parameters for a scaled batch process, leading to less intermediate scale-up optimisation, and a more rapid process delivery. 2.3. Continuous-Flow Copper-Catalyzed Grignard Reagent Transmetalation Chemistry. 2.3.1. Highly Regioselective Copper-Catalyzed Conjugate Addition to an Unsaturated Ketone. Transmetalation of a Grignard reagent or organolithium reagent with a soluble copper salt to form an organocopper reagent (RCu), a Gilman cuprate (R2CuM), or other cupric species (e.g., R2CuM2L for example) drastically alters the regioselectivity observed in addition reactions to unsaturated carbonyl compounds, where the ‘soft’ copper nucleophile prefers to attack at the conjugate position (1,4-) as opposed to the ‘hard’ magnesium nucleophile which attacks the G

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

Scheme 5. Continuous-flow copper-catalyzed epoxide ring-opening reaction of epoxide 6 with vinylmagnesium bromide displays high chemoselectivity for the homoallyic alcohol product (7) over the bromohydrin (8), and high regioselectivity for 7 over the other possible regioisomer (9)

forming a halohydrin, often as the major product, with difficult to reproduce results if good selectivity is observed. Transmetalation to the organocopper reagent in situ leads to much improved yields and chemoselectivity for the carbon nucleophile. A large-scale batch process has been carefully optimised by chemists at Merck, where it was found that slow addition of the Grignard reagent to the epoxide and copper catalyst was required to give high yield of the desired homoallylic alcohol product.28 We attempted to adapt this method into a continuous-flow preparation of homoallylic alcohol 7 from racemic epoxide 6 (Scheme 5). As was the case with the conjugate addition example (Scheme 4), we were able to carry the electrophilic epoxide substrate with the copper catalyst in one stream, and mix directly with the vinylmagnesium bromide reagent in a matched stoichiometry, and obtain the product alcohol (7) in a high and comparable yield to the optimised result from the batch preparation. The merit of this preparation over the batch procedure is the decreased reaction time of 4 min compared to over 1 h in batch. Over 8 h of continuous collection, we obtained 29.2 g of homoallylic alcohol 7. The possible bromohydrin byproduct (8), or other regioisomeric homoallylic alcohol (9) were completely suppressed. We hope to further explore this area in order to more fully understand the nature of the organocopper intermediates in relation to solvent, temperature, cuprate type and stability. We envisage that the greater control provided by flow chemistry will enable us to achieve this goal, especially toward enantioselective processes. 2.4. Controlled DIBAL-H Reduction of a Methyl Ester in Flow: Starting Material for Natural Product Synthesis. The diisobutylaluminium hydride (DIBAL-H) reduction reaction of methyl esters to aldehydes at low temperatures is a well-known transformation but one which is complicated in batch due to problems of over-reduction. Researchers often prefer to fully reduce the system and follow this with a selective partial-oxidation to form the desired aldehyde, despite the obvious lower redox and step economy.29 The advantages offered by flow chemistry to overcome these shortcomings have already been well demonstrated under segmented flow

quantities of the copper reagent and Grignard reagents in general. An inline methanolic quench of the enolate intermediate at −10 °C was employed, as we found this substantially reduced the amount of aldol-type byproducts observed when compared to an offline aqueous ammonium chloride quench. The soluble copper species was prepared by taking a suspension of commercial, unpurified copper(I) bromide dimethyl sulfide complex in THF and adding to it a slight excess of the phosphine ligand. In this way, after about 1 h of stirring at room temperature under argon, all of the solid material dissolved, which enabled easy pumping of these solutions. A second operator, who was familiar with the operation of the flow machine, but unfamiliar with details of the optimisation process required to develop this conjugate addition reaction, followed the optimised procedure as a test of the reproducibility of the method. We can report that essentially identical conversion, yield and regioselectivity was observed between the two operators (100% conversion in both experiments, a 95% yield achieved over 8 h continuous product collection compared to a 98% yield over 3 h continuous product collection, and a >99: 1 regioselectivity for the 1,4product in both experiments). A control experiment was also run where the copper complex and phosphine ligand were removed from the system, taking care to first purge the system with 5% aqueous ammonium hydroxide solution, to remove any deposited copper species from the flow machine. This provided predominantly the direct addition product 5 as expected, in good yield (77%) and good regioselectivity (4:1). 2.3.2. Highly Chemoselective Copper-Catalyzed RingOpening of an Epoxide. Ring-opening reactions of epoxides with organometallic carbon nucleophiles are a powerful class of carbon−carbon bond-forming reactions, generating an αsubstituted alcohol, often with high regioselectivity for the more sterically free terminus. Also, given the myriad of reliable methods for asymmetric epoxidation,27 this is an important protocol for creating enantioenriched materials for natural product synthesis and medicinal chemistry programs. However, these reactions typically suffer from competing opening of the epoxide by the halide counteranion of a Grignard reagent, H

