A Well‐Defined Anionic Dicopper(I) Monohydride Complex that Reacts like a Cluster

Abstract Low‐nuclearity copper hydrides are rare and few well‐defined dicopper hydrides have been reported. Herein, we describe the first example of a structurally characterized anionic dicopper hydride complex. This complex does not display typical reactivity associated with low‐nuclearity copper hydrides, such as alcoholysis or insertion reactions. Instead, its stoichiometric and catalytic reactivity is akin to that of copper hydride clusters. The distinct reactivity is ascribed to the robust dinuclear core that is bound tightly within the dinucleating ligand scaffold.


Structure
Reference Goeden et al. 1986 [1] Cu Frey et al. 2011 [3] Wyss et al. 2013 [4] Nakamae et al. 2014 [5] Zall et al. 2016 [6] Zhang et al. 2016 [7] Romero et al. 2017 [8] Jordan et al. 2016 [9] Tran et al. 2020 (crystal structure) [ Speelman et al. 2021 [11] These structures were found with a Conquest search of the CCDC database. [12] A combination of two queries was used. The first is of the type "drawn", where two copper atoms bridged by a hydrogen atom that is connected to each Cu atom with bond type "any" were searched. The second query is of type "formula" and searches for Cu2 (excluding any structure with 1 or more than 2 copper atoms). Structures with transition metals other than copper in the Cu2H core were omitted as well as structures with bridging borohydrides. This search did not yield the 2004 structure of Sadighi and coworkers [2] due to the absence of the hydride ligands in the crystallographic data. Figure S1: Spacefill model of complex A generated from the reported crystal structure. [13] Hydride Experimental methods

General considerations
All manipulations were performed under N2 atmosphere in a glovebox or using Schlenk techniques, with anhydrous reagents unless mentioned otherwise. Ambient temperature in our glovebox is on average 27°C. Glassware was oven dried at 130°C overnight or flame dried before use. Solvents except THF were collected from an SPS system and degassed by bubbling N2 gas through it, subsequently they were dried over molecular sieves. The THF was distilled over Na/benzophenone ketyl (purple), subsequently it was degassed by bubbling N2 gas and further dried over molecular sieves. Water content in the solvents was tested using Karl-Fisher titration and titration with a ketyl solution. Deuterated solvents were degassed using three freeze pump thaw cycles and dried over molecular sieves. The PNNP ligand as well as complex A and [ t-Bu PNNP**Cu2Mes]K18-crown-6 were prepared according to literature procedure. [13,14] HBArF24•2Et2O (Brookhart's acid) was prepared according to literature procedure. [15] All commercial reagents were obtained from Sigma Aldrich or Acros, liquids were degassed using three freeze pump thaw cycles and dried over molecular sieves before use. 4chlorobenzaldehyde was sublimed to eliminate the benzoic acid oxidation product and subsequently stored in the glovebox. NMR spectra were recorded on an Varian MRF 400 equipped with a OneNMR probe and Optima Tune system or a Varian VNMR-S-400 equipped with a PFG probe. All resonances in 1 H NMR were referenced to residual solvent peaks. [16] IR-data was recorded on a PerkinElmer SpectrumTwo Infrared Spectrophotometer equipped with an ATR-probe. Elemental analysis was performed by MEDAC ltd. in the United Kingdom.

Synthesis of complex 1
A solution of KO t Bu (11.4 mg, 94.1 µmol, 2.3 eq) and 18-C-6 (24.7 mg, 93.4 µmol, 2.3 eq) in THF (2 mL) was added dropwise to a stirred red solution of A (47.5 mg, 41.4 µmol) in THF (5 mL) at ambient temperature. The reaction mixture gradually turned orange and was left to stir for one hour. Subsequently, the solvent was evaporated and the residual orange solid was washed with n-pentane (7 ml). The residue was extracted with THF until washings were colorless (~7 mL), extracts were combined and all volatiles were evaporated under a dynamic vacuum to give 1 as an orange solid in 75% yield (63.3 mg, 61.9 µmol). Complex 1 can be stored as a solid at -40°C under N2 atmosphere for 2-3 weeks without a visual change in color and morphology. The NMR spectra of 1 contain a minor impurity of A, likely formed during sample preparation due to the high sensitivity of 1.
Single crystals of 1 suitable for X-ray diffraction were grown by vapor diffusion of pentane into a concentrated THF solution of the compound.

