The elusive abnormal CO2 insertion enabled by metal-ligand cooperative photochemical selectivity inversion

Direct hydrogenation of CO2 to CO, the reverse water–gas shift reaction, is an attractive route to CO2 utilization. However, the use of molecular catalysts is impeded by the general reactivity of metal hydrides with CO2. Insertion into M–H bonds results in formates (MO(O)CH), whereas the abnormal insertion to the hydroxycarbonyl isomer (MC(O)OH), which is the key intermediate for CO-selective catalysis, has never been directly observed. We here report that the selectivity of CO2 insertion into a Ni–H bond can be inverted from normal to abnormal insertion upon switching from thermal to photochemical conditions. Mechanistic examination for abnormal insertion indicates photochemical N–H reductive elimination as the pivotal step that leads to an umpolung of the hydride ligand. This study conceptually introduces metal-ligand cooperation for selectivity control in photochemical transformations.


Supplementary Methods
All experiments were performed under inert conditions using standard Schlenk and glove-box techniques (argon atmosphere). Solvents were purchased in HPLC quality (Sigma Aldrich) and dried using an MBraun Solvent Purification System. THF was additionally dried over Na/K. Deuterated solvents were obtained from Deutero GmbH and dried over Na/K (d6-benzene, THF-d8). Magnesium, potassium hydroxide, lithium aluminum hydride (95%), 13 CO2 and D2 were purchased from Sigma Aldrich and used without further purification. Lithium aluminum deuteride (98% isotopic purity) was purchased from Strem chemicals and used without further purification. Hexamethyldisiloxane (HMDSO) was purchased from Sigma Aldrich, purified by distillation and stored over molecular sieves (3 Å). CO (≥99.997% purity) was purchased from Air Liquide. CO2 was purchased from Linde. [NiBr{N(CHCHPtBu2)2}] was prepared according to the literature.(1) TEMPO-D was prepared according to literature, but with use of deuterated acetone and water.(2) NMR spectra were recorded on Bruker Avance III 300, Avance III 400 or Avance 500 spectrometer with a Prodigy broadband cryoprobe. Spectra were calibrated to the residual solvent signals (C6D6: δH = 7.16 ppm, δC = 128.06 ppm; THF-d8: δH = 3.58 ppm, δC = 67.21 ppm). 31 P-NMR data was referenced externally to phosphoric acid (δ = 0.0 ppm). The following abbreviations were used for signal multiplicities: s (singlet), d (doublet), t (triplet), p (pentet), m (multiplet), br (broad). LIFDI (Linden CMS) mass spectra were measured by the Zentrale Massenabteilung, Fakultät für Chemie, Georg-August-Universität. Elemental analyses were obtained from the Analytisches Labor, Georg-August-Universität using an Elementar Vario EL 3 analyzer. IR spectra were obtained as KBr pellets or in solution with a Thermo Science Nicolet iZ10. EPR spectra were measured using a Bruker ELEXSYS E500 spectrometer, equipped with the digital temperature control system ER 4131VT using nitrogen as coolant. All spectra at 150 K were recorded at about 9.4 GHz microwave frequency and 4 G field modulation amplitude, 100 kHz field modulation frequency, and around 10 mW microwave power. Photolysis experiments were performed using a 150 W Hg(Xe) arc lamp with a lamp housing and arc lamp power supply from LOT-QuantumDesign GmbH. IR irradiation was cut off by a water filter and the photolyzed sample was kept at room temperature by cooling with a water bath. If not stated otherwise, a white-glass filter with a cut off wavelength of 305 nm was used. Gas phase analysis was performed by a Shimadzu GC-2014 equipped with a TCD detector and a molecular sieve 5Å, 80/100 column. Kinetic data analysis was performed using the program package COPASI.  Wavelength dependence of photochemical CO2 activation by 1: 3.3 mg (0.0079 mmol, 1.00 eq) 1 and 2 µL HMDSO are dissolved in 0.5 mL THF-d8 and filled into a J-Young NMR tube. The solution is degassed by three pump-freeze-thaw cycles, 1 atm CO2 pressure (≥99.5% purity, no further purification) is applied and the sample is photolyzed using a white-glass (λexc. > 305 nm) or green-glass filter (>420 nm). Photolysis of 1 under Ar: 4.5 mg (0.011 mmol, 1.00 eq) 1 are dissolved in 0.5 mL C6D6 and 3 µL HMDSO are added as internal standard. The sample is filled in a J-Young NMR tube and the sample is photolyzed (λexc. > 305 nm). 1 H NMR spectroscopy is measured repetitively to determine the conversion of 1. conversion vs. c0 (7) for the photochemical reaction of 1 with CO2. Concentrations are determined by 1 H{ 31 P} NMR spectra by integration vs. the internal standard.