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

Scheme 6. Continuous-flow preparation of Garner’s aldehyde (11) via the DIBAL-H reduction of methyl ester 10

Scheme 7. Synthetic approach toward (E/Z)-tamoxifen (12) using organometallic reagent-mediated transformations

conditions, with examples of telescoping of the aldehyde product directly; for example we have demonstrated reduction/ Roush crotylation sequences for the generation of key intermediates in the synthesis of complex natural products,30 whilst Jamison has demonstrated a reduction/olefination sequence in flow microreactors.31 Flow synthesis with DIBAL-H on a process scale has been well documented by Lonza chemists, where it was employed in the reduction of methyl butyrate at impressive flow rates of up to 43 g min−1 of the starting material.32 Here we aimed to achieve the selective reduction of a methyl ester to the aldehyde mediated by DIBAL-H under continuousflow conditions to prepare gram quantities of Garner’s aldehyde (11), an L-serine derived, configurationally stable α-chiral aldehyde widely used as the starting material for over 100 natural product syntheses.33 The material can be costly from commercial suppliers if purchased on the meso-scale and thus a reproducible continuous-flow procedure for its preparation would be of value. The peristaltic tube choices in this example were unimportant, as all solvents and reagents were compatible with either type of fluoropolymer flow tube (Scheme 6). The key parameters considered in the optimisation of this process were the temperature and residence time. We found long residence times at −70 °C were required to drive the reduction of methyl ester 10 toward completion. Higher temperatures led to decreased selectivity for the aldehyde as expected. However, at this temperature, we were able to carry an excess of the reducing agent with our substrate and still achieve the desired selectivity. The inline methanolic quench to destroy the excess DIBAL-H at this temperature was also critical to the selectivity, with an offline Rochelle’s salt (sodium potassium tartrate) work up to remove aluminium salts. The system was pressurised to 6

bar using a back-pressure regulator, as it was found that, at this pressure, the hydrogen gas formed from the excess of hydride reagent did not evolve from solution, thus allowing a stable pressure to hold and the pumps to pump at a constant rate. Under these employed conditions, we were able to prepare 9.78 g of the aldehyde product (11) over 8 h of continuous collection, representing an 81% isolated yield. Over-reduction to the alcohol in this experiment was controlled to 95% purity was prepared from

L-serine

N

dx.doi.org/10.1021/op4001548 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development

Article

triturated with boiling hexane, the hot mixture decanted, and the solvent removed to afford pure (E/Z)-tamoxifen (12) as a slow-to-crystallise white solid (12.43 g, 33.5 mmol, 84%) with an (E:Z) ratio of 25:75. The analytical data matched those previously reported for each individual isolated isomer.39

output was transferred back to waste, and both lines switched to pumping solvent simultaneously. The clean-down procedure for the system was then begun. The collected volume was separated, the acidic aqueous phase was extracted with diethyl ether (3 × 100 mL), the organic phases were combined, washed with brine (100 mL), and dried over magnesium sulfate, and all volatiles were removed in vacuo. Ketone 15 was isolated as a white solid (40.0 g, 178.3 mmol, 97%) in 98% purity (NMR), the other identified materials being starting material (