1D
The deuterated analogue of A (AD) was prepared according to literature procedure. [13] Without isolating AD, 18-crown-6 (19.2 mg, 72.6 µmol, 2.4 eq) and KOtBu (7.8 mg, 69.5 µmol, 2.3 eq) in 0.2 ml THF were added dropwise to a THF (0.5ml) solution of AD (1eq, 30.9 µmol). The mixture was left to stir for 20 mins and was subsequently analyzed by 2 H NMR spectroscopy.  Exposure of complex 1 to air Complex 1 (5.6 mg, 5.5 µmol) was dissolved in THF (0.6 ml) and sealed in a J-Young tube. On the Schlenk line, the solution was degassed with three freeze-pump-thaw cycles. The tube was backfilled with air by opening the tube and closing it again. NMR of the solution shows mostly the peaks reported for A along with some smaller unidentified species which are hypothesized to be oxidation products.  Complex 1 (1 eq, 4.8 mg, 4.7 µmol) was dissolved in 0.65 ml THF and transferred to a J-Young tube. D2SiPh2 (4.6 eq, 4 µl, 21.6 µmol) was added to the tube and the solution was mixed by shaking. After 1.5 hours the mixture was analyzed by 1 H and 2 H NMR spectroscopy, which showed the formation of 1D concomitantly with the formation of HDSiPh2.  Complex 1 (1 eq, 20.0 mg, 19.6 µmol) was dissolved in THF (1.5 ml), and HBArF24•2Et2O (1 eq, 21.2 mg, 19.6 µmol) in THF (1.5 ml) was added dropwise to the stirring solution of 1. During the addition, the mixture turned from orange to red. Both 1 H NMR and 31 P NMR showed the formation of complex A, the NMR data of A matches those reported in literature. [13] Figure S12: 1 H NMR spectrum of the product of 1 with 1eq HBArF24 in THF-h8 at 25°C with PRESAT solvent suppression for the THF peaks. Figure S13: 31  According to modified literature procedure [17] , benzhydrol (204.99 mg, 1.11 mmol, 1 eq) was dissolved in THF (15 ml). Diphenylchlorosilane (0.22 ml, 1.13 mmol, 1 eq) was added dropwise to the stirring solution of benzhydrol. The mixture was stirred for 3 hours after which NMR showed only minimal conversion. Triethylamine (0.15 ml, 1.08 mmol, 0.95 eq) was added dropwise to the stirring solution, which led to the immediate formation of a white precipitate. The solvent was evaporated and the white solid was extracted with hexane (6 ml). The extract was dried and recrystallized from 3.3 ml hexane at -40 °C yielding a white powder (48%, 195.9 mg, 0.53 mmol). NMR analysis is consistent with the reported values. [18]  Complex 1 (15.1 mg, 14.7 µmol) was dissolved in THF-d8 (0.6 ml) and transferred to a J-Young tube. 4fluoro phenylacetylene (1.7 µL, 14.7 µmol, 1 eq) was added. The tube was closed. The sample was heated stepwise and monitored with NMR. First at 40°C overnight, then 3 hours at 50°C, then 60°C overnight, then 70°C for 3 hours and finally 80°C overnight.

Reaction of 1 with CO2
Complex 1 (5.2 mg, ) was dissolved in 0.6 ml THF and transferred to a J-Young tube. The tube was attached to a gas addition setup and degassed with three freeze-pump-thaw cycles. The tube was filled with CO2 at room temperature. The mixture was analysed with NMR after 1 hour, 1 day and 2 days. After two days, a colourless precipitate had formed which was filtered off, washed with THF extensively until no orange colour was visible, and dissolved in D2O showing the presence of formate.
We hypothesize that initially, the CO2 is attacked by the backbone leading to the resonances at 4.70 and 4.75 ppm (dashed circle) after 1 hour. Resonances at similar position have been reported for a related mononuclear PNP CO2 complexes. [19][20][21] After this, conversion to A and an unidentified species that also contains a partially dearomatized ligand backbone is observed. Moreover formate (presumably KO2CH) starts visually precipitating, which was confirmed by 1H NMR analysis of the precipitate ( Figure S19). The composition of the solution after 2 days does not change further, even after heating analogous samples at 80 °C.

Proposed mechanism for ketone hydrosilylation
From the deuterium labeling experiments it can be inferred that in the mechanism, the CH on the former carbonyl position is coming from the silane and not from the hydride in 1. Hence, if the silane is activated, the hydride from the incoming silane cannot become chemically equivalent to the hydride of 1. Taking inspiration from the proposals by Bellemin-Laponnaz and coworkers [22] , we considered a modified Ojima mechanism and a hypervalent silicon mechanism ( Figure S27). The first is the bimetallic Ojima mechanism in which the first step is a bimetallic oxidative addition of the Si-H bond over the two copper centers. Considering the sterics of the formed intermediate of this cis addition (A), it seems unlikely that a carbonyl would insert into the Cu-H bond of the hydride coming from the silane. Therefore we propose that a hypervalent silicon mechanism is most likely for 1 . Variations of this type of mechanism have been proposed for copper and other transition metal based catalysts. [22,23] Also base catalyzed hydrosilylation of ketones are thought to follow this mechanism. [24] The first step (A) of the hypervalent silicon mechanism is the nucleophilic attack of the hydride of 1 on silicon, forming a pentavalent Si intermediate. This is followed by the coordination of the ketone to Si (B) and the insertion of the ketone into the Si-H bond (C).
Step D closes the cycle by releasing the siloxane product. At this point we cannot exclude either pathway, but future research is aimed at elucidation of the operational mechanism.

