Continuous‐Flow Synthesis and Derivatization of Aziridines through Palladium‐Catalyzed C(sp3)−H Activation

Abstract A continuous‐flow synthesis of aziridines by palladium‐catalyzed C(sp3)−H activation is described. The new flow reaction could be combined with an aziridine‐ring‐opening reaction to give highly functionalized aliphatic amines through a consecutive process. A predictive mechanistic model was developed and used to design the C−H activation flow process and illustrates an approach towards first‐principles design based on novel catalytic reactions.

Continuous flow processes represent aparadigm shift in the manufacture of fine chemicals,s pecialties,a nd pharmaceuticals owing to the demonstrable gains in efficiency and product quality. [1] Acrucial aspect of realizing this ideal is the effective crossover from fundamental chemistry advances to process engineering.I ti sn oticeable that the transition of novel catalytic transformations of potential industrial interest into continuous flow processes is often slow owing to the inherent complexity of the reaction systems.W ith often limited mechanistic understanding of new reactions,i ti s rarely straightforward to determine the best reactor configuration or predict their behavior at scale,which can make the design of an optimal process difficult regardless of whether it is batch, semibatch, or continuous.The ultimate desired target for process chemistry is afully predictive process model that accounts for aw ide range of operating conditions and scale effects.H owever,t he development of such models for complex reactions represents as ignificant methodological challenge. [2,3] Ac entral field in modern synthetic organic chemistry is transition-metal-catalyzed CÀHa ctivation. Thel ast 15 years have seen tremendous advances in the use of many different metal catalysts to functionalize traditionally unreactive C À H bonds;palladium salts,inparticular, have enjoyed agreat deal of success in effecting CÀHactivation reactions. [4] Despite the potential of these seemingly ideal strategic bond-forming reactions,t he uptake of CÀHa ctivation in pharmaceutical and agrochemical processes and manufacture is limited to arelatively small number of examples. [5] Part of the reason for this deficiency is the limited mechanistic understanding of these complex reactions,which frequently are heterogeneous under operating conditions;t his characteristic can preclude industrial applications of either batch or continuous CÀH activation processes.
Recently,o ne of our groups reported an ew palladiumcatalyzed C(sp 3 ) À Hactivation reaction on hindered aliphatic amines to give aziridines (Scheme 1a). [6] Having performed some initial mechanistic studies on this transformation, [7] we considered the possibility of utilizing this reaction as ap latform for aC ÀHa ctivation reaction in flow.G iven the prevalence of aliphatic amines in biologically active molecules,aflow-based C(sp 3 ) À Ha ctivation process on these molecules would represent an important advance.T here are very few examples of flow processes for C(sp 3 )ÀHa ctivation, [8] and no flow C(sp 3 )ÀHa ctivation reactions have been conducted on ag ram scale.
Herein we report the successful translation of firstprinciples mechanistic understanding of aC (sp 3 ) À Ha ctivation reaction for the formation of aziridines into ap ractical flow process.Apredictive process model was used to explore in silico the behavior of asemibatch and aflow reactor prior to testing of the reactors at microreactor and gram scales (Scheme 1b-c). Additionally,w es how that the aziridine products can be used directly as part of as equential flow process,thus enabling the rapid synthesis of complex aliphatic amines with minimal purification.
During our previous studies,w ed iscovered that the addition of acetic acid to the reaction mixture for the palladium-catalyzed C À Ha ctivation to form aziridines resulted in as ignificant rate enhancement. [7] We chose these conditions from which to develop ap redictive kinetic model that would facilitate the transfer from batch to flow processes. Aproposed catalytic cycle for the CÀHactivation, accounting for the influence of acetic acid, is shown in Figure 1a.W e began our studies by examining this pathway by quantummechanical and kinetic approaches.
