Continuous‐flow Synthesis of Aryl Aldehydes by Pd‐catalyzed Formylation of Aryl Bromides Using Carbon Monoxide and Hydrogen

Abstract A continuous‐flow protocol utilizing syngas (CO and H2) was developed for the palladium‐catalyzed reductive carbonylation of (hetero)aryl bromides to their corresponding (hetero)aryl aldehydes. The optimization of temperature, pressure, catalyst and ligand loading, and residence time resulted in process‐intensified flow conditions for the transformation. In addition, a key benefit of investigating the reaction in flow is the ability to precisely control the CO‐to‐H2 stoichiometric ratio, which was identified as having a critical influence on yield. The protocol proceeds with low catalyst and ligand loadings: palladium acetate (1 mol % or below) and cataCXium A (3 mol % or below). A variety of (hetero)aryl bromides at a 3 mmol scale were converted to their corresponding (hetero)aryl aldehydes at 12 bar pressure (CO/H2=1:3) and 120 °C reaction temperature within 45 min residence time to afford products mostly in good‐to‐excellent yields (17 examples). In particular, a successful scale‐up was achieved over 415 min operation time for the reductive carbonylation of 2‐bromo‐6‐methoxynaphthalene to synthesize 3.8 g of 6‐methoxy‐2‐naphthaldehyde in 85 % isolated yield. Studies were conducted to understand catalyst decomposition within the reactor by using inductively coupled plasma–mass spectrometry (ICP–MS) analysis. The palladium could easily be recovered using an aqueous nitric acid wash post reaction. Mechanistic aspects and the scope of the transformation are discussed.


Introduction
Aryl and heteroaryl aldehydes are important intermediates in the synthesis of biologically activem olecules ( Figure 1). There are an umber of synthetic strategies to form aryl aldehydes from their correspondinga ryl bromides.O ne strategy is to use halogen-lithium exchange and subsequently react the lithium intermediate with dimethylformamide (DMF) (Scheme 1a). [1] However,t his protocol requires stoichiometrica mountso f metal and has limited substrate scope due to the sensitivity of some substrates to decomposition by as trong base such as nbutyllithium (nBuLi). Pd-catalyzed formylation of aryl bromides has emerged as ap owerful methodology in the organic synthesis of carbonyl compounds. [2] In the case of reductive carbonylations, CO is used in combination with ah ydrogen donor source,s uch as silyl and tin hydrides, or formate salts, to achieve formylation at low pressure (Scheme 1b). [3,4] However, these protocols often requirehigh catalystloadings, and silicon and tin hydrides are relativelye xpensive, which limits their use. Ac ontinuous-flow protocolu tilizing syngas( CO and H 2 )w as developedf or the palladium-catalyzed reductivec arbonylation of (hetero)aryl bromides to their corresponding (hetero)aryl aldehydes. The optimizationo ft emperature, pressure, catalyst and ligand loading, and residence time resultedi np rocess-intensified flowc onditions for the transformation. In addition, a key benefit of investigating the reaction in flow is the ability to precisely control the CO-to-H 2 stoichiometric ratio, which was identified as having ac ritical influence on yield. The protocol proceeds with low catalyst and ligand loadings:p alladium acetate (1 mol %o rb elow) and cataCXium A( 3mol %o rb elow). A variety of (hetero)aryl bromidesa ta3mmol scale were converted to their corresponding (hetero)aryl aldehydes at 12 bar pressure (CO/H 2 = 1:3) and1 20 8Cr eaction temperaturew ithin 45 min residence time to afford products mostly in good-to-excellent yields (17 examples). In particular,asuccessful scale-up was achievedo ver 415 mino peration timef or the reductive carbonylation of 2-bromo-6-methoxynaphthalenet os ynthesize 3.8 go f6 -methoxy-2-naphthaldehyde in 85 %i solated yield. Studies were conducted to understand catalystd ecomposition within the reactor by using inductively coupled plasma-mass spectrometry (ICP-MS) analysis. The palladium could easily be recovered using an aqueous nitric acid wash post reaction. Mechanistic aspects and the scopeo ft he transformation are discussed.
In particular,S ni sh ighly undesirable for pharmaceutical manufacture due to contamination and toxicity.I na ddition, in the case of silyl hydrides, the silyl hydridei st ypically added in excess ( % 2-3 equiv), which increases costs and complicates post-reaction processing.
