Why does the synthesis of N-phenylbenzamide from benzenesulfinate and phenylisocyanate via the palladium-mediated Extrusion–Insertion pathway not work? A mechanistic exploration

Keywords: biaryl coupling, desulfination, DFT calculations, extrusion, insertion, mass spectrometry, palladium mediated reactions, reaction mechanisms.

There has been considerable interest in developing transition metal catalysed reactions for organic synthesis that avoid a transmetalation step requiring the use of stoichiometric and often toxic organometallic/organometalloid reagents.^[1] Thus alternative reagents that allow for formation of the key organotransition metal intermediate have been sought. Two key classes of alternative reagents have emerged as front runners: carboxylic acids, which undergo metal-catalysed decarboxylation reactions,^[2–9] and sulfinic acids or their salts, which undergo related desulfination reactions.^[10–13] Based on our gas-phase studies over the past two decades, where we have examined a wide range of metal-catalysed decarboxylation reactions^[14–22] and some metal-catalysed desulfination reactions,^[23,24] recent efforts have focussed on combining gas-phase (Scheme 1a) and solution-phase mechanistic studies coupled with DFT calculations to develop a new class of reactions for the synthesis of amides, thioamides, amidines and alkenes. These studies involve palladium-mediated/catalysed extrusion of CO₂ to form an organopalladium intermediate followed by insertion of an appropriate (hetero)cumulene (Scheme 1b).^[25–28] Given that these synthetic methods were limited to the use of 2,6-dimethoxybenzoic acid as a substrate, we recently explored the use of silver carbonate for the synthesis of N-phenyl-benzamide starting from benzoic acid and phenyl isocyanate.^[29] While the desired ExIn mechanism operates in the gas-phase (Scheme 1c, Eqns 1, 2), a different base-catalysed condensation mechanism not requiring silver operates in solution. Given that phenylsulfinic acid has been shown to readily undergo desulfination by palladium complexes in both the gas-phase^[24] and in synthetic protocols,^[30,31] here we demonstrate that while the desired palladium-mediated ExIn mechanism operates in the gas-phase (Scheme 1c, Eqns 1, 2), a palladium-mediated biaryl coupling side reaction dominates in solution.^[32–35]

Scheme 1.  Mechanism-based approaches that provide the guidance to develop new synthetic methods: (a) gas-phase Pd-mediated extrusion–insertion (ExIn) reactions, (b) one-pot approaches to synthesise thioamides, amidines and amides from carboxylic acids and (c) gas-phase investigation of M(phen)-mediated ExIn for amide synthesis where M = Ag^[29] or Pd (this work). The extrusion reaction is shown in Eqn 1 while the insertion reaction is given in Eqn 2.

Electrospray ionisation (ESI) of a methanolic solution of 1,10-phenanthroline, palladium trifluoroacetate and benzoic acid gave rise to the cationic complexes, [(phen)_nPd(O₂SC₆H₅)]⁺. As noted previously, these complexes undergo desulfination under collision-induced dissociation (CID) conditions to form [(phen)_nPd(C₆H₅)]⁺, as illustrated for [(phen)₂Pd(O₂SC₆H₅)]⁺ in Fig. 1a, Eqn 3. Another alternative route to prepare [(phen)Pd(C₆H₅)]⁺ in the gas phase is via ligand loss from [(phen)_nPd(C₆H₅)]⁺ (Fig. 1b, Eqn 4). While [(phen)₂Pd(C₆H₅)]⁺ was found to be unreactive towards phenylisocyanate in the gas phase (Fig. 1c), [(phen)Pd(C₆H₅)]⁺ undergoes an ion–molecule reaction (IMR) with phenylisocyanate to yield a product ion at m/z 482 (Fig. 1d, Eqn 5), a reaction previously observed for [(phen)Pd(C₆H₅)]⁺ formed via decarboxylation instead. The resultant [(phen)Pd(NPhC(O)C₆H₅)]⁺ (m/z 482) fragments undergo both deinsertion (Eqn 6) and loss of benzene (Eqn 7), as previously reported for the ExIn product formed via decarboxylation.^[27]

