|Home | About | Journals | Submit | Contact Us | Français|
We report the efficient N-arylation of acyclic secondary amides and related nucleophiles with aryl nonaflates, triflates, and chlorides. This method allows for easy variation of the aromatic component in tertiary aryl amides. A new biaryl phosphine with P-bound 3,5-(bis)trifluoromethylphenyl groups was found to be uniquely effective for this amidation. The critical aspects of the ligand were explored through synthetic, mechanistic, and computational studies. Systematic variation of the ligand revealed the importance of (1) a methoxy group on the aromatic carbon of the “top ring” ortho to the phosphorus and (2) two highly electron-withdrawing P-bound 3,5-(bis)trifluoromethylphenyl groups. Computational studies suggest the electron-deficient nature of the ligand is important in facilitating amide binding to the LPd(II)(Ph)(X) intermediate.
Methods for the cross-coupling of amides with aryl halides and pseudohalides have matured to a point that they may be used to reliably prepare a wide variety of substances.i Advances in this area are due, in large part, to the identification of improved supporting ligands. The classic copper-mediated coupling of an amide with an aryl iodide (the Goldberg reaction) proceeds catalytically and at lower temperatures when diamine ligands are employed.ii Efficient Pd-catalyzed coupling of amides and related nucleophiles (ureas, carbamates, sulfonamides) with aryl bromidesiii and aryl or vinyl sulfonatesiv has largely been accomplished by the use of xantphos or XPhos as the supporting ligand. Most recently, the use of monodentate biarylphosphines bearing a methyl or methoxy group ortho to the phosphorus (L3) has enabled the coupling of amides with heteroaryl and aryl chlorides.v,vi A prominent limitation of all amidation methods is they are mostly restricted to primary amides or lactams—a general method for the intermolecular cross-coupling of acyclic secondary amides with aryl halides and/or pseudohalides has not yet been reported.vii–ix The large size of acyclic secondary amides,3a combined with their relatively low nucleophilicity (compared to amines), has made secondary amides a particularly challenging substrate class for cross-coupling.
Tertiary aryl amides and sulfonamides, the primary products described in this paper, are found in a variety of biologically active molecules (Figure 1).x Access to this class of compounds by arylation of the secondary amides would have advantages over traditional acylation strategies because it would allow for easy variation of the aromatic component, facilitating the generation of analogs for SAR studies. Additionally, secondary amides and carbamates are often intermediates in synthetic sequencesxi and the ability to arylate these directly would improve the efficiency of synthetic routes.
Herein, we describe the rational development of a new biaryl phosphine (JackiePhos, L6) with P-bound 3,5-(bis)trifluoromethylphenyl groups, that facilitates amide binding by creating a more electrophilic Pd(II) intermediate. We have examined aspects of this catalyst system by synthetic, mechanistic, and computational investigations. The synthetic investigations will focus on: (1) substrate scope for the cross-coupling of secondary amides with aryl triflate and nonaflates, (2) cross-coupling of related nucleophiles (ureas, carbamates, sulfonamides), (3) extension of this method to include aryl chlorides, and (4) development of a diarylation process. The mechanistic investigations will focus on: (1) correlation of ligand structure to catalyst stability and (2) trends in aryl halide reactivity and the mechanistic implications. Finally, a computational study will examine each catalytic step for a series of related ligands in order to understand the effect of structural substitutions.
In designing a new ligand for the N-arylation of secondary amides, we took into account existing knowledge of ligand effects and the mechanism of amidation. Xantphos, the standard ligand for the amidation of aryl iodides, bromides, and triflates, was ineffective as a ligand for the reaction in Chart 1. Instead we chose to focus our efforts on ligands derived from the BrettPhos biaryl motifxii (Chart 1), which contains two methoxy groups on the upper ring of the biaryl. This scaffold imparts properties that give catalysts superior performance in a number of C–N bond-forming transformations. Reactions employing BrettPhos13 (L2) and t-Bu2BrettPhos5c (L3) provided unsatisfactory yields of the desired product (27% and 21%, respectively). Based on our knowledge that “transmetallation” was rate-limiting for primary amides5a (and this step would likely be even more difficult for the larger secondary amides) we undertook the development of ligands that would facilitate this step. We hypothesized that a more electrophilic and less sterically-demanding Pd(II) intermediate was necessary to facilitate amide binding. Moreover, a similar strategy was employed by Beletskaya for the N-arylation of ureas in which an electron-deficient xantphos ligand possessing P-bound 3,5-(bis)trifluoromethylphenyl groups was shown to give the most active catalyst.xiii Toward this goal we prepared ligands L4–L6, in which the P-bound alkyl groups in L2 and L3 were replaced with aryl groups. The strongly electron-withdrawing 3,5-(bis)trifluoromethylphenyl substituent in ligand L6 provided the most efficient ligand and the catalyst derived from L6 provided the desired product in 87% yield. Consistent with our supposition, the conversion and product yields increased as the ligand became more electron-deficient (R = 3,5-CF3Ph > 4-CF3Ph > Ph).