We focused on the turnover-limiting step,w hich was believed to be CÀHb ond cleavage.T his assumption was supported by experimental kinetic isotope effect (KIE) studies, [7] which gave av alue of 3.9. Calculations performed at the density functional theory (DFT) level by the use of Gaussian 09 (Figure 1b)showed that the bis(amine) complex A was the catalyst resting state,a si tw as thermodynamically the most stable species in the reaction mixture. [9] Thek ey reactive intermediate (mono(amine)-Pd complex B)w as found to be less stable by 6.2 kcal mol À1 .C yclopalladation from the (mono)amine-Pd complex B to form intermediate D was characterized as as ingle elementary step (through concerted metalation-deprotonation,CMD) [10] with abarrier of 22.9 kcal mol À1 .W ealso considered that the higher acetate concentration resulting from the addition of acetic acid may lead to CÀHb ond cleavage involving an external acetate ligand. However,w ef ound that this pathway was kinetically unfavorable (Gibbs activation energy 38 kcal mol À1 ; TS E). Instead, we found that protonation of the amine 1 with acetic acid to give 1·HOAc was energetically beneficial by 0.8 kcal mol À1 .T herefore,w eb elieve that acetic acid protonates the free amine 1 to give 1·HOAc,thereby reducing the free-amine concentration, which in turn reduces the concentration of the off-cycle bis(amine)-Pd complex A.T hus,p alladium is maintained in the catalytic cycle instead of being held in the off-cycle complex A.
Therefore,t he mechanism of the acid-accelerated C À H activation reaction was reduced to ak inetic model incorporating three equilibria (K 0 , K 1 ,a nd K 2 )a nd one irreversible reaction (k 3 ;F igure 1a). Thee lementary steps following the turnover-limiting step were ignored in setting up the model. A set of unique,kinetically controlled experiments was designed and performed to build ad ataset for parameter estimation. Va lues of kinetic parameters for equilibria at 70 8 8Cw ere found to be and for the turnover-limiting step (70 8 8C), k 3 = 9.23 [min À1 ]. Kinetic parameters were fitted to the experimental data using Figure 1. a) Working model for the reaction mechanism. Kinetic parameters corresponding to the reactions precedingt he rate-limiting step are marked "k". The region in the box illustrates the differenceb etween the mechanism under the original reaction conditions and the mechanism under the new fully optimized conditions described herein. b) Energy profile of cyclopalladation proceedingthrough aCMD mechanism. Values of the kinetic parameters for equilibria at 70 8 8C: K 0 = 3.14 Lmol À1 min À1 , K 1 = 2963 160.29 Lmol À1 min À1 , K 2 = 7415.93 Lmol À1 min À1 ;and for the turnover-limiting step (70 8 8C): k 3 = 9.23 min À1 . gPROMS ModelBuilder.T he values obtained from parameter estimation are in very good agreement with the values predicted by DFT calculations (see the Supporting Information). Theweighted residual was found to be smaller than the c 2 value corresponding to the 95 %c onfidence region, thus indicating that the mechanism shown in Figure 1a is ag ood representation of the reaction pathway.S mall correlations between most of the kinetic parameters (see the Supporting Information) proved their independence.T he larger correlation between K 0 and K 2 can be explained by their competitive nature with respect to the starting material.
Af ully predictive kinetic model allows the selection of optimal conditions to maximize conversion and yield as well as minimize by-product formation, with am inimum of experimental work. Them odel also helps to predict the behavior of the reaction system on previously untested scales. All of these factors can result in the faster design of ac ontinuous process.B efore transition to the flow setup,w e further tested the kinetic model to assess its accuracy under dynamic conditions of asemibatch process.W efound that the kinetic model accurately predicted the results for two different semibatch experiments,i nw hich ab atch reactor was fed continuously with the starting material, and ideal mass and energy transfer were assumed ( Figure 2). Aslight deviation of the model from the experimental data was observed in the early stages of the experiment. This deviation can be explained by the fact that the semibatch reactions were performed without preheating of the reaction mixture,a nd subsequently the reactor operated at lower temperature until the system reached thermal equilibrium. Thegood agreement between the experimental data and the predicted kinetics provides further support for the model as an accurate representation of the molecular processes.