Fig. 1.  Multistage mass spectra (MSⁿ) of uni- and bi-molecular reactions associated with key steps of the ExIn reaction: (a) MS² CID experiment showing the extrusion of SO₂ from [(phen)₂Pd(O₂SPh)]⁺ (m/z 670, with normalised collision energy (NCE) = 17, Eqn 3); (b) MS³ CID experiment showing the loss of a phen ligand from [(phen)₂Pd(Ph)]⁺ (m/z 543, NCE = 22, Eqn 4); (c) MS³ IMR experiment between the organometallic cation [(phen)₂Pd(Ph)]⁺ (m/z 543) and phenyl isocyanate at 2000 ms activation time and (d) MS⁴ IMR experiment between the organometallic cation [(phen)Pd(Ph)]⁺ (m/z 363) and phenyl isocyanate at 10 ms activation time. The concentration of phenyl isocyanate is 1.22 × 10¹⁰ molecule cm⁻³ inside the ion trap under the IMR. Asterisks are used to designate the mass-selected precursor ions.

Encouraged by the gas-phase studies, suggesting a palladium-mediated stepwise extrusion of SO₂ followed by insertion of phenyl isocyanate could be a viable approach to synthesise amides, the palladium-catalysed approach was explored using a one-pot method under ligand free conditions and in the presence of neutral ligands using different solvents (DMSO or NMP) and with and without additives (base or acid). As in our previous work, the crude reaction mixtures were analysed via GC-MS. No amide product was observed under a range of reaction conditions (all attempts are listed in Table 1). Instead, in all cases the dominant side product was the biaryl species arising from double desulfination followed by homocoupling. The formation of the desulfinated intermediate was detected by electrospray ionisation HRMS (entry 4 in Table 1, Fig. 2), which showed the arylpalladium complexes coordinated with one acetonitrile and one 6mbpy ligand and one with two 6mbpy ligands (at low abundance). However, the coordination with phenyl isocyanate was not detected, nor was the [M + H]⁺ ion of N-phenyl-benzamide. The GC-MS data revealed that there was still a significant amount of unreacted phenyl isocyanate while the ESI-HRMS revealed that some phenyl isocyanate has transformed into its urea analogue.

Table 1.  Attempts for the ExIn reaction between aromatic sulfinate salt and isocyanates.

Fig. 2.  The ESI-HRMS spectrum (+ve, Thermo Obitrap MS, CH₃CN solvent) showing the full spectrum for entry 4 of Table 1. [(6mbpy)¹⁰⁶Pd(Ph)(NCCH₃)]⁺ at m/z 396 and [(6mbpy)₂¹⁰⁶Pd(Ph)]⁺ at m/z 523. The inset shows the isotope distribution pattern of the [(6mbpy)Pd(C₆H₅)(NCCH₃)]⁺ from entry 4 in Table 1. Observed (top) and calculated (bottom). The ions at m/z 425 and 447 are due to urea side products arising from hydrolysis of the phenylisocyanate.

The formation of biaryl products is likely due to coordination of a second phenyl sulfinate to the arylpalladium followed by desulfination and reductive elimination of biphenyl. It is worth noting that related sulfinate coordinated arylpalladium intermediates have been formed via oxidative addition of aryl iodide into PCy₃ ligated palladium(0) followed by coordination with a sulfinate anion.^[35] The resultant binuclear palladium complex has sulfinates as bridging ligands and could be transformed to a mononuclear palladium complex via the addition of another equivalent of the ligand PCy₃ (Scheme 2a). Both the binuclear and mononuclear palladium complexes were structurally characterised via X-ray crystallography (Scheme 2b, c). Unfortunately such species were not detected using ESI-MS as they have no net charge.

