Evidence for Suzuki–Miyaura cross-couplings catalyzed by ligated Pd3-clusters: from cradle to grave

Pdn clusters offer unique selectivity and exploitable reactivity in catalysis. Understanding the behavior of Pdn clusters is thus critical for catalysis, applied synthetic organic chemistry and greener outcomes for precious Pd. The Pd3 cluster, [Pd3(μ-Cl)(μ-PPh2)2(PPh3)3][Cl] (denoted as Pd3Cl2), which exhibits distinctive reactivity, was synthesized and immobilized on a phosphine-functionalized polystyrene resin (denoted as immob-Pd3Cl2). The resultant material served as a tool to study closely the role of Pd3 clusters in a prototypical Suzuki–Miyaura cross-coupling of 4-fluoro-1-bromobenzene and 4-methoxyphenyl boronic acid at varying low Pd ppm concentrations (24, 45, and 68 ppm). Advanced heterogeneity tests such as Hg poisoning and the three-phase test showed that leached mononuclear or nanoparticulate Pd are unlikely to be the major active catalyst species under the reaction conditions tested. EXAFS/XANES analysis from (pre)catalyst and filtered catalysts during and after catalysis has shown the intactness of the triangular structure of the Pd3X2 cluster, with exchange of chloride (X) by bromide during catalytic turnover of bromoarene substrate. This finding is further corroborated by treatment of immob-Pd3Cl2 after catalyzing the Suzuki–Miyaura reaction with excess PPh3, which releases the cluster from the polymer support and so permits direct observation of [Pd3(μ-Br)(μ-PPh2)2(PPh3)3]+ ions by ESI-MS. No evidence is seen for a proposed intermediate in which the bridging halogen on the Pd3 motif is replaced by an aryl group from the organoboronic acid, i.e. formed by a transmetallation-first process. Our findings taken together indicate that the ‘Pd3X2’ motif is an active catalyst species, which is stabilized by being immobilized, providing a more robust Pd3 cluster catalyst system. Non-immobilized Pd3Cl2 is less stable, as is followed by stepwise XAFS of the non-immobilized Pd3Cl2, which gradually changes to a species consistent with ‘Pdx(PPh3)y’ type material. Our findings have far-reaching future implications for Pd3 cluster involvement in catalysis, showing that immobilization of Pd3 cluster species offers advantages for rigorous mechanistic examination and applied chemistries.


Introduction
Suzuki-Miyaura cross-coupling (SMCC) reactions are ubiquitous in applied chemical synthesis laboratories around the world, forming a keystone in the synthesis of an eclectic array of high value organic products. 1Palladium is the metal catalyst of choice for many of these transformations, although to be sustainable this precious metal ought to be used responsibly, i.e. used at low catalyst loadings, and be efficiently recovered and recycled, particularly in large scale chemical processes. 2Advances in the design of well-dened ligand systems (e.g. the Buchwald-type ligands such as XPhos, SPhos and many others) have positively contributed towards applied SMCC reaction processes. 3On the other hand, multinuclear Pd species, from small Pd n clusters to Pd nanoparticles, offer unique reactivity and activity in SMCC reactions, as well as other cross-coupling chemistries. 4Immobilized-Pd 5 or the use of supported Pd nanoparticles allow for more effective 6 catalyst recycling, particularly when used in conjunction with magnetic co-additives such as iron. 7Moving towards 2030, one can argue that simply improving the activity and recyclability of Pd catalysts is not enough.However, if one can alter reaction outcomes of cross-coupling reactions based on unique reactivity of a reproducible, well-dened multinuclear Pd species, then that could be a highly exploitable tool in synthetic chemistry. 8In this context, the identication of Pd 3 clusters of the type [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X (X can be a variety of coordinating or noncoordinating anions; structures originally discovered and characterized by Coulson and Dixon) 9 as competent, highly active and selective catalysts for cross-coupling reactions has prompted several research groups to exploit their distinctive and unique behavior.4d,10 SMCC reactions with di-and tetra-nuclear Pd complexes have also been reported, although with less detailed mechanistic evaluation. 11A guiding example with Pd 3 is the crosscoupling of 2,4-dibromopyridine, where typical mononuclear Pd species containing phosphine ligands exhibit C2 site-selectivity. 12n the other hand, [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X species (X = Cl or OAc) exhibit C4 site-selectivity.10b Indeed, higher order stabilized Pd nanoparticles exhibit similar behavior. 13Although the mechanism for this site-selectivity switch is not fully known, the ability to program either reaction outcome by the choice of catalyst provides synthetic chemists with a powerful tool for wider applications.
In 2017, Li et al. reported the reactivity and behavior of [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X species in SMCC reactions, suggesting it to be a highly active catalyst for the cross-coupling of an array of coupling partners, including aryl chlorides.4d The latter point is remarkable, as clearly the aggregated cyclic Pd 3 cluster increases the reactivity of this Pd-phosphorus based catalyst system.Moreover, the authors suggested that there was an inversion of the steps in SMCC reactions catalyzed by [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X species (X = SbF 6 ).Despite the mechanistic complexity of typical SMCC reactions, i.e. the difficulty in elucidating Pd nuclearity and speciation (both Pd and B), the scientic community broadly agrees with a sequence of oxidative addition, transmetallation and reductive elimination steps.However, a transmetallation-rst step was proposed for the action of [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X species, which is followed by oxidative addition (s-bond metathesis suggested) and reductive elimination at the intact Pd 3 cluster center.Thus it would appear that [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X operates in a mechanistically distinct manner to typical mononuclear Pd catalysts, particularly Pd 0 (PPh 3 )n catalyst systems.Indeed, Fairlamb et al. demonstrated the normally expected sequence is only observed for Pd 0 (PPh 3 ) n where n > 3 in reactions of 2-bromopyridine 10a or 2,4-dibromopyridine 10b with aryl boronic acids.When n < 3 higher order Pd n clusters play a key role.
