Catalytic Reversible (De)hydrogenation To Rotate a Chemically Fueled Molecular Switch

Abstract We report here the development of a rotating molecular switch based on metal‐catalyzed reversible (de)‐hydrogenation. Under an argon atmosphere, acceptorless dehydrogenation induces a switch from an alcohol to a ketone, while reversing to a hydrogen pressure switches back the system to the alcohol. Based on a tolane scaffold, such reversible (de)‐hydrogenation enables 180° rotation. The absence of waste accumulation in a switch relying on chemical stimuli is of great significance and could potentially be applied to the design of efficient complex molecular machines.


(2-ethynylphenyl)(phenyl)methanone:
According to a modified procedure 4 an 50 mL round bottom flask was charged with phenyl(2-((trimethylsilyl)ethynyl)phenyl)methanone (125 mg, 0.45 mmol, 1.00 equiv) in MeOH (4 mL) under argon atmosphere. Then K2CO3 (62.0 mg, 0.45 mmol, 1.00 equiv) was added at once and the reaction mixture was allowed to stir for 90 min. H2O (10 mL) was added to the reaction mixture, followed by a precipitate formation, then the solid was filtered on a fritted funnel to remove the solution. The precipitate was dissolved in EtOAc (20 mL) dried over Na2SO4, filtered and the solvent evaporated and the crude product S8 was used without further purification. The product was isolated as a dark solid (92.0 mg, 0.45 mmol, 99%).

(4-ethynylphenyl)(phenyl)methanone:
According to a modified procedure 4 an 50 mL round bottom flask was charged with Phenyl(4-((trimethylsilyl)ethynyl)phenyl) (278 mg, 1.00 mmol, 1.00 equiv) in MeOH (9 mL) under argon atmosphere. Then K2CO3 (138.5 mg, 1.00 mmol, 1.00 equiv) was added at once and the reaction mixture was allowed to stir for 90 min. H2O (30 mL) was added to the reaction mixture, followed by a precipitate formation, then the solid was filtered on a fritted funnel to remove the solution. The precipitate was dissolved in EtOAc (60 mL) dried over Na2SO4, filtered and the solvent evaporated, the crude product was used without further purification. The product was isolated as a dark solid (206 mg, 1.00 mmol,
TLC (cyclohexane/EtOAc 9:1) Rf 0. When performing the hydrogenation on this substrate using the iridium catalyst, rapid decomposition was observed.

Molecular switch through multiple cycles:
To a 10 mL Schlenk equipped with a cooler ( Flushing is mandatory to perform a rapid equilibrium between the gas/solution phase. S14 NMR monitoring of the molecular switch cycles: Figure S2: 1 H NMR monitoring of switch cycles. Experiment starts at t = 0, with 100% 2alcohol and 0% 2ketone.

Effect of catalyst deactivation over multiple cycles:
To a 10 mL Schlenk equipped with a cooler (

NOESY analysis of 2alcohol and 2ketone:
NOESY NMR sequence were recorded in CDCl3 solutions at 300K using a Bruker AVANCE III 600MHz spectrometer equipped with a Triple Resonance 1 H/ 13 C/ 15 N 5mm TXI probe with Z-gradient using Bruker Topspin 3.5.PL7 software version. The NMR sequence was "noesygpphzs" . The NOESY spectrum was obtained with an F2 spectral width of 10ppm and 2K data points and an F1 spectral width of 512 t1 increments and 2 scans. Therefore, to reduce data acquisition time, we implemented NUS (50% random sampling). The spectra were processed by IRLS (iteratively re-weighted least squares) protocol, with SI2=4K and SI1=4k and pure cosine squared sine window functions applied to both dimensions.

S16
On the full NOESY spectrum of 2alcohol, two signals were found to be important for the conformational determination ( Figure S). Indeed, proton 16 can perceive in its spatial environment protons 8 and proton

1.
A strong NOE signal is observed for the 16/8 interaction, while only a slightly distinguishable signal is observed for the 16/1, confirming the predominance of the conformation where the alcohol function is on the same side as the NMe2 moiety. Figure S4: NOESY spectra of 2alcohol.
For 2ketone, the NOESY spectra did not provide any additional interactions between the lower and upper part of the switch that could be used for the conformational analysis of the compound (Figure S). S17 Figure S5: NOESY spectra of 2ketone.

