Introduction

The introduction of fluorine atoms into organic molecules has attracted chemists’ great attention over the years due to the enhanced physicochemical properties of the fluorinated compounds1,2,3,4,5,6,7. Acyl fluorides are of great importance in terms of synthetic utility, which are wildly applied in the solution and solid-phase peptide synthesis as the use of α-amino acid fluorides generates the corresponding amides in good yields without racemization8,9,10,11. Acyl fluorides are also commonly used in the transition-metal catalyzed cross-coupling reaction as the acylating reagents, fluorination reagents, and so on12,13,14,15 (Fig. 1A). Synthesis toward acyl fluorides has been developed in the past few decades which were majorly categorized into two parts, nucleophilic deoxyfluorination of carboxylic acid16,17,18,19,20,21,22,23,24 and Pd-catalyzed fluorocarbonylation of aryl halides with CO gas or equivalents25,26,27,28. However, the direct fluorocarbonylation of aryl/alkyl boron compounds has not been reported before. As the complementary method, herein, we present the palladium/phosphine synergistic redox catalysis to achieve fluorocarbonylation of potassium aryl/alkyl trifluoroborate using trifluoromethyl arylsulfonate (TFMS) as COF2 source through the synergistic redox of Pd(II)/Pd(0) and P(III)/P(V).

Fig. 1: Reaction design.
Fig. 1: Reaction design.The alternative text for this image may have been generated using AI.
Full size image

A Applications of acyl fluorides. B Decomposition of TFMS reagent and difluoro P(V) species and CO via the phosphine nucleophilic attack COF2. C Our blueprint for palladium/phosphine synergistic redox catalysis of fluorocarbonylation of potassium aryltrifluoroborate. D Proposed mechanism.

Due to the high corrosivity and toxicity, COF2 is rarely studied29,30,31,32 until now. Only a few cases showed the potential for the laboratory-scale synthesis of COF2 through the fluoride substitution of (tri)phosgene29,30,32 or decomposition of trifluoromethoxy anion and radicals33,34,35,36. The previous studies in our group suggested that trifluoromethyl arylsulfonate (TFMS) could generate trifluoromethoxy anions in situ, which could decompose into COF2 and fluoride anions37,38,39 (Fig. 1B). Inspired by the pioneering work of the Shreeve group, COF2 could be easily reduced into CO and afforded the difluorophosphine as a byproduct40 (Fig. 1B). Meanwhile, palladium-catalyzed fluorocarbonylation of aryl halides from CO source25,26,27,28 and Lewis acidic phosphonium accelerated migratory insertion of CO41,42 was well-documented. Taking these discoveries and the broad synthetic utilities of acyl fluorides into account, it is meaningful and promising to explore the palladium/phosphine synergistic redox catalysis fluorocarbonylation of potassium aryl/alkyl trifluoroborate with TFMS; we assumed that two interwoven catalytic cycles Pd(II)/(0) and P(III)/(V) might be engineered to generate simultaneously (1) LnArPd(II)F species and COF2 via the β-F elimination of LnArPd(II)OCF3 intermediate, and (2) difluoro P(V) species and CO via the phosphine nucleophilic attack COF2. Through the different spatial coordination of fluoride ions, the reducibility of difluoro P(V) species may be tuned for matching the oxidative addition of LnPd(0) species to regenerate catalytic LnPd(II) species (Fig. 1C).

Our proposed mechanistic cycle for palladium/phosphine synergistic redox catalysis of fluorocarbonylation of potassium aryl/alkyl trifluoroborate 1 is outlined in Fig. 1D. We presumed that the transmetallation of 1 would happen to form complex I-2, which underwent carbonyl insertion to generate intermediate I-3, followed by reductive elimination to release the product 2 and generate the Pd(0) species I-4. Concurrently with this palladium cycle, we envisioned that phosphine ligand L4 reacts with COF2 via nucleophilic attack to form I-5, followed by fluoride ion migration to afford five-coordinated P(V) intermediate I-6, which might proceed through CO release to afford activated difluoro P(V) species I-7. And then, a synergistic redox mechanism with Pd(0) species I-4 may occur via ligand exchange and oxidative addition, regenerating catalyst Pd(II) species I-1 and phosphine ligand L4.

