Reversible Oxidative Addition of Non-Activated C-H Bonds to Structurally Constrained Phosphenium Ions

Abstract

A series of structurally constrained phosphenium ions based on pyridinylmethylamidophenolate scaffolds are shown to undergo P(III)/P(V) oxidative addition with C-H bonds of alkynes, alkenes and arenes. Non-activated substrates such as benzene, toluene, and deactivated chlorobenzene are phosphorylated in quantitative yields. Computational and spectroscopic studies suggest a low-barrier isomerization from a bent to a T-shaped isomer that initiates a phosphorous-ligand cooperative pathway and subsequent ring-chain tautomerism. Remarkably, the C-H-bond activations occur reversibly, allowing for reductive elimination back to P(III) at elevated temperatures or the exchange with other substrates.

Graphical Abstract

Introduction

The selective cleavage and functionalization of non-activated carbon-hydrogen (C-H) bonds represents a pivotal transformation in organic synthesis, bypassing the requirement for prefunctionalization and improving the atom-economy and sustainability of a given synthesis.¹ New methodologies developed towards this goal have relied on precious transition metals, whereby oxidative addition (OA) constitutes a prominent mechanism by which the C-H activation operates. In main group chemistry, C-H bond cleavage occurs more commonly by deprotonation or hydride abstraction, while examples of OA to main group elements have been mostly limited to more polar bonds or activated hydrocarbons.² Moreover, the reported OA are facilitated by a substantial thermodynamic driving force to the more stable higher oxidation state, generally rendering these reactions irreversible.

The interest in non-trigonal phosphorus(III) compounds as a platform for main group redox catalysis has come into focus due to the possibility of cycling between the P(III)/P(V) redox couple without substantial bias for either OA or reductive elimination (RE).³ Accordingly, reversible OA of O-H and N-H bonds to a main group element has been achieved for a series of structurally constrained P(III) compounds (fig. 1A & B).⁴ Enhanced biphilic reactivity arises from altering the local symmetry at phosphorus enforced by tethered ligand systems, which in turn lowers the HOMO-LUMO gap. While previously restricted to polar E-H bonds (E = O, S, N),^{4a, c, 5} a single-center OA of dihydrogen to phosphorus was recently reported employing a rigid, carborate-based ligand framework.⁶ Fewer examples for inserting P(III) compounds into C-H bonds exist, but were demonstrated with electrophilic, two-coordinate phosphenium ions.⁷ However, these reactions required activated C-H bonds of metallocenes, xanthene and cycloheptatriene. Our group has shown the possibility of activating C-H bonds for different heteroarenes by P(V) Lewis superacids in cooperativity with catecholate- and amidophenolate ligands.⁸

Figure 1:

A Previous examples of structurally constrained P(III) compounds and B their E-H bond activation by oxidative addition. C The phosphenium ions capable of reversible C-H bonds activation presented in this work.

Herein, we report a class of structurally constrained phosphenium ions based on easily preparable and modifiable pyridinylmethylamidophenolate ligands (fig. 1C). Leveraging an increase in Lewis acidity at phosphorus combined with a dative P-N bond enables reversible OA of C-H bonds, including those of unactivated arenes. It represents the first reversible C-H oxidative addition at a main group element and a critical step toward p-block elements with transition metal like behavior.⁹

Results and Discussion

The pyridinylmethylaminophenol ligands LZ (Z = 1–4) were prepared on a multi-gram scale by reductive amination of the corresponding commercially available 2-pyridinecarboxaldehydes and 2-aminophenols (fig. 2A). A convenient workup by washing with diethyl ether led to analytically pure compounds. Subjecting these aminophenols to phosphorus trichloride in the presence of triethylamine furnished the phosphorus chlorides Z-Cl in good yields (fig. 2A).

Figure 2:

A Synthesis of phosphenium salts [Z][WCA] and ³¹P NMR shifts. (R^F = OC(CF₃)₃, WCA = weakly coordinating anion, X = 1–4). B Molecular structures of 1-Cl and 4⁺. Thermal ellipsoid at 30 % probability. The counteranion [Al(OR^F)₄]⁻ and all carbon-based hydrogen atoms were omitted for clarity. Selected bond distances [Å] and angles [deg] of 1-Cl: d(P1-N1) = 2.373, d(P1-N2) = 1.7032(16), d(P1-O1) = 1.6732(15), ∠O1-P1-N1 = 166.67, ∠O1-P1-N2 = 90.39(7), ∠N2-P1-N1 = 76.76, and 4⁺: d(P1-N1) = 1.922(2), d(P1-N2) = 1.702(3), d(P1-O1) = 1.620(2), ∠O1-P1-N1 = 108.05(9), ∠O1-P1-N2 = 94.38(11), ∠N2-P1-N1 = 85.86(10).

