Highly selective hydrogenation of amides catalysed by a molybdenum pincer complex: scope and mechanism

A series of molybdenum pincer complexes has been shown for the first time to be active in the catalytic hydrogenation of amides.


Synthesis of N-methyl-N-(4-(methylamino)phenyl)benzamide
H N N O N 1 ,N 4 -dimethylbenzene-1,4-diamine (500 mg, 3.67 mmol, 1.00 equiv) and (445.8 mg, 0.61 mL, 4.40 mmol, 1.20 equiv) were dissolved in 10 mL of dry DCM and subsequently cooled to 0 °C in an ice bath. Benzoyl chloride (515.9 mg, 0.43 mL, 3.67 mmol, 1.00 equiv) was slowly added dropwise over 10 minutes and the reaction mixture was allowed to warm to room temperature overnight. The reaction mixture was quenched with 50 mL of water, stirred for 20 minutes at room temperature and afterwards diluted with 50 mL of DCM. The phases were separated and the aqueous phase was washed two times with 15 mL of DCM. The organic phases were combined, washed with brine and finally dried over Na 2 SO 4 . The drying agent was filtered off and the solvent was removed under reduced pressure. The crude product in form of a greenish solid was purified by column chromatography (Ethylacetate/Heptane, gradient 0:100  100:0) on silica gel to afford the title compound (285.58 mg, 32 %) as a white solid.

Computational Details
General Computational Information.
DFT calculations were carried out with the Gaussian09 software package. 10 The hybrid meta-GGA M06 11 functional was selected on the basis of geometry (Figure S1 Table S1) benchmark, using X-Ray crystal structures as references. Structures were fully optimized without any geometry or symmetry constraints, combining the double-z LANL2DZ (on Mo, including relativistic effects) 12 and 6-31+G** (on all other elements) 13 basis sets. Vibrational frequencies were computed at the same level of theory to classify all stationary points as either saddle points (transition states, with a single imaginary frequency) or energy minima (reactants, intermediates and products, with only real frequencies). These calculations were also used to obtain the thermochemistry corrections (zero-point, thermal and entropy energies) at the experimental p = 50 atm and T = 373 K. The energy of the optimized geometries was refined by single point calculations with triple-z quality basis sets, including the LANL2TZ 12 on Mo and the 6-311+G** on all other elements. 14 The energies reported in the text were obtained by adding the thermochemistry corrections to the refined potential energies. The solvation effects of toluene were included in both the geometry optimizations and energy refinements using the continuum SMD model. 15 The ultrafine (99,590) grid was used in all calculations to increase numerical accuracy and to facilitate convergence. A data set collection of input files and computational results is available in the ioChem-BD repository and can be accessed online via https://iochembd.bsc.es/browse/handle/100/193698. 16 The complex reaction mechanisms inferred from the calculations were interpreted by means of quantitative microkinetic models ( Figure S4, Figure S5 and Table S2), simulated with the COPASI software. 17 Time course simulation were carried with the LSODA algorithm.

DFT functional benchmark
In a previous work of the group, the hydrogenation of amides by an iron (II) Noyori-type bifunctional catalyst was studied by using the M06 functional. 6 This method was selected based on a method benchmark using X-ray geometries and CCSD(T) energies. In order to obtain comparable results, the same functional was initially chosen for this study. This functional was found to give geometries in good agreement with those experimentally obtained for complexes Mo-1a (RMSD = 0.037 Å), Mo-1c (RMSD = 0.031 Å), and Mo-4 (RMSD = 0.030 Å) and therefore was selected for this study. The geometry optimization and energies of the possible spin states for these species were consistent with a doublet for Mo-1a, and a singlet ground state for Mo-1c and Mo-4, respectively.  Figure S1. Mo-complexes used for the geometry benchmark using M06 with the labels used in Table S1, and the corresponding free energies for the first and second excited states. Mo-4 quintuplet did not converge.

Mo-1a
Mo  Table S1. Root mean square deviation of distances (in Å) of optimized geometries with respect experimental single crystal X-ray diffraction geometries, for Mo-1a, Mo-1c and Mo-4 molecules.
Comparison Iron system vs Molybdenum system.
In this work, a mechanism in which a methoxide intermediate is involved in the hemiaminal C-N bond cleavage (Mo-ts-12-13) has been proposed with Mo. This mechanism differs from the one previously proposed with Fe, in which the N of the hemiaminal is coordinated to Fe during the C-N bond cleavage (Fe-ts-C H N H ). We have calculated ts-12-13 with Fe (see Figure  S2) and has a higher energy than ts-C H N H , indicating that the methoxide mechanism is not preferred with Fe.  Figure S2. Computed free energies, in kcal mol -1 , for selected TSs and minima involved in the hemiaminal C-N bond cleavage step with Mo and Fe-systems.
The mechanism of catalyst recovery by addition of H 2 to the methoxide complex Mo-9a is shown in Fig. S3. In this pathway, methanol assists the activation of the Mo-H2 complex (Mo-14) by acting as a proton-shuttle. The global energy barrier for the catalyst recovery mechanism is 23.0 kcal mol -1 , which is similar to the global barrier for the hydride transfer with N-methylformanilide (

Hydrogenation of formaldehyde
The free energy profile for the formaldehyde reduction is represented in Figure S4.  Figure S4 Free energy profile in kcal mol -1 for the formaldehyde hydrogenation to methanol by Mo-5.

S93
Mo-9a and Mo-ts-6-7 with Li + , Na + and K +  Figure S5. Free energies (kcal mol -1 ) for the comparative isodesmic reaction between Li + , Na + and K + in Mo-9a and Mo-ts-6-7 Two microkinetic models were constructed: 1) assuming a barrierless catalyst activation; 2) including a catalyst activation process with an energy barrier estimated to fit the experimental conversions. We have not studied computationally the catalyst activation process due to the complexity and little experimental information obtained for this reaction.

1) N-methylformanilide with EtOH poisoning assuming barrierless catalyst activation.
The N-methylformanilide conversion vs time traces using Mo-5 as catalyst were obtained by running a microkinetic model described below. A concentration of 12.5 mM of Mo-5 was used. The elementary steps of the mechanism underlying the microkinetic model are given in Figure S7 and Figure S8, together with the ΔG ‡ values derived from the DFT calculations in Table S2.

2) N-methylformanilide with EtOH poisoning assuming a catalyst activation.
The N-methylformanilide conversion vs time traces using Mo-1a as catalyst were obtained by running a microkinetic model described below. A concentration of 12.5 mM of Mo-1a and 12.5 mM of NaHBEt 3 were used. The elementary steps of the mechanism underlying the microkinetic model are given in Figure S7 and Figure S8, together with the ΔG ‡ values S96 derived from the DFT calculations in Table S2.

Mo-10
The cation location was determined by computing the energy of selected species (Mo-3, Mo-4, Mo-5, and Mo-9a) with the cation in different positions (interacting with two CO ligands, P 2CO ; or interacting with CO and a lone pair, P CO/LP ; see scheme S9). The location yielding the lowest energy was the one used in the energy profiles, and is the one represented in the Schemes of the manuscript. In most cases, small energy differences (<2 kcal/mol) are obtained when comparing P 2CO and P CO/LP structures.