Scheme 2.  (a) Previous report on the synthesis of arylpalladium sulfinate intermediates and their X-ray crystal structures showing (b) binuclear and (c) mononuclear complexes.

Having established that the gas-phase ExIn reaction occurs but that the biphenyl side product is formed in solution, we were interested in using DFT calculations to explore the mechanistic aspects (reaction pathways and energetics) associated with the biphenyl side product and to compare the energetics to that for the insertion of phenylisocyanate (Fig. 3). We first examined the relative stabilities of the three coordinate complex [(phen)Pd(C₆H₅)]⁺, 4, and the DMSO solvated complex, [(phen)Pd(C₆H₅)(S(O)Me₂)]⁺, 5. The latter was found to be more stable and was thus used as the key complex to calculate the insertion versus biphenyl side reaction pathways. The key energy barriers for the insertion manifold are associated with the transition states for displacement of the DMSO ligand by the isocyanate ligand TS5–6 (17.8 kcal mol^–1) and the insertion reaction TS6–7 (13.4 kcal mol^–1). In contrast, replacement of the coordinated DMSO ligand with the anionic sulfinate proceeds with a low barrier of 2.4 kcal mol^–1 (TS9–10) and is a highly exothermic reaction to form the highly stable S-coordinated palladium complex, 10. Even though the desulfination reaction through TS10–11 has a relatively high energy barrier of 29.4 kcal mol^–1, it is lower than the energetics required for formation of 6 from 10 (33.9 kcal mol^–1). Thus desulfination is the favoured pathway. After losing SO₂, the homocoupling reaction of the double desulfinated palladium complex, 12, leads to the formation of a biaryl coordinated complex, 13, via a low energy barrier (8.9 kcal mol^–1). The fact that both key transition states in the homocoupling pathway are lower in energy than the one from the insertion pathway supports the hypothesis regarding the failure for the amide synthesis in the condensed phase.

Fig. 3.  DFT calculated potential energy diagram (energies in kcal mol^–1) showing the competition between insertion of phenyl isocyanate (left) and desulfinative homocoupling reaction (right). Method used: CAM-B3LYP-D3BJ/BS2//M06/BS1 level of theory using solvent (DMSO) continuum (CPCM) approach.

While we have observed the protodecarboxylation side reaction in our previous studies of ExIn reactions involving the 2,6-dimethoxybenzoic acid substrate,^[25–28] we never observed the formation of the biaryl side product. Thus we were interested in establishing the energetics of the biaryl side reaction relative to the insertion of phenylisocyanate (Fig. 4). The key energy barriers for the insertion manifold are associated with the transition states for displacement of the solvent ligand by the isocyanate ligand TS5–6b (17.9 kcal mol^–1) and the insertion reaction TS6–7b (10.9 kcal mol^–1). In contrast, replacement of the coordinated solvent ligand with the anionic carboxylate requires more energy (TS9–10b, 20.6 kcal mol^–1 relative to 5b), while the decarboxylation step through TS10–11 has an energy barrier of 28.6 kcal mol^–1. After losing CO₂, the homocoupling reaction of the palladium complex, 13b, leads to the formation of a biaryl coordinated complex, 14b, via an energy barrier (19.3 kcal mol^–1). The fact that both key transition states in the homocoupling pathway are higher in energy than those from the insertion pathway is consistent with the lack of biaryl formation in the experiments.

Fig. 4.  DFT calculated potential energy diagram (energies in kcal mol^–1) showing the competition between insertion of phenyl isocyanate (left) and decarboxylative homocoupling reaction (right). Method used: CAM-B3LYP-D3BJ/BS2//M06/BS1 level of theory using solvent (DMA) continuum (CPCM) approach.

