Reaction Mechanism of Pd‐Catalyzed “CO‐Free” Carbonylation Reaction Uncovered by In Situ Spectroscopy: The Formyl Mechanism

Abstract “CO‐free” carbonylation reactions, where synthesis gas (CO/H2) is substituted by C1 surrogate molecules like formaldehyde or formic acid, have received widespread attention in homogeneous catalysis lately. Although a broad range of organics is available via this method, still relatively little is known about the precise reaction mechanism. In this work, we used in situ nuclear magnetic resonance (NMR) spectroscopy to unravel the mechanism of the alkoxycarbonylation of alkenes using different surrogate molecules. In contrast to previous hypotheses no carbon monoxide could be found during the reaction. Instead the reaction proceeds via the C−H activation of in situ generated methyl formate. On the basis of quantitative NMR experiments, a kinetic model involving all major intermediates is built which enables the knowledge‐driven optimization of the reaction. Finally, a new reaction mechanism is proposed on the basis of in situ observed Pd‐hydride, Pd‐formyl and Pd‐acyl species.


Experimental section Chemicals
All manipulations were carried under argon atmosphere using standard Schlenk technique. 1,2-Bis(di-tert-butylphosphino)xylene (d t bpx, L) was purchased from Sigma Aldrich and recrystallized from MeOH. d4-MeOH and 13 C-paraformaldehyde ( 13 C-PFA) were purchased from Eurisotope. All other chemicals were purchased from Sigma Aldrich and used without further purification. 1-Octene was degassed via freeze-pump-thaw cycles, d4-MeOH was dried over Mg/I2, methyl formate was dried over K2CO3 then molsieves 3 Å, D2CCl2 was dried over CaCl2 then molsieves 3 Å, and formic acid was dried over anhydrous CuSO4.
To determine the limit of detection for CO gas via NMR different CO solutions in MeOH were prepared. The solutions were made by pressurizing degassed d4-MeOH with CO gas (611, 708, 804, 899 and 1013 mbar). According to Henry's law = where c is the concentration, p the pressure and K the Henry coefficient, the concentration of a gas dissolved in a solvent scales linearly with its pressure above the solution. For the system CO/MeOH the Henry coefficient K at 25 °C is 12100 Pa L mol -1 [2] translating the aforementioned CO pressures to concentrations of 5.0, 5.9, 6.6, 7.4 and 8.4 mmol L -1 , respectively. As normal CO gas was used only 1 % of the gas molecules were NMR active 13 CO molecules. To nevertheless record a reasonable 13 C-NMR signal 1024 scans were used. Subsequently the NMR spectra were normalized to the d4-MeOH signal and the CO signals at 185.4 ppm were integrated from 185.3 to 185.5 ppm. The resulting calibration curve is defined by = • + Where PCO is the integrated peak area, cCO the CO concentration, m the slope of the calibration curve and b the intercept. The calibration curve can be seen in Figure SX. To determine the limit of detection for CO in MeOH a blank d4-MeOH sample was measured. Subsequently the standard deviation σB of the noise integrated inside the same spectral window as the CO signal (0.2 ppm) of the blank sample was calculated. From σB the limit of detection LOD can be calculated by using the calibration curve parameters m and b [3]  This result can be translated to the chosen ex-situ measurement conditions (256 scans, 100 % 13 C-PFA as potential 13 CO source) via the following equation [4] ∝ � 3 0 3 where n is the number of nuclear spins being observed, γe is the gyromagnetic ratio of the spin being excited, γd is the gyromagnetic ratio of the spin being detected, B0 is the magnetic field strength, and t is the experiment acquisition time. As can be seen from the equation the SNR scales linearly with the number of NMR-active nuclei and with a square-root dependency for the measurement time. Thus an enhancement factor can be derived This enhancement factor becomes 50 if the above described experimental conditions are used in the equation and the spectra were recorded on the same NMR spectrometer (the number of scans is proportional to the measurement time): Thus with the chosen parameters NMR spectroscopy is 50 times more sensitive for the detection of CO. With this the LOD for the conducted ex-situ experiments can be directly calculated. The LOD becomes 0.1 mM, 4 weight ppm or 13 mbar when 100 % 13 CO and 256 scans at 25 °C are utilized for the 13 C-NMR experiments.