Despite the work conducted on [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X catalyst species, there is limited information available on the pathways for catalyst activation, understanding productive catalyst turnover and potential pathways for catalyst deactivation, i.e. from cradle to grave.Indeed, apart from our preliminary kinetic studies, 10a there is very little reported on the speciation and kinetic behavior of [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X species in catalytic cross-coupling reactions.Preliminary X-ray Absorption Spectroscopy (XAS) data on the behavior of [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X species reported by Li et al. show changes, 4d however the EXAFS data were not tted to structures proposed making speciation assignments difficult.This prior work, in addition to Pd 3 Cl 2 's distinct selectivity and high activity in cross couplings, prompted us to conduct further XAS studies and mechanistic experiments on the Pd 3 cluster system.
Herein, we report our ndings on the behavior of a phosphine-supported (immobilized) [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 2 (-PPh 2 )-Ar-supported]Cl species in SMCC reactions.Various immobilized systems have been applied to SMCC reactions and are reviewed elsewhere, 14 however, we have focused on a supported catalyst system to facilitate our structural study of the intact Pd 3 cluster species in a manner that [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]X cannot alone.We have further conducted studies using [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]Cl as the catalyst system, so that there is a degree of comparability enabling broader conclusions to be reached.In a wider context, the combination of catalyst immobilization strategies and in solution structural characterization (XAS) presented in this study provides a robust approach for discriminating reaction participants and evaluating Pd speciation in SMCC reactions.In doing so we signpost how this approach can be invaluable for furthering our understanding of this important class of reactions, while recognizing the complexity of the catalyst speciation (summarized in Fig. 1).

Results and discussion
Synthesis and characterization of immobilized-Pd 3 Cl 2 3 The synthesis of [Pd 3 (m-Cl)(m-PPh 2 ) 2 (PPh 3 ) 3 ]Cl 1 (referred to as Pd 3 Cl 2 1) can be accomplished by the reaction of trans-PdCl 2 (-PPh 3 ) 2 with H 2 in aniline at 90 °C.10a The immobilization of 1 used of an (aminomethyl)polystyrene resin (∼1.5 mmol g −1 amine loading, 200-400 mesh), which was treated with a linkable N-succinimidyl 3-(diphenylphosphino)-propionate phosphine ligand to afford 2, which was then directly reacted with Pd 3 Cl 2 1, giving the immobilized-Pd 3 Cl 2 3 (referred to as immob-Pd 3 Cl 2 3) (Scheme 1).The non-polar polystyrene resin was hypothesized to provide a suitable environment in which to interact with the Pd 3 cluster, since the PPh 3 /PPh 2 groups give it a non-polar hydrophobic chemical environment, particularly above and below the plane of the triangular Pd 3 structural motif.Furthermore, we anticipated that these interactions protect the Pd 3 cluster from the harshness of the basic environment induced by the use of n-Bu 4 NOH.
The connectivity of the Pd 3 Cl 2 cluster motif during the immobilization process was conrmed by a solid-state 31 P NMR experiment (Fig. 2a).The 31 P NMR spectra of Pd 3 Cl 2 1 and immob-Pd 3 Cl 2 3 conrm the presence of the triangular Pd 3 motif anchored with functionalized (aminomethyl)polystyrene resin; similar 31 P chemical shis are seen for the phosphorus chemical environments.For Pd 3 Cl 2 1 we observed a broad peak with some ne structure at ca. d −7 to 43 ppm, which possesses spinning side bands at d −44 and 80 ppm.These signals are attributed to the distal and proximal phosphines ligated to Pd.A similar but broadened 31 P NMR spectrum is seen for immob-Pd 3 Cl 2 3 with the same main feature around −7 to 43 ppm, which given the sensitivity of 31 P NMR points to the presence of the same phosphorous environments.The phosphine functionalized resin prior to addition of the Pd cluster is quite different with a singlet centered at −13 ppm with two side bands (Fig. 2a).None of this signal at −13 ppm remains in immob-Pd 3 Cl 2 , showing that all the resin bound phosphine is interacting with the Pd cluster.As expected, free PPh 3 and excess Pd 3 Cl 2 are observed in the solution phase by 1 H and 31 P NMR aer the immobilization step (Fig. S3 and S4 †).
A similar, but broader peak is seen for immob-Pd 3 Cl 2 3, along with spinning side bands.broader signals, which we associate with the subtly different pore environments of the resin support, and the substituted possibility of phosphine being distal or proximal to chloride.However, there is no evidence to show which phosphine (i.e.distal or proximal position to chloride) undergoes substitution or how many substitutions occur within the cluster.Importantly, the reectance UV-vis spectrum of immob-Pd 3 Cl 2 3 (red-shied) shares similarity with Pd 3 Cl 2 1 (Fig. 2b), conrming the structure of the triangular Pd 3 motif following immobilization.There is a physical difference in the color of these materialsimmob-Pd 3 Cl 2 3 is dark purple whereas Pd 3 Cl 2 1 is of a dark red appearance (Fig. 2d).Attenuated total reectance infrared spectroscopic analysis of immob-Pd 3 Cl 2 3 revealed the presence of the vibrational peaks associated with the amide carbonyl group, that overlay the amide group in the spectrum of the phosphinefunctionalized resin 2 between 1640-1680 cm −1 (Fig. 2c).This implies the preservation of the amide linkage in immob-Pd 3 Cl 2 3 and that the phosphine remained tethered to the polystyrene during the Pd 3 immobilization reaction.These are not present in the stating resin material (no amide carbonyl).The retention of the amide linkage is important as it conrms the tethered phosphine was not released from the polystyrene anchor such that it was able to act as a new solution phase ligand and that any exchange really was the result of interaction of Pd 3 Cl 2 1 with the resin bound phosphine 2.
Summarizing the initial characterization of the immob-Pd 3 Cl 2 3, 31 P SSNMR provides evidence of the Pd 3 motif remaining, and no signicant other P containing species being present, UV-vis data show a lack of change to the metal-ligand interactions and ATR-IR conrms, via presence of the carbonyl, that the linker unit is intact.