Results

Mechanistic investigation

The initial mechanism investigation was focused on stoichiometric experiments. Following Buchwald’s protocol43, the aryl–Pd bromide 4 was easily synthesized. To our delight, the subsequent anion exchange succeeded to afford the target Pd(II)–OCF3 complex 5a. Formation of 5a was demonstrated by the 19F NMR, which showed a broad peak appeared at δ −30 ppm, although it completely decomposed in the NMR tube under room temperature overnight through the β-F elimination (Fig. 2A). During the preparation of our manuscript, Shen and co-workers reported the preparation of trifluoromethoxylated Pd(II) complexes, see ref. 44. The structure of 5a was also confirmed by X-ray crystallography44. Furthermore, the thermal decomposition of 5a favorably generated the aroyl fluoride product 2a, suggesting that the Pd(II)–OCF3 complex 5a was the key intermediate in the reaction. The DFT calculation (see Supplementary Information for more details) also demonstrated that β-F elimination (ΔG‡ = 8.4 kcal/mol) was much more preferential than the direct reductive elimination (ΔG‡ = 40.5 kcal/mol) and explained the exclusive formation of aroyl fluoride.

Fig. 2: Mechanism investigation.
Fig. 2: Mechanism investigation.The alternative text for this image may have been generated using AI.
Full size image

A Synthesis of Pd(II)–OCF3 complex. B Reaction course study. C Synthesis of difluoro-Brettphos. D Catalytic method for palladium/phosphine synergistic redox catalysis of fluorocarbonylation of potassium aryltrifluoroborate. E Isotope labeling of CO.

The aryl–Pd trifluoromethesulfonate 6 was synthesized for the reaction course study of stoichiometric transformation by simultaneous 19F NMR characterization. To avoid the influence of heterogeneity, the soluble fluoride TASF45 was used in place of inorganic fluorides. Pd(II)–OCF3 complex 5b was generated immediately after the injection of TFMS, and desired product 2y was observed subsequently (Fig. 2B). A concomitant intermediate was also signified in the spectrum at δ −47.06 (d, J = 684.8 Hz) and kept increasing with product 2y over the reaction. Accordingly, the 31P NMR signal was a triplet and appeared at δ −20.08 (t, J = 681.6 Hz). The splitting pattern and slitting constant of around 680 Hz indicated a new P–F bond was formed.

Based on the structural analysis from 19F NMR, we believed the fluorine atom in the ligand was transferred from COF2 gas40, and the synthetic method of intermediate 7 was then proposed by treating Brettphos ligand with COF2 which could be generated in situ from TFMS triggered by the catalytical amount of fluoride. As expected, the difluoro-Brettphos 7 was quantitatively obtained in 19F NMR and the precipitation generated after placing overnight was qualified for further spectroscopic characterization, including X-ray crystallography (Fig. 2C). Two identical fluorides were trans-positioned to each other at perpendicular axial and matched the NMR result that two F-atoms on the phosphorus center were equivalent. Meanwhile, COF2 was reduced into CO. Encouraged by the stoichiometric experiments, the catalytic method for palladium/phosphine synergistic redox catalysis of fluorocarbonylation of potassium aryltrifluoroborate was investigated (see Supplementary Information for more details), and Xantphos was found to be the best ligand (Fig. 2D). The reason might be that once Xantphos was difluorinated, one vacant coordination site was generated and could be occupied by CO facilely, followed by the favorable migratory insertion to afford the aroyl–Pd complex, which was beneficial for the catalysis. Besides generating a new coordination site, Lewis acidic fluoro-phosphonium cation appended to a palladium complex was also proved to promote CO insertion into a Pd–C bond41,42 and lowered the activation barrier. To demonstrate the role of difluorophosphines in the reaction, the difluoro-Xantphos L5 was synthesized and identified by X-ray crystallography; subsequently, it was used as a ligand and behaved similarly to Xantphos. An isotope tracing experiment was performed to verify the CO insertion instead of COF2. When the complex 5a was heated under a 13CO atmosphere, the isotope-labeled aroyl fluoride was observed (Fig. 2E), suggesting the free CO molecule participated in the reaction. These results revealed the migratory insertion of CO was one of the critical steps in the catalytic cycle.