Gas-phase studies provide valuable information on elementary steps relevant to organometallic chemistry used in organic synthesis. Here we have shown that desulfination reactions can be used to form the organopalladium intermediates [(phen)_nPd(C₆H₅)]⁺ and that in the case of n = 1, phenyl isocyanate inserts to yield [(phen)Pd(NPhC(O)C₆H₅)]⁺. A key challenge we have found in translating these gas-phase ExIn reactions to solution phase protocols which produced the desired product in high yield is the formation of unwanted side products. Thus while gas-phase studies provide exquisite control in a ‘pristine environment’, they do not capture the rich milieux of the condensed phase where other reagents that are absent in the gas-phase can facilitate the formation of side products. In the case of decarboxylation of benzoates, a side reaction often encountered in the condensed phase is protodecarboxylation in which the arylorganometallic intermediate is protonated by an acid to form the arene. Here we have encountered a different side reaction in the desulfination ExIn approach: biaryl formation via double desulfination followed by reductive elimination. This series of reactions cannot be observed in our gas-phase studies since the organopalladium intermediate cation [(phen)Pd(C₆H₅)]⁺ cannot react with a second phenylsulfinate, C₆H₅SO₂⁻. The DFT calculations on the solution phase competition between insertion of isocyanate and homocoupling via desulfination confirmed the experimental results. Lower energy barriers identified in the homocoupling reaction via double desulfination on the palladium centre account for the biaryl species being the dominant product in the condensed phase.

Reagents, purchased from various commercial sources, were used as received. Chromatographic silica media (Davisil, 40−63 μm), was used as the stationary phase in flash column chromatography.

Ligated palladium cations, [(L)_nPd(O₂XR)]⁺, (n = 1 or 2, X = C or S) were subjected to CID to form aryl-palladium cations [(L)_nPd(R)]⁺, which were then mass selected for subsequent ion–molecule reaction studies with phenyl isocyanate. We followed the protocols outlined in previous work.^[20,21] For instance, methanolic solutions of palladium(ii) salt (10 mM), carboxylic acid or sodium benzenesulfinate (10 mM) and 1,10-phenanthroline (10 mM) were mixed based on a ratio of 1:1:2 and then diluted to 10 µM in palladium salt. A syringe pump (flowrate of 5 μL min⁻¹) was used to inject the diluted solution into a modified linear ion-trap mass spectrometer (Thermo Finnigan LTQ) via the ESI source. The modified system allows ion–molecule reactions between mass selected ions and neutral molecules such as phenylisocyanate within the linear ion trap.^[36,37] All spectroscopic data were acquired between 20 and 100 duplicate spectra with 3–5 microscans in each scan.

The sheath gas setting was 10 AU, auxiliary gas was set to 5 AU, the sweep gas was 0 AU, the spray voltage was 4 kV, the capillary temperature was set to 250°C, the capillary voltage was 2 V and the tube lens voltage was set to 75 V.

The precursor ion was mass selected with a window of 1 m/z and subjected to collisional activation via collisions with the helium bath gas using a 10 ms activation time. The normalised collision energy (NCE) was set so as to achieve a precursor ion depletion to 10%.

The precursor ion was mass selected with a window of 1 m/z and subjected to IMRs with phenylisocyanate. The NCE was set to 0% so as not to activate the precursor ion.

¹H and ¹³C NMR spectra were recorded using a Jeol 400 MHz NMR spectrometer at 298 K, and were referenced to the ¹H shift in CDCl₃ (7.24 ppm) and ¹³C shift in CDCl₃ (77 ppm). All the NMR spectra are reported in parts per million (ppm) and coupling constants (J) are reported in Hertz (Hz). Multiplicities are recorded as: t = triplet, d = doublet, s = singlet.

High-resolution mass spectra (ESI-HRMS) were obtained on a Thermo Scientific Exactive Plus Orbitrap mass spectrometer (Thermo, Bremen, Germany) via positive ion ESI and were used to examine species present in reactions mixtures and to confirm the molecular formulas of purified isolated products.