Suzuki-Miyaura cross-coupling (SMCC) catalysis
The reactivity of immob-Pd 3 Cl 2 3 was examined in a benchmark SMCC reaction of 4-uoro-bromobenzene 4 and 4-methoxyphenyl boronic acid 5 under optimized reaction conditions  (Table 1 and Fig. 3).The conditions were identied from our earlier study on the regioselective SMCC reaction of 2,4-dibromopyridine with aryl boronic acids, which revealed the unusual C4 site-selectivity for Pd 3 Cl 2 1 and Pd(OAc) 2 /#2PPh 3 precatalyst systems.10b As an aside, the same atypical C4 site selectivity is observed with immob-Pd 3 Cl 2 3 as catalyst, demonstrating the generalizability of immob-Pd 3 Cl 2 3 as a direct analogue for nontethered Pd 3 Cl 2 1 in SMCC catalysis (Table S4 †).In the present mechanistic study, however, we chose to focus on the simpler reaction of 4 with 5 to avoid possible regioselectivity complications.The base, n-Bu 4 NOH, is employed in a THF/H 2 O solvent mixture (1 : 1, v/v).While it is a highly basic, mixed solvent system, there is one single phase, where the organoboron Where there is a question about Pd speciation, it is imperative to limit speciation in respect to boron, which makes the conditions advantageous for studying Pd catalyst speciation.We selected to employ 68 ppm (by moles) of Pd (note: total ppm Pd is three times catalyst ppm; this Pd concentration equates to a 0.34 mol% catalyst loading with respect to 4).The SMCC reaction catalyzed by immob-Pd 3 Cl 2 3 reaches completion within 15 minutes to give cross-coupled product 6 (entry 1, Table 1).Lowering the Pd catalyst loadings to 0.22 mol% (immob-Pd 3 Cl 2 3, 45 ppm by moles Pd) and 0.12 mol% (immob-Pd 3 Cl 2 3, 24 ppm by moles Pd) led to extended reaction times (18 and 28 minutes, respectively) (entries 2 and 3, Table 1).
Remarkably, under comparable conditions using Pd 3 Cl 2 1, the reaction proceeded to completion within 5 minutes (entry 4, Table 1).We anticipated a higher rate of catalysis for Pd 3 Cl 2 1 since the accessibility to the "Pd 3 Cl 2 " core structure in immob-Pd 3 Cl 2 3 would be subject to greater substrate diffusion limitations.Nevertheless, the high rate of this reaction is remarkable, which supports the claims made about high reactivity using aryl cross-coupling partners (including aryl chlorides) made by Li and co-workers in their independent study.4d By way of comparison, the "Pd(OAc) 2 /2PPh 3 " pre-catalyst system, known to invoke Pd 3 cluster catalysis in SMCC and Kumada cross-coupling reactions, 10b also effectively catalyzed the reaction (entry 5, Table 1).
The SMCC reaction progress was monitored using 1 H/ 19 F NMR spectroscopic analysis, enabling kinetic proles to be obtained.We examined both immob-Pd 3 Cl 2 3 and Pd 3 Cl 2 1, at different catalyst loadings.Interestingly, sigmoidal kinetic proles were seen for reactions catalyzed by both immob-Pd 3 Cl 2 3 and Pd 3 Cl 2 1, suggesting that a pre-catalyst activation step is necessary.The induction period extends to slightly longer times on reducing the catalyst loading, which is longer for the immob-Pd 3 Cl 2 3 system than for Pd 3 Cl 2 1, presumably because of diffusion issues for the former catalyst.For the Pd 3 Cl 2 1 catalyst, one needs to go down to very low catalyst loadings (6.8 ppm Pd, 0.034 mol%) to reveal the full extent of the induction period.
Given the high catalytic activity associated with Pd 3 Cl 2 1 we hypothesized that the inuence of the exogenous phosphine resin support 2 also merited assessment.The kinetic proles in Fig. 3c for the reaction catalyzed by Pd 3 Cl 2 1: exogenous phosphine resin 2 in a 1 : 1, 1 : 2 and 1 : 3 ratio revealed that the activity is reduced versus Pd 3 Cl 2 1 alone and further reduced with increasing equivalents of resin (Fig. 3c).The model SMCC reaction was also recharged with both reactants and additives at t = 15 min (based on when the reaction ends as judged from the kinetic curve shown in Fig. 3) to assess catalyst reactivity and catalyst deactivation. 1H NMR analysis of samples taken at the end of each run revealed the temporal reactivity of 3 (six runs) with enduring good conversions to cross-coupled product (see Fig. 4).The result of this experiment suggests that the active catalyst can re-enter into the catalytic cycle in the presence of fresh reactants.

Probing the heterogeneous behavior in the SMCC reactions
To assess whether the SMCC reaction occurs heterogeneously, or through a Pd leaching mechanism from the polymer resin, experiments were conducted to gauge the heterogeneity/ homogeneity of the processes involving the immob-Pd 3 Cl 2 3 catalyst system.The mercury-drop test has been one of the common tools for examining heterogeneous catalyst behavior, where a positive test can result in poisoning of any aggregated Pd 0 present (inferred as being the active catalyst species).The alternative outcome is that mercury has no effect, indicating a role for soluble molecular catalyst species (likely of low metal nuclearity). 15However, the applicability of the test for reactions involving different molecular complexes of Pd, 16 Pt 17 and Rh 18 has been recently questioned, due to direct reactions of mercury with homogenous metal species such as palladacycles, 19 or a high dependence of the test outcomes to the operational conditions (e.g.stirring rate/size of the reaction vessel). 20The validity of the mercury poisoning test for M 0 /NHC or M II /NHC (M = Pd, Pt) complexes was brought into question by Ananikov et al. 20,21 Certain mononuclear Pd II L n complexes react with metallic Hg through an oxidative-reductive transmetalation process to form Hg II L n and Hg x Pd y species, whereas Pd 0 L n complexes decompose under exposure to metallic mercury. 20hus, the issue is that a poisoned catalytic reaction system Fig. 4 Recharging SMCC reaction promoted by immob-Pd 3 Cl 2 3 (1 mol%, 68 ppm Pd) with fresh batch of substrates and base at the end of catalysis (taken to be at 15 minutes intervals based on earlier reaction profiles in Fig. 3) and monitoring the reaction progress up to six runs (ST = sampling time during the overall experiment).The reaction was analyzed by 1 H NMR using 1,3,5-trimethoxybenzene as internal standard.
could have (potentially) involved mononuclear PdL n complexes, bringing about an incorrect interpretation when poisoning does occur.Conversely, if a catalyst system is unperturbed by mercury addition, where the operational conditions are carefully selected to ensure the test can poison reactivity (see ESI †), the attribution of activity to aggregated Pd(0) may still be reasonably ruled out.That is, if the test is conducted correctly to ensure mercury contacts the entire system in adequate excess, an outcome of unperturbed reactivity in the presence of Hg is sufficient to exclude Pd(0) particles, which would suffer from Hg poisoning (while the converse test outcome is inconclusive for the reasons above).