NMR Analysis of enantiopure alcohols:
Enantiomers of the molecular switch 2alcohol were obtained by chiral HPLC separation.

NMR analysis of 2ketone and 2alcohol at 300 K, 314 K and 348K.
NMR spectra were recorded on a Bruker 500 (500 MHz) spectrometer at 3 different temperatures.
In both 2ketone and 2alcohol temperature elevation has induced a partial rotation of the benzamide moiety ( Figure S).

IR analysis and Theoretical calculations Experimental details
The infrared (IR) spectra were measured on a Bruker VERTEX70 FTIR spectrometer. A transmission cell equipped with BaF2 windows and of 200 μm of optical pathlength was used. Solutions were prepared by dissolving the solid samples in CD2Cl2 or in C6D6. Each spectrum was recorded using 30 scans with a scan frequency of 10KHz. The apodization function used is Blackman-Harris 3 terms. The spectrometer is continuously purged with dry CO2-free air. The cell filled with solvent served as a reference. The temperature has been kept constant at 298K.

Theoretical calculations
Calculations were performed on 2alcohol and 2ketone. The conformational study was done using a stochastic exploration of the potential energy surface (PES) using simulated annealing with RM1 semiempirical level as implemented in Ampac10. 12 For each molecule, a set of almost 40 geometries of conformations were generated with various simulated annealing, each done with a different geometry or parameterized with a different initial temperature. A geometry optimized with SMD(C6D6)/B3LYP/6-311++G(d,p) level has been used as starting structure. Only the dihedral angles of this initial geometry were allowed to relax during the annealing, the bonds lengths and the valences angles were kept constant. Then, the conformations with energy lower than 2.5 kcal mol -1 compared to the lower energy conformation were kept and fully optimized. The geometry optimizations, vibrational frequencies, IR absorption and VCD intensities were calculated with Density Functional Theory (DFT) using B3LYP functional combined with 6-311++G(d,p) basis set. The average solvent effects were modeled implicitly by the dielectric continuum model SMD implemented in Gaussian 16. 13 The vibrational frequencies and IR absorption intensities were calculated using the same theoretical level as for geometry optimization SMD(C6D6 or CD2Cl2)/B3LYP/6-311G++(d,p). Computed harmonic frequencies are generally larger than those experimentally observed. They have been calibrated using a standard scaling factor of 0.98. IR absorption and VCD averaged spectra were constructed from calculated dipole and rotational strengths assuming Lorentzian band shape with a half-width at half maximum of 8 cm -1 .
All calculations were performed using Gaussian 16 package. 13

Analysis of 2alcohol: Solvent effect on 2alcohol
IR spectra of 2alcohol were measured in two solvents: C6D6 and CD2Cl2 ( Figure S14). The C=O stretching band shifts by 6 cm -1 to the far infrared region when the solvent changes from C6D6 to CD2Cl2. This shift is mostly due to the difference in polarity between these two solvents.
On the IR spectrum of the racemic mixture in C6D6, a shoulder is observed at 1675 cm -1 . This shoulder is less intense for the enantiopure molecule in C6D6 and disappears in CD2Cl2. Two hypotheses can explain the presence of this additional stretching C=O band: a conformational effect (i.e there are at least two populated conformations for which the C=O group is in sufficiently different environments to induce a shift in the vibration frequency) or the presence of dimers in equilibrium with the free molecule.