GC-MS (Agilent 7890A/5975C GC-MS) analyses were carried out in an HP-5ms capillary column (Agilent Technologies, phenyl methyl siloxane, 30 m × 0.25 mm × 0.25 μm). To achieve a good separation, the time program was used by beginning with 5 min at 70°C, followed by a 15°C min^–1 ramp to 300°C and then 10 min at this temperature.

To a solution of palladium trifluoroacetate (0.1 mmol, 0.5 equiv.) and bidentate ligand (0.11 mmol, 0.55 equiv.) in DMSO or NMP (2 mL) was added sodium benzenesulfinate (0.2 mmol, 1 equiv.), phenyl isocyanate (0.4 mmol, 2 equiv.) and additive. The mixture was heated at 110°C for 2 h and quenched with 1 M HCl (1 mL) and water (50 mL), followed by liquid–liquid extraction with diethyl ether (3 × 50 mL). The combined organic fractions were washed with water (100 mL) and brine (100 mL) and dried over anhydrous MgSO₄. The sample was then subjected to analysis via GC-MS as described above. In the case of entry 12 of Table 1, the solvent was removed, and the residue was purified by column chromatography to give the biaryl.

1,1′-Biphenyl (B) was prepared using the general method (entry 12 of Table 1): yield 7 mg (46%) as white solid. Column chromatography (silica gel, diethyl ether/n-hexane: 1/10). ¹H NMR (400 MHz, CDCl₃): δ 7. 59 (d, J = 6.9 Hz, 4H), 7.44 (t, J = 7.3 Hz, 4H), 7.34 (d, J = 7.5 Hz, 2H). ¹³C NMR (400 MHz, CDCl₃): δ141.34, 128.84, 127.34, 127.26. GC-MS (EI): m/z [M]⁺ calcd. for C₁₂H₁₀ 154.1, found 154.1.

The Gaussian 16 suite of programs was used to fully optimise all reactants, intermediates, transition states and products at the M06 level of density functional theory (DFT).^[38,39] The effective-core potential of Hay and Wadt with a double-ξ valence basis set (LANL2DZ) was used to describe Pd^[40,41] and the 6-31G(d) basis set was chosen for the other atoms.^[42] In addition, a polarisation function (ξf = 1.472) was solely added for Pd.^[43,44] BS1 was used to designate this combination of basis sets. In order to account for the solvation effects (DMSO in Fig. 3 and DMA in Fig. 4) on the optimised structures the CPCM model was used.^[45] Frequency calculations were carried out at the same level of theory as those for the structural optimisation. The Berny algorithm was used to locate each of the transition structures. Intrinsic reaction coordinate (IRC) calculations were used to confirm the connectivity between structures of transition states and minima.^[46,47]

Single-point energy calculations were carried out to further refine the energies. Thus the energies of the structures obtained from the M06/BS1 calculations were recalculated with a larger basis set (BS2) at the B3LYP-D3BJ or CAM- B3LYP-D3BJ level of theory.^[48–51] BS2 utilises def2-TZVP11 for all atoms along with the effective core potential including scalar relativistic effects for Pd.^[52] The solvation effect of DMSO and DMA were also considered in the single-point calculations using the CPCM model. Relative enthalpy (ΔH) and Gibbs energies (ΔG) at the BS2 level of theory were calculated using the correction values calculated from M06/BS1. Based on the method reported by Okuno, extra corrections for entropy calculations were considered in the solvent system.^[53] When DMSO or DMA participate in the equilibrium of a certain transformation step, an additional correction was considered based on the concentration of the DMSO or DMA using the method proposed by Keith and Carter (Eqn 6 of their paper was used).^[54] Unless otherwise stated, all the enthalpy and Gibbs free energies were calculated and corrected from the B3LYP-D3BJ/BS2//M06/BS1 level of theory.

Supplementary material is available online.

The data that support this study are available in the article and accompanying online supplementary material.

The authors declare no conflicts of interest.

The authors thank the ARC for financial support (DP180101187 funding to AJC and RAJO), and acknowledge a generous allocation of time from the National Computing Infrastructure.