Anastas and Kirchhoff developed the 12 principles of green chemistry as ar esponse for the necessity to reduce the environmentali mpact of chemicals. [5] Perhaps the greenest and most atom-economic source of CO and H 2 is synthesis gas (syngas, CO/H 2 ). Syngasi sahighly abundant and inexpensive feedstock that is availablef rom many sources in the chemicals industry. [6] Syngasc an be produced from the splitting of water and activation of carbon dioxide by electrolysis.F urthermore, virtually all hydrocarbons, derived from natural gas, petroleum, and coal, can be used as af eedstock for the production of syngas through partial oxidation, steam reforming, or gasification. Future sustainable energy policies are likelyt os ee an increase in the use of biomass or municipal waste for syngas production.T he first palladium-catalyzed formylation utilizing syngas was reported in 1974 by Schoenberga nd Heck using [Pd(PPh 3 ) 2 Cl 2 ]a sc atalysta tv ery high pressures (80-100 bar) and elevated temperatures (80-150 8C). [7] The protocol was not widely adopted for conventional organic synthesis due to the high pressures required. In 2006, Beller and co-workersr eported the formylation of aryl and heteroaryl bromides by Pd(OAc) 2 ,d i(1-adamantyl)-n-butylphosphine (cataCXium A), tetramethylethylenediamine (TMEDA) as base, using synthesis gas (CO/H 2 = 1:1) at relativelyl ow pressures (5-7 bar) with 16 hr eaction time (Scheme1c). [8] The protocol utilized very low loadings of catalyst( 0.25 mol %) and ligand (0.75 mol %) in most cases and was demonstrated on aw ide substrate scope.
There are many process challenges associated with handling gas-liquid transformations in batch reactors, particularly at large scales. The interfacial area between the gas andl iquid phases becomes proportionally smaller with increasing reactor size;t herefore, the reaction is more likely to be mass transfer limited, which leads to reproducibility problemsd uring scale up. In addition, most of the gas is in the headspace and therefore the reactor needs to be pressurized to maximize the amount of gas in solution and reduce mass transfer effects. A large inventory of highly poisonous CO and extremely flammable H 2 needs to be loaded and pressurized into the batch vessel from ag as cylinder.T ypical commercial batch reactors can operate between 2a nd 6bar;t heseh igherp ressures require more specialized and expensive equipment. These challenges and that the reaction utilizes toxic and flammable gases unfortunately renders this transformation increasingly unacceptable in contemporary organic synthesis within ab atch environment. One solution is to form the gas or gases in situ or ex situ from solid or liquid reagents (gas surrogate) that liberate CO/H 2 ,w hich addresses some of the challenges associated with handling gases. The problem is that in situ formation generally requires the presence of at ransition-metal catalysta nd strong base in combination with high temperatures (> 100 8C) to releaseC O, therefore often resulting in compatibility issues between the CO-producing and CO-consumingr eaction. [9] Pioneering research by the Skrydstrup group resulted in the development of at wo-chamber batch solution for forming gases ex situ. [10] In particular,9 -methylfluorene-9-carbonyl chloride was employedt og enerates toichiometric amounts of CO and potassium formate as the hydride source within at wo-chamber system (COware) for reductivec arbonylation of aryl iodides. [11] Madsen and co-workersa lso demonstrated at wo batch chamber configurationf or the ex situ formationo fC O and H 2 using an iridium-catalyzed dehydrogenative decarbonylation of hexane-1,6-diol, which was fed into asecond chamber for the formylation of aryl bromides. [12] The Ley group pioneered the tube-in-tube reactor gas-loading concept to enable the safer introduction of gases into the liquid-phase from gas cylinders. [13] Teflon AF-2400 (a fluoropolymer) is used as as emipermeable membrane, which is permeable to gases but impermeable to liquids. The tube-in-tube flow reactor was successfully appliedf or hydroformylation and some carbonylation reactions butn ot specifically for reductivec arbonylation reactions. [14] Furthermore, Ley and co-workersr ecently showed that oxalyl chloridec an be hydrolyzed by using NaOH to form CO in situ in flow and subsequently used the generatedC Oi nc arbonylation reactions. [15] The aforementioned strategies are good options for research-scale experimentation; however, both approaches suffer from limited scalability in terms of atom inefficiency,p oorer performance at scale-up, or are simply too expensive. [16,17] Tubular plug flow reactors have emerged as ap latform for the safe, efficient, and scalable utilization of gases direct from gas cylinders by using mass-flow controllers. [17] Gas-liquid reactions have successfully been applied for the synthesis of active pharmaceutical intermediates (APIs)b yu sing continuous-flow reactors. [18] The improved safety features of continuous-flow reactors enable the safe operation at higher pressures and temperatures, including above the boilingp oint of the solvent. [19] Acontinuous-flow reactor only needs arelatively small pressurized reactor volumec ontaining the reaction mixture and, when properly designed,c an sustain