With all these studies in mind, we conducted a mercury poisoning test for the model SMCC reaction catalyzed by immob-Pd 3 Cl 2 3 (68 ppm, 0.34 mol%), which reached 50% product conversion aer 8 minutes (see Fig. 2a).Addition of metallic mercury (300 equiv.with fast agitation on a standard magnetic stirrer hot platesee Section 2.8 in the ESI †) at t = 8 minutes had no effect on the reaction rate or product conversion, reaching completion within 15 minutes (Fig. 5).Interestingly, in a control reaction we determined that Pd 3 Cl 2 1 does react with metallic mercury, but only on longer timescales (see Fig. S21 †), leading to decomposition.Our results, taken together show that: (a) Pd nanoparticles (Pd 0 species) are not operative in SMCC reactions catalyzed by immob-Pd 3 Cl 2 3; (b) the Pd 3 Cl 2 cluster motif is likely protected from metallic Hg in immob-Pd 3 Cl 2 3, under the reaction conditions tested.
In another experiment, the immob-Pd 3 Cl 2 3 catalyst was separated from the model SMCC reaction aer 4 minutes to monitor the reaction progress.The analysis showed that the reaction proceeds although at a signicantly slower rate (see Fig. S22 †).While the result could suggest the leaching of soluble Pd 0 species into the solution (note: not supported by the Hg drop test), some of the supported Pd catalyst particles evaded removal by ltration, even using syringe lters (with 0.22micron pore size).Moreover, the presence of small particles were observed in the reaction ask post-ltration.Nevertheless, the removal of the majority of immob-Pd 3 Cl 2 3 showed reaction retardation.Such retardation (taken in isolation) on ltration of the solid species can't explicitly exclude the possibility of leached Pd nanoparticles or leached molecular species becoming bound to the lter, but the much-decreased rate on ltration (given the difficulty of excluding all the solid material) is consistent with the attribution of reactivity drop to the removed solid immob-Pd 3 Cl 2 3.
The three-phase test, devised by Rebek and Gavina, 22 was used to test for potentially reactive soluble catalytic Pd species that detach from a heterogenous (pre)catalyst and leach into the solution (being the active catalyst species).This is typically done by immobilizing one of the reactants and probing its reaction with another soluble coupling partner, in the presence of the immobilized catalyst. 23The diffusion of two solids or solidbound reaction components should be inhibited versus diffusion of soluble component(s).For this experiment, we prepared a polystyrene-bound aryl iodide 7. SMCC reactions of 7 with 5 were examined using either immob-Pd 3 Cl 2 3 or soluble Pd 3 Cl 2 1, under otherwise identical conditions (Table 2).
The immob-Pd 3 Cl 2 3 catalyst was effectively inactive in the three-phase test (less than 1% conversion to cross-coupled product 8 aer 24 h).Immobilization of the aryl halide restricts the accessibility of the immob-Pd 3 Cl 2 3 catalyst; thus, any catalytic reaction cannot proceed under the employed reaction conditions.The outcome conrms that leached soluble Pd species are not playing a signicant role under the reaction conditions tested.The equivalent reaction using Pd 3 Cl 2 1 as the catalyst was sluggish, as might be expected given the introduction of a diffusion constraint on one species involved, and the possibility the amide linker in immobilized derivative 7 might be considered inhibitory to the aryl halide (28% product  conversion noted aer 24 h). 24It should be noted that in an effort to keep the reactions more similar in timescale we have used the more reactive aryl iodide for these experiments (weak C-I bond, facile oxidative addition), which, while not perfect, removes the risk of speciation differences over longer timescales of reaction.
We therefore also conducted a control to determine that immob-Pd 3 Cl 2 3 was able to effectively cross-couple 4-iodo-N-methylbenzenamide 9 with p-methoxy-phenyl boronic acid 5 (gave 86% conversion to 10 aer 1 h) (Scheme 2).It was important to conrm this, as it shows similar reactivity to that of our normal test reaction between 4 and 5, but with a similar inhibitory group/aryl iodide combination to that in immobilized derivative 7. Table 2 three-phase test experiment for the SMCC reaction of an immobilized aryl iodide 7 with 4methoxyphenyl boronic acid 5 to give immobilized product 8.
The three-phase test results cannot rule out the potential release of a "Pd 3 Cl 2 " species (or derivative) cluster from the support into the solution following a reaction with substrate(s), which could be subsequently redeposited on to the support.Supported Pd catalysts are known to exhibit releaseredeposition (catch) mechanisms in catalysis. 25There is no evidence for loss of the "Pd 3 Cl 2 " species from immob-Pd 3 Cl 2 3. Indeed, the only way to trigger release is by the reaction of exogenous (excess) PPh 3 with immob-Pd 3 Cl 2 3 in CH 2 Cl 2 , which Scheme 2 Control SMCC of 4-iodo-N-methylbenzeneamide 9 with 4-methoxyphenyl boronic acid 5, catalyzed by immob-Pd 3 Cl 2 3. occurs instantaneously at room temperature to deliver Pd 3 Cl 2 1 into solution; conrmed by 31 P NMR and ESI-MS (+ve mode) (Fig. 6a).