To disclose the full mechanism, especially for the synergistic redox of P(III/V)/ Pd(II/0) and the role of AgF, density functional theory (DFT) calculations were carried out by using Brettphos as the model ligand consistent with the control experiment in Fig. 3. DFT calculations were performed using Gaussian 09 at the SMD(1,4-Dioxane)-M06L/6-311+G(d,p)/Def2-TZVP(Pd,Ag) level of theory (see Supplementary Information for more details).

Fig. 3: Computational studies.
Fig. 3: Computational studies.The alternative text for this image may have been generated using AI.
Full size image

A Free energy profile for the catalytic cycle of the P(III)-P(V) mechanism releasing CO. Computed at the SMD(1,4-Dioxane)-M06L/6-311+G (d,p)//B3LYP-D3 /6–31G(d). B Free energy profile for the Pd(II)/Pd(0) catalytic cycle releasing product 2a. Computed at the SMD(1,4-Dioxane)-M06L/6–311+G (d,p)/Def2-TZVP(Pd,Ag) //B3LYP-D3 /6–31G(d)/SDD(Pd,Ag).

The dual catalysis enables the fluorocarbonylation reaction illustrated in the free energy profiles in Fig. 3. Our investigations indicate that the phosphine ligand indeed plays a bifunctional role, which can not only be a ligand to coordinate palladium but also can be a reactant with COF2 to release CO. In Fig. 3A, the Brettphos ligand L4 is facile to react with COF2 via nucleophilic attack to form INT1, when the valence state of P would be oxidized from III to V valence. Then, fluoride ion migration of INT1 occurs to afford five-coordinated INT2 based on the P(V) center with the geometry of distorted trigonal bipyramidal configuration. The intermediate INT2 then would be attacked by fluoride ions to form octahedral intermediate INT3 or INT3’, which would release CO via TS4 or TS4’ to generate the trans-difluoro INT4 or cis-difluoro INT6, respectively. And the cis-difluoro INT6 could be isomerized to the stable trans-difluoro INT4 supported by the isolated crystal structure 7. Furthermore, F- ligand dynamic coordination/dissociation during the transformation between INT4 and INT6 indicated the tuning effect of F- anion for the spatial coordination of difluoro P(V) species. The turnover-limiting step of this phosphine cycle step has to overcome the 26.4 kcal/mol energy barrier that is accessible for a reaction that proceeds at 80 °C. Our calculated mechanism of the phosphine cycle rationalizes the generation of CO from COF2, which is in agreement with our experimental observation in Fig. 2B.