Rate dependence of abnormal CO2 insertion on photon flux: 0.5 mL of a 9.69 mM solution of 1 in THF-d8 containing HMDSO as internal standard are filled into a J-Young NMR tube. The solution is degassed by three pump-freeze-thaw cycles, 1 atm CO2 pressure (≥99.9993% purity, purification by passing through P4O10, Drierite and cooling to -40°C) is applied and the sample is photolyzed (λexc. > 305 nm) with different lamp intensity. Deuteration is determined by comparison of the integrals with the integral of the tBu groups.
Part of the output energy was used to generate pump pulses at 400 nm by second harmonic generation. Pulse energies of about 0.4 μJ focused to a diameter of about 200 μm were used to excite the sample. Tuneable mid-IR probe pulses were generated by difference frequency mixing of idler and signal pulses from a home-build two stage optical parametric amplifier (OPA)(11) pumped by 0.5 mJ of the regenerative amplifier output. The mid-IR beam was split into a reference and a probe beam. The probe pulse passed a translation stage and was superimposed with the pump pulse in the sample cell. To eliminate over-all molecular rotational effects to the signal the relative plane of polarization between pump and probe was set to the magic angle of 54.7°. Probe and reference pulses were directed to a polychromator and separately detected by a liquid-nitrogen cooled HgCdTe-detector of 2 x 32 pixels. To minimize CO2 and water absorptions the mid-IR beam path was purged with dry nitrogen. The hermetically sealed stainless steel sample cell equipped with two CaF2 windows of 1 mm thickness and a magnetic stirrer had an optical path length of 0.6 mm. The cell was filled under argon atmosphere with 11 mM sample solutions in THF-d8. photolyzed for a certain time with a certain lamp intensity. A 1 mL aliquot was taken, 1 mL of a 0.0055 M solution of 2,2'-phenanthroline was added and the reaction was quenched using 1 mL of a Na2OCCH3 solution at pH =3.5 (pH was adjusted using H2SO4). The sample was then diluted to a total volume of 10 mL and left in the dark for 1 hour to equilibrate. Afterwards a UV/Vis spectrum was recorded.
The photon flux of the lamp was determined using the following equation: Where I is the photon flux, A is the absorbance of the photolyzed sample corrected for the hours with an applied current of 7 A and therefore a photon flux of 4.782x10 -7 mol min -1 .
Afterwards an aliquot of the photolyzed solution was taken and a 31 P{ 1 H} NMR spectrum containing the PPh3 solution as internal standard was measured to determine the conversion of 1 by comparison with the spectrum recorded before photolysis. 46% conversion (2.456x10 -8 mol/min) could be observed giving a quantum yield of 5.3% by the following equations: Structure optimizations and free energies: All calculations were performed within the ORCA program package. (13)(14) The molecular structures were optimized using the PBE(15) functional, Grimme's dispersion correction with Becke-Johnson damping (D3(BJ)) ( [16][17] and the Resolution of Identity (RI-J) (18)(19)(20)(21) approach to minimize computational costs and Ahlrichs' revised all-electron def2-SVP basis set and the corresponding def2/J auxiliary basis set. (21)(22)(23)(24) Tight convergence criteria in the SCF procedure and optimization and a fine integration grid (Grid 5) were applied throughout.