The ³¹P NMR chemical shifts for these compounds range from 133.7 ppm for 3-Cl, to 162.1 ppm for 2-Cl. The substantial differences are attributed to varying degrees of pyridine coordination to phosphorus in solution, thereby changing the shielding at phosphorus. Single crystals suitable for X-ray diffraction of 1-Cl to 4-Cl deposited from concentrated solutions in toluene after cooling to −40 °C. The solid-state structures show the pyridine moieties oriented towards the phosphorus atom and in-plane with the amidophenolate moiety (fig. 2B for representative structure of 1-Cl). The P1-N1 bond distances mostly follow the trend set by the ³¹P NMR shifts, with longer bonds corresponding to more downfield shifted signals. Chloride abstraction with a suitable salt of a weakly coordinating anion (WCA), such as NaB(C₆F₅)₄, quantitatively yielded the desired phosphenium salts [Z][WCA]. Using LiAl(OR^F)₄ (R^F = C(CF₃)₃), analytically pure salts of 2⁺ and 4⁺ could be isolated with the aluminate counterion. Slow cooling of a concentrated solution of [4][Al(OR^F)₄] in dichloromethane afforded single crystals suitable for X-ray diffraction. The phosphenium ion (fig. 2B) adopts a C₁ symmetric structure in the solid state, with significant folding along the P1-N2 axis, resulting in an ∠O1-P1-N1 angle of 108.05(9)°. The other two angles spanned by the heteroatoms connected to phosphorus are also appreciably different, reflecting the asymmetry of the cation (∠O1-P1-N2 = 94.38(11)° and ∠N2-P1-N1 = 85.86(10)°). The P1-N2 and P1-O1 bond lengths are on the shorter side but comparable to related geometrically distorted P(III) complexes (d(P1-N2) = 1.702(3) Å; d(P1-O1) = 1.620(2) Å). By contrast, the dative nature of the phosphorus-pyridine bond gives rise to an elongated P1-N1 bond at 1.922(2) Å.^{4a, 5c, 10} The value surpasses the phosphorus-pyridine bond lengths of previously reported constrained phosphenium ions III and IV (fig. 1A, 1.81 and 1.752(3) Å).^{4c, 5c}

NMR spectra of 1⁺-4⁺ show varying degrees of broadening of the ³¹P NMR signals. Contrasting the inherent diastereotopic nature of the methylene protons, they appeared magnetically equivalent in the ¹H NMR for all cations, displaying a doublet from coupling with the ³¹P nucleus. These observations indicate fast dynamic conformational processes in solution. Decoalescence of the diastereotopic protons by low-temperature NMR was not observed for 1⁺ even at −80 °C, suggesting a relatively low isomerization barrier. Thus, the cations and the isomerization reaction of 1⁺ were studied by DFT computations. The lowest energy conformers of cations 1⁺-4⁺ adopt bent structures in accordance with the crystal structure of 4⁺. However, a T-shaped planar isomer, e.g., IM-I for 1⁺, is a metastable intermediate only 4.1 kcal mol⁻¹ higher in energy (fig. 3a), and even lower for the other derivatives (table S9 in the ESI).

Figure 3:

A Reaction coordinate profile for the inversion of 1⁺ at the DSD-BLYP(D3BJ)/def2-QZVPP-SMD(DCM)//r²-SCAN-3c level of theory.¹³ B Frontier orbitals of IM-I, computed at the r²-SCAN-3c level of theory.

The low P edge-inversion barrier of 6.0 kcal mol⁻¹ connecting the two enantiomers of 1⁺ via IM-I agrees with the experimentally observed rapid isomerization (fig. 3A). Pyridine dissociation/reassociation was also considered, but no minimum structure with a disconnected pyridine could be found on the PES, and the gradients along the dissociation pathway appeared much steeper than along the edge-inversion coordinate. By way of comparison, for the structurally related NNNP pincer compound II a higher P edge-inversion barrier of 10.7(5) kcal mol⁻¹ was determined by VT-NMR decoalescence.^4a We reasoned that the thermally accessible, structurally constrained IM-I could dominate the reaction properties of the cations and studied its electronic structure. Upon planarization of the bent global minimum towards IM-I, the LUMO energy gets significantly lowered (fig S21, ESI). While residing at the ligands for the bent form, the frontier orbitals become located on phosphorus in IM-I and are separated by a very small HOMO-LUMO gap (2.5 eV, fig. 3B). The structures of 1-Cl – 4-Cl can be considered as formal chloride adducts of the Lewis acidic cations. It supports the perspective that the energy penalty for planarization is easily compensated upon Lewis base coordination.