This ability to trigger release of the Pd 3 cluster from the resin support affords an additional test as to whether the Pd 3 cluster remains intact on the resin during reaction.An SMCC reaction of 4-uoro-bromobenzene 4 and 4-methoxyphenyl boronic acid 5 was again performed under the normal reaction conditions with 68 ppm Pd, but stopped aer 10 minutes by ltration (where Fig. 3 shows ∼60% conversion, but catalytically relevant species are still present).The ltered resin was then washed thoroughly and reacted with excess PPh 3 to trigger release of the cluster into clean solvent (details given in the ESI †).ESI-MS of the released Pd 3 shows not only the initial [Pd 3 Cl] + ion, but also [Pd 3 Br] + (Fig. 6b), pointing to the resin bound species having been involved in a process that swaps out the bridging halide.
The results taken together suggest that the induction period observed from the cross-coupling catalysis is not a consequence of leached mononuclear or nanoparticulate Pd catalyst formation.At 10 minutes where >50% conversion has occurred ICP showed only leaching of 0.58% of the initial Pd into the liquid phase for a sample ltered to remove the solid (and some of the polymer bound material undoubtedly evades the lter, so this is a worst-case scenario).This is further corroborated by the quantication of the XAS step-edge height for the pre-reaction immob-Pd 3 Cl 2 3 and 'during reaction' samples-proportional to concentration-implying the solid contained the same Pd amount, within error, in both cases.In addition to these tests performed to exclude catalysis by leached molecular or nanoparticle species, the overall rates of reaction imply a minor component would have to exhibit far greater performance than those seen for SMCC reactions at these temperatures ∼40 °C (see Table S11; † it should be noted that typical rates of "ppb Pd" SMCCs or catalysis by "impurity Pd" are generally identied only at signicantly higher temperatures, e.g.110 °C), 26a in keeping with relating Pd-catalyzed cross-coupling reactions where ppm-Pd descriptors have been rigorously examined.26b A possible explanation for the induction period is the interplay of planar aromatic [Pd 3 X] + species with structurally distorted non-aromatic [Pd 3 X/X'] species, both of which are plausible species characterized by single crystal XRD methods.10a The possible importance of d-orbital aromatic stability metal aromaticity 27 in rendering, for example the [Pd 3 (m-SR) 3 (PR 3 ) 3 ] + Y − cluster complexes as being oxygen and moisturestable, 28 has been previously highlighted, 29 and its potential relevance to Pd 3 cross-coupling discussed. 30Here the postulated change from more stable aromatic Pd 3 Cl 2 to the structuraldistorted bromide bridged cluster being accompanied by an induction and rate increase is consistent with this being an important effect in these systems.The primary point, however, is that the immob-Pd 3 Cl 2 3 is a viable tool to study the component mediating Pd n cluster catalysis.

Studies by X-ray absorption spectroscopy (XAS: EXAFS/XANES)
XAS has been used widely to study the physical form of Pd catalysts, 31 particularly immobilized Pd catalyst systems 25b,32 and well-dened and/or evolving Pd nanoparticle species. 33The pros of the approach are: (1) X-ray absorption is unambiguously specic to a given element, in this case Pd; (2) the penetrating nature of X-rays makes it possible to obtain measurements in complex matrices without perturbing the system, here we do not remove from the solvent for analysis; (3) the spectral information obtained is rich, with information about oxidation state, coordination environment and geometries being possible to obtain with appropriate modelling; and (4) it is inherently quantitative with normalized spectra of different species allowing their relative contributions to a mixture to be ascribed.The cons of the XAS approach are: (1) it is an averaging method so small components can be overlooked (but this is true of many techniques); and (2) high intensity beam damage to the Pd samples is possible (reassuringly in the present case excellent agreement is seen for the as prepared immob-Pd 3 Cl 2 3; tting models are based on expected structure).The reference standards selected in our study include combinations of bromide, phosphines and nBu 4 N + species, the latter from the base used in reaction, and these have largely not been reported previously.PdBr 2 is consistent with spectra reported in the literature. 34o explore the identity of the catalytic Pd species during the SMCC reactions, and to identify the fate of the catalyst aer reaction completion, samples of reactions during and aer catalysis were analyzed using Pd K-edge extended X-ray absorption ne structure (EXAFS) and X-ray absorption near edge structure (XANES) techniques.For preparation of the samples, the catalyst particles were separated from the reaction mixture by cannula ltration (paper lter), aer 5 minutes (during productive catalyst turnover, i.e. catalyst activated) and also aer reaction completed.We further examined the concentrated ltrate of the reaction solution aer separation of the supported catalyst system, which was evaporated to dryness both aer 5 minutes and aer reaction completion (Fig. 7).Comparison of these samples allows us to understand whether Pd 3 Cl 2 clusters remained intact and potentially give information about any leached Pd species present.
The Pd K-edge XAS spectra of the ltered catalyst, both during and aer reaction in both the XANES region (Fig. 7a) and based on ts of the EXAFS data to the Pd 3 motif (Fig. 7b), are in excellent agreement with the data obtained for immob-Pd 3 Cl 2 3 before reaction (tting parameters and details are given in the ESI †).The invariant XANES suggest no marked change in either oxidation state or local co-ordination environment, and the fact that the Pd 3 motif (with Pd, P and halogen nearest neighbors) is preserved well in the tting of the EXAFS data strongly points to the fact that the structure of solid-supported Pd 3 clusters remains similar during the catalysis and even aer catalysis.As was highlighted in Fig. 1, this situation afforded by our immobilization strategy is an unusual and valuable occurrence in mechanistic studies of Pd cross-coupling catalysts, where deactivated Pd species are typically all that remains upon postmortem analysis.The structure of immob-Pd 3 Cl 2 3 therefore remains a realistic model for the species bound to the polymer throughout the reaction.Note: XAS and the use of nearest neighbor ts cannot strictly distinguish species 1 and 3, but the difference in reactivity between them, three-phase test and fact that this species is extracted for analysis by ltration of the polymer all imply this is likely to be 3 rather than 1.
