Cobalt-Catalyzed Hydrosilylation of Carbon Dioxide to the Formic Acid, Formaldehyde, and Methanol Level—How to Control the Catalytic Network?

The selective hydrosilylation of carbon dioxide (CO2) to either the formic acid, formaldehyde, or methanol level using a molecular cobalt(II) triazine complex can be controlled based on reaction parameters such as temperature, CO2 pressure, and concentration. Here, we rationalize the catalytic mechanism that enables the selective arrival at each product platform. Key reactive intermediates were prepared and spectroscopically characterized, while the catalytic mechanism and the energy profile were analyzed with density functional theory (DFT) methods and microkinetic modeling. It transpired that the stepwise reduction of CO2 involves three consecutive catalytic cycles, including the same cobalt(I) triazine hydride complex as the active species. The increasing kinetic barriers associated with each reduction step and the competing hydride transfer steps in the three cycles corroborate the strong influence of the catalyst environment on the product selectivity. The fundamental mechanistic insights provide a consistent description of the catalytic system and rationalize, in particular, the experimentally verified opportunity to steer the reaction toward the formaldehyde product as the chemically most challenging reduction level.


■ INTRODUCTION
The catalytic reduction of carbon dioxide (CO 2 ) to valueadded products is essential for utilizing renewable energy sources and starting materials in sustainable energy systems and chemical industries. Transition metal complexes can facilitate the transformation of CO 2 into valuable chemicals on multiple oxidation levels. Many catalytic systems are reported to selectively catalyze the two-electron reduction of CO 2 yielding derivatives of formic acid (HCOOH), 1−12 while significantly fewer enable further reduction to the formaldehyde 13−20 (H 2 CO) or methanol 21−29 (H 3 COH) product platforms. With photosynthesis, nature has developed an ingenious but highly complex mechanism to exploit the intermediate reduction level of formaldehyde in the form of carbohydrates (C n H 2n O n ). In contrast, the thermodynamically favored over-reduction of formaldehyde and its derivatives to the methanol level often prevails with chemical catalysts, making this important product platform particularly difficult to exploit via direct reduction routes. Controlling selectivity in catalytic CO 2 reduction by unraveling the precise interaction of elementary steps and reaction parameters thus remains a significant chemical challenge.
While large-scale applications for CO 2 reduction require the use of "green" hydrogen, catalytic reductions using activated hydrides such as boranes or silanes are very useful as molecular probes to expand our knowledge in this field. 30−38 Their structures inherently enable heterolytic activation pathways to deliver hydrides for the two-electron reduction together with an oxophilic counterpart. 39 Examples that facilitate the transformation of CO 2 with such reducing agents to control the selectivity to two different reduction levels are shown in Figure 1A−C. The manganese pincer complex 1 reported by Gonsalvi was found capable of reducing CO 2 to either the formate level at 25°C or the methanol level at 80°C. 40 Selective formation of either formate or methanol derivatives was also reported for the hydrosilylation of CO 2 by the iridium complex 2, whereby the CO 2 pressure determined the product preference. 41 The tethered ruthenium−sulfur complex 3 was found to facilitate hydrosilylation of CO 2 to the formaldehyde or the methanol level, as reported by the group of Oestreich. 42 Here, the reaction temperature was the main parameter determining the selectivity. Selective arrival at three different reduction levels was reported for the nickel hydride complex 4 used in the hydroboration of CO 2 but required adjustment of the reductant ( Figure 1D). 43 Our group recently reported that the cobalt triazine pincer complex 5 is a suitable catalyst for the selective hydrosilylation of CO 2 to either formic acid, formaldehyde, or methanol derivatives using phenylsilane (PhSiH 3 ) as a reducing agent ( Figure 2). The catalyst was active at mild reaction conditions (r.t. to 80°C, 1−40 bar), low catalyst loadings (0.2 mol %), and various solvents in the presence of potassium tertbutanolate (KO t Bu). The formation of silylated products could be steered precisely by adjusting the temperature,  solvent, and pressure as control parameters. 44 In general, higher temperatures, lower CO 2 pressures, and higher concentrations favored the reduction to the formaldehyde and ultimately toward the methanol level. Higher CO 2 pressures stopped the reaction at the formate level, however. Formate and methanol derivatives were obtained with selectivities of 98 and 99%, respectively. Notably, the most challenging formaldehyde level could also be reached with high selectivity of 71%, demonstrating the possibility of targeting the three different product platforms in high yields without changing the nature of the reducing agent.
