A Stable N‐Heterocyclic Silylene with a 1,1′‐Ferrocenediyl Backbone

Abstract The N‐heterocyclic silylene [{Fe(η5‐C5H4‐NDipp)2}Si] (1DippSi, Dipp=2,6‐diisopropylphenyl) shows an excellent combination of pronounced thermal stability and high reactivity towards small molecules. It reacts readily with CO2 and N2O, respectively affording (1DippSiO2)2C and (1DippSiO)2 as follow‐up products of the silanone 1DippSiO. Its reactions with H2O, NH3, and FcPH2 (Fc=ferrocenyl) furnish the respective oxidative addition products 1DippSi(H)X (X=OH, NH2, PHFc). Its reaction with H3BNH3 unexpectedly results in B−H, instead of N−H, bond activation, affording 1DippSi(H)(BH2NH3). DFT results suggest that dramatically different mechanisms are operative for these H−X insertions.


Synthesis of 1DippSi(H)OH
The product is formed when benzene solutions of 1DippSi are exposed to an inert gas atmosphere containing trace amounts of water, which may happen simply by serendipity. The following procedure, which was originally intended to be used for the reaction of the silylene with carbon suboxide, was found to afford the water addition product conveniently in reproducible yields. A 100 mL flask was charged with a mixture of malonic acid (150 mg, C34H44N2FeOSi (580.66): calcd. C 70. 33, H 7.64, N 4.82 %; found C 70.22, H 7.44, N 4.40 %.

B X-Ray Crystallography
For each data collection a single crystal was mounted on a micro-mount and all geometric and intensity data were taken from this sample. Data collections were carried out at 100(2) K using MoKα radiation (λ = 0.71073 Å) either on a Stoe IPDS2 diffractometer equipped with a 2-circle goniometer and an area detector, or on a Stoe StadiVari diffractometer equipped with a 4-circle goniometer and a DECTRIS Pilatus 200K detector. The data sets were corrected for absorption, Lorentz and polarisation effects. The structures were solved by direct methods (SHELXT) and refined using alternating cycles of least-squares refinements against F 2 (SHELXL2014/7). [S5] C-bonded H atoms were included to the models in calculated positions, heteroatom-bonded H atoms have been found in the difference Fourier lists. All H atoms were treated with the 1.2 fold or 1.5 fold isotropic displacement parameter of their bonding partner.

II Computational Details
Geometry optimizations and harmonic frequency calculations were performed using the ORCA program package [S7] (Version 4.1.2) employing the PBEh-3c density functional composite method [S8] combined with a higher integration grid (Grid5) to avoid spurious imaginary frequencies. Optimized structures were characterized as minima or first order saddle points by eigenvalue analysis of the computed Hessians. Connectivities between minima and the corresponding transition states were validated by intrinsic reaction coordinate (IRC) calculations [S9] or by displacing the transition-state geometries along both directions of the transition mode, followed by unconstrained optimizations to the respective minima.
For selected structures, relative energies from correlated wavefunction theory were computed on a smaller model system as well as on the full system. For a smaller system with H-truncated nitrogen atoms single-point energies were obtained using the explicitly correlated coupled-cluster ansatz [S10] CCSD(T)-F12b [S11] as implemented in the Molpro2015.1 program [S12] in combination with the F12optimized correlation consistent polarized triple-zeta orbital [S13] and auxiliary [S14-S16] basis sets of the cc-pVTZ-F12 family for all non-metal atoms. For the Fe atom, the aug-cc-pVTZ orbital basis [S13] was used in conjunction with the universal JKfit option for integral fitting in the Fock matrix construction along with the aug-cc-pVTZ/MP2Fit auxiliary basis used for the many-electron integrals and CABS representation. For the full system single-point energies were obtained from correlated DLPNO-CCSD(T)-F12 [S11, S17] computations as implemented in ORCA 4.2.1 using the F12-optimized correlation consistent polarized triple-zeta orbital and auxiliary basis sets for non-metal atoms. For the Fe atom, the aug-cc-pVTZ orbital basis and associated auxiliary sets were employed.
For convenience, the following Chart shows the numbers of the computed structures for experimentally observed compounds:

Comparison of frontier molecular orbitals and spin densities
For comparison with 1 HOMO-LUMO gaps and singlet-triplet splittings have been calculated for three experimentally characterized silylenes at the PBEh-3c level of theory. Figure S45 shows the respective Lewis structures, frontier molecular orbitals and spin densities of the triplet state. All energies given are in eV.