Several other aspects of the reaction conditions utilized for the N-arylation of secondary amides deserve comment.xiv Aryl nonaflates (ArONf = ArOSO2(CF2)3CF3)xv were chosen as a cross-coupling partner because they are less susceptible to hydrolysis than aryl triflates, while still providing a similar, highly electrophilic, Pd(II) intermediate.xvi,xvii The use of a non-polar solvent such as toluene was found to be critical in obtaining high yields, whereas reactions in polar solvents (e.g., t-BuOH, DME) were less efficient. Additionally, the inclusion of molecular sieves in the reaction mixture improved the yields by inhibiting the formation of phenols and related side-products.xviii The [(allyl)PdCl]2 precatalyst was chosen due to its ease of use and lack of coordinating ligands (such as dba).xix However, Pd(OAc)2 could also be used if a preactivation step was employed in order to ensure formation of the Pd(0) complex.5b
With optimized reaction conditions in hand, we examined the scope of the method. Secondary amides coupled efficiently with a variety of electron-deficient, electron-neutral, and moderately electron-rich aryl nonaflates—with electron-deficient aryl nonaflates typically giving the highest yields (Table 1, entries 1–3). α-Branched amides reacted smoothly (entries 5 and 6), but branched alkyl substitution on the nitrogen was not tolerated. In addition to nonaflates, aryl triflates were also successfully converted to product (entries 7 and 8). In difunctionalized substrates, it was found (unsurprisingly) that the nonaflate group reacted in preference to a chloride substituent (entries 9 and 10). A relatively small group at the ortho position of the aryl sulfonate such as Cl was tolerated, but larger substituents were not. Acetanilide and other N-aryl amides are traditionally difficult substrates for Pd-catalyzed cross-coupling reactions, but by using a slight modification of our general protocol, employing K3PO4 as the base, acetanilide could be N-arylated in 78% yield (entry 11).xx We note that attempts to realize the same product from the aryl bromide using xantphos as the supporting ligand gave a very low yield of the product.xxi
Given the successful coupling of secondary amides, we next examined the reaction of carbamates, ureas, and sulfonamides. Substituted ureas proved to be difficult substrates and could only be coupled with electron-deficient nonaflates, and then only in moderate yields (Table 2, entries 1 and 2). In contrast, secondary carbamates, which are less sterically demanding, were excellent substrates. The coupling of Boc- and Cbz-protected amines proceeded in 87% and 90% yield, respectively (entries 3 and 6). The use of secondary sulfonamides also provided good yields of coupled products (entries 7–9). These examples are significant because there are only isolated examples of the successful cross-coupling of N-alkyl carbamatesxxii and to the best of our knowledge, the intermolecular coupling of N-alkyl acyclic sulfonamides or N-alkyl, N′,N′-dialkyl ureas has not yet been reported.
In order to expand the utility of this method, we sought to develop conditions that would allow for the coupling of secondary amides with aryl chlorides. Many aryl chlorides are commercially available and the ability to utilize these substrates would improve the generality of this method.xxiii Increasing the reaction temperature and utilizing Cs2CO3 as the base provided good yields of amidation products. The substrate scope of the aryl chloride coupling was briefly examined as shown in Table 3. A variety of aryl chlorides were successfully coupled with secondary amides as well as with a secondary carbamate (entry 4) and a secondary sulfonamide (entry 5). The reaction of more hindered amides proved to be difficult with aryl chlorides; for those substrates in particular, aryl triflates/nonaflates are recommended as the coupling partners.
Given our success in coupling N-aryl amides with aryl nonaflates and chlorides, we examined the diarylation of a primary amide.xxiv By combining a primary amide with 2.2 equivalents of aryl halide or pseudohalide and three equivalents of base under otherwise identical reaction conditions to that described above, good yields of the desired N,N-diaryl amides could be obtained (Table 4). This diarylation method allows for the rapid synthesis of diaryl benzamide derivatives. These compounds exhibit interesting biological activity as nuclear receptor binding agents.xxv An example is 3-hydroxy-N,N-diphenylbenzamide, which alters pS2/TFF1 expression levels. (pS2/TFF1 is important in maintaining gastric mucosal integrity and may act as a tumor suppressor signal in the stomach.xxvi) Through the diarylation of 3-methoxybenzamide (entries 4–6), a series of 3-hydroxy-N,N-diphenylbenzamide analogs were readily available.