For the Pd(II)/Pd(0) catalytic cycle, L4 and 1a were chosen as the starting materials in order to keep consistent with the zero point of P(III)-P(V) cycle shown in Fig. 3B. The Brettphos bound Pd(II) difluoro intermediate INT8 was generated by favorable ligand exchange with Pd complex with exothermic −13.1 kcal/mol (1). And the additive AgF salt was introduced to activate 1a through SEAr mechanism via TS5 with a +3.8 kcal/mol low energy barrier to form the activated Ag-Ar intermediate INT7 (2). The overall thermodynamic energy was exothermic −11.0 kcal/mol energy considering both ligand exchange (−13.1) and SEAr (+2.1). The formed Ag-Ar intermediate INT7 could then react with Pd(II) species INT8 through the key transmetallation, transferring the aryl group from Ag to Pd. This energy barrier of Ag assisted model via TS6 was calculated to be favorable comparing other aryl group transfers in Supplementary Fig. S4, which was also supported by the study of aryl silver species46,47,48,49 by Ritter et al. The following CO insertion step was initiated by Ar–Pd intermediate INT10 and underwent three-membered ring transition state TS8, in which the Brettphos ligand displayed a slight ligand dissociation of η2-aryl interaction to accept extra CO coordination. And this dynamic coordination of the bi-phenyl group in Brettphos ligand would re-bind to Pallidum in the next intermediate INT13. In the structure of TS8, the closest distance between Pd and CAr on the ligand is 3.37 Å, while it turns into 2.53 Å and 2.63 Å in INT11 and INT13, respectively. The following reductive elimination was a low barrier step releasing activated Pd(0) species INT13 and the final product 2a. Then only cis-difluoro P(V) intermediate INT6 was suggested to react Pd(0) via ligand exchange and oxidative addition regenerating catalyst species INT8. Note that the most stable trans-difluoro INT4 cannot easily undergo oxidative addition via TS11, which is +4.6 kcal/mol relative energy higher than TS10 with respect to the energy of zero point (1a+L4) (see Supplementary Information for details), probably due to the trans-effect contribution of F− to hardly weaken C–F bond. So we speculate that cis-difluoro P(V) INT6 is an active intermediate involved in the catalytic cycle. And the isomerization of cis- and trans- difluoro P(V) under F− can control the reactivity to achieve this synergistic redox transformation. After the oxidative addition of the P–F bond to produce INT15, the second fluorine transfer tend to be facile through TS12 with an even low energy barrier. Overall, the turnover-limiting step for Pd(II)/Pd(0) catalytic cycle would be the first oxidative addition via TS10, overcoming a reasonable 24.8 kcal/mol at 80 °C (Fig. 3B).

The substrate scope of the transformation was explored, and the generality of the reaction was covered in Fig. 4, demonstrating the efficient fluorocarbonylation of various (hetero)aryl/alkyl trifluoroborates, including some pharmaceutically relevant structures. The unstable acyl fluorines 2 were transformed into more valuable amides derivatives 3 immediately after the reaction following Manabe’s protocol26 due to the handy hydrolysis during chromatography purification. In general, for (hetero)aryl trifluoroborates, electronic and steric effects showed little influence on the reaction, while the electron-withdrawing groups and meta-substituents decreased the yields to some extent. Various substrates bearing electron-donating to electron-withdrawing substituents such as halides (1a-1d, 1v), alkyl and aryl groups (1h, 1k, 1o, 1u), alkoxyl (1g, 1i, 1m), ketone (1n), nitro (1p) and nitrile (1q) were successfully converted to the desired acyl fluorides with yields ranging from 50 to 99%. To our delight, most ortho-substituted substrates (1e, 1j, 1l, 1o, 1t) were well tolerated in this protocol and gave moderate yields (42 to 64%) with increasing steric hindrance. More importantly, the multi-substituted aryl trifluoroborates (1d, 1j, 1o, 1v) could be smoothly fluorocarbonylated. In addition, alkyl trifluoroborates (1w-1ae) were successfully converted to the corresponding desired alkyl fluorides with moderate yields under standard conditions. To prove the high compatibility of the reaction, the complicated pharmaceutical skeletons, Indometacin (1af), Fenofibrate (1ag), Gibberellic acid (1ah), Triclosan (1ai), Deoxycholic acid (1aj), and Estrone (1ak) derivatives were examined in the transformation and afforded the products (2af, 2ag, 2ah, 2ai, 2aj, 2ak) with moderate yields without touching the resting functional groups. Especially, 2aj and 2ak were obtained with isolated yields of 35 and 55%, respectively. These results showed the ability to allow the late-stage fluorocarbonylation of bioactive molecules and natural product derivatives.