the pressureo fa nu nexpected combustion. [20] Gas-liquid segmented (Taylor) flow regimes generated in flow microchannels provide ah igh interfacial area between the gas and liquid phases within at ubular flow reactor,t herefore mass-transfer effects are minimized. The utilization of CO in flow for organic synthesis has been demonstrated by an umber of groups. [21] In particular, Ryu and coworkers demonstrated am icroflowp rocess for ar adical-based carbonylation reaction of alkyl iodides and bromides to aldehydes and ketones. [22] However,t his procedure required environmentally unfriendly and expensive Bu 3 SnH and very high CO pressures (80 bar). At Eli Lilly,t he successful and safe scaleup of hydroformylation and reductive amination was demonstrated by using CO and H 2 within al arge-scale tubular reactor, with one example demonstrated at a2MT scale. [23][24][25] We were inspired by the low pressure batch protocol reported by Beller and co-workers for the reductive carbonylation of (hetero)aryl bromides with synthesis gas as as ustainable and cost effective reagent. [8] We herein report the development of ac ontinuous-flow protocol for the reductivec arbonylation of (hetero)aryl bromides to (hetero)aryl aldehydes using synthesis gas. To our knowledge,t his is the first reported flow procedure for the reductive carbonylation of (hetero)aryl bromides using syngas.

Results and Discussion
The gas-liquid continuous-flow reactor setup consisted of two high-pressure liquid pumps (HPLC)( P, Uniqsis) for introducing the liquid feeds (see Figure 2a;s ee also Figure S1 in the Sup-porting Information). CO gas and H 2 gas were introduced in a controlled manner into the system from gas cylinders using calibrated mass flow controllers (MFC, Bronkhorst-EL). The liquid and gaseous streamsw ere combined in as imple fourway inlet mixer (M) at room temperature.T he mixer was connectedt ot he tubular reactor through af luoropolymer tubing (PFA, 1/8 in (1 in = 2.54 cm) outer diameter( OD), 1/16 in internal diameter (ID)). The PFAt ubing allowed visual inspection of the flow profile. The residence time (t res )r eactor was a6 0mL stainless-steel coil reactor (1/8 in OD, 1/16 in ID)h eated on an aluminumh eatingb lock (Uniqsis FlowSyn). The reaction mixture exited the system through ashort cooling loop and an adjustableb ack pressure regulator (BPR, Swagelok, 0-25 bar), which maintained ac onstants ystem pressure. An itrogen purge was installed at the outlet.Apressure sensor (PS1) was integrated directly after one of the liquid pumps before entering the mixer to measure the system pressure.
Optimization experiments were performed with 4-bromoanisole (1a) as am odel substrate under conditions close to those reported by Beller and co-workers( Ta ble 1) using Pd(OAc) 2 and cataCXium A [8] but at residence times more appropriate to flow processing (< 1h). We expectedt hat 4-bromoanisole (1a) would display relativelyl ow reactivity towards the transformation due to the electron-donating effect of the methoxy substituent because oxidative addition of the aryl bromide to the active palladium(0) speciesi st ypically the rate-determining step in this transformation. [26] For these reactions, 2.5 mmolo f substrate and 0.75 equiv of base were dissolved in toluenea nd introduced as one feed, and Pd(OAc) 2 (5 mol %) and cataCXiu-mA (15 mol %) were dissolved in toluene and introduced as the second feed, to provideh omogeneous solutions. Pd 0 precipitate formation occurred over time if the palladium catalyst and base were introduced in the same feed, from reduction of Pd II to Pd 0 particles. [14b] Sample loops and injection valves were used to load the liquid feeds. The liquid feeds were each pumped at equal flow rates. When the reaction was started, the injection valves were switched from the carrier solventt ot he feeds for the reaction, and the feed mixtures were carried into the mixer,w here they combinedw ith CO and H 2 to give as egmented gasliquid (Taylor) flow regime under the flow rates used in this study ( Figure 2b). Initially,t he influence of temperature, pressure and gas flow rate were investigated to identify appropriate reaction conditions for the continuousflow reductive carbonylation of 4-bromoanisole (1a). For temperature and pressure optimization( Ta ble 1), CO and H 2 were fed in excess at equal flow rates to give % 2.2 equiv of each gas relative to the substrate. The flow rates were adjusted at differentp ressures to provide comparative residence times. The conversion was relatively low at 5bar pressure (entry 1), which was most likely caused by insufficientm ass transfer of CO and H 2 from the gas phase to the liquid phase.H owever,adrop-in conversion was observed at 15 bar,p robablyd ue to catalystd eactivation by CO (entry 3). 10 bar pressure was identified to provide the best compromise between reaction rate and avoidingu nwantedc atalyst deactivation (entry 2). The reaction proceeded smoothly,g iving 95 %c onversion and 89 %d esired product yield at 120 8C, 10 bar pressure, and 36 min residence time (entry 4). Conversionw as significantly lower at 100 8C (entry 5) from as lower reaction rate. Ah igher reactiont emperature resulted in higher conversion but did not improvey ield due to accelerated catalystd ecomposition (entry 6).