Our recent studies on C–N bond formation have shown that ligands containing methoxy groups on the upper ring of the biaryl backbone provide significantly better results than those without the substituent. In order to probe the importance of the methoxy groups, a series of JackiePhos analogs were prepared and tested. Ligands L7–L9 (Table 5), containing variable substitution on the upper biaryl ring, were examined under standard reaction conditions for the coupling of N-butylacetamide and of 4-t-butylphenyl nonaflate. The catalyst derived from ligand L9, containing only one methoxy substituent ortho to the phosphorus led to successful amidation. Yields for amidation reactions using L9 were similar to those obtained with JackiePhos, and monitoring the amidation with in situ IR indicated that the reaction profiles were nearly identical. However, the catalyst derived from ligand L7, lacking both of the methoxy groups, provided only 5% of the desired product and quickly degraded to palladium black. This indicates that the upper methoxy ortho to the phosphorus is of critical importance to catalyst stability. To determine if this was simply a steric effect, ligand L8 containing an ortho methyl group was synthesized. Interestingly, L8 was not a useful ligand for amidation reactions despite NMR experiments showing it undergoes oxidative addition to PhCl. Thus, the methoxy group ortho to the phosphorus of JackiePhos cannot be substituted with a slightly larger, non-coordinating group.
To better understand the relative reactivity of the aryl halide or pseudohalide in these amidation reactions, we compared the reaction profiles of phenyl triflate, iodobenzene, bromobenzene, and chlorobenzene with N-butyl acetamide. The reactions were monitored for product formation by in situ IR spectroscopy (Chart 2). The reaction of N-butyl acetamide with phenyl triflate proceeded to completion in just over 5 hours. In contrast, the reaction of aryl halides did not reach completion in 22 hours. We observed some unexpected trends in the reactions of the aryl halides—the reaction of chlorobenzene was significantly faster than that of bromobenzene, which in turn was much faster than iodobenzene, which produced essentially no product. The general trend of ArOTf/ArONf > ArCl > ArBr > ArI was also observed for the other substituted aryl halides/pseudohalides examined.
To gain a more thorough understanding of the origin of these trends, we carried out several competition experiments. In Table 6 a comparison of the yields for the cross-coupling of N-butyl acetamide with aryl electrophiles under the optimal reaction conditions developed for aryl chlorides is shown. The yields given in this table are consistent with the relative reactivity of this series of substrates. Given that aryl chlorides react faster under these conditions than the corresponding aryl bromides, we performed a competition experiment between 1-n-butyl-4-chlorobenzene and 4-bromotoluene (eq 1). As expected, exclusive cross-coupling with the 4-bromotoluene was observed, consistent with a mechanism in which oxidative addition occurs preferentially with 4-bromotoluene. A similar competition experiment between 1-n-butyl-4-chlorobenzene and 4-iodotoluene provided no cross-coupled product (eq 2), demonstrating that oxidative addition occurred with 4-iodotoluene in preference to 1-n-butyl-4-chlorobenzene. These results strongly disfavor a mechanism in which oxidative addition affects the observed rate, but instead are consistent with the idea that “transmetallation” step is responsible for the observed trend in reactivity. In our previous kinetic studies on amidation, the faster rate of “transmetallation” for aryl triflates was attributed to a more labile LPd(Ph)OTf complex compared to LPd(Ph)halide complexes.5a Analogously, the rate of amide/amidate attack on LPd(Ph)Cl may be faster than attack on LPd(Ph)Br due to the more electrophilic nature of palladium in LPd(Ph)Cl. This is supported by the trend in rates of associative ligand exchange of [Pt(dien)X]n+ complexes, following the order Cl > Br > I.xxvii
Future advances in the design of biarylmonophosphine ligands would benefit from a firm understanding of the effects bestowed by substituents appended to the biaryl backbone. Although we were pleased with the superior performance of JackiePhos in comparison to other biarylmonophosphine ligands, we sought further mechanistic support for our original hypothesis that its electron withdrawing nature allows for faster transmetallation; this prompted us to investigate the role of the structural features present in this ligand. For these studies, we chose to focus on two points: (1) determination of the effect of substitution of different P-bound R groups (i.e., 3,5-CF3Ph, t-Bu, Cy, Ph) and (2) the development of an understanding of the role of the methoxy group bound to the upper ring ortho to the phosphorus.