Data processing
NMR spectra were referenced, background and phase corrected with MestReNova (10.0.2-15465). [5] 1 H and 13 C-NMR spectra were referenced to the residual solvent signal (MeOH in d 4 -MeOH: 3.35 or 49.85 ppm respectively). [6] All NMR spectra were background corrected with a third order Bernstein polynomial. [7] The subsequent kinetic analysis and data visualization was done using R (4.0.2) [8] and the R packages deSolve [9] , FME [10] , RColorBrewer [11] and colorspace [12] . For the kinetic analysis the 1 H and 13 C spectra were normalized to one of the aromatic OTssignals between 7.680 and 7.738 or 129.58 and 129.92 ppm, respectively. The tosylate anion OTsdoes not participate in the catalytic hydroesterification reaction and thus can be used as an internal standard. The following integration borders were used to extract the relative amounts of substance and therefore the concentration profiles: 2. Supporting Information S5

Kinetic model
The concentration data extracted from the 1 H NMR measurements was fitted with the following system of differential equations: where PFA stands for paraformaldehyde, MM for methoxy methanol (H2C(OMe)(OH)), MF for methyl formate, P for methyl nonanoate, O for 1-octene, CO2 for carbon dioxide and DMM for dimethoxy methane (H2C(OMe)2). k1-k5 were extracted by fitting the experimental concentration profiles for each of the seven species with the system of differential equations. As the concentration of PFA was not directly available from the NMR spectra the concentration profile of PFA was simulated by using the previously extracted kinetic rate constant for the PFA depolymerization (0.001997 s -1 ). [13] The concentration of CO2 was also not available from 1 H NMR spectra. Thus, the relative concentration from its 13 C peak was used, adjusted by a semi-quantitative factor of 0.52 to account for its increased 13 C content compared to OTs -(which is used to normalize the 13 C spectra.). To account for the excess of 1-octene used the model was fitted with a modified 1-octene concentration: The final 1-octene concentration at 1080 min was subtracted from the general 1-octene concentrations for the fitting process. For visualization the final 1-octene concentration was added to the experimental and fitted concentration profiles again. All experimental concentration profiles and the resulting fits and the extracted kinetic rate constants are shown in Figure S44.

Quantum chemical calculations
To further investigate the reaction mechanism following the C-H activation of methyl formate, quantum chemical calculations were performed. To model the chemical structure of 1-octene at reasonable computational demand the alkene was approximated by 1-butene. All quantum chemical simulations were performed using the Gaussian16 software. [14] The ground state equilibrium structures and electronic properties were obtained at the density functional (DFT) level of theory utilizing the B3LYP XC functional; [15] i.e. of the substrate  Figure S49). The def2-SVP basis set as well as the respective core potentials were applied for all atoms. [16] A subsequent vibrational analysis was carried out for each optimized ground state structure to verify that a minimum on the potential energy (hyper-)surface (PES) was obtained. All calculations were performed including D3 dispersion correction with Becke-Johnson damping. [17] An analogous computational setup was applied for the optimization of transition states (TSs), while an initial guess in the vicinity of the saddle point was at first obtained via the Nudged Elastic Band (NEB) [18] method as implemented in pysisyphus [19] with xtb. [20] Thereafter, the TSs were obtained in Gaussian16 via the Berny algorithm, [21] followed by a vibrational analysis to verify that a first-order saddle point on the PES was obtained. All optimized structures (xyz files) and the corresponding Gaussian output (log files) can be found online at the open-source repository of the European Commission Zenodo (DOI: 10.5281/zenodo.4153003).

4.
In-situ NMR spectra 13 Figure S49. Reaction sequence as well as the respective Gibbs free energies and activation energies for the alternative alkoxy carbonylation reaction pathway (compare Scheme 3). The activation energy for the final elimination step is very high with 141.1 kJ mol -1 .