We knew from the outsett hat aw ell-known phenomenon, and often unavoidable process, is the aggregation of Pd 0 to form clusters that ultimately and irreversibly precipitate in the form of Pd black, which can then deposit onto the reactor wall. [27] Deposited Pd can be recovered from as tainless-steel coil by washing with 20 %a queous nitric acid solution at 60 8C. The reactor coil was always washed between experiments (unless otherwise specified) with aqueousn itric acid solution to remove residual Pd. Washing with aqueous nitric acid solution demonstrated that considerable amounts of Pd were lost from solution andd eposited onto the reactor wall. No Pd 0 black particles were observed in the collected reactionm ixtures, but Pd 0 particles wereo bserved when the solutions were kept overnight, indicating that not all Pd was deposited on the reactor channels. Running ar eaction without fresh Pd(OAc) 2 and without washing the deposited Pd indicated that the deposited Pd is catalytically inactive for the desired transformation( entry 7). The amount of deposited Pd was measured by inductively coupled plasma-mass spectrometry (ICP-MS) for the optimized conditions (vide infra). No desired transformationo ccurred in the absence of catalyst (entry 8).
We next investigated different catalytic systemst hat might provide better performance under the process-intensified conditions utilized in ac ontinuous-flow environment( Ta ble S1). Pd(OAc) 2 /cataCXium Ag ave the highestc onversion and yield compared to as election of other phosphine ligands. The structure of the ligand appears to be very specifict ot he successo f the reaction:t wo bulky alkyl groups and the long aliphatic tail are very important for the high efficiency of the catalytic system in terms of electron richness and steric shielding of the complexes. [28] The catalytic system is criticalt ot he reaction given the limited stability of the corresponding palladium(0) complexes in presence of base and CO, especially under high temperatures and pressures.
The solubility of CO and H 2 in toluene is relativelyp oor, 7.59 10 À3 mol L À1 and 3.31 10 À3 mol L À1 ,r espectively,u nder standard conditions (20 8Ca nd 1bar). Thus, elevated pressures are necessary to dissolve as ufficient amount of the gases in solution to provide ar easonable reactionr ate and therefore an appropriate residence time (< 1h)f or processing within at ubular reactor.There are an umber of reports measuring the solubility of CO and H 2 in organic solvents;h owever data available for high temperature/pressure regimes are limited. [29] Delmas and co-workers measured the solubility of CO and H 2 in toluene at relatively high temperatures (up to 100 8C) and pressures (up to 15 bar). [29c] The solubility of each gas in the liquid phase obeys Henry's law whereby the amounto fd issolved gas is proportionalt oits partial pressure in the gas phase; [30] therefore, the reactionr ate should be directly proportional tot he amount of gas dissolved in solution.T he data from Delmas and co-workers was simulated in DynoChem (Scaleup Systems) to predictt he solubility of CO and H 2 as af unctiono fp ressure. The solubility of CO in the liquid phase was approximately double that of H 2 ( Figure 3a). H 2 is competing with CO for dissolution in the liquid phase, and when these are used in equimolar amountst hen the amount of H 2 dissolved in solution will be lower than CO, so the effective concentration of CO will be higher.I ncreasing the reactiont emperature would only have as mall influenceo nt he amount of dissolved H 2 (Figure 3b)w hereas the solubility of CO in the liquid phase displayed only marginal temperature dependence (not shown). Conv. 1a [b] [%] Yield 1b [b] [%] Selec The ability to carefully control the relative stoichiometric ratio of CO to H 2 through varying the gas flow rates by using mass-flow controllers is ak ey benefito fu sing continuous-flow reactors( Ta ble 2). The influence of gas stoichiometry has not previously been studied for reductivec