Based on mechanistic investigations of related hydrosilylation 40 and hydroboration 45 catalysts for CO 2 reduction to arrive ultimately at methanol derivatives, it can be suggested that the stepwise six-electron reduction achieved with catalyst 5 proceeds via a cascade of three individual catalytic cycles. CO 2 is reduced first to silyl formate in this consecutive process, which then acts as the substrate to be reduced to the formaldehyde as silylated acetal, which is ultimately reduced to the methanol level, as seen in Figure 2. We, therefore, aimed to rationalize the selectivity control through temperature, CO 2 pressure, and concentration in a combined experimental and theoretical effort by determining the detailed influence of these parameters on the relative kinetic barriers of each of the three consecutive cycles. The nature of the active catalyst species formed from complex 5 and the substrate activation processes were elucidated by synthesis, spectroscopy, and reactivity studies.
Based on the obtained insights, we subsequently investigated the catalytic mechanism with density functional theory (DFT) methods analyzing the three catalytic cycles individually and combining them to the energy profile for the overall reaction sequence. Finally, we simulated the product distribution under different sets of reaction parameters using microkinetic modeling of the reaction sequence compared to the experimental observations. The data provide a coherent description of the energy landscape, explaining in detail how fine adjustments of the control parameters lead to a high selectivity for the three different CO 2 reduction levels and suggesting general guidelines on how to optimize for the formaldehyde product platform. This fundamental knowledge provides tools for designing and developing new adaptive catalysts that allow selective access to the different C 1reduction products.

Identification of the Active Species
We first addressed the question of possible activation pathways of the precatalyst 5. In many examples of catalytic CO 2 reduction with molecular cobalt complexes, Co(I) hydride complexes are considered a significant part of the catalytic cycle as the active species. 46,7,47−50 For example, complex 6 was reported as an active catalyst for the hydrogenation of CO 2 to formate in the presence of Verkade's base, achieving high turnover frequencies of 74 000 h −1 at room temperature and 10 bar of CO 2 /10 bar of H 2 (Figure 3, left). 7 Computational and thermodynamic studies indicated an outer-sphere activation of CO 2 by the cobalt hydride complex with hydride transfer as a rate-limiting step of the reaction. 51,52 Complex 7 (Figure 3, middle) was used as a catalyst for the hydrosilylation of CO 2 , leading to a mixture of silyl formates, bis(silyl)acetals, and methoxysilanes. 47 More recently, cobalt-catalyzed N-formyla-tion of amines in the presence of CO 2 and H 2 was reported. 50 The authors suggested in situ formation of the hydride complex 8 (Figure 3, right) as the active species.
Since complex 5 used as a precursor complex contains Co in the oxidation state +II, the formation of a cobalt(I) complex requires its reduction prior to catalysis. This seems well in line with the catalytic results, where the activity of 5 was strongly enhanced in the presence of potassium tert-butanolate (KO t Bu). 44 Alkali tert-butanolates were previously reported to activate silanes to form transient MH/alkoxysilane complexes as strong reducing agents. 53 Following these lines, the in situ formation of cobalt(I) hydride complex 9 under catalytic conditions was considered. Therefore, we set out to synthesize 9 to investigate its performance as a kinetically competent species in CO 2 hydrosilylation.
The Co(I) hydride complex 9 was prepared by following a modified literature procedure reported for the synthesis of the structurally related complex 7. 54 The addition of 2.1 equiv of sodium triethylborohydride to complex 5 at room temperature in C 6 D 6 led to the formation of a dark green solution containing 9 (Scheme 1).
The structure of complex 9 was confirmed by NMR spectroscopy. The 1 H NMR spectrum of 9 corroborates a diamagnetic C 2v -symmetric compound, and the resonances corresponding to the pincer ligand are found in the expected ranges. A high field signal at −13.31 ppm appears as a triplet with a 2 J P,H coupling constant of 56 Hz and can be assigned to the hydride ligand ( Figure S1). The presence of a terminal cobalt hydride bond was further confirmed by infrared spectroscopy showing a medium intensity band at 1746 cm −1 , which agrees with previously reported values for related complexes. 55,54 The identity of 9 was further supported by HR-MS analysis showing the mass corresponding to the protonated molecular complex [C 27 H 49 N 9 P 2 Co] + .