Another signicant nding from EXAFS analysis was revealed by comparing the goodness of t for bridging -Br, -Cl and -Ar (a carbon atom in the 1st co-ordination shell, which scatters less than Cl) for each sample.As expected in the prereaction catalyst, Cl provides a suitable t.For the postreaction samples placing Br in the bridging position between two palladiums (average co-ordination number of Pd by Br = 0.67) provides a distinctly better t, and in neither the during reaction or post-reaction sample will placing a carbon in the bridging position, representing a bridging aryl, provide an acceptable t (see ESI †).The better t for the bromide-bridged Pd 3 cluster suggested a substitution of chloride by bromide in the structure of the Pd 3 cluster.This is supported by the reaction clusters released from the resin post-ltration aer 10 minutes of reaction in Fig. 6b being a mixture of Cl and Br bridged Pd 3 clusters (also seen by Li et al. using ESI-MS with both Br and I containing aryl halide reactants).4d The precise steps by which the bromide-containing Pd 3 cluster forms likely involve reactions with substrates (depicted in Scheme 3).Repetition of the reaction kinetic prole for immob-Pd 3 Cl 2 3, but in the presence of a large excess of Br − (1 equiv.n Bu 4 NBr with respect to aryl halide) showed no signicant change in kinetics, suggesting this incorporation must occur reactively (see ESI †).
Plots of the ltrate samples (where polymer resin containing immob-Pd 3 Cl 2 3 was ltered out and the solvent removed to dryness) obtained during and aer catalysis against a variety of reference samples in both XANES (Fig. 7c) and EXAFS (Fig. 7d) regions show no match to the spectral features for palladium metal, Pd 3 clusters, nor other reference samples, including PdBr 2 and [Pd 2 Br 6 ][n-Bu 4 N] 2 11.It should be noted at this point that only very subtle differences are seen experimentally (Pd 7nm -Pd foil ) 35 or predicted theoretically as a function of metallic palladium nanoparticle size (Pd 13-atom -Pd foil ), 36 and so the Pd foil standard is a suitable reference for metallic Pd particles of any size.Attempts to use mass spectrometry to identify the Pd species in the ltrate were unsuccessful, either due to other species present swamping the signal or there being a poorly dened mixture of end product species.The nature of this species is discussed later, but at this point we can condently rule out this leached Pd being present as Pd 3 clusters, which would readily be seen by HRMS, metallic palladium particles (readily seen in XAS) or the other reference samples considered.
The reference sample [Pd 2 Br 6 ][n-Bu 4 N] 2 11 is reported to be catalytically active in cross-coupling reactions and we were therefore careful to consider its possible formation from the Pd 3 X 2 catalyst system.25b,37 Here, complex 11 was prepared through treatment of n-Bu 4 NBr and PdBr 2 in THF and H 2 O (1 : 1) under inert atmosphere at room temperature and was characterized by far infrared, elemental analysis and X-ray single crystal diffraction (Fig. 8).
The pre-catalyst 11 showed very high reactivity in forming cross-coupled product 6.The result of this experiment supports the XAS outcome that leached Pd species (which were relatively inactive) do not share any similarity with 11.It is also notable that even if such species were present at low concentration (below those that we could detect), they would not be capable of promoting the cross-coupling reaction at this temperature at the rate seen by the present highly reactive immob-Pd 3 Cl 2 3 catalytic system.
The catalytic activity of 11 was examined in the model SMCC reaction of 4 and 5 (using 0.1 mol%, 68 ppm Pd), under our working reaction conditions, showing quantitative conversion to 6 within 3 minutes (Scheme 4), cf. 15 minutes for immob-Pd 3 Cl 2 3 (Table 1).
XAS analysis of Pd 3 Cl 2 under SMCC reaction conditions.To provide further insight into the changes to the catalyst occurring, an analogous XAS study of the homogeneous Pd 3 Cl 2 1 undergoing the SMCC reaction under the same conditions was conducted.Unlike immob-Pd 3 Cl 2 3, the homogeneous Pd 3 Cl 2 1 cannot be separated from any leached or reaction products by simple ltration (i.e.without risking signicant perturbation of the catalyst species).Instead, at each stage of the reaction (Pd 3 Cl 2 1 alone; Pd 3 Cl 2 1 in THF/H 2 O 1 : 1 v/v solvent aer n-Bu 4 NOH base added; aer aryl boronic acid added; aer aryl bromide added) an aliquot of the reaction mixture containing all components was rapidly quenched to LN 2 temperatures in a Nalgene cryovial for XAS analysis.The resulting sequence of XANES spectra are shown in Fig. 9a.The initial spectrum of Pd 3 Cl 2 1 is an excellent t to the expected rst shell coordination of Pd in the cluster (Fig. 9b).
However, it is immediately clear from the XANES that even combining with THF/H 2 O/base begins to alter the Pd speciation.Further time and/or reagent addition caused further changes in the spectra (it is difficult to discriminate which, as both time and addition of reagents took place).The changes in the spectra mirror the visual observation that there was an immediate change to a black color upon adding the aqueous n-Bu 4 NOH base to a THF solution of Pd 3 Cl 2 1.While our rst thoughts turned to precipitation of Pd black, no distinct large Pd particles were visible by eye from this point throughout the subsequent reagent addition steps performed (during which the appearance of the reaction solution remained unchanged).The impression of there being a single liquid phase can be misleading if the Pd is nanoparticulate in nature and does not precipitate.However, as will be seen shortly the conclusion that  no Pd metal is being formed (nanoparticulate or otherwise) is supported by further analysis of the XAS data.
Qualitatively the XANES spectra through the reaction steps appear to become progressively more similar to the ltrate during reaction from the immob-Pd 3 Cl 2 3, while bearing little similarity to Pd metal (spectra both shown for reference in Fig. 9a).It is important to note that the above work on immob-Pd 3 Cl 2 3 showed that the leached species seen in the concentrated post-reaction ltrate were inactive for the test SMCC reaction versus either Pd 3 Cl 2 1 or immob-Pd 3 Cl 2 3.
