Encapsulation Enhances the Catalytic Activity of C‐N Coupling: Reaction Mechanism of a Cu(I)/Calix[8]arene Supramolecular Catalyst

Abstract Development of C−N coupling methodologies based on Earth‐abundant metals is a promising strategy in homogeneous catalysis for sustainable processes. However, such systems suffer from deactivation and low catalytic activity. We here report that encapsulation of Cu(I) within the phenanthroyl‐containing calix[8]arene derivative 1,5‐(2,9‐dimethyl‐1,10‐phenanthroyl)‐2,3,4,6,7,8‐hexamethyl‐p‐tert‐butylcalix[8]arene (C8PhenMe6 ) significantly enhances C−N coupling activity up to 92 % yield in the reaction of aryl halides and aryl amines, with low catalyst loadings (2.5 % mol). A tailored multiscale computational protocol based on Molecular Dynamics simulations and DFT investigations revealed an oxidative addition/reductive elimination process of the supramolecular catalyst [Cu(C8PhenMe6)I]. The computational investigations uncovered the origins of the enhanced catalytic activity over its molecular analogues: Catalyst deactivation through dimerization is prevented, and product release facilitated. Capturing the dynamic profile of the macrocycle and the impact of non‐covalent interactions on reactivity allows for the rationalization of the behavior of the flexible supramolecular catalysts employed.

C8Phen was obtained according to the previously reported protocol, 2 with modifications in the purification procedure. Previously synthesized p-tert-butylcalix [8]arene (obtained according to the literature procedure, 3 0.80 g, 0.62 mmol) and CsF (0.94 g, 6.22 mmol) were dried in a round-bottom Schlenk flask at 120 °C for 2 h. After cooling to room temperature, the solids were dissolved in 15 mL of THF, the formation of a white suspension was observed. The temperature was increased to 50 °C for 12 h while stirring, resulting in a yellow solution. PhenBr2 (0.28 g, 0.73 mmol) was then added, and the mixture was stirred at room temperature for 36 h. Then, volatiles were evaporated under reduced pressure, and the crude solid was dissolved in 10 mL of chloroform/toluene (10:1). The organic phase was washed with 15 mL of 0.1 M HCl, followed by 50 mL of saturated NaHCO3 solution, and extracted with 20 mL of chloroform. The organic phase was dried over CaCO3, and after slow solvent evaporation an orange solid was obtained. The product was purified by column chromatography on silica gel with dichloromethane as eluant, and the product was washed with hexanes, resulting in a white crystalline solid in 86% yield (0.78 g, 0.52 mmol), m.

S5.1.1 Free Energy Calculation
Harmonic frequency calculations were performed on the optimized structures to verify that they are indeed energy minima. The Goodvibes python script 5 was used to obtain the zero-

S5.1.2 Molecular Dynamics Simulations
Structures were equilibrated using a modified procedure by Wallnöfer at al. 7 that involves extensive heating and cooling to achieve equilibration. The production runs (in explicit chloroform) were carried out in NpT ensembles at 300 K, using Amber18. 8 Temperature was regulated with the Langevin thermostat, 9 whereas the pressure was kept at 1 bar using the Berendsen barostat. 10 The SHAKE algorithm 11 was used to restrain hydrogen bonds allowing for a time step of 2 fs; coordinates were saved every 10 ps, simulating a total of 1 µs. Accelerated MD (aMD) simulations 12 were carried out using the dual-boost algorithm implemented in Amber18, 8 where a bias was applied on the total potential and an additional boost on the dihedral term. 8 A total of 12 aMD simulations, each with 1 µs, were performed with various boosting parameters. These settings were derived as proposed by Pierce et al. 13 by performing 100 ns classical MD simulations. All trajectories of the production runs were combined, and the structures were aligned on the phenanthroyl bridge. A hierarchical clustering was applied on the combined trajectories to obtain a structurally divers ensemble.
We did not re-weight the trajectories because (i) accurate re-weighting is difficult to achieve within aMD and energy minima remain such regardless of their relative energies, (ii) our goal was to obtain structures that are as diverse as possible, since relative energies of the conformers will change upon Cu(I) coordination.

S5.2 Computational protocol to determine the dissociation transition state TS3-4
The transition state TS3-4 presents a particular difficulty in obtaining a valid structure, as the electronic energy is rising in a monotonic fashion, as the dissociation coordinate is elongated. While this makes the determination of the exact structure difficult, the energy barrier of this TS can still be estimated. By manually creating structures along the reaction coordinate, 0.25 Å apart, and performing a restrained optimization as well as frequency calculations, the initial steps of the reaction can be characterized ( Figure S44). When another datapoint is added, with the product structures infinitely separated, a jump in free energy can be observed. By fitting a sigmoid function on the values of -TΔS (black dotted line), the free energy along the dissociation (red line) can be estimated from adding the calculated enthalpy (blue dotted line) to -TΔS (black dotted line). The highest point is taken as the estimated for TS3-4. The approximated transition-state structure is depicted in Figure   S50.

S5.3 Structures of reactive species
The transition states TS2-3 (proton abstraction) and TS4-5 (iodide abstraction) are diffusion controlled, hence, a barrier of 20 kJ/mol is assumed and no structures are reported.

S5.4 Subsequent cycles of the reaction
The first cycle of the reaction begins with iodine and ends with bromine attached to the Cu atom. All further cycles of the reaction take place starting with the Br analogue of structure.
An investigation of the electronic energy path of the Br equivalents of structures 1 to 4 is shown in Figure S57. The reaction energies are largely similar, with the main differences being in the case of 3 and 4 with the bromine intermediates being 7 and 20 kJ mol -1 higher in energy.

S5.5 Investigation of the supramolecular complexe
The supramolecular catalyst is a difficult structure to fully investigate, as QM calculations are computationally intensive and do not take into account the flexibility of the calixarene ring. MD simulations can not only describe the flexibility of the cage, but they allow for the inclusion of explicit solvent, which is essential for preventing the calixarene cavity from collapsing. This comes at the cost of an inability to take into account any bond formation or dissociation. To this end, we developed a protocol, as shown in Figure S58, where we combine MD simulations with QM optimizations, in such a way that we are able to obtain snapshots along the reaction pathway. By utilizing DFT for elucidating the reaction path on    Figure S59), it is evident that the structures are very similar. Indeed, even CuI coordination yielding 1calix (bottom panel in Figure S59) did not change much, but very similar clusters were obtained for the simulation of 1calix in chloroform compared to simulation of the ligand alone with cavities of similar size. As results for chloroform and toluene were similar, all subsequent calculations were performed in chloroform only to reduce computational costs.
As the simulations yielded well-defined cavities, fitting of structures 2 to 7 deemed possible without too many clashes. To this end, structures 2 through 7 were fitted into the cavity, using the phenantroyl bridge as a base for alignment, yielding 2calix through 7calix. Using the number of steric clashes as the criterion, each fitted structure was ranked. The most populated cluster of the simulation of 1calix showed the least amount of clashing and thus was chosen as the template for subsequent optimizations. The resulting structures were subsequently optimized with PBE0/def2-SVP/D3+COSMO, followed by single point calculations with the larger def2-TZVP basis set and the energetic landscape was evaluated. Final structures are depicted in Figure S60.
As simulations of structure 1calix as well as 6calix were performed, the flexibility of the calixarene ring was also investigated. It was observed that the calixarene units exhibits less movement during the simulation of 6calix. This was quantified as the relative entropy of the diherdrals between the individial units calculated with the X-entropy script. 14 In Figure S61, it can be seen that the distributions of the dihedrals in the simulated structure 6calix are considerably narrower than those of 1calix.