The catalytic competence of 9 was examined in the hydrosilylation of CO 2 with phenylsilane at 1 bar of CO 2 applied as a continuous gas stream and 1 mol % catalyst loading at 80°C in comparison with 5 in the presence of KO t Bu or NaBEt 3 H as an additive (Scheme 2). The quantities of silyl formates, bis(silyl)acetals, and methoxysilanes were determined by 13 C{ 1 H} NMR spectroscopy, 47 and the observed turnover numbers (TONs) are shown in Figure 4. Within 4 h, a conversion of 35% of the Si−H units was . Thus, compound 9 reaches even higher turnover numbers for the reduction process than complex 5 in the presence of 4 mol % KO t Bu under the same reaction conditions (column 1). The isolated complex 9 also shows higher activity than an in situ preparation from 5 and NaBEt 3 H (column 2). These results indicate that 9 acts as the active species in the catalytic hydrosilylation of CO 2 and that the presence of KO t Bu is required mainly for its in situ formation from 5 but does not contribute to the actual hydride transfer during catalytic turnover.

Activation of Phenylsilane and CO 2 by Co−H Complex 9
First, we evaluated the possibility that the activation of phenylsilane by 9 occurs before the reaction with CO 2 , since the structurally related complex 7 was reported to activate hydrosilanes by oxidative addition to form Co(III) dihydride complexes. 47 In full analogy, adding phenylsilane to the in situ prepared complex 9 at room temperature led to the formation of the cobalt(III) complex 10 (Scheme 3, see the SI for details).
In the 1 H NMR spectrum of 10, the triplet resonance at −9.39 ppm exhibits a 2 J P,H coupling constant of 47 Hz. The chemical shift is similar to related cobalt(III) complexes and can be assigned to the two hydride ligands at the metal center in a mutual trans-position. 47 A second triplet at 4.40 ppm with a smaller 3 J P,H coupling constant of 11 Hz corresponds to the hydrosilyl ligand. The 31 P{ 1 H} NMR spectrum features a resonance at 174.0 ppm corresponding to the triazine ligand. The symmetric and asymmetric Co−H stretching frequencies appear at 2026 and 1758 cm −1 in the IR spectrum, respectively. This reactivity indicates that cobalt(III) is accessible via oxidative addition of Si−H bonds with the triazine-based ligand platform.
The reactivity of 10 toward CO 2 was found to be rather sluggish, however. When 10 was exposed to 1 bar of 13 CO 2 , NMR spectroscopic analysis revealed that conversion of CO 2 into a formate group required a reaction time of 2 days (see the SI for details). Considering that conversion of CO 2 into formate derivatives was previously observed to be completed after 4 h at room temperature with 0.2 mol % catalyst loading, 44 the low reactivity of 10 toward CO 2 indicates that an outer-sphere CO 2 activation pathway at the Co(III) species is unlikely to account for the catalytic process. Therefore, 10 is considered as an off-cycle species that can regenerate the active species 9 rather than being a catalytically competent intermediate.
Next, we investigated the activation of CO 2 by complex 9 as the initial step of the catalytic cycle. The insertion of CO 2 into metal hydride bonds to furnish formate complexes is wellknown and considered as an important step in CO 2 reduction catalysis. [46][47][48][49]52,50,56,40,41 Complex 9 reacts readily under 1 bar of CO 2 , leading to the formate complex 11 (Scheme 4, see the SI for details).
The NMR spectra of 11 reveal a diamagnetic C 2v -symmetric product, and the resonances of the ligand backbone appear in the expected ranges. The 1 H and 13 C resonances of the formate ligand were located at 7.83 and 173.9 ppm, respectively, and were assigned by 2D NMR experiments and 13 C-labeling. The 13 C isotopologue prepared from 5 (see the SI for details) shows a 1 J C,H coupling of 202 Hz in the 1 H NMR spectrum, while the 13 C resonance at 173.9 ppm associated with the formate carbon atom is strongly increased in intensity (Figures S13 and S14). The symmetric and asymmetric C−O vibrations are visible at 1380 and 1593 cm −1 in the IR spectrum ( Figure  S15).
Subsequently, we reacted the 13 C isotopologue of 11 with PhSiH 3 to explore the possibility of silylation of the formate group (Scheme 5, path A). When 11 was reacted with a Scheme 2. Hydrosilylation of CO 2 Catalyzed by Complexes 5 and 9  Subsequently, activation of the Si−H bond of the product in a second oxidative addition step leads to 12.