arbonylations. The conversion and yield significantly drops when CO and H 2 were used at a3 :1 ratio, corresponding to 3.3 equiv of CO and 1.1 equiv of H 2 (entries1and 2). Pd becamee asily deactivated by the CO because CO is as trong p-acceptorl igand, forming palladium carbonyl clusters that can irreversibly form Pd black. The rate of oxidative additioni ss ignificantly reducedd ue to the loss of active catalyst. The clustering of Pd atoms is facile in the presence of CO resulting in nonactive palladiumc arbonyl complexes. [31] One approach to prevent catalystd eactivation is to utilize CO at ac lose-to-stoichiometric amount. [32] Very good conversion and yields were obtained at a1 :1 CO-to-H 2 ratio ( % 2.2 equiv of each gas). AC O-to-H 2 ratio of 1:3r esulted in even better results with 99 %c onversion and 98 %y ield at 35 min residence time (entry 6). Under these conditions, the system becomes highly starved on CO towards the end of reaction. In these circumstances, at the beginning of the reactor there is ah igherC Oc oncentration while towards the end the concentration is very lowb ecause almost all CO has been consumed, thus improving process safety at the outlet due to the low concentration of CO.
The main limitation of cataCXium Ai st hat it is ap roprietary ligand and therefore relatively expensive comparedt om any other phosphine ligands. Consequently,i tw as important to identify flow conditions that providedl ow ligand loadings to reduce costs and minimize waste. The catalyst and cataCXiu-mAloadings werelowered to more commercially viable levels, 1a nd 3mol %, respectively,f or subsequent optimization. The conversion and yield dropped significantly on reducing the catalysta nd ligand loadings, whichc ould be improved by increasingt he base to 3equiv (Table S2, entry 6). Further optimization at al ower catalyst loading demonstrated that increasing the pressure from 10 to 12 bar resulted in as ignificant improvement in conversion and yield (Table 3, entries 3a nd 5). However,t here was ad rop-in conversion at 14 bar,i ndicating elevated catalyst deactivation from CO poisoning at higher pressures (entry 6). The optimal system pressure was identified as 12 bar (entry 7). The CO-to-H 2 ratio becamee ven more important when the catalystand ligand loadings were lowered to more commerciallyv iable levels,1and 3mol %r espectively (entries 1-3), compared to when higherc atalyst and ligand loadings were used (Table 2). For the reaction in as egmented gas-liquid flow pattern, only as mall excess of CO and H 2 are neededw hereas the reaction in ab atch autoclave would require much more due to the reactor headspace. [8] The low dosing of gases using continuous-flowr eactors is ak ey benefit of continuous-flow reactorsi nt erms of reducing usage and wastageand improving safety.  Conv. 1a [b] [%] Yield 1b [b] [%] Selec.
[ Another strategy to prevent deactivation by CO is to utilize phosphine ligands at high ligand-Pd ratios (Table 4). [33] Al igand-to-catalyst ratio of 2:1r esulted in ad rop in conversion and yield (entry 1), probably due to catalyst poisoning from overcoordination of CO. The utilizationo faligand-to-catalyst ratio of 4:1o nly resulted in am arginal improvement in yield (entry 4) compared to a3:1 ratio. In contrast, an increase in the ligand-to-catalyst ratio to 5:1r etards the reaction and decreased the aldehyde yield to 71 %( entry 5). The slight increase in yield obtained when using al igand-to-catalyst ratio of 4:1c ompared to 3:1o ften cannotb ej ustifiedb ased on the increasei nc ost associated with using ah igher ligand excess. We also investigated the use of stabilizing solvents to prevent the formation of black Pd 0 particles. Even thoughe nvironmentally their use shouldb em inimized, [34] polar aprotic solvents, for example, DMF and dimethylacetamide (DMA), can stabilize Pd 0 species in solution. [27b] However,u sing these solvents as co-solvents did not improvec onversion or yield (Table S3) and therefore not investigated further.