Thekinetic model was used to design aprocess model on the assumption that an ideal plug flow reactor was used. We defined two restrictive parameters:The reaction temperature could not be higher than 120 8 8Co wing to the thermal instability of palladium acetate, [13] and full conversion must be reached within 10 min for al arge space-time yield to be attained. Them odel was used to select operating conditions that fulfilled both these requirements,while also considering that high acid concentrations may lead to degradation of the aziridine product (see Scheme 1b). [7] Optimization in silico by using the process model was performed with respect to the concentration of the starting material (amine 1), Pd(OAc) 2 , and AcOH, and with respect to the temperature.T he results led us to several possible sets of conditions,w hich were then tested experimentally.O ur final process allowed ar eduction of the catalyst loading to 0.5 mol %w ith ar esidence time of 10 min (Figure 3). We did not notice any by-product formation over this time period.
Following optimization of the flow reaction, the design of as uitable process involving ap urification method was undertaken. Theaim was agram-scale synthesis characterized by al arge space-time yield and the design of ac onvenient separation method for both the product and the catalyst (if possible,a lso suitable for catalyst recycling). Thew hole process should operate autonomously to be desirable for large-scale applications.Since the catalytic system in batch is homogeneous,the same approach was initially applied to the flow process.W efirst tested the reaction on the microscale by using at emperature-controlledc ustom-made silicon-glass microfluidic reactor. Theu se of ah omogenous catalyst is beneficial in providing well-defined active sites and, given effective mixing, rapid diffusivity to the catalyst. However,an efficient way of separating the catalyst from the products is required. Owing to the nature of the reaction system, ab iphasic separation model was not possible.F urthermore, Figure 3. Results of continuous-flow experiments under optimal conditions; C SM = 0.05 mol L À1 , C Cat = 2.5 10 À4 mol L À1 , C AcOH = 0.5 mol L À1 . Concentrations were measured at the outlet of the reactor for different residence times in order to find the optimal residence time and compare experimental results with the process model. Reaction A: C SM0 = 0.05 mol L À1 ; C SMfeed = 0.1 mol L À1 ;R eaction B: C SM0 = 0mol L À1 , C SMfeed = 0.025 mol L À1 ;SM: starting material;C at: palladium acetate;0:concentrationi nr eactor before reaction start; feed:c oncentration in the feeding stream. Time "0" corresponds to when the pumping started.

Angewandte Chemie
Communications the most powerful, acid-based, metal scavengers were not suitable for separating the Pd(OAc) 2 from the reaction mixture because they would scavenge both the catalyst and the aziridine product. It was found that commercially available QuadraSil AP (with apendant primary amine) [11] successfully coordinated only to the palladium species,thus enabling removal of the catalyst from the spent reaction mixture.I nasecond separation step,the amine product could be separated from the reaction mixture by the use of as ilica gel functionalized with strongly acidic groups.Inthis study we used the Isolute SCX-3 gel (with pendant sulfonic acid groups) [12] as asuccessful amine scavenger (Figure 4a). Thep roduct can be washed on the gel and subsequently removed by the use of an eluent with ah igh pH value.I mplementation of this "catch and release" technique yielded ap roduct of high purity (Figure 4b).
Following the successful development and optimization of as mall-scale process,w ef urther tested the reaction on alarger scale by using aV apourtec R-Series system [14] equipped with a10mL, thermostated poly(tetrafluoroethylene) (PTFE) tubular reactor. Full conversion was achieved with af low rate equal to 1mLmin À1 ,w hich produced 4.68 mmol (771.9 mg) of aziridine product per hour, thus resulting in as pace-time yield of 0.463 [kg L(reactor) À1 h À1 ]. TheV apourtec R-Series is designed to operate autonomously,and the operational time is only limited by the capacity of the scavenger columns.T his limitation can be easily overcome with as etup including two sets of columns:one operational and one in astand-by mode.T othe best of our knowledge,noC ÀHactivation of amines has been carried out previously in aflow system, and no CÀHactivation process has been performed on am ultigram scale. [15] Both experimental designs resulted in very good yields (isolation of products in 90 % yield), high purity of products (> 90 %according to 1 HNMR spectroscopy of the crude product against an internal standard), and, on the basis of inductively coupled plasma optical emission spectroscopy (ICP OES), concentrations of palladium in the mixture leaving the scavenger column below 1ppm.