In contrast, upon addition of a slight excess of phenylsilane to 11 at room temperature, the immediate formation of complex 10 as the main product can be observed (Scheme 5, path B), while complex 12 was not visible. The isotope-labeled formate carbon atom can be assigned to a 13 C resonance at 162.2 ppm. The 1 H− 13 C HMBC NMR spectrum exhibits 3 J C,H coupling with a 1 H resonance at 5.82 ppm that corresponds to a hydrosilyl group, suggesting silylation of the formate group in addition to the formation of 10. Complex 10 could form in a similar pathway to 12, while the last step involves oxidative addition of phenylsilane instead of silyl formate. Thus, the reactivity of 11 toward phenylsilane indicates the transformation of the formate group into a silyl formate unit and the intermediate generation of 9, supporting the key role of 9 and 11 in the catalytic cycle.
We tested complexes 10 and 11 as catalysts for CO 2 hydrosilylation in a closed Schlenk tube containing 1 bar of 13 CO 2 (see the SI for details). The products were quantified by 13 C{ 1 H} NMR spectroscopy, and the CO 2 conversion was determined from the amount of silylated products formed. 44 After 4 h at 80°C, 35% of 13 CO 2 conversion and 83% selective formation of silyl formate were observed in the case of 10, whereas 78% 13 CO 2 conversion and 53% selective formation of silyl formate could be detected for 11 ( Figure 5). The catalytic performance of 10 is thus low compared to 11, while the latter exhibits an activity comparable to complex 5 in the presence of KO t Bu. 44 These results are consistent with the slow reactivity of 10 with CO 2 (vide supra) and the rapid formation of 12 from 11 (Scheme 5).
Complex 12 is only observed in the presence of stoichiometric amounts of PhSiH 3 (Scheme 5), which suggests that it is not formed under catalytic conditions under a large excess of PhSiH 3 with respect to the catalyst. The significantly lower activity of 10 compared to 9 and 11 indicates that 10 is not contributing to a major pathway, since otherwise lower conversion would be obtained. Furthermore, NMR spectroscopic analysis of the reaction mixture under catalytic conditions did not reveal any signals associated with the presence of either 10 or 12. Low relative energies calculated for 10 and 12 as compared to 9 and 11, respectively, further suggest that they are not catalytically competent due to their high stability (see the SI for details). Therefore, complexes 9 and 11 were chosen as the key reactive intermediates for the description of the catalytic cycle, while 10 and 12 were considered as off-cycle species that are not directly involved in the product formation.
The high conversions observed when using 5 and KO t Bu indicate the formation of 9 in the reaction mixture. The strongly reducing conditions in the presence of tert-butanolates and PhSiH 3 might favor a reduction of Co(II) to Co(I). 53 NMR spectroscopic analysis of the reaction mixture containing 5, PhSiH 3 (2 equiv), and KO t Bu (2 equiv) indicated the presence of 10 (see the SI for details), which suggests a reaction pathway that involves the initial reduction of 5 to 9, followed by oxidative addition of PhSiH 3 resulting in 10. Such an alternative reaction pathway generating 10 would rationalize the high catalytic activity of 5 in the presence of KO t Bu.

Computational Mechanistic Studies
Based on the experimentally verified intermediates 9 and 11 and their reactivity toward PhSiH 3 and CO 2 , a possible catalytic cycle was investigated with density functional methods on the B3LYP/D3BJ level of theory for optimization and frequency calculations with the def2-TZVP and the def2-SVP basis sets. Under conditions with an excess of phenylsilane compared to CO 2 , we considered the Si−H activation of phenylsilane to be the dominant pathway compared to Si−H activation of 13 and other hydrosilanes generated during the reaction (see the SI for details). Therefore, we chose to study the formation of the monosubstituted product 13 in the computational analysis as a model to describe the complex reaction network. The calculations were performed using a conductor-like polarizable continuum model (CPCM) with benzene as a typical low-polarity solvent, in which all three reduction levels can be obtained experimentally. 44 It was also used in the experimental analysis of the mechanism for the synthesis of 9, 11, and 14.