Evaluation of the PBEh-3c method for the full and the H-truncated molecular system
The PBEh-3c method has primarily been designed for the efficient assessment of molecular structures but it also provides reasonable relative energies with respect to benchmark databases. [S8] To test the suitability of PBEh-3c for the reactions investigated here, the thermochemistry for two representative reactions were compared to single-point energies from correlated wavefunction theory (Schemes S1 and S2). Unless stated otherwise in the following, relative free energies ∆G 298 are given in kcal mol −1 (imaginary wave numbers of transition modes in parentheses, cm -1 ). Bonds formed or broken in transition states are shown dashed, unreactive H atoms are omitted and the orientation of Dippsubstituents in transition state structures are indicated by showing the C ipso atom only. Scheme S1. Computed reaction paths for the H−H bond insertion in the H-truncated molecular model 1HSi; PBEh-3c relative total energies ∆E in kcal mol -1 (CCSD(T)-F12b/cc-pVTZ//PBEh-3c, DLPNO-CCSD(T)-F12/cc-pVTZ//PBEh-3c in parentheses, in this order).

Additional Information on reaction paths investigated
The silylene 1DippSi does not react with H 2 to the corresponding silane, which is in line with reaction barriers above 50 kcal mol -1 calculated for 1 for both addition pathways identified (Scheme S3).

Scheme S3. Computed reaction paths for the H−H bond insertion by 1.
For the reaction of 1 with H 2 O multiple paths involving one or two additional H 2 O molecules acting as proton shuttles were investigated. For the rate determining transition state of the most favorable path the corresponding kinetic isotope effect was calculated. In order to estimate the effect of hydrogen tunneling, a 1-D Wigner correction [S36] was applied. The direct oxidative addition of water to 1 is kinetically least favorable (Scheme S4 path d) while O−H addition across a Si−N bond with subsequent proton shift is effectively catalyzed by a proton-shuttle mechanism with additional water molecules (Scheme S4, a-c and Scheme 3 in the main text).
Overall the application of the 1-D Wigner correction lowers activation barriers by ca. 2-3 kcal mol -1 (Schemes S4 and S5). This is accompanied with an increase of the predicted KIEs (Scheme S4a).
As for the reaction with H 2 O, two pathways, i.e. direct N-H bond insertion as well as N-H bond addition across an N−Si bond of 1, were examined for ammonia and, as before, proton shuttling involving an additional NH 3 molecule reduces the activation barrier substantially (Scheme S6).

Scheme S6.
Computed reaction paths for the N−H bond cleavage in NH3 by 1 and subsequent hydrogen migration.

S41
A reaction involving the silaimine S8 can clearly be ruled out because of the very high computed activation barrier of 76 kcal mol -1 (Scheme S7).
For the reaction of 1 with H 3 BNH 3 both, insertion into an N−H or a B−H bond, are kinetically strongly disfavored (Scheme S8).

Scheme S8.
Computed reaction paths for the direct B/N−H bond insertion in H3BNH3 by 1.
In contrast to expectation silane 7 is not the primary product of the reaction of 1 with H 3 BNH 3 . Extrusion of H 2 BNH 2 from 10 is hindered by an unreasonably high effective barrier of 47 kcal mol -1 (Scheme S9).

Scheme S9.
Computed reaction path for the release of H2BNH2 from 10, effective barrier relative to 10. Scheme S11. Computed reaction paths for the SN-like reaction of H3BNH3 with 1.
Finally, following a reviewer's suggestion, we elucidated the role of the BH 3 adduct S12 as resting state in the catalytic cycle sketched in Scheme S12: low-barrier B-H insertion leads to S14 and NH 3 transfer from H 3 BNH 3 yields the product 10. The catalytic cycle is closed by regeneration of S12 with another equivalent of 1 and the liberated BH 3 . The rate-limiting ammonia transfer via TSS27 is, however, connected with a prohibitively high activation barrier of 30 kcal mol -1 . Hence this scenario cannot compete with the initial dehydrogenation of H 3 BNH 3 by 1 (Scheme 4 in the main text). Scheme S12. Computed reaction path for the reaction of B2H6 with 1.