Overall, these properties suggested high Lewis acidity of the cations, which were thus evaluated experimentally. The effective Lewis acidity for cations 1⁺-4⁺ was assessed using the Gutmann-Beckett method.¹¹ The resulting shift of the ³¹P NMR signals of bound triethylphosphine oxide (OPEt₃) relative to free OPEt₃ ranged between 36.2 and 40.9 ppm in the order 4⁺>2⁺>3⁺>1⁺. These values indicate Lewis acidities that are remarkably high for a P(III) compound, and even comparable to Stephan’s highly Lewis acidic P(V) cation FP(C₆F₅)₃⁺ (40.4 ppm) or that of B(C₆F₅)₃ (30.6 ppm).¹²

Intrigued by the electronic structure and the facile thermal accessibility of a putatively reactive T-shaped isomer, the behavior of 1⁺ in the OA of E-H bonds was investigated. Thus, 1⁺ was reacted with diphenylamine, an amine of low basicity that had not been shown to undergo OA to phosphorus in previous studies. Mixing 1⁺ and diphenylamine in CD₂Cl₂, a new species with a doublet at −45.2 ppm in the ³¹P NMR and a large ¹J_PH coupling constant of 858.3 Hz formed immediate- and quantitatively, consistent with the product [1]•[H][NPh₂]⁺ of N-H oxidative addition (fig. 4). Upon treatment with 1-methylindole, multinuclear NMR data confirmed OA within minutes, this time of the C-H bond at 3-position of the indole to 1⁺ (δ −44.3 ppm, J=763.3 Hz). The sp-hybridized C-H bond of phenylacetylene was also selectively cleaved within minutes at room temperature (δ −65.6 ppm, J=813.4 Hz). Small amounts of a minor product (9 %) were observed in this case, tentatively assigned as the product of cooperative addition of the alkyne across the phosphorus pyridine bond as a competing pathway. While all these addition reactions to 1⁺ occurred quickly and quantitatively at room temperature, no reaction occurred with more electron-poor arenes such as thiophene, where even heating to 80 °C did not yield a reaction.

Figure 4:

Oxidative addition of E-H bonds to 1⁺, with ³¹P NMR data, all reactions are quantitative. [B] = [B(C₆F₅)₄].

To understand these limitations, the mechanism of the oxidative addition of 1-methylindole was studied using DFT calculations at the DSD-BLYP(D3BJ)/def2-QZVPP-SMD(DCM)//r²-SCAN-3c level of theory.¹³ For the OA of amines and alcohols, previous studies of the neutral NNNP and NNSP pincer compounds ruled out the single-site, concerted additions with computed barriers exceeding 50 kcal mol⁻¹, but disclosed a ligand-assisted deprotonation pathway.^{4a, d} Accordingly, a prohibitive barrier of 55 kcal mol⁻¹ was calculated for the concerted addition of 1-methylindole to 1⁺, and a ligand-cooperative reaction pathway was investigated. Scanning the substrate approach to the reactive form IM-I, the pyridine arm was found to dissociate until deprotonation of the now acidified proton via a seven-membered transition state becomes viable (fig. 5A). A subsequent isomerization of the pyridinium intermediate IM-II and IM-III is followed by σ³-P/σ⁵-P tautomerization through proton transfer from nitrogen to phosphorus via TS-III, completing the overall process. The sequence shows similarities with the electrophilic concerted metalation deprotonation or the ambiphilic metal-ligand assistance common to some transition metal-mediated C-H activations, but the overall transformation can be classified as oxidative addition.¹⁴ While TS-II and TS-III are located almost equally high in energy for 1-methylpyrrole as a substrate, TS-II becomes increasingly higher relative to TS-III with decreasing nucleophilicity of the arene (ΔG^‡_sol(TS-II [TS-III]) = 1-methyl pyrrole: 22.5 [22.4] kcal mol⁻¹, thiophene: 30.2 [27.4] kcal mol⁻¹, benzene = 36.2 [28.7] kcal mol⁻¹, fig. 5A/C). The results align with the lack of reaction of 1⁺ with thiophene and the fact that an intermediate IM-II was never observed, corroborating relatively fast conversion of IM-II to the final product.

