Revisiting the mechanism of the Fujiwara – Moritani reaction †

The Fujiwara – Moritani reaction between p -methylacetanilide and n -butyl acrylate, catalysed by Pd Ĳ OAc) 2 in the presence of toluenesulfonic acid and benzoquinone, was (re-)investigated using reaction calorimetry and complementary spectroscopic methods. The (most) active catalyst was identified and the catalytic turnover rate was found to be independent of all stoichiometric reagents. Catalyst regeneration and deactivation pathways are discussed.


Introduction
The Fujiwara-Moritani (F-M) reaction constitutes one of the earliest examples of a C-H activation reaction catalyzed by transition metals. The original report in 1967 described the reaction between a styrene-palladium chloride complex and benzene derivatives to form stilbenes. 1 Catalytic reactions were later accomplished by using PdĲOAc) 2 as a precursor for the olefination of simple (hetero)aromatic substrates (Scheme 1, eqn i and ii). 2,3 In the early reports, oxygen gas was used in combination with copper or silver acetate to facilitate the dehydrogenative process, forming water as a byproduct. In principle, the F-M reaction is more desirable than the Heck arylation reaction in terms of atom-and stepeconomies. In practice, however, the lack of a reactive centre poses problems such as poor regioselectivity and/or multiple substitutions of the olefin. Consequently, the F-M reaction was largely eschewed by the synthetic chemistry community, in favour of the Heck arylation reaction towards the end of the 20th century.
In 1979, Horino and Inoue reported that anilides could undergo regioselective ortho-palladation with PdĲOAc) 2 to form palladacycles that can react with alkenes to form 'Heck' products. 25 More than two decades later, the catalytic reaction was demonstrated successfully by an industrial research team, 26 where the addition of benzoquinone (BQ) and toluenesulfonic acid (TsOH) effected the coupling between acetanilides (1) and acrylates (2) under mild reaction conditions (Scheme 2).
The F-M reaction is generally believed to proceed via a redox cycle (Scheme 3): initiating with a reaction between the PdĲII) salt and the acetanilide to produce a palladacycle II. 25 In the presence of an alkene, the complex undergoes migratory insertion and β-H elimination to afford the C-C coupled product and a hydridopalladiumĲII) intermediate (III). At this juncture, it is postulated complex III undergoes reductive elimination to produce acetic acid and a Pd(0) species, which is re-oxidised by BQ under acidic conditions, 27 to regenerate the active PdĲII) catalyst.

Results and discussion
In this work, we initiated a series of studies on the F-M reaction between 4-methylacetanilide (1, where Z = para Me) and butyl acrylate 2, under similar conditions 28 to those reported by Boele et al. 26 (Scheme 2), with the aim to delineate the roles of, and synergistic effects between, the TsOH and BQ in the catalytic cycle, as well as the deactivation processĲes).

Reaction order in PdĲOAc) 2
In this study, reaction progresses were monitored using reaction calorimetry for reactions catalysed by ≥2.5 mol% of PdĲOAc) 2 . At lower catalyst loadings (or where reactions were slow), aliquots were extracted and analysed by HPLC.
PdĲOAc) 2 is known to exist in monomeric, dimeric, trimeric forms in solution, depending on the solvent, and amount of additives and water present. 29 In acetic acid, the reaction solvent employed in F-M reaction, it is expected to be largely trimeric, 30 which can dissociate into dimeric and monomeric species in a rapid equilibrating process (Scheme 4). 31 To determine whether all three forms of PdĲOAc) 2 could be catalytically active in the F-M reaction, 32 the dependence of reaction rate on the [Pd] concentration was studied. By varying the catalyst loading between 2.5-10 mol% in the presence of 1 equivalent of TsOH, a linear plot of initial rates vs. [PdĲOAc) 2 ] (Fig. 1A) and an overlay of the plots of ν obs / [PdĲOAc) 2 ] vs. [1] (Fig. 1B) were obtained, showing that the reaction is first order with respect to [Pd]. The zero-intercept observed in the Fig. 1A suggests there is only one catalytically active species that is kinetically significant.