A more quantitative analysis of the XANES data in Fig. 9a and k-space EXAFS data can be achieved with principal component analysis (PCA) and least squares tting.PCA of the XANES series in Fig. 9a shows signicance above the noise in the rst two components (Fig. S34, † the edge region is slightly present in the third component, but noise at the edge where a large change is present leads to articial variation because of slight energy misalignments).A scree plot (Fig. S34b †) also has an elbow aer the rst two points, pointing to only two principal components that contribute signicantly.Similar analysis using an approach used elsewhere 34 on the k-space EXAFS spectrum region between k = 0 and k = 11.5 is again largely comprised of only two components that are signicantly above the noise level of the data (Fig. S36 †).It should be noted that the sample aer adding base, but no other reagents, is slightly less well tted in the EXAFS region by only two components, which may hint at transitional species being present, but there is insufficient data to infer more than this being a possibility.
Given the visual similarity between the post-reaction ltrate sample (leached from immob-Pd 3 Cl 2 3) to the later stages of the reaction of Pd 3 Cl 2 1, a two-component linear combination t was undertaken for each spectrum in the sequence in Fig. 9a.This produced a good t in the XANES region of the data (Fig. S40 †), with the corresponding quantities of Pd 3 Cl 2 1 and the ltrate-like species being shown in Fig. 9c.This pleasingly shows a similarity between the chemical fate of the homogenous and albeit more stable immobilized Pd 3 Cl 2 catalysts, supporting the idea that the immobilized system can facilitate broader studies of Pd 3 Cl 2 1.Given the above inactivity of the leached species in the ltrate that appears over time and the fast kinetics of the reaction, we have no evidence for other species besides Pd 3 Cl 2 1 being involved in the reaction process.
While the precise chemical identity of the species from the ltrate is somewhat unimportant if the species is ascribed as inactive, understanding the fate of Pd 3 Cl 2 1 is still of interest to understand the reactions that bring about its formation.As Spectra offset for clarity, the lowest XANES spectrum is correctly placed on the y-axis with all XANES spectra normalized between 0 and 1 and EXAFS oscillations centered around y = 0 prior to any offset.
noted earlier the XANES spectra of the species in the ltrate did not t visually with spectra of any common standards, such as PdBr 2 , Pd metal, or [Pd 2 Br 6 ] 2− .While a visual similarity in the spectra is a good rst check, in the ESI † we consider a more quantitative approach using principal component analysis, which has been employed in XANES studies for some time (Section S2.16 †). 38This speculatively points to some form of Pd x (PPh 3 ) y -type species being the end fate of Pd in this system.
Finally, the stepwise approach used in this experiment allows us to re-examine the mechanistically surprising claim of Li and co-workers that an aryl-bridge Pd 3 cluster forms as a stable transmetallation intermediate that can be observed by XAS.4d The details of the experiments conducted previously are unclear, but the concentrations used in transmission XAS appear to have been around 0.01 wt% Pd from the details given in their publication, lower than is typically suitable for transmission experiments and likely produce noisy data.The claim concerning transmetallation is on the basis of reduced amplitude at low values of k (<5 Å), but this is extremely susceptible to the background functions and the energy spectra from which the k-space data were obtained were not given.The R-space data presented by Li and coworkers seem distinctly different from those we have consistently obtained for the cluster (homogenous or immobilized) with weaker than expected amplitudes for the Pd-Pd scatter (which dominates the region 2.5-3 Å), and an absence of the Pd-Cl scatter contribution (easier to identify when tting the real part of the data than by inspection of the magnitude R-space plots).Repeating the stepwise experiment with the 1 : 1 v/v 2-propanol/H 2 O solvent mixture they report and using K 2 CO 3 base as they did (which is only moderately soluble, although can form methoxide with alcohols 39 ) we were able to see the Pd 3 Cl 2 1 cluster intact at each step of the process, with only a halogen exchange on adding bromobenzene (Fig. S36-S40 †).Crucially in the sample aer all other reagents were added except bromobenzene the spectrum is a signicantly better t to chloride than carbon (aryl) in the bridging position.In short, obtaining higher quality data at the low concentrations of Pd typically used in SMCC reactions by taking advantage of modern uorescence detection capabilities (B18 Diamond Light Source) and tting this using a summation of theoretically calculated scattering paths (ARTEMIS 40 ) we saw no evidence with either the immobilized cluster or attempts to reproduce their experiment using Pd 3 Cl 2 1 for the proposed [Pd 3 Ar] + intermediate.The other evidence they presented for this was based on ESI-MS of CH 2 Cl 2 solutions of the reagents.4d In our hands we found that the base (K 2 CO 3 ) is insoluble in this solvent/it is not the medium in which the reactions were conducted (note: limited experimental details given in the original publication).It also appears the signal associated with [Pd 3 Ar] + is still <20% of the Pd 3 Cl 2 1 cluster signal, and so even if present in the same amount, XAS being an average technique the Cl bridge would still yield a better t (and have dominated their data also).It is also noteworthy that using some routes to prepare the Pd 3 Cl 2 1 cluster, we have sometimes seen the aryl bridged Pd 3 cluster as a synthetic biproduct (rather than reactive intermediate -Fig.S27 †), which is a possible alternative origin for such species.We have repeated the synthetic route to 1 described by Li and co-workers (yield = 2%) and have not been able to obtain evidence showing the intermediacy of [Pd 3 Ar] + .One nal point is a further potential for confusion in their study arises from the fact that the phenyl group in the PhB(OH) 2 substrate used is the same as that found in the PPh 3 /PPh 2 ligands in Pd 3 Cl 2 1, avoided in the present study through the use of the methoxy-functionalized boronic acid 5.In short, the current study does not provide any evidence to support the existence of the aryl-bridged intermediate suggested by Li and co-workers.4d However, additional work is required to fully investigate this hypothesis and establish if such a species is kinetically relevant, using a series of organoboron compounds to probe the transmetalation step rst hypothesis.