Figure 5:

A Computed reaction coordinate profile for the oxidative addition of 1-methylpyrrole to 1⁺ at the DSD-BLYP(D3BJ)/def2-QZVPP-SMD(DCM)//r²-SCAN-3c level of theory.¹³ B Solvent free energies of selected transition states and oxidative addition of 1-methylpyrrole with varying substituents X,Y. C Solvent free energies of selected transition states and oxidative addition of thiophene (benzene) with varying substituents X,Y.

We then systematically studied the effects of ligand backbone modifications on the transition state energies and reaction thermodynamics by DFT (fig. 5B/C). Placing both an electron-withdrawing bromine or an electron-donating methoxy group into the pyridine ortho-position (−X) lowered the transition state TS-II for 1-methyl pyrrole deprotonation by 4.9 kcal mol⁻¹ and 3.4 kcal mol⁻¹, respectively (fig. 5B). Lesser effects on TS-II were computed for substituents at the other positions of the pyridine ring (see table S9, ESI). Alteration of the amidophenolate backbone with two chlorine atoms also brought about only a minor lowering of TS-II (2.1 kcal mol⁻¹) but made the overall thermodynamics of the reaction more favorable. The most substantial product stabilization was seen with the electron-donating methoxy group in the pyridine ortho-position. Overall, the substituent effects appear additive in nature, as the relative stabilization of products and transition states for oxidative addition are approximately the sum of stabilizations afforded by each substituent (fig. 5B/C). This combinatorial nature allows both the kinetics and thermodynamics of the OA to be precisely tuned by variation of the substitution pattern. Further, phosphenium ions 2⁺ and 4⁺ featured the lowest overall computed transition states, even for the more challenging substrates.

Thus, we turned back to our experimental efforts. In accordance with the computationally determined trends, 2⁺ and 4⁺ reacted readily with thiophene already at room temperature and the reactions were complete within two days and five hours, respectively (fig. 6). Single crystals of [2]•[H][C₄H₃S]⁺ were grown from a concentrated dichloromethane solution and afforded the solid-state structure by X-ray diffraction (fig. 7), unequivocally showing the product of oxidative addition. The substituents are arranged in a distorted trigonal bipyramid around phosphorus, with the thiophene located at the equatorial positions. Related 2-bromothiophene and furan also underwent oxidative addition, furnishing the respective phosphonium salts [2]•[H][C₄H₂BrS][Al] and [4]•[H][C₄H₃O][Al]. For the former, an excess of substrate was required to drive the reaction to completion.

Figure 6:

Extended scope of the C-H oxidative addition, R^F = C(CF₃)₃, reactions are all quantitative. *The individual coupling constants could not be determined reliably due to an overlap of peaks.

Figure 7:

Thermal ellipsoid plots of [2]•[H][C₄H₃S]⁺, [4]•[H][C₆H₅]⁺ and [3]•[H][C₆H₄Cl]⁺ at 30 % probability. The counterion [B(C₆F₅)₄]⁻ and all carbon-based hydrogen atoms were omitted for all structures for clarity. Selected bond distances [Å] and angles [deg] of [2]•[H][C₄H₃S]⁺: d(P1-N1) = 1.9760(14), d(P1-N2) = 1.6769(14), ∠O1-P1-N1 = 169.21(6), ∠O1-P1-N2 = 91.07(7), ∠N2-P1-N1 = 82.66(7); [4]•[H][C₆H₅]⁺: d(P1-N1) = 1.9966(18), d(P1-N2) = 1.6735(17), ∠O1-P1-N1 = 170.90(7), ∠O1-P1-N2 = 91.22(8), ∠N2-P1-N1 = 82.34(8); [3]•[H][C₆H₄Cl]⁺: d(P1-N1) = 1.918(2), d(P1-N2) = 1.670(2), ∠O1-P1-N1 = 170.04(9), ∠O1-P1-N2 = 89.92(9), ∠N2-P1-N1 = 83.44(9).

Next, non-activated benzene was probed. A solution of 4⁺ in a 1:5 mixture of dichloromethane-d₂ and benzene heated to 80 °C for a day showed ~30 % oxidative addition, while full conversion was attained after another day at 100 °C.