General remarks
Calculations were performed using Gaussian 16 rev. C01 software. [25] The Becke 1988 exchange functional [26] was used in combination with the Perdew 1986 correlation functional(BP86) [27] . The redefinition of Ahlrichs triple-zeta split valence basis set (def2-TZVP) was used on all atoms. [28] No empirical dispersion was employed since this has been shown to provide less accurate results for A due to an overestimation of the attractive dispersive energy, which is likely caused by the steric crowding in that molecule. [13,29] NBO [30] calculations were performed using the NBO 6.0 software. [31] Natural population analysis [32,33] , Mulliken population analysis [34] and computation of Wiberg bond indices [35] was also performed using the NBO 6.0 software. QTAIM analysis [36] and visualization was performed using the MultiWFN program. [37] Generating QTAIM basins and integration of the electron density in them was performed with MultiWFN using a 'high grid' (with a spacing of 0.06 Bohr). Starting geometries for the optimizations were obtained from the coordinates of the crystal structures if possible, or by modification of the optimized geometry of the most similar complex.

Geometry optimizations
The calculated structure of 1 is a close match in geometry with the crystal structure (figure S28). The Cu-H bond lengths are 1.649 Å, which is the same distance found in complex A for the copper centers on the dearomatized side of the naphthyridine pocket. [13] The highest occupied molecular orbital of 1 is located mainly on the backbone (figure S29). The position of the hydride in complex 1 was probed with a relaxed potential energy surface scan to verify that there is not a secondary local minimum corresponding to a terminally bound hydride. The Cu1-Cu2-H34 angle was scanned from its original position (42.5°) in 12 steps of 5°C to its final position of 97.5°. The scan shows a constant increase of energy, confirming the absence of a secondary minimum ( Figure S30).

Calculated NMR shifts
The 1 H shifts of the hydride ligand in 1 where calculated and a predicted chemical shift is 2.67 ppm was found. This is ~1.8 ppm higher than the experimental value. This difference can be ascribed to the lack of relativistic treatment at the level of theory used (BP86/def2-TZVP). When an SMD solvent model (THF) is added, a value of 2.03 ppm is found which is more accurate as expected since solvent effects are known to strongly influence the calculation of 1 H shifts. The remaining difference between the experiment and the theoretical calculations can be ascribed to the lack of relativistic treatment at the level of theory used. It has been reported that these relativistic effects are often the reason for the up-field shift of hydrides, hence it fits that a non-relativistic method predicts a shift too downfield. [38] In these calculations the H-P coupling was also probed. Here, a close match with the experiments both with (calc: 24.7 Hz, exp: 25.3 Hz) and without solvent model (calc: 25.8 Hz, exp: 25.3 Hz) is found since the calculation of coupling constants is much less affected by the aforementioned solvent and relativistic effects.

Complex 1
NBO analysis assigns a lone-pair on the hydride ligand (H34). This lone-pair delocalizes to both copper centers evenly with a second order perturbation energy of 56.30 kcal/mol each. This is consistent with the C2V symmetry of the complex. There is no donation of copper based orbitals toward the hydride (H34) observed consistent with a 3center 2-electron bond. The Wiberg bond index between the two copper centers is small (0.1150) consistent with an interaction via the hydride ligand. In the natural binding index, this value is 0.3391, which we propose is also due to the 3-center 2-electron bond in the dicopper hydride core since there is no indication for a direct Cu1-Cu2 interaction in the second order perturbation analysis. This is characteristic for an open 3-center 2-electron bond. [39] Quantum theory of atoms in molecules (QTAIM) analysis is consistent with the NBO analysis in showing no bond critical point between the two copper centers ( Figure S31). As expected, there are bond critical points between the hydride (H34) and the copper centers. These bond critical points are located in the positive area of the Laplacian, indicating a bond with mostly dative character from the hydride to the two copper centers. This is in line with the assignment of an open 3-center 2-electron bond. [IPr2Cu2H]BF4 [4] For the NBO calculations, the optimized coordinates from the report of Sadighi and co-workers were used. [4] The calculations were performed using the originally reported level of theory (PBEPBE [40] /ZTVP [41,42] with GD2 [43] empirical dispersion), as well as the level of theory used for complex 1 (BP86/def2-TZVP). Only the results for the latter will be discussed here since the difference was minimal and does not influence the conclusions.
The hydride (H33) donates towards the copper centers, in this case into the Cu-C anti-bonding orbitals with a second order perturbation energy of 82.41 kcal/mol each. This bonding situation is analogous to what is observed for 1. Also similarly 1, a Cu-Cu Wiberg bond index of 0.1100 and a natural binding index of 0.3316 is found for [IPr2Cu2H]BF4. This is again consistent with an open type of 3-center 2-electron bond since the second order perturbation analysis shows no indication of direct interaction between the two copper(I) centers. This open type of three center bonding was also already proposed by Sadighi and coworkers based on their calculations. [4] NBO analysis, however, does not indicate the unconventional 3-center 6-electron bond that was originally proposed by the authors.