Conclusions
In conclusion, we studied 'Pd 3 Cl 2 ' catalytic behavior in a typical SMCC reaction by tethering it to a resin support to afford a valuable tool for mechanistic studies.Experimental heterogeneity tests, such as Hg drop test and three-phase test, showed no active Pd leached from the immobilized Pd catalyst within the reaction's time scale.Analysis of XAS data from the ltered catalyst during and aer reaction showed the triangular Pd 3 cluster motif intact, with only exchange of chloride for bromide in the bridging position during the substrate turnover.XAS of the ltrate aer catalyst removal showed nonactive Pd species, with no match to a series of reference samples such as PdBr 2 , [Pd 2 Br 6 ][n-Bu 4 N] 2 , and Pd foil.XAS analysis of non-tethered (non-mobilized) Pd 3 Cl 2 1 catalyst in the same SMCC reaction, conducted stepwise, indicates the Pd speciation at the later stages of the reaction is like that in the reaction ltrate when immob-Pd 3 Cl 2 3 was used as catalyst.Using principal component analysis and target transformation from a range of potential standards along with the available EXAFS data, the identity of the inactive Pd species in the ltrate sample from reaction of immob-Pd 3 Cl 2 or later stage reaction of Pd 3 Cl 2 was suggested to be Pd x (PPh 3 ) y , exhibiting a blend of Pd-Pd and Pd-P interactions.This study provides the rst experimentally compelling insight 30 into the speciation of Pd involving the catalytic behavior of Pd 3 type clusters stabilized by phosphorus containing ligands within cross-coupling reactions.Finally, during the proof stage for our manuscript we became aware of a study reported by Scott et al. involving the temperature-inducated activation of the Coulson-cluster on a carbon support.The results are complementary to those described here, and provide insight into how changes in the 'Pd 3 Cl 2 ' cluster can occur at signicantly higher temperatures (between 150 and 250 °C). 41ANES X-ray absorption near edge structure XAS X-ray absorption spectroscopy

Fig. 1
Fig.1(a) General Pd-catalyzed SMCC reaction scheme, (b) the mechanistic approach on Pd catalyst speciation during and after catalysis, (c) and current work showing the higher order Pd 3 catalyst robustness during and after catalysis.

Fig. 5
Fig. 5 Kinetic profile of the model SMCC reaction obtained when excess metallic Hg (300 eq.) added at t = 8 min and reaction progress monitored by 19 F NMR analysis.The Pd concentration of 68 ppm (moles) equates to a 0.34 mol% catalyst loading with respect to 4 [the ArBr].

Table 2 a
SMCC on an immobilized aryl iodide substrate 7 Catalyst Conversion a,b,c to 8 (Conversion monitored by 1 H NMR (consumption of 5).b <3% crosscoupled byproduct formed as result of amide hydrolysis.c The Pd concentration of 68 ppm (moles) equates to a 0.34 mol% catalyst loading with respect to 7 [the ArI].

Fig. 6
Fig. 6 (a) Reaction of immob-Pd 3 Cl 2 3 with 5 equiv.PPh 3 in CH 2 Cl 2 at 22 °C; 31 P NMR spectrum was recorded upon dissolution (within 10 min).ESI-MS (+ve mode) confirms pseudo-molecular ion at m/z 1509, with the correct isotopic distribution pattern centered at m/z 1511 (bottom right); (b) ESI-MS (+ve mode) analysis of the solution after reaction of excess PPh 3 with post-reaction immob-Pd 3 Cl 2 3 cluster which was filtered off 10 minutes into Suzuki-Miyaura catalytic turnover and thoroughly washed, showing both [Pd 3 Cl] + and [Pd 3 Br] + ions shown, with their correct isotopic distribution patterns.

Fig. 7
Fig. 7 Pd K-edge XANES spectra of the immobilized-[Pd 3 Cl 2 ] cluster 3 as reference (orange), during reaction (green, reaction stopped at 5 min) and post-reaction (cyan) (a); Pd K-edge R-space EXAFS data and fits of the immobilized-[Pd 3 Cl 2 ] cluster 3 before reaction, during reaction and post-reaction, solid colored lines denote the data, dotted black lines the fits, with the fitting window indicated in grey (full details given in the ESI †) (b); Pd K-edge XAS spectra (XANES (c) and extended (d)) of the concentrated filtrate during reaction and post-reaction, compared to immobilized-[Pd 3 Cl 2 ] 3, and PdBr 2 and [Pd 2 Br 6 ][ n Bu 4 N] 2 as references compounds.Spectra offset for clarity, the lowest spectrum correctly placed on the y-axis in each case, with XANES spectra all normalized between 0 and 1 and R-space EXAFS data offset by 5 units per spectrum.

Fig. 9
Fig. 9 Pd K-edge XANES spectra of the Pd 3 Cl 2 cluster 1 alone (blue), and after stepwise addition of H 2 O/THF and nBu 4 NOH base (pink), phenylboronic acid (brown), and phenyl bromide (platinate purple) (a); Pd K-edge k-space and R-space EXAFS data and fits of the initial Pd 3 Cl 2 cluster 1, dashed blue lines denote the data, solid blue lines show the fit, with the fitting window shown above and the contributing paths from Pd-Pd, Pd-P and Pd-Cl shown below (b); plot of the contributions determined by least squares fitting the spectra shown in (a) for the stepwise reaction of the initial Pd 3 Cl 2 cluster 1 and the concentrated post reaction filtrate from the immob-Pd 3 Cl 2 experiments as explained in the text(c).Spectra offset for clarity, the lowest XANES spectrum is correctly placed on the y-axis with all XANES spectra normalized between 0 and 1 and EXAFS oscillations centered around y = 0 prior to any offset.

Table 1
Varying content of the palladium catalyst system in a standard SMCC reaction Pd ppm (moles) is three times catalyst ppm (moles) for the Pd 3 cluster; the Pd concentration of 68 ppm (moles) equates to a 0.34 mol% catalyst loading with respect to 4 the [ArBr].b Product 6 conversion ( 19 F NMR; note product conversions mirrored in 1 H NMR spectral data) of crude reaction sample with internal standards, 1,3,5-trimethoxybenzene and 4,4 0 -diuorobenzophenone. c Conversion to 6 obtained at room temperature. a