Since we were unable to measure the relative rates of each mechanistic step due to difficulties in isolating catalytic intermediates, we turned instead to a computational approach using DFT methods. Computational investigations of Pd-catalyzed cross-coupling reactions have become increasingly routine, but an examination of the literature reveals that the majority of these analyses are limited to catalysts containing small phosphines (e.g., PH3, PMe3, H2PCH2CH2PH2) that serve as models for larger ligands (e.g., PPh3, P(t-Bu)3, Ph2PCH2CH2PPh2).xxviii While the conclusions from these studies have provided much insight into the operative mechanism of these processes, many difficult cross-coupling reactions do not proceed with these simple ligands. Therefore, caution should be exercised when applying approximations to ligands when examining such reactions.xxix
With the rapid increases in computational power and concurrent decrease in cost, we favor a more rigorous approach, which entails modeling the complete catalyst with all-atom DFT. We believe that studying the subtle structural features of the ligand is critical for a thorough mechanistic understanding. Furthermore, in the following discussion we compare several related biaryl monophosphines that behave markedly different in promoting catalysis. The direct comparison of ancillary phosphine ligands by computational methods represents an infrequently used but powerful approach to understanding intricate details of critical mechanistic processes.28a,xxx
Our strategy to unraveling the structure/activity relationship of these ligands was to systematically change one portion of the ligand and calculate the effect on each step of the catalytic cycle. Scheme 1 shows the general outline of the possible steps in the catalytic cycle we would examine—oxidative addition (eq 1), associative ligand exchange (eq 2), dissociative ligand exchange (eq 3), formation of the κ2-bound amidate (eq 4), and reductive elimination (eq 5). The ligands we opted to focus on were L1 (xantphos), L6 (JackiePhos), L2 (BrettPhos), L3 (t-Bu2BrettPhos), L4 (Ph2BrettPhos), and L10 (a simplified version of L8), which are shown in Scheme 1. Xantphos was included in this comparison since it is a very effective ligand for many amidation reactions3a–h, 4a,b and its role in Pd-catalyzed cross-couplings has never been studied computationally.xxxi It also serves as a benchmark for comparison to our biarylmonophosphine ligands, and to validate our computational results by further correlation with experiment. Our justification for modeling L10 instead of L8 rests on previous studies that demonstrate only the methyl group ortho to the phosphorus influences the properties of these ligands in amidation reactions.5a
All calculations were performed using the Gaussian ’03 suite of programs,xxxii primarily with Dell PowerEdge 1950 servers containing two quad-core Intel Xeon processors or Sun servers containing dual-core Opteron processors, running the sun grid engine (SGE) queuing program. We also made use of supercomputing resources for potential energy scans. Ground-state geometry optimizations were performed with the Berny algorithm using all-atom DFT (B3LYP/6-31(d)) and the LANL2DZ basis set with the Hay–Wadtxxxiii effective core potential (ECP) for all Pd and Br atoms. Frequency calculations were performed on all optimized structures to confirm that the minima had no negative frequencies and transition states had a single imaginary frequency. The Gibbs free energies were calculated at 298.15 K and 1 atm. Single point calculations were then performed with the 6-311+g(2d,p) basis set to obtain higher accuracy electronic energies. Solvation effects were calculated at the 6-311+g(2d,p) level with the CPCM methodxxxiv in conjunction with the UAKS cavity and employing a dielectric constant of ε = 2.374 for toluene. Basis set superposition error (BSSE) was corrected for by use of the counterpoise method of Boys and Bernardi.xxxv Although BSSE is a general phenomenon when using incomplete basis sets, correcting for it is critical when constructing potential energy diagrams.xxxvi
As stated earlier, it is unlikely that oxidative addition is the rate-limiting step in most Pd-catalyzed amide arylation reactions based on previous work performed in our group.5a Nevertheless, the barriers for this step were calculated (Table 7), as the rate-limiting step could change for some ligands. We began by finding the barrier of oxidative addition to bromobenzene (1-TS, entry 1) and chlorobenzene (2-TS, entry 2) with xantphos·Pd (1). Interestingly, the geometry of these transitions states were pseudotetrahedral about the Pd center as opposed to the more common square planar conformation, probably a result of the large bite angle of xantphos.xxxvii Alternative transition states with cis or O-bound geometries could not be located. A moderate barrier of 26.8 kcal/mol was calculated for 1-TS, but a high barrier of 36.6 kcal/mol was calculated for 2-TS. This is consistent with the experimental observation that the use of xantphos does not allow for the Pd-catalyzed cross-coupling of unactivated aryl chlorides.5a,xxxviii
We found four energy minimized oxidative addition complexes for xantphos·Pd(Ph)(Br), shown in Table 8. The tetrahedral transition state 1-TS initially leads to the highest energy tetrahedral isomer 2a (confirmed by an IRC calculation), which then can isomerize to the lower energy transxxxix (2b), O-bound (2c) or cis (2d) complexes. The bond angles and bond lengths about the Pd center are listed and illustrate the large range of conformations the xantphos catalyst can adopt, underscoring the dynamic nature of this ligand. The O5–Pd–P3 bond angle of 2c is only 76.0° while the P1–Pd–P2 bond angle ranges from 104.9° in 2d to 113.3° in 2a to 146.2° in 2b.