Cycle 1: Catalytic Formation of Silyl Formate
Two possible pathways are in principle conceivable for the generation of formate species from CO 2 in homogeneous catalysis. 49,52,56 In the inner-sphere process, activation of CO 2 occurs via coordination to the metal and subsequent migratory insertion. Alternatively, the outer-sphere pathway involves nucleophilic attack of the metal hydride at the weakly Lewis acidic carbon atom of noncoordinated CO 2 . In the case of complex 9, all attempts to locate a local minimum corresponding to an outer-sphere activation of CO 2 resulted in dissociation of the two reactive components during the geometry optimization, and an inner-sphere process was thus considered more favorable. Scheme 6 depicts the corresponding catalytic cycle used as a basis for structural and energetical analysis.
The first step in the formation of 11 was found to proceed via coordination of CO 2 in a side-on η 2 -mode at complex 15 as a local minimum on the potential energy surface (PES, Figure  6). The alternative η 1 -O and η 1 -C coordination modes were considered but either led to dissociation during the geometry optimizations or were not connected to the transition states relevant for the hydride transfer, respectively. The coordination is slightly endothermic and proceeds via the transition state TS1, in which the imaginary frequency involves bending of the CO 2 .
Complex 15 has a slightly distorted square pyramidal coordination geometry, whereby CO 2 is aligned along the P− Co−P axis and exhibits significant bending (137.6°). The coordinated C−O bond is slightly elongated at 124 pm as compared to the noncoordinated one at 121 pm. The Co−C and Co−O1 distances are 200 and 223 pm, while the second Co−O2 distance is much longer in agreement with the lack of an additional interaction (297 pm). Analysis of the frontier orbitals confirms a mixing of the electron-donating cobalt d z 2 orbital and the electron-accepting in-plane π*-orbital of the CO 2 (Figure 7). The donation of electron density from the cobalt d z 2 orbital into the antibonding in-plane π*-orbital (LUMO) is also manifested in the negative sum of the Mulliken gross atomic charges of CO 2 (−0.48). Analysis of the orbital contributions reveals that both the C and the O atom participate in the interaction. However, only a partial electron transfer to the CO 2 occurs, and cobalt can be considered to remain largely in its oxidation state of +I.
A decrease of the OCO angle is often associated with CO 2 reduction processes, since it lowers the LUMO's energy and also leads to more pronounced localization at the carbon atom. Both of these factors lead to enhanced electrophilicity of the carbon atom and thus facilitate nucleophilic attack. 58 The enhanced carbon weight of the in-plane π*-orbital at the carbon atom also increases the spatial overlap with the d z 2 orbital and thus further lowers the LUMO's energy. Consequently, the nucleophilic attack of the hydride to the carbon atom corresponding to the migratory insertion occurs as a slightly exothermic spontaneous process ( Figure 6). The initially formed structure 16 relaxes to the experimentally identified formate complex 11.
The subsequent silane activation and product release could occur via a σ-metathesis pathway, as discussed in previous reports. 59,60,35,41 However, thorough scans of the potential energy surfaces indicated far higher energy of intermediates and transition states than the oxidative addition/reductive elimination pathway. Additionally, the reactivity of 9 and 11 with hydrosilanes yielding 10 and 12 (Scheme 5) demonstrates the possible oxidative addition to the cobalt(I) complexes. The oxidative addition at 11 to form 14 is exergonic and occurs without a noticeable barrier. The primary product 17 arising from the reductive elimination contains the η 2 -C=O-bound silyl formate ester, whose release of product 13 regenerating the Co−H complex 9 is again slightly exergonic.
The driving force for the catalytic turnover of the low-lying intermediate 14 is provided by the exergonic overall catalytic process to convert CO 2 and PhSiH 3 to 13 with a Gibbs free energy of −4.5 kcal mol −1 . This is very different from CO 2 hydrogenation, where the formic acid formation is uphill under standard conditions. 3,61−63 The strong Si−O bond provides the additional driving force. The possibility of reductive elimination of formic acid from an alternative isomer of 14 was found to be less favorable than the generation of silyl formate 13 (see the SI for details).
The rate of catalytic turnover can be estimated by the energy span δE that is calculated from the most stable intermediate (turnover determining intermediate, TDI) and the highest transition state on the potential energy surface (turnover determining transition state, TDTS). If the TDI follows after the TDTS, the Gibbs free energy of the reaction is added to their energetic difference. 64 (TDI) if TDTS after TDI (2) The energy span from the first cycle can be calculated from 14 as the TDI and TS1 as the TDTS. Since the TDI follows after the TDTS, the Gibbs free energy of ΔG R = −4.5 kcal mol −1 is taken into account, and an energy span of 24.2 kcal mol −1 can be calculated using eq 1, which is in line with the experimental conditions for catalytic turnover.