Concentration effect and DFT computations
On figure S15 are displayed the IR spectra of the racemic mixture measured in C6D6 for various concentrations obtained by successive dilutions of a primary solution of concentration 0.1 mol.L -1 . In this concentration range, no significant modification of the spectra is observed. This result further confirms that the shoulder observed on the spectra measured in C6D6 would result from a conformational effect. The conformational analysis of the free molecule carried out in DFT in both solvents C6D6 and CD2Cl2 leads to the same two major conformations A1 (or B1) and A2 (or B2) (Tables S2 and S3, Figure   S16). In CD2Cl2, there are also two other minor conformations A3 and A4.  The conformations A1 (or B1) and A2 (or B2) are geometrically very close and owe their greater stability to the synergy of several stabilizing interactions (OH...NMe2 hydrogen bonding, π-stacking and Van der Walls interactions). Among the conformations found in our analysis, we find that those for which there is hydrogen bonding between the alcohol and amide functions (A7 for instance) have a calculated enthalpy much higher than that of the most stable conformation and are therefore not populated (figure S15 and table S2). Regarding these results, a conformational effect could only occur from an explicit interaction between the solute and the solvent (this effect not being taken into account in the theoretical SMD model used which treats the solvent implicitly). If we cannot exclude this possibility, the comparison of the measured and calculated spectra is more in line with the hypothesis of an equilibrium between free molecules and dimers. Indeed, the resulting average spectra calculated in CD2Cl2 and in C6D6 are similar and consistent with the experimental measurements. In particular, the solvent effect observed experimentally at the C=O elongation band is correctly predicted by the calculations (Figures S17 and   S2). We can see the shift towards the far IR spectral range of the C=O stretching band in CD2Cl2. The calculated shift is however larger than the measured one which suggests that the calculation method used seems to slightly overestimates this effect. In the spectral range 1100-2000 cm -1 a satisfying agreement is obtained between the measured and the calculated spectra in CD2Cl2 for the free molecule (figure S18). Similarly, the measured and the calculated IR spectra for the free molecule in C6D6 are also in good agreement (figure S19). It is therefore reasonable to conclude that the dominant conformations in solution are those in which a hydrogen bond is formed between the alcohol function and the dimethylamino moiety (figure S16).
The calculated average IR spectrum was built from the weighted sum of the spectra of the conformations that were found. The weighting coefficient is the Boltzmann population of each conformation, calculated using the enthalpies at 298 K from the DFT calculations (figure S17 and table S2).  This result allows us to assume that the free molecule is the predominant state in solution. Despite πstacking interactions with a C6D6 molecules cannot be ruled out, it is reasonable to assume that they are less favorable than dimer formation. In order to support this hypothesis, we have optimized the S31 geometry of a dimer in C6D6, extracted from the structure obtained by single crystal X-ray diffraction (figure S18a). For these calculations the dispersion potential GD3BJ of Grimme's dispersion with Becke-Johnson damping was introduced to improve the description of the dimer. For a consistent comparison, GD3BJ was also introduced for the calculation of the free molecule spectrum. The comparison of the IR spectrum calculated for this dimer geometry with the one calculated for the free molecule (figure S18b) shows that in the presence of an equilibrium between these two forms of the molecule and in the hypothesis of an equilibrium shifted to the free molecule, only the position of the C=O elongation band would distinguish them. This is consistent with the experimental observation. Due to the polarity of dichloromethane, the formation of dimer is less favorable than in C6D6, which is what we observe on the IR spectra measured in this solvent.

Analysis of 2-ketone:
The same study was carried out with 2ketone in C6D6. The conformational analysis of this molecule led us to calculate 13 conformations (table S4). Among these conformations, 5 have calculated Boltzmann populations higher than 5% and were thus retained to build the calculated IR spectrum ( figure S21).
According to the interactions between the ketone function and the benzamide moiety, we can classify these 5 conformations. Firstly, there are the conformations C1 and C3 characterized by a hydrogen bond between the ketone function and the NH group of the amide as well as by a CH-π type interaction between a CH of the phenyl moiety of the ketone and the π system of the benzamide moiety. The C2 and C4 conformations are characterized by a double interaction of the ketone function with the NH and a C-H(benz.) of the benzamide moiety. Finally, the least populated conformation C5 is characterized by the fact that the ketone function has no interaction with the benzamide moiety. The comparison of the measured and calculated IR spectra shows a satisfactory agreement ( Figure   S22). As before, it is reasonable to conclude that the dominant conformations of 2ketone in C6D6 solution are those in which the ketone interacts with the benzamide moiety although the calculations predict a minor conformation (<1%) without this kind of interactions.