Although aryl chlorides, triflates, and nonaflates are all capable coupling partners when using JackiePhos, we chose to focus on modeling oxidative addition with chlorobenzene as it would simplify the calculations and corresponds to the most difficult case in the series that we studied. In Table 9, the calculated barriers for oxidative addition to chlorobenzene with Pd(0) complexes of JackiePhos (entry 1), BrettPhos (entry 2), t-Bu2BrettPhos (entry 3), Ph2BrettPhos (entry 4), and L10 (entry 5) are listed. They are quite low—even for electron-deficient ligands—ranging between 10.8 and 20.4 kcal/mol, and are in agreement with our previous computational study29b on the reactivity of SPhos·Pd and XPhos·Pd. The low barrier calculated for JackiePhos (20.1 kcal/mol, entry 1) supports our assertion that this step is not rate-limiting. It is interesting that even though JackiePhos is significantly more electron-deficient than xantphos, the barrier to oxidative addition of chlorobenzene with the former is over 16 kcal/mol lower in energy. This is consistent with previous findings from our laboratory concerning the relative importance of steric and electronic factors on the rates of cross-coupling reactions.xl
The step following oxidative addition is “transmetallation”, in which ligand exchange of an amide for a halide probably occurs either by an associative or dissociative mechanism. Given the pKa difference between amides (15.1 for N–H) and carbonate (10.3), we believe that deprotonation occurs following amide binding to the Lewis acidic Pd(II) center. If an associative mechanism is operative, then the amide must first bind to Pd, giving rise to a five-coordinate intermediate in the case of xantphos and a four-coordinate intermediate in the case of monophosphines, followed by deprotonation and halide dissociation. An examination of the literature reveals very few mechanistic details concerning ligand substitution with amides either experimentally or theoretically.
The amide included in our calculations was N-methyl acetamide, which was chosen to be representative of secondary, acyclic, alkyl amides. The calculated geometry of the four-coordinate trans-xantphos·Pd(Ph)(Br) complex 2b, and five-coordinate square pyramidal complex 8 with an O-bound amide are shown in Scheme 2. This binding is endothermic by 28.7 kcal/mol of energy and results in a fairly weak Pd–Br interaction in the axial position with a bond length of 3.22 Å. A similar structure with a triflate counterion could not be located.
The binding energies of N-methyl acetamide to each biaryl monophosphine·Pd(Ph)(Cl) complex were next determined (Table 10). We first calculated the energy of amide binding to the JackiePhos·Pd(Ph)(Cl) complex in cases where the Pd center is pointed away from the lower biaryl ring (complexes 9b–d). The most favorable interaction, being endothermic by 18.2 kcal/mol,xli occurred when the amide was bound to Pd through its oxygen atom and oriented cis to the phosphorous (9b). The energy of amide binding with the Pd but centered over the bottom ring was also calculated and was significantly higher in energy in both possible configurations (9f, 26.4 kcal/mol and 9g, 25.1 kcal/mol). The energies of the other L·Pd(Ph)(Cl)(amide) adducts in configuration b were next calculated. We also verified that the lowest energy configuration for JackiePhos held for other ligands and they did with the exception of isomer f. For BrettPhos and t-Bu2BrettPhos, this isomer was not a stationary state but we found that isomer g oriented with the amide trans to the phosphorus did represent a stable minimum. Larger energies for amide binding were calculated for BrettPhos-(23.0 kcal/mol, 10b) and t-Bu2BrettPhos-(23.5 kcal/mol, 11b) ligated complexes. In contrast, with Ph2BrettPhos, a significantly lower energy for this step (19.8 kcal/mol, 12b) was calculated. As in the oxidative addition step, L10 is calculated to behave similarly to JackiePhos; its oxidative addition complex requires only 17.5 kcal/mol for amide binding.
Further information on the mechanism of this step would be gained from examining transition states for amide attack and amide deprotonation. However, our efforts to locate such structures were unsuccessful and at this point we can only qualitatively say that they will be higher in energy than the corresponding ground states.