Fig. 4: Substrate scope for fluorocarbonylation of potassium aryl/alkyl trifluoroborate.
Fig. 4: Substrate scope for fluorocarbonylation of potassium aryl/alkyl trifluoroborate.The alternative text for this image may have been generated using AI.
Full size image

Reaction conditions: 1 (0.250 mmol, 1.0 equiv), TFMS (0.875 mmol, 3.5 equiv), Pd(cod)Cl2 (0.0125 mmol, 5 mol%), Xantphos (0.025 mmol, 10 mol%), AgF (0.250 mmol, 1.0 equiv), dioxane (0.125 mol/L), 100 °C, 12 h, N2 atmosphere. Yields of acyl fluorides 2 were determined by 19F NMR with tribromofluoromethane or phenylsulfonyl fluoride as an internal standard unless otherwise noted. Yields of isolated amide derivatives 3 are given in parentheses. *Yields of isolated product 2#. TFMS (4.0 equiv) was used.

Discussion

In conclusion, we have developed the palladium/phosphine synergistic redox catalysis of fluorocarbonylation of potassium aryl/alkyl trifluoroborate with TFMS as a COF2 source. The new method realized the direct conversion of potassium aryl/alkyl trifluoroborate to acyl fluorides, which allows the late-stage fluorocarbonylation of a variety of organic molecules and known drugs. Furthermore, the key intermediates of labile Pd(II)–OCF3 complex 5a and difluoro-Brettphos 7 were synthesized and spectroscopically characterized, including X-ray crystallography. In addition, the proposed reaction mechanism involving the interwoven catalytic cycles Pd(II)/(0) and P(III)/(V) was supported by experimental and computational studies and indicated that phosphine ligand plays a bifunctional role in the catalysis, which might stimulate the development of new reactions involving the synergistic redox of Pd(II)/Pd(0) and P(III)/P(V). Our studies provide a potential strategy to control the reducibility of P(V) species by the spatial coordination of fluoride ions, which might innovate more catalytic possibilities for widely used phosphine ligands.

Methods

General procedure for the synthesis of acyl fluorines

In an N2 glovebox, to aryl fluoroboric acid potassium salt (1) (0.25 mmol, 1.00 equiv.), Pd(cod)Cl2 (3.5 mg, 0.0125 mmol, 0.05 equiv.), AgF (31.7 mg, 0.25 mmol, 1.00 equiv.) and Xantphos (14.5 mg, 0.025 mmol, 0.10 equiv.) in a 4.00 mL sealed vial tube were added 1,4-dioxane (2.00 mL) and TFMS (trifluoromethyl 4-fluorobenzenesulfonate) (140.0 µL, 0.875 mmol, 3.50 equiv.). Then the sealed vial was taken outside the glovebox, and the reaction mixture was stirred for 12 h at 100 °C. The system was filtered and concentrated in vacuo. The residue was purified by preparative HPLC with acetonitrile/water as an eluent.

General procedure for the synthesis of amides derivatives

In an N2 glovebox, to aryl/alkyl fluoroboric acid potassium salt (1) (0.25 mmol, 1.00 equiv.), Pd(cod)Cl2 (3.5 mg, 0.0125 mmol, 0.05 equiv.), AgF (31.7 mg, 0.25 mmol, 1.00 equiv.) and Xantphos (14.5 mg, 0.025 mmol, 0.10 equiv.) in a 4.00 mL sealed vial tube were added 1,4-dioxane (2.00 mL) and TFMS (trifluoromethyl 4-fluorobenzenesulfonate) (140.0 µL, 0.875 mmol, 3.50 equiv.). Then the sealed vial was taken outside the glovebox, and the reaction mixture was stirred for 12 h at 100 °C. The system was cooled to room temperature, bubbled with argon for 10 min, and added BnNH2 (88.4 mg, 0.825 mmol, 3.3 equiv.) and diisopropylethylamine (DIPEA) (106.6 mg, 0.825 mmol, 3.3 equiv.), the reaction mixture was stirred for 3 h at room temperature. The system was filtered and concentrated in vacuo. The residue was purified by preparative TLC eluting with n-hexane/EA.