Cycle 2: Catalytic Formation of Bis(silyl)acetal
To further reduce the silyl formate to the formaldehyde level, we considered insertion into the Co−H bond of complex 9 as a starting point for the catalytic cycle (Scheme 7, Figure 8).
Initially, this leads back to 17 where the silyl formate is coordinated via the C−O bond and forms a π-complex. The forward reaction under the formation of the η 2 -complex 18 leads to the second two-electron reduction of the carbon atom in competition with dissociation of silyl formate 13 from 17 to regenerate 9. Finalizing the insertion step to form the C−H and Co−O bonds leads to 19 in a strongly exothermic process. Oxidative addition of phenylsilane results in the formation of 20, followed by extrusion to reform the Co−H unit in 21 that constitutes the lowest energy structure at −20.0 kcal mol −1 relative to the starting point of the overall sequence. Finally, product release generates 9 and bis(silyl)formate 22. The fourelectron reduction of CO 2 to bis(silyl)formate 22 has a Gibbs free energy of −11.7 kcal mol −1 and provides the driving force for the catalytic turnover. However, the second reduction step is hindered by a higher kinetic barrier than the first twoelectron reduction step (Figure 6), as 7.3 kcal mol −1 destabilizes 18 compared to 9. Thus, the high barrier that involves 18 hampers the bis(silyl)acetal formation at lower temperatures. Although 18 was characterized as a local minimum by frequency calculations, it is a high-lying structure on the potential energy surface. Surface scans indicate a rather smooth energy surface, which hampered localization of the transition state connecting 17 and 18. Structure 18 thus provides a reasonable point of reference for the estimation of the energy span. This way, with the overall lowest intermediate 14 as TDI, an energy span of 28.9 kcal mol −1 can be estimated using eq 2. Ligand rearrangement within 20 leads to 23, which bears 1,3diphenyldisiloxane and formaldehyde in its coordination sphere. This results in a change of the cobalt oxidation state from +III to +I. Geometry scans indicated a concerted formation of the Si−O bond and cleavage of the C−O bond, while no geometry corresponding to an energy maximum was visible, indicating a smooth PES and a transition state close to 23. With a pentacoordinated silicon atom, 23 is the highest lying structure of the catalytic cycle and has a relative Gibbs free energy of 8.3 kcal mol −1 . Similar to 18, the high-lying intermediate 23 was used as a proxy for the transition state that could not be located due to the smooth energy surface. The high energy rapidly relaxes by dissociation of formaldehyde to 24 as a strongly exothermic step, followed by dissociation of the siloxane resulting in the regeneration of 9. The generation of formaldehyde as an organic intermediate was also reported for the hydrosilylation of CO 2 with the manganese complex 1 40 and proposed based on computational studies for the hydroboration of CO 2 to a methanol derivative catalyzed by a nickel pincer complex. 60 The hydrosilylation of formaldehyde to methoxysilane 30 involves a series of insertion, oxidative addition, and reductive elimination similar to the two previous reduction steps ( Figure  9). Notably, the coordination of formaldehyde to 9 resulting in 25 occurs as an exergonic process and thus is more facile than the coordination of CO 2 or silyl formate 13 to 9. The hydride transfer affecting the third two-electron reduction to 26 and the rearrangement to the alkoxy complex 27 proceeds by the transition states TS2 and TS3 with moderate activation barriers, while the reaction to 27 has a strong thermodynamic driving force of −31 kcal mol −1 relative to the starting point of this cycle. The reductive elimination of the methoxysilane from 28 to 29 involves the transition state TS4. The third reduction level provides the strongest thermodynamic driving force of −50.7 kcal mol −1 . However, due to the high energy of 23, the third reduction level is associated with a high energy span of ca. 29.9 kcal mol −1 (eq 2).

Selectivity Control in the Reaction Cascade of Cycles 1−3
The energy landscape for the consecutive catalytic CO 2 hydrosilylation to the formate, formaldehyde, and methanol level is composed of a sequence of the three catalytic cycles discussed above, as shown in Figure 10. The stepwise 2e − reductions result from three mechanistically distinct hydride transfer steps that are all mediated by the same Co−H intermediate 9. The resulting energy profile summarized in Figure 10 enables us to rationalize the selectivity control in a qualitative and semiquantitative way. The relative values of the energy spans δE(formate) ≪ δE(formaldehyde) < δE(methoxide) reflect that silyl formate can be obtained very selectively at mild temperatures. While higher temperatures allow for further reduction to bis(silyl)acetal, over-reduction to methoxide is more difficult to avoid due to the second and third cycle's very similar energy spans. The high kinetic hindrance associated with the rearrangement of bis(silyl)acetal 22 to formaldehyde and 1,3-diphenylsiloxane makes the isolation of the silylated formaldehyde derivatives possible, while the subsequent reduction of formaldehyde to methoxide 30 is a very fast process.