The accessibility of isomer b in Table 10 relies on the free rotation about the Caryl–P bond, which has been shown in previous computational studies to be quite difficult for Me4t-Bu2XPhos·Pd(Ph)(amidate).5a The JackiePhos·Pd(Ph)(amidate) complex is likely to have a lower barrier of rotation for two reasons: first, the methoxy group ortho to phosphorous is smaller than the methyl group in Me4t-Bu2XPhos, and second, the P-bound arenes of JackiePhos are much smaller than the bulky t-Bu groups of Me4t-Bu2XPhos. In order to support this assertion, we performed a potential energy scan about the Caryl–P bond in which a restricted structure optimization was executed at every 10°. We chose to calculate this with the JackiePhos·Pd(Ph)(amidate) complex because it, too, would need to freely rotate once formed and, being larger than the chloride complex, would be the more difficult case. The maxima of the resulting energy curve, shown in Figure 2, corresponds to a barrier of rotation of 19.8 kcal/mol—an amount of energy that is readily overcome even at room temperature.
We next performed potential energy scans about the C2–P3 bond with analogous complexes containing the other ligands and very similar curves were obtained (not shown). However, the maximum for the BrettPhos complex lies at a dihedral angle of 90° while all of the other ligands gave maxima at 80°. The barrier heights for these complexes are listed in Table 11: JackiePhos (entry 1), BrettPhos (entry 2), and Ph2BrettPhos (entry 4) give barriers of very similar energies (19.2, 20.5, and 19.1 kcal/mol, respectively). Of note is that switching the methoxy group of the upper ring to a methyl group in JackiePhos increases the barrier significantly to 23.3 kcal/mol (entry 5), presumably because of its larger size. We also found that the barrier to rotation for the most sterically congested ligand, t-Bu2BrettPhos (entry 3), is only 27.7 kcal/mol (compared to 33 kcal/mol for Me4t-Bu2XPhos5a). It follows from the Eyring equation that this barrier corresponds to a half life of only 9.16 min at 110 °C assuming a first order rate law. Based on these data, we conclude that under the reaction conditions (110–130 °C) all the ligands including t-Bu2BrettPhos can freely rotate with the Pd moiety being either over or away from the lower biaryl ring.
The loss of a halide/pseudohalide to form cationic Pd(II) intermediates is quite common in Heck-type processes,xlii but is not necessary for most cross-coupling reactions in which an organometallic component participates as the nucleophile. However, this pathway should be considered for reactions involving weakly-nucleophilic, non-metallic partners, such as hindered amides. Furthermore, dissociative ligand exchange through a cationic Pd(II) intermediate has been suggested by van Leeuwen to occur for xantphos-Pd(II) complexes undergoing cis/trans isomerization39a and in the amination of aryl triflatesxliii.
For the following calculations, we deemed it important to model both the bromide and triflate, as the latter are so much more effective as leaving groups.xliv It was critical for the inclusion of solvation effects for this step as dissociation was highly endothermic in the gas phase. As depicted in Table 12, the energy difference between cationic complex 14 and both the trans-xantphos·Pd(Ph)(Br) (2b) and trans-xantphos·Pd(Ph)(OTf) (15) complexes were determined. The loss of the bromide is endothermic by 17.5 kcal/mol but loss of triflate is endothermic by only 0.8 kcal/mol. It should be pointed out that correction for BSSE was critical for Br dissociation as it was originally in error by 16.5 kcal/mol, while triflate dissociation was only in error by 3.2 kcal/mol (a more typical value). The unusually large error in the first case is likely a result of the use of the approximations implicit in the LANL2DZ basis set for an entire fragment.xlv
The ionization energy requirement for the bromide complex could be low enough to outcompete the associative pathway. For aryl triflates, the nearly equal energies of 14 and 15, our inability to locate a 5-coordinate amide-bound Pd-triflate complex as well as structural characterization by van Leeuwen39a of a cationic xantphos·Pd(Ar)(OTf) complex are good evidence that ligand exchange occurs by a dissociative pathway. However, based on the experimental results that more hindered amides (e.g., those that are the focus of this paper) are not readily arylated with a xantphos/Pd catalyst and reductive elimination should not be difficult (vide infra), we must conclude that there is also a significant energy cost to approach of the amide and/or isomerization from O- to N-bound isomers.