The analogous reaction with 2⁺ was complete after two days at 110 °C. From this reaction, single crystals could be grown by vapor diffusion of pentane. The structure shows the oxidative addition of benzene, with the arrangement of the substituents around phosphorus analogous to [2]•[H][C₄H₃S]⁺ (fig. 7). These products represent rare examples of an intermolecular, single-site oxidative addition of benzene to a main group compound and the first operating via an electrophilic pathway. The only other examples of main-group oxidative addition of benzene were reported using highly nucleophilic aluminyl anions.¹⁵

The addition of toluene proceeded under milder conditions and ran to completion within one day at 70 °C. A mixture of ortho-, meta- and para-C-H activation products in a 3:1:6 ratio was obtained, showing a clear preference for the ortho- and para-positions in accordance with an electrophilic mechanism. Even the OA of deactivated chlorobenzene to 3⁺ was achieved at 140 °C, resulting in a mixture of isomers. Moreover, the C-H bond of an alkene in 1,1-diphenylethylene was selectively cleaved at room temperature (fig. 6 and fig. S20 for scXRD-derived connectivity). Even pristine alkenes such as 1-hexene are converted, although poor selectivity and competing site reactions have prevented unambiguous product analysis in this latter case. These results indicate the vast potential of this novel C-H functionalization chemistry.

Next, the properties of the OA reaction products were considered. As the CH-containing P(V)-phosphonium products were computed to not be overly stabilized compared to the P(III) starting materials, we assumed that reductive elimination should be thermally accessible. Direct evidence for reductive elimination was acquired by heating a solution of [2]•[H][C₄H₂SBr]⁺ in dichloromethane, for which the oxidative addition was assumed to be close to thermoneutral (fig. 8A,i). Indeed, phosphenium ion and free 2-bromothiophene continually formed at 60 °C until the amount of 2⁺ and [2]•[H][C₄H₂SBr]⁺ stabilized at a ratio of 71:29 after one day (fig. 8B). We reasoned that this partial reductive elimination would further enable exchange between several substrates. Upon adding 2-methylthiophene to a solution of 2⁺ or 4⁺, immediate formation of the OA products was observed. After adding 1-methylindole to these mixtures, the more stable indole addition product was formed after one day at room temperature in both cases (fig. 8A,ii). Close to complete conversion occurred after heating to 60 °C for two hours. Exchange of benzene was also observed upon heating [4]•[H][Ph]⁺ in toluene-d₈ at 130 °C. The products resulted from C-D addition of toluene-d₈ to phosphorus and benzene formation by reductive elimination (fig 8C). The absence of H/D scrambling in the products gives firm support that the process happens via reductive elimination, and not by alternative proton transfer channels.

Figure 8:

A Reductive elimination from [2]•[H][C₄H₂SBr]⁺ until an equilibrium was reached with 71 % of 2⁺ in solution, with ³¹P NMR spectra tracking the reaction progress in CD₂Cl₂ after (a) zero, (b) one, (c) five, (d) twenty-four hours at 60 °C, B. Exchange of substrates at [Z]⁺ at rt (Z = 2⁺, 4⁺) or 80 °C (Z = 1⁺), C. Exchange of benzene from 4⁺ by toluene-d₈.

Finally, we became interested in extending the CH-phosphorylation protocol toward the post-functionalization of arenes. Benzene was reacted with [4][Al(OR^F)₄] under the optimized reaction conditions, and the formed [4]•[H][Ph][Al(OR^F)₄] was hydrolyzed, furnishing phenylphosphinic acid in 91% yield with the protonated ligand as byproduct (fig. 9). This simple reaction illustrates the general possibility to generate functionalized products of non-activated arenes and opens a wealth of future directions.

Figure 9:

Phosphorylation of benzene to phenylphosphinic acid and protonated ligand. NMR yield is given. *A 5:1 mixture with CH₂Cl₂ was used for better solubility.

CONCLUSIONS

The structurally constrained phosphenium ions prepared in this work activate C-H bonds by oxidative addition. Modifying the ligand around P allows precise control over reaction kinetics and thermodynamics. Even non-activated, strong C_sp-H and C_sp2-H bonds, including those of benzene and chlorobenzene undergo oxidative addition. The reactions were shown to be reversible, leading to reductive elimination at elevated temperature and leveraging pathways for the exchange of activated substrates. The mechanism was elucidated by computations, and the critical step for the oxidative addition was determined to be a cooperative C-H deprotonation. The structurally constraint, T-shaped planar isomer of around 4 kcal mol⁻¹ above the bent ground state represents the putative key intermediate in the reaction. The inherent asymmetry of the phosphenium ions lends itself to future use in stereo- and enantioselective bond activations. Experiments to steer the reactivity towards catalytic functionalization of arenes or olefins and other substrates are currently underway in our laboratories.