[  Conv. [b] [%] Yield [b] [%] Selec. [b] [%] Dehal.(c) [b] [%]  The crude product wasp urified by column chromatography to give pure product in 85 %i solated yield, which enabled preparation of 3.8 gp roduct, giving at hroughputo f0 .7 gh À1 from the continuous-flow process. Reactorc ontamination( fouling) is ac riticali ssue in flow chemistry that is often overlooked. [38] As discussed above,P d 0 speciesa ggregate to generate Pd clusters, which ultimately ir-reversibly precipitate in the form of Pd black,w hich deposits on metal surfaces. ICP-MSa nalysisw as conducted to measure the amount of Pd deposited on the reactorw alls compared to the amount of Pd remaining in solution and to obtain am ore thorough understanding of the slow decrease in conversion and yield over operation timef or some substrates. ICP-MS analysisw as conducted on samples fractionated at 20 min intervals for as eparate experimental run for the reductivec arbonylation of 4a over % 2h.C ontrol experiments confirmed that the untreated steel material itself cannot catalyze the reductive carbonylation (vide supra). The amount of Pdm easured in the collected reaction solution decreasedo ver operation time, which indicated that the presence of existing "inactive" deposited catalysta cceleratedt he deposition of further catalyst ( Table 6). ICP-MS measurements, along with the slow decrease in conversion and yield throught he run, demonstrated that the rate of decomposition of catalyst increases over the duration of the run. The catalyst can either enter the catalytic cycle or aggregate to form initially soluble Pd clusters, which at some point will turn into insoluble Pd black. ICP-MS Figure 4. Long-run profile. For reaction conditionsa nd analytics, seeT able 4, entry4,with 0.5 mol %Pd(OAc) 2 and 1.5 mol %c ataCXium Au sed for the long run. No samples were taken until color (reaction mixture) was observed at the BPR. Samples were fractionated at 20 minintervals. The first and last samples were less concentratedd ue to dilution by the "push-out" solvent. analysisc onfirmed that % 80 %o ft he Pd was deposited on the walls over the duration of the experiment and that Pd is easily recovered by washing the reactor with 20 %a queousn itric acid. Thef ormation of Pd black is self-catalyzed and leads to the withdrawal of Pd from the catalytic cycle. ICP-MS analysis also confirmed that no Fe or Co leachedf rom the stainless steel into the reaction solution or from the aqueous nitric acid wash (Table S5). The catalytic cycle for Pd-catalyzedf ormylation between aryl bromidesa nd synthesis gas (CO/H 2 1:1) proposed by Beller and co-workersi ss hown in Scheme2. [26] The catalytic cycle involves the oxidative addition of the aryl bromide with the active palladium(0) species,m igratory insertion of CO into the ArÀPd bond, coordination of ah ydrogen molecule, and subsequent base-mediated hydrogenolysis of the resulting acyl complex to give the desired aldehyde. The catalytic cyclei sc ompleted by the reactiono ft he palladium hydrobromide complex with base to regenerate Pd 0 .I nt he study by Beller and co-workers, the carbonylpalladium(0) complex [Pd n (CO) m L n ]a nd hydrobromide complex [Pd(Br)(H)L 2 ]were identified as catalytic resting states;t hese complexes were not directly involved in the catalytic cycle. Consequently,t he active catalyst[ PdL] is always at low levels throughoutt he reaction, thus making the oxidative addition the rate-determining step;t herefore, aromatics containing an electron-donating group are slower to react than the corresponding aromatics containing ane lectron-withdrawing group. The efficiency of the Pd 0 catalystd epends on the rate of the oxidative addition relative to the decomposition of Pd 0 ;t he agglomeration eventually leads to the formationo fP db lack, whichc oats the reactor channels. The rate of the agglomeration process is second order in Pd or higher,w hereas oxidative addition is usuallyf irst order in palladium(0),t herefore the rate of Pd decompositiona ccelerates throughout operation time due to the presence of Pd on the reactorc hannels. ICP-MS showed that increasing amounts of catalystw ere lost from solution over operation time, indicating that the presence of existing deposited Pd catalyzes the agglomeration process. The very slow decrease in conversion and yield observed over time for some substrates is caused by the increasing rate of catalyst decomposition over operation time.