With a handle on the difference in energy for ionization in xantphos·Pd(Ph)(OTf/Br) complexes, we were prompted to perform similar calculations on the JackiePhos·Pd(Ph)(OTf/Cl) complexes (Table 13). The empty coordination site in the cationic product may be cis or trans to the phosphorous, but structures in which the empty site is trans to the phosphorous were much lower in energy and are the only ones considered here. Additionally, the Pd may be pointed away from the bottom ring, in which it has a dative bond with the oxygen, or towards the lower ring, in which it interacts with the π-electrons of the arene. Surprisingly, C-bound isomer 18 is significantly lower in energy than O-bound isomer 17. Unlike the xantphos system, however, ionization of both the chloride (42.9 kcal/mol to reach 18) and triflate (34.9 kcal/mol to reach 18) were prohibitively high in energy. The resistance to ionization of the JackiePhos system is probably a consequence of the Pd being both coordinatively unsaturated and electron poor.
Previous work has demonstrated that Pd-catalyzed amidation reactions are complicated by the formation of κ2-amidate complexes in which the amide is simultaneously bound to the palladium center at both the nitrogen and oxygen atoms.39b In these studies, it was shown that κ2-amidate complexes do not undergo reductive elimination as readily as κ1-amidate complexes. While some monophosphine ligands may not be effective ligands for amidation as they allow κ2-amidate formation, we have previously shown that bulky monodentate biaryl phosphines can form effective catalysts for amidation.4c,5 In light of the experimental evidence presented,39b suggesting that catalyst inhibition and inactivation results from κ2-binding, we calculated the relative energies of κ1- and κ2-bound isomers of both xantphos·Pd(Ph)(amidate) and JackiePhos·Pd(Ph)(amidate) complexes. Quite unexpectedly, the κ2-bound xantphos structure (20) is calculated to be lower in energy than the cis-κ1-bound structure (19) by 9.3 kcal/mol (Scheme 3). This energy is seemingly in disagreement with the previously reported IR and NMR data obtained for xantphos·(Ph)(N-phenylacetamide) complexes that support κ1-bound structures as being more stable.39b However, the amide-bound complexes in this study were prepared by the addition of potassium amides to xantphos·Pd(Ph)(Br) at room temperature, and the thermodynamically more stable isomer may have not been kinetically accessible. Similarly, JackiePhos·Pd(Ph)(amidate) favors κ2-bound structures, pointed both towards (22, −2.3 kcal/mol) and away (23, −3.7 kcal/mol) from the lower biaryl ring (Scheme 4). In contrast to the xantphos system, the barrier to reach these structures is probably not as high since there is an open coordination site on Pd. Based on these results, we believe κ2-binding does not necessarily lead to decomposition or inhibit catalysis (i.e., it can form reversibly), but rather, it may lie on the path to catalyst decomposition for unhindered monodentate ligands.
Reductive elimination is often the most difficult step for the arylation of weaker nucleophiles such as amides.xlvi We first calculated the barrier for C–N bond formation from the xantphos·Pd(Ph)(amidate) complex and found it was 20.4 kcal/mol (8-TS, Scheme 5). In contrast to the tetrahedral geometry in the transition state for oxidative addition, this geometry is square planar, proceeding from cis-complex 19. We believe that a tetrahedral transition state similar in geometry to those found for oxidative addition (Table 7) is no longer accessible due to the larger size of the amide compared to bromide.
Next, we compared the barrier to C–N bond formation for each of the biaryl monophosphine ligands (Table 14).xlvii The ligands that allowed for the lowest barriers of reductive elimination (16.6 kcal/mol for both) were t-Bu2BrettPhos (11-TS, entry 2) and L10 (13-TS, entry 5), while the barrier for this step was slightly higher with JackiePhos (17.7 kcal/mol, entry 1). Both BrettPhos and Ph2BrettPhos were predicted to be less effective at promoting this step, with barriers of 19.7 (entry 2) and 20.0 kcal/mol (entry 4), respectively.
Five-membered cyclic transition states were also found in which an O-bound amidate pivots back towards the reactive Ph group to form a C–N bond (Figure 3). In both the case of xantphos (14-TS, 23.0 kcal/mol) and JackiePhos (15-TS, 22.0 kcal/mol), this transition state was higher in energy, leading us to believe that it constitutes a minor pathway for this reaction. However, it may be more important in the arylation of other less hindered amides or if O to N isomerization proves difficult.
Without an analysis of a continuous potential energy diagram for the catalytic cycle, the relative contributions from each mechanistic step are not entirely clear. Accordingly, the relative energies of each step through the amide binding event were connected in Figures 4 and and55 so that such an analysis could be performed. We did not construct a potential energy diagram for the entire catalytic cycle because of the uncertainties in the barriers to transmetallation as well as the energy change for salt precipitation and dissolution, as this is a heterogeneous reaction.