Effect of additive (p-toluenesulfonic acid)
Brønsted acid additives are commonly employed in many transition metal-catalysed C-H functionalisation reactions (for e.g. Cu, 33 Ru, 34 Co, 35 Rh, 36 Ir, 37   catalytic cycles are rarely studied. In the F-M reaction, the incorporation of 0.5-1 equivalent of p-toluenesulfonic acid (TsOH) has a dramatic effect on the catalytic turnover, allowing the reaction temperature to be lowered from 80 to 20°C. 26 Indeed, during this work, the calorimetric studies were initiated by the addition of TsOH to mixtures of 1, 2 and PdĲOAc) 2 which were thermally-equilibrated at 30°C. It was suggested that the addition of the strong acid protonates an acetate ligand to generate catalytically active PdĲOAc) + ; a hypothesis subsequently supported by a computational study by MacGregor and co-workers, who proposed an energetically-feasible acetate-assisted, concerted, metalationdeprotonation (CMD) mechanism for the C-H metalation step. 40 More recently, it has also been shown that the addition of carboxylic acids to PdĲOAc) 2 promotes the C-H activation step kinetically and thermodynamically. 41 Interestingly, the F-M reaction can also proceed in the absence of carboxylate ions. In 2010, Nishikata and Lipshutz reported that oxidative ortho-olefination of electron-rich acetanilides can be effected at room temperature using cationic [PdĲMeCN) 4 ]ĳBF 4 ] 2 as a catalyst precursor. 42,43 In the same year, Brown and co-workers also reported that an isolated cationic cyclometalated complex containing a η 1bound toluenesulfonate anion is catalytically active in the F-M reaction between acetanilides and methyl acrylate. 44 This led the authors to propose that [PdOTs] + is able to activate the C-H bond of an acetanilide in the absence of acetate, via a 'general base catalysis' (GBC). 45 TsOH (pK a −7) is a much stronger acid than AcOH (pK a 4.7). Hence, we envisaged that TsOH can protonate Pd-bound acetate ligands to generate a mixture of 1 : 1 46 Fig. 2 suggests that while there may be more than one catalytically active species (Scheme 5), there is only one that may be kinetically relevant.
Above 50 mol% (reaction conditions), however, the reaction rate became independent of [TsOH] (saturation kinetics), implying a rapid equilibrium is reached between these homoleptic and heteroleptic species, compared to the turnover-limiting step. 26 Substituting TsOH with different para-substituted aryl sulfonic acids p-Z-C 6 H 4 SO 3 H (where Z = CF 3 , Me, OMe, 1 equivalent) has no discernible effects on the reaction rate (Fig. S1, ESI †). The addition of p-trifluoromethyl benzenesulfonic acid (Z = CF 3 ) to PdĲOAc) 2 in acetone-d 6 induced a δ 19 F shift (Fig. S6, ESI †), which can be used to construct a Job plot (continuous variation method): the resultant plot of Δδ vs. the mole fraction of the reactants displayed a maximum at 0.5 (Fig. 3), suggesting that the formation of a 1 : 1 complex is the most thermodynamically favourable; the curvature of the plot suggests a K eq value > 100. 48 As may be expected, an attempt to reproduce the titration experiment in acetic acid did not produce any noticeable change in δ 19 F, as the equilibrium will favour the diacetate complex. Hence, in order for either PdĲOAc) + or PdĲOTs) + to Scheme 5 Ligand exchange between acetate and toluenesulfonate ions.  operate as active catalysts under the reaction conditions (where acetic acid is employed as a reaction solvent), they will need to be substantially more reactive than PdĲOAc) 2 (Curtin-Hammett principle). This is further supported by comparing kinetic rates, where the reaction catalysed by PdĲOTs) + under acetate-free conditions (reported to be 0.115 mM min −1 in acetone-d 6 ) 44 is approximately 20 times slower than the reaction performed with PdĲOAc) 2 /TsOH in acetic acid (2 mM min −1 , present work). Combined with the result of the Job plot, and the absence of any para-substituent effect on the aryl sulfonate anion, we conclude that PdĲOAc) + is likely to be a more active catalyst than PdĲOTs) + under the reaction conditions.