Hydride charge comparison
To compare the charge of hydride ligand between 1 and [IPr2Cu2H]BF4, the charges obtained with the natural population analysis were used. To verify that these charges were representative, we decided to also compare the Mulliken charges and the QTAIM charges (obtained by integrating the electron density over the basins) of these two complexes. All three of these methods indicated a more negative charge on the hydride ligand in 1 than in [IPr2Cu2H]BF4 (Table S2).

Tracking charge upon deprotonation
To see how the negative charge that is introduced by the deprotonation is distributed, we considered the theoretical scheme in Figure S32. Each of the three structures (A, Amono and 1) were optimized and an NBO calculation for each structure was performed. The natural population analysis shows the electron density on each atom, comparing this value between the structures shows the movement of electron density upon deprotonation. Comparing A with Amono shows the influence of dimerization on the distribution of the charge whereas the comparison between Amono and 1 indicates how the negative charge introduced by deprotonation is distributed.  Figure S32: The deprotonation of A to form 1 in two theoretical steps, first is the formation of a monomer and second the deprotonation of this theoretical monomer Amono to form 1.
In the first step going from A to Amono, the main change in electron density is visible in the core of the molecule ( Figure S33). In the core the electron density previously located on the hydride atom (and also to some extend on the phosphorus atoms) is relocated to the two copper centers that have become more electron rich. We rationalize this by considering that the negative hydride is only stabilized by two copper centers in Amono rather than three in A. This destabilizes the high electron density on the hydride ligand and hence it is found more on the copper centers. When comparing Amono to 1 ( Figure S34), it becomes clear that the largest part of the introduced negative charge is distributed over the backbone with only a slight (0.02 e -) increase in the electron density on the hydride ligand. The specific positions in which the electron density increases/decreases can be rationalized by the possible resonance structures of the backbone ( Figure S35) Surprisingly, the electron density on the copper centers remains unaffected upon deprotonation.   Examining the two changes that A undergoes upon deprotonation (losing a proton and becoming a monomer) rationalizes the change in electronic structure that is observed. The loss of electron density observed on the hydride despite the addition of a net anionic charge is caused by the formation of a monomer rather than a dimer. This change also causes the increase in electron density on the copper centers. The change in charges on the backbone are mainly due to the deprotonation and delocalization of the negative charge over the conjugated naphthyridine structure. X-ray crystal structure determination [C20H40KO8][C26H43Cu2N2P2] • C4H8O, Fw = 1092.36, orange plate, 0.35  0.29  0.07 mm 3 , monoclinic, P21/c (no. 14), a = 9.6631(2), b = 25.2853(6), c = 23.0542(6) Å, β = 90.136(1) °, V = 5632.9(2) Å 3 , Z = 4, Dx = 1.288 g/cm 3 ,  = 0.94 mm -1 . The diffraction experiment was performed on a Bruker Kappa ApexII diffractometer with sealed tube and Triumph monochromator ( = 0.71073 Å) at a temperature of 150(2) K up to a resolution of (sin /)max = 0.65 Å -1 . The Eval15 software [44] was used for the intensity integration. For the prediction of the reflection profiles a split-mosaic model was used. A multi-scan absorption correction and scaling was performed with SADABS [45] (correction range 0.54-0.75). A total of 102303 reflections was measured, 12932 reflections were unique (Rint = 0.056), 9412 reflections were observed [I>2(I)]. The structure was solved with Patterson superposition methods using SHELXT. [46] Structure refinement was performed with SHELXL-2018 [47] on F 2 of all reflections. Nonhydrogen atoms were refined freely with anisotropic displacement parameters. One coordinated and one non-coordinated THF molecule were refined with disorder models. The metal-coordinated hydride atom was located in difference Fourier maps and kept fixed. All other hydrogen atoms were introduced in calculated positions and refined with a riding model. 699 Parameters were refined with 291 restraints (distances, angles, and displacement parameters in the disordered THF molecules). R1/wR2 [I > 2(I)]: 0.0398 / 0.0955. R1/wR2 [all refl.]: 0.0650 / 0.1089. S = 1.020. Residual electron density between -0.34 and 0.53 e/Å 3 . Geometry calculations and checking for higher symmetry was performed with the PLATON program. [48] CCDC 2145253 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.