First, in Figure 4, the potential energy diagram for the xantphos-Pd catalyst is depicted. Although aryl triflates likely react through a dissociative pathway, we will examine the energy pathway for the aryl bromide case because it will be more comparable to the mechanism involving the biarylmonophosphine-derived catalysts. In the catalytic cycle with bromobenzene, oxidative addition is slightly endothermic by +3.3 kcal/mol based on the difference between 1 with bromobenzene and the most stable trans-xantphos-Pd(II) isomer (2b). Assuming transmetallation operates by associative ligand exchange, amide binding (8) brings the total energy needed to 32.0 kcal/mol.
The information contained in the potential energy diagram of the biarylmonophosphine-Pd catalysts in Figure 5 is not as readily apparent as in Figure 4. First, the thermodynamics of oxidative addition to reach C is drastically different for each catalyst, all being exothermic but with an energy change ranging between −0.2 and −13.8 kcal/mol. As a consequence, the more exothermic this step becomes, the more energy that will be required to scale the subsequent reaction barriers. In this scenario, the energy requirement for catalysis will not be the energy change from starting materials to the highest barrier but the difference between the lowest point (C) and the highest barrier. In the cases of t-Bu2BrettPhos and L10, the energy for bond rotation (D) is the highest peak in Figure 5, but we believe that with consideration of the barriers for transmetallation and reductive elimination it will not be rate-limiting.
Although we lack the information to give predicted energy requirements and overall rates for each catalyst, we find that the energy differences from the lowest energy point and the amide-bound complex correlate well with ligand performance in most cases. These energies are listed in Table 15 and the smallest energy requirement is associated with JackiePhos (18.2 kcal/mol) and L10 (17.5 kcal/mol). We feel that the trends shown here will be consistent but amplified when considering the barriers for amide attack and deprotonation. Since L10 is calculated to perform equally well or better than JackiePhos in each mechanistic step but is completely ineffective in promoting the reaction, we believe the catalyst formed from it is not stable under the reaction conditions, degrading by an as yet undetermined pathway.
In summary, we have reported an efficient method for the N-arylation of acyclic secondary amides with aryl nonaflates, triflates, and chlorides. The catalyst system is also useful for the N-arylation secondary carbamates, ureas, and sulfonamides and the diarylation of primary amides. A ligand for this transformation, JackiePhos (L6), was designed based on the knowledge that “transmetallation” was rate-limiting for the arylation of primary amides and the hypothesis that a more electrophilic and less sterically demanding Pd(II) intermediate would facilitate the “transmetallation” of secondary amides. Our synthetic, mechanistic, and computational studies showed the success of JackiePhos is due to: (1) the methoxy group ortho to the phosphorus and (2) the strongly electron-withdrawing P-bound 3,5-(bis)trifluoromethylphenyl groups.
Theoretical studies have provided insight into the mechanism of “transmetallation” and show that the lowest energy conformation for amide binding occurs while Pd is pointed away from the lower biaryl ring. A potential energy scan about the Caryl–P bond indicates that this conformation, in which Pd interacts with the methoxy group ortho to the phosphorus, is quite accessible. Our synthetic studies show that the ortho methoxy group is critical for catalyst stability and cannot removed or substituted with a methyl group. Additionally, computational studies predict the most favorable energies of amide binding for ligands containing the highly electron-withdrawing P-bound 3,5-(bis)trifluoromethylphenyl groups. Overall, it seems the “transmetallation” of secondary amides is facilitated by the ability of the ligand to access a highly electrophilic Pd(II) intermediate where the Pd is oriented away from the bottom ring of the biaryl backbone and stabilized by the ortho methoxy group. Additionally, the barrier to reductive elimination was lower for the electron-deficient JackiePhos compared to other sterically similar ligands (Ph2BrettPhos, L4). However, there remain many questions on the precise mechanism and energetic requirements of the transmetallation step, preventing the construction of a full potential energy diagram. Overall, these studies further expand our understanding of structural features of biarylmonophosphine ligands and how each of these contributes to the catalysts derived from them.
We thank the National Institutes of Health (NIH) for funding this project (Grant GM-58160). We thank Merck, Boehringer Ingelheim, BASF (Pd compounds), Chemetall (Cs2CO3) and Nippon Chemical for additional support and the National Science Foundation through the National Center for Supercomputing Applications (NCSA), which provided supercomputing resources. We are grateful to B. P. Fors for carrying out additional experiments to address reviewer comments.
Supporting Information Available: Experimental procedures and spectral data for all compounds, computational data, and kinetic plots. This material is available free of charge via the internet at http://pubs.acs.org.