Catalyst deactivation
A series of 'same excess' experiments 49 were performed to probe the catalyst stability towards different reaction components (Table 1). Experiments 1 and 2 were performed with different initial concentrations of acrylate 2 (0.25 and 0.175 M, respectively), maintaining the same excess of the acetanilide 1 (0.05 M). The resultant plots (Fig. 4A) showed that experiment 2 proceeds with a faster reaction rate than experiment 1, indicative of the catalyst deactivation. This can be attributed entirely to the presence of products (experiment 5, Fig. 4B). Subsequent addition of each of the two reaction products, hydroquinone (HQ) and 3 (experiments 3 and 4, respectively), showed that the HQ has a stronger inhibitory effect than the substituted alkene 3 (Fig. S5, ESI †).
The good overlap between the curves in Fig. 4B also shows that other catalyst deactivation pathways, such as the formation of Pd black, is not significant under these reaction conditions.
Completing the cycle: regeneration of the active catalyst BQ is believed to have dual roles in this process, 51 acting as an oxidant and a ligand to stabilize the low valent-Pd, preventing it from degradation to Pd black. It was also proposed that following β-hydride elimination, the resultant AcO-Pd-H intermediate III (Scheme 3) is reduced to a BQstabilised Pd(0) species, which is regenerated to PdĲII) by BQ in the presence of acetic acida computational study of this process by López and Maseras showed that the transformation of (η E -BQ)PdĲAcOH) 2 to PdĲOAc) 2 can occur via consecutive hydrogen transfer from AcOH to BQ (Scheme 6)the energies were computed to be 92.1 kJ mol −1 , and 65.7 kJ mol −1 for the reverse (dehydrogenative) process. 50 The low energy associated with the dehydrogenative step will account for the observed HQ inhibition of the catalyst.

Rate dependence on the concentration of reactants
In the previous study by the de Vries group, 26 the electrophilic attack of the cationic PdĲOAc) + on the arene ring was proposed to be the turnover-limiting step, based on the following observations: i) the F-M reaction proceeds faster with electron-rich acetanilides (Hammett correlation); ii) observation of a significant kinetic isotope effect (k D /k H = 3; established by comparing k obs obtained from using acetanilide-h 5 and d 5 ); and iii) the reaction between the isolated palladacycle complex II with the olefin is an order of magnitude faster than the equivalent reaction using the catalyst generated in situ. In the study by Brown et al., conducted in the absence of acetate-ions, 44 the reaction also exhibited first order dependency on acetanilide, zero order in BQ and a modest effect of the alkenes, which also supports a turnover-limiting metalation step. Thus, we were rather surprised to find that the rate of the reaction is, in fact, independent of the initial acetanilide concentration (Fig. 5). Subsequently, the reaction rate was also shown to be independent of the initial concentrations of the alkene and BQ (Fig. S2 and S3, ESI †).
The apparent zero-order in acetanilide may be explained by the existence of an acetanilide-bound complex (IV) immediately prior to the rate-limiting cyclometalation step (Scheme 7). The proposal will also explain why the heteroleptic [PdĲOAc)]ĳOTs] complex is more active than the corresponding ditosylate or diacetate salts, as it contains a labile OTs that can be displaced easily by the acetanilide substrate to form complex IV.

Kinetic modelling
The initiating steps of the updated catalytic cycle shown in Scheme 7 is reminiscent of the Michaelis-Menten mechanism, where the reaction is dominated by a reversible binding of the substrate (acetanilide) to the active metal centre (K 1 ) to form complex IV, prior to slow, irreversible C-H bond cleavage to form the cyclometalated complex II (k 2 ), which undergoes spontaneous C-C coupling (zero order in alkene) to generate the product 3 and the hydridopalladium intermediate (k 3 ). BQ then participates in the regeneration of the active catalyst, which can be inhibited by the HQ byproduct (K 4 ).
A microkinetic model of these elementary steps can be constructed from the reaction intermediates and reactants, incorporating six discreet rate constants (Scheme 8). Using an ordinary differential equation (ODE) solver, the data gathered from 16 independent kinetic experiments were fitted to the model, from which the rate constants can be extracted (Fig. S7-S9, ESI †).
A good fit between the simulated and experimental data suggests that catalyst decomposition (e.g. formation of inactive Pd black) is not a significant process under these reaction condition, as was previously verified by the 'same excess' experiments ( Fig. 4 and S5 †). While the rate constants confirmed k 2 as the turnover-limiting step (as expected), it is interesting to note the magnitudes of the two equilibrium processes: K 4 (0.05) ≪ K 1 (9.17 × 10 4 ), suggesting that the stronger binding of the acetanilide (compared to HQ) to the [PdĲOAc)]ĳOTs] provides the driving force for the turnover.