Conclusions
Green and sustainable chemical processes rely not only on effectivec hemistry but also on the implementationo fr eactor technologiest hat enhancer eaction performance and overall safety.W eh ave developed ac ontinuous-flow protocolf or Pdcatalyzedr eductivec arbonylation of (hetero)aryl bromides to aldehydes, with syngas as an inexpensive, atom-economic,a nd environmentally friendly source of CO and H 2 .R elativelyl ow catalystl oadings (0.5-1 mol %) andl igand loadings( 1.5-3mol %) provided moderate-to-excellent product yields. The reactionc onsumes only CO, H 2 ,a nd base as stoichiometric reagents. The continuous-flow protocol enabled the reaction time to be significantly reduced compared to the batch protocols available. For continuous-flow reactions, gaseous reagents can be easily and accurately dosed into the system by using mass-flow controllers, thus enablingp recise control over the CO-to-H 2 stoichiometric ratio. The investigationo fg as stoichiometricr atio demonstrated that using CO/H 2 at a1 :3 ratio preventedt he formation of inactive Pd carbonyl clusters and therefore increased product yield. The flow reaction uses pure gases as feedstock to generate gas-liquid segmented flow patterns, which allows the reaction to be completed within 45 min residence time with much smaller excess( 1.1 equiv of CO, 3.3 equiv of H 2 )o fg ases than required forbatch processes. Undert he flow conditions, at the end of the reactor the CO concentrationi sv ery lowb ecause almost all CO has been consumed, thus improving process safety at the outlet due to the low amounto fC Op resent.T he presence of deposited catalyst within the reactor was shown to have an egative effect on the reductivec arbonylation. Inductively coupled plasma-mass spectrometry (ICP-MS) analysis demonstrated that the amount of deposited catalyst on the reactorc hannels increased over the duration of ar un. The deposited catalyst couldb er ecovered using an aqueous nitric acid wash. To improves afety, recentb atch examples attemptedt ou se liquid and solid reagentsa sg as surrogates for CO and H 2 .T he continuous-flow protocolw ith H 2 and CO offers as afe, atom-economic, and environmentally benign alternative to these gas-surrogatep rocedures.T he process developed herein is especially appealing for industrial applications, where atom economy,s ustainability,r eagent cost, andr eagent availability and safety are important factors. Severalk ey active pharmaceutical intermediates (APIs) were synthesized in ac ontinuous ande nvironmentally benign manner.I np articular, ac ontinuous-flow protocol was operated for a6hr un time to produce 3.8 go fa ni mportant active pharmaceutical intermediate. Am ajor advantage of the continuous-flow protocol is the ability to handlep ure H 2 and CO under process-intensified conditions in as afe and scalable manner.N evertheless,t he long run and ICP-MS analysis dem- onstrated thatt here are challenges associated with catalystdecomposition over time. Further work is necessaryt oi dentify improved catalytic systems that allow the reactions to occur withouta ny decomposition over time under process-intensified conditions.

Experimental Section
General methods NMR spectra were recorded on a3 00 MHz instrument (75 MHz for 13 C). Chemical shifts (d)are expressed in ppm downfield from tetramethylsilane (TMS) as internal standard. The letters s, d, t, q, and m stand for singlet, doublet, triplet, quadruplet, and multiplet. Gas chromatography coupled with af lame ionization detector (GC-FID) analysis was performed using aH P5 column (30 m 0.250 mm 0.025 mm). After 1min at 50 8Ct he temperature was increased stepped up to 80 8Ca t28Cmin À1 ,t hen up to 300 8Ca t2 58Cmin À1 , and kept at 300 8Cf or 4min. The detector gas for the flame ionization was H 2 and compressed air (5.0 quality). GC-MS spectra were recorded using aH P5-MS column (30 m 0.250 mm 0.25 mm) with helium as carrier gas (1 mL min À1 constant flow) coupled with a mass spectrometer (EI, 70 eV). After 1min at 50 8C, the temperature was increased in 25 8Cmin À1 steps up to 300 8Ca nd kept at 300 8C for 1min. All solvents and chemicals were obtained from standard commercial vendors and were used without any further purification. All compounds synthesized herein are known in the literature. CAUTION:C Oi sh ighly toxic and flammable, therefore extreme care must be taken when handling. H 2 is extremely flammable. CO alarms must be installed and N 2 purge used at the outlet. All equipment must be set up in aw ell-ventilated fume hood. At horough safety assessment should be made before conducting any experiments.