To extend the synthetic utility of this flow C À Hactivation procedure,w ei nvestigated the coupling of the CÀHa ctivation step to as ubsequent reaction in as ingle flow process.I t had been previously shown that the aziridine products could be functionalized through ring-opening reactions with various nucleophiles,s uch as carboxylic acids,a zides,t hiols,a nd halogens. [6,16] To further highlight the benefits of translating this batch process into flow,w ea imed to develop ar obust procedure suitable for opening aziridines with weak nucleophiles and reagents that could be potentially hazardous when used in batch.
In our previous study,the ring-opening reaction required the action of either astrong Brønsted or Lewis acid to activate the aziridine.I nt he current system, we questioned whether the strong sulfonic acid scavenging resin might also function Figure 4. a) Flow process for the aziridination CÀHa ctivation reaction. Flow reactor:s iliconglass microfluidic reactor or PTFE tubular reactor. The first column separates the catalyst from the spent mixture;the second column separates the product. Reactants were pumped by: pump A( starting material, oxidant), pump B(catalyst, acetic acid, acetic anhydride). A6bar pressure regulatorw as fitted. b) Elution of the product from the "catch" column. c) Aziridine ring opening and elution from the "catch" column to give functionalizedm orpholinones.
as an activating agent for aziridine ring opening, thereby reducing the number of chemical operations and streamlining the overall transformation. Thef irst step towards this goal was achieved by retaining the aziridine on the "catch" Isolute SCX-3 column. While on the column, the aziridine is immobilized by sulfonic acid protonation, which makes it susceptible to nucleophilic attack:h eating to just 60 8 8Ci s required to facilitate such ar eaction. When an ucleophilewater, methanol, or hydrazoic acid generated in situ-was pumped through the resulting column (Figure 4c), the derivatized products 3a-c were obtained in good yields ( Figure 5). Theu se of as olid-state support significantly facilitated the reaction procedure,m inimized the number of purification steps,a nd shortened the reaction time considerably.
Although several reactions have been described previously for aziridine ring opening in continuous flow, [17] to the best of our knowledge,none have been performed by using an immobilization-activation strategy.T he high purity of the crude mixture (> 90 %, 1 HNMR) allows direct transformations of the synthesized amines without any additional treatment. This protocol could be used more generally for the opening of various aziridines through an effective "catchrelease" flow process.F urthermore,o therwise time-consuming and hazardous reactions with hydrazoic acid precursors can be conducted much more safely.
Following the successful application of homogeneous catalysis,w eb riefly investigated the use of ah eterogeneous catalyst. Theu se of palladium(II) salts for the aziridination reaction significantly narrowed the choice of commercially available heterogeneous catalysts.However, satisfying results were observed with polymer-bound dichlorobis-(triphenylphosphine)palladium(II);I CP OES measurements of the spent reaction mixture showed palladium leaching below 5mol %. There are very few reports of heterogeneous Pd II catalysts,e specially those that are resistant to leaching, which is ag eneral limitation of this aspect of flow chemistry. Therefore,although this level of leaching cannot exclude the possibility that the reaction is catalyzed by solubilized species, the preliminary result suggests that heterogeneous catalysis is viable for the flow C À Ha ctivation process.
In summary,a saresult of extensive kinetic and mechanistic investigations involving DFT studies,w eh ave successfully performed at wo-step continuous-flow synthesis of substituted morpholinones on the basis of CÀHa ctivation. By designing the process from first principles,wesignificantly shortened the overall reaction time and increased the spacetime yield, while also reducing the number of purification steps.Athorough understanding of the reaction system allowed us to shorten the optimization time and efficiently design the flow process.Novel aspects of the process include the implementation of aC (sp 3 )ÀHa ctivation reaction in acontinuous-flow system and its performance on amultigram scale.Our study further illustrates the utility of flow chemistry as at ool for functionalization through consecutive reactions. Afurther highlight is the nucleophilic opening of immobilized aziridines in flow.