View Article Online
Towards the end of the reaction, as the concentration of acetanilide is depleted, the binding of the HQ by-product become competitive, leading to catalyst inhibition.

Role of the terminal oxidant (BQ vs. O 2 )
Benzoquinone is widely implicated in the catalyst regeneration step. It has an important role in stabilising the monometallic Pd species, thus preventing the aggregation of Pd that leads to the formation of catalytically inactive Pd black. In the studies described above, a stoichiometric amount of BQ was employed. Under these conditions, the formation of Pd black was not a competitive process. However, the use of BQ as a stoichiometric reagent is unattractive in terms of the atom-economy. BQ and HQ are known irritants and possible carcinogens, so it will be desirable to reduce or replace them with safer oxidants. Previously, Pd-catalyzed F-M reactions have been reported to proceed under high O 2 pressures (>8 bar) and elevated temperatures with an excess of the arene reactant. [52][53][54] Encouraged by these accounts, we attempted to replace BQ with O 2 . However, our efforts were thwarted by highly capricious results: in the absence of BQ, a large variation of reaction yields between 11 to >95% were obtained from twenty-two individual, but identical reaction mixtures conducted under aerobic conditions. Speculating that diffusion of gaseous O 2 into the liquid-phase may be the cause of the irreproducibility, reactions were repeated up to five times, in parallel, on a larger scale with mechanical (overhead) stirrers (STEM Integrity10 Reaction Station), varying the stirring rate between 350-1200 rpm. In all cases, no correlation between reaction yield and stirring speed/ reaction vessel can be established. In this part of the study, the formation of Pd black was observed in the early stages of the reaction, irrespective of the reaction outcome. This led us to postulate that, in the absence of BQ, Pd aggregation occurs readily to form a 'cocktail' of interchanging Pd species, including complexes, colloids/clusters and nanoparticles 55a highly stochastic system that is difficult to control and replicate using stirred glass reactors and ambient temperature and pressure. 56 Indeed, reproducibility of the reaction could only be achieved by the addition of at least 50 mol% of BQ or 20 mol% of CuĲOAc) 2 (as a redox mediator).

Conclusions
The work provides new insights to the kinetic profile of the F-M reaction (Schemes 7 vs. 3). In the presence of TsOH, a heteroleptic [PdĲOAc)]ĳOTs] complex is generated, which is catalytically more active than either PdĲOAc) 2 or PdĲOTs) 2 . The F-M reaction was found to operate under saturation (Michaelis-Menten) kinetics, where zero order rate dependencies were observed for all the stoichiometric reagents (acetanilide, acrylate and BQ). This is attributed to the reversible binding of the acetanilide substrate to the catalytically active Pd species, prior to a slow intramolecular C-H activation step. Subsequent kinetic modelling revealed that competitive binding between the acetanilide and HQ to the active catalyst is also important in achieving catalytic turnover.
Irreversible catalyst deactivation occurs through the Pd aggregation, which can be prevented by the addition of BQ or Cu salts. In the present work, the coupling between 1 and 2 were attempted under aerobic conditions (in the absence of BQ or redox mediators). Under ambient conditions, quantitative conversions can be obtained in some instances, but the reaction outcome was found to be hugely irreproducible, possibly due to the highly stochastic nature of the Pd-Pd aggregation process and the difficulty in controlling the delivery of O 2 to the solution under ambient conditions. Understanding the causes and control of these processes will be important in the future development of atom-economical and sustainable oxidative coupling reactions under ambient, aerobic conditions.

Conflicts of interest
There are no conflicts to declare.