Representative procedure for reductivec arbonylation of (hetero)arylb romides Data are reported in Ta ble 5. Flow experiments were performed using the continuous-flow setup depicted in Figure 2( also see Figure S1 for alabeled image). The continuous-flow setup is described in detail in the Results and Discussion section. The solution of substrate (0.5 m in PhMe, corresponding to 0.25 m within the reactor), tetramethylethylenediamine (TMEDA) (3 equiv), and diphenylether (15 mol %) as an internal standard (stream 1) and Pd(OAc) 2 (1 mol %o r0 .5 mol %) and cataCXium A( 3mol %o r1 .5 mol %) in PhMe (stream 2) were loaded into their corresponding sample loops. The liquid feeds were pumped using two high-pressure liquid pumps (HPLC) (P,U niqsis) with af low rate of 0.2 mL min À1 for each pump, using toluene as ac arrier solvent. The flow rates of the gas streams were measured and controlled by two calibrated mass-flow controllers (MFCs) using flow rates of 2.5 (CO) and 7.5 mL n min À1 (gas flow rates were measured in units of mL n min À1 , where nr epresents measurement under standard conditions, i.e., T n = 0 8C, P n = 1.01 bar) (H 2 ). The system was maintained at 120 8C and 12 bar pressure to provide % 45 min residence time. The residence time was measured from the four streams mixing at the mixer until color was observed at the BPR. The liquid pump flow rates, temperature, and pressure were measured and monitored by the control platform of the pumping system. Once color was observed at the BPR, fractions were collected for 10 min intervals over a4 0min period. Collection was stopped once no color was observed at the BPR. Yields and conversion were determined by GC-FID using diphenylether as internal standard and the reported values are an average from 30 min collection time. In some cases fractions were combined for purification by silica gel chromatography.
Flow procedure:F low experiments were performed using the continuous-flow setup depicted in Figure 1( also see Figure S1 for al abeled image). Before the reaction, the entire flow system was washed with 20 %a queous nitric acid solution at 60 8Ct oe nsure that no residual Pd was still deposited on the reactor channels and then subsequently washed with acetonitrile and then toluene. Calibrated MFCs (EL-Bronkhorst) were set to the desired flow rates (2.5 mL n min À1 for CO and 7.5 mL n min À1 for H 2 ), and gases started to flow into the reactor.T he pressure was slowly increased at the BPR (Swagelok). When the system reached 3bar,t he liquid pumps were started, each liquid pump was operated at 0.2 mL min À1 ,c orresponding to at otal liquid flow rate of 0.4 mL min À1 ,w ith both pumping toluene. The pressure was slowly increased to 12 bar, and the temperature set to 120 8C. Once at the desired temperature and pressure, the streams were switched to the feed solutions. The feeds were introduced directly through the pumps. The streams were mixed using af our-way inlet mixer at room temperature to give as egmented flow regime and then flowed through the reactor.T he residence time % 45 min was the time measured from the four streams mixing at the mixer until color was observed at the BPR. The tubing after the stainless-steel coil and the Swagelok BPR were immersed in an ultrasound bath and heated at 80 8C for the duration of the experiment to prevent accumulation of solids in front of and within the BPR. The feed solutions were pumped for 350 min, then toluene was pumped for the remaining time as ac arrier solvent. At otal of 18 fractions were collected, a fraction was collected every 20 min (approx. 8mL), and conversion and yield measured by GC-FID using diphenylether as an internal standard (see Figure 4f or conversion and yield over operation time).
ICP-MS:T he amount of Pd deposited within the reactor compared to remaining in solution was determined by ICP-MS analysis. The crude reaction solution collected from the reactor was evaporated under reduced pressure to remove all volatile compounds. The resulting residue was dissolved in acetonitrile/concentrated nitric acid to give ah omogeneous solution. The deposited Pd from the reactor channels was collected by washing with 20 %a queous nitric acid at 60 8C. The solutions were diluted with nitric acid to 40 mL and placed in av ial for microwave digestion. Microwave-assisted acid digestion was carried out in an MLS UltraClave IV instrument. The temperature was ramped up in 30 min to 250 8Ca nd kept at this temperature for af urther 30 min. After appropriate dilution Pd was quantitatively determined at m/z 105 with an Agilent 7500ce inductively coupled plasma-mass spectrometer.Acalibration was performed with an external calibration curve established from 1000 g Pd L À1 standard (CPI International). Indium served as the internal standard.