No symmetry restrains were imposed and the optimized structures were defined as minima (no G sol = G gas + RTln(24.47) G sol = G gas + 1.89 kcal/mol (23) Therefore, a correction of 1.89 kcal/mol results which has been added in the steps that involve coordination or release of CO2. The G value of the THF molecule has been further corrected by applying the actual concentration in the pure solvent which is ( = 0.889 g/ml (25°C), c = 12.3 mol/l): G′ sol = G solv + 1.49 kcal/mol (25) That gives an overall correction factor of 3.38 kcal/mol for THF.
The total free energies have thus been obtained by addition of the single point SCF energies with the free energy contributions from the thermal analyses at the D3(BJ)-RI-J-PBE/def2-SVP level and the above mentioned thermal corrections.

DLPNO-CCSD(T) benchmark studies:
To assess the quality of the DFT calculations, domain-based local pair natural orbital based coupled cluster (DLPNO-CCSD(T)) single point calculations (28)(29)(30)(31)(32) have been performed for selected molecules employing a truncated model system in which the tert-butyl groups of the pincer ligand have been replaced by methyl groups. The model complexes were fully optimized on the D3(BJ)-RI-J-PBE/def2-TZVP level and resemble strongly the evaluated structures of the full system. In the DLPNO-CCSD(T) calculations, we applied the all-electron correlation consistent cc-pVTZ and cc-pVQZ basis sets (33)(34)(35)(36) in combination with the cc-pVTZ/C and cc-pVQZ/C density fitting basis sets (37)(38)(39) and tight SCF convergence criteria. The complete basis set (CBS) limit has been estimated for the Hartree-Fock energy from the SCF energy of the cc-pVQZ basis ( ) by (40): Where  is 5.46 (41) and A has been extrapolated from the SCF energies obtained with the cc-pVTZ ( ) and cc-pVQZ bases: 3 and 4 represent the cardinal numbers of the basis sets cc-pVTZ and cc-pVQZ, respectively.
The correlation energy has been extrapolated to the complete basis set limit by (42) (47) were also included as a correction within the scalar relativistic framework. The density-fitting procedure was used within the RIJCOSX approximation (48) in conjunction with the corresponding def2-Coulomb fit basis sets (49) and large integration grids (Grid6 and GridX6 in ORCA convention). Grimme's dispersion correction with Becke-Johnson damping (D3(BJ)) was employed as well.
The UV/Vis spectra of selected compounds were obtained at the same level but employing the B3LYP functional (50) within the time-dependent DFT framework as implemented in ORCA.
In contrast to the EPR calculations, spin-orbit coupling has been neglected but solvent effects were included by application of the COSMO model (THF, = 7.25, n= 1.407).
It is well known that the PBE0 and the B3LYP functionals perform well for the calculation of EPR and UV/Vis data, respectively (51).        (3), CCDC 1561990 (4), CCDC 1561991 (5), CCDC 1561992 (6), CCDC 1561993 (7), CCDC 1561994 (9), CCDC 1574302 (10)  Suitable single crystals for X-ray structure determination were selected from the mother liquor under an inert gas atmosphere and transferred in protective perfluoro polyether oil on a microscope slide. The selected and mounted crystals were transferred to the cold gas stream on the diffractometer. The diffraction data were obtained at 100 K on a Bruker D8 three-circle diffractometer, equipped with a PHOTON 100 CMOS detector and an INCOATEC microfocus source with Quazar mirror optics (Mo-Kα radiation, λ= 0.71073 Å).
The data obtained were integrated with SAINT and a semi-empirical absorption correction from equivalents with SADABS was applied. The structure was solved and refined using the Bruker Ni-H hydrogen atom was found from residual density map and isotropically refined. The asymmetric unit contains only a half complex molecule.
Supplementary Table 19. Crystal data and structure refinement for 1.
Disorder of O3 and O3a was refined with fixed population of 0.5 for both sites using PART and EADP commands.

Supplementary
Supplementary