The possibility to control the selectivity by the CO 2 pressure results from the competing insertion processes of either CO 2 (9 to 11) or silyl formate (9 to 19) into the Co−H bond. Higher CO 2 pressure makes the CO 2 insertion the dominant process and hinders the silyl formate insertion. Thus, it prevents the forward reaction after the first reduction, making the formate species the strongly preferred product. In contrast, reducing the solvent amount increases the concentration of silyl formate, while the chemical potential of CO 2 is dependent on the partial pressure and thus remains constant. This agrees with the observation that the reduction beyond the formate level is favored under neat conditions without solvent. Similar considerations about the influence of the CO 2 pressure and the reactant concentration on the selectivity in the hydrosilylation 41 and hydroboration 43 of CO 2 have been discussed previously.
To validate the qualitative conclusions, the calculated Gibbs free energy profiles for the three reduction steps were used to perform kinetic simulations of the product distributions (Figures 11 and 12). Bearing in mind the strong sensitivity of the rate constants on small variations of the Gibbs free energies, the relative distributions are considered rather than absolute values. Simulation of a reaction at 80°C after 4 h results in nearly quantitative conversion of CO 2 (98%) as observed in the experiment (97%). The absolute values of the simulated product distribution silyl formate ≫ bis(silyl)acetal > methoxysilane corroborate better with the experimental data obtained at lower temperatures indicating that the calculated energy spans are slightly overestimated (Figure 11, panel A). The simulation reflects the further consumption of PhSiH 3 correctly at higher temperatures and extended reaction times whereby the bis(silyl)acetal and methoxysilane yields are increased consecutively (Figure 11, panel B). Simulating an increase of the CO 2 pressure to 20 bar at the same temperature results in a higher yield of the silyl formate and shows a lower rate of further reduction ( Figure 11, panel C). Therefore, the   pronounced influence of variations in temperature and pressure on the product distribution is reproduced correctly in the simulations based on the relative energy profiles of the three consecutive catalytic cycles.
The temperature regimes favoring the selectivity for the individual products as obtained from the kinetic simulations are depicted in Figure 12 (left). The reaction conditions (solvent, catalyst concentration, pressure, reaction time) were chosen close to those used in the experiments (Figure 12, right). 44 The absolute temperatures associated with high selectivity for each product are higher than in the experiment. For example, while maximum selectivity for the formaldehyde level (53 and 52%) was observed at 80 and 100°C, 44 the simulation shows a maximum selectivity of 73% at 140°C. Still, the temperature windows and the selectivities obtained in the experiments are reproduced very well. It can be seen that high selectivity for the silyl formate can be obtained readily at milder temperatures, and nearly full conversion to methoxide can be achieved at optimized conditions. The intermediate formaldehyde level cannot be reached with the quantitative conversion of the formate and full suppression of the over reduction to methoxide simultaneously, however. The maximum selectivities of the three product levels obtained from the simulations compare remarkably well with the highest selectivities obtained after careful optimization experimentally.

■ CONCLUSION
In conclusion, the present combined experimental and computational study rationalizes the selective reduction of carbon dioxide to silylated formic acid, formaldehyde, and methanol derivatives catalyzed by cobalt triazine complex 5. The experimentally observed control factors are reflected directly in the energy profiles of the individual cycles and their connection to the full energy surface of the reaction network. The increasing energy spans δE(silyl formate) ≪ δE(bis-(silyl)acetal) < δE(methoxysilane) rationalize that higher temperature is required for reduction to formaldehyde and methanol derivatives. High CO 2 pressures lock in the reaction on the formic acid level as the rate is defined by the competing hydride transfer to either CO 2 or silyl formate mediated by the active species 9. However, high concentrations of catalyst and reductants favor the forward reactions leading to formaldehyde or methanol derivatives.
The turnover determining transition state for the reduction beyond the formaldehyde level is not associated with the C=O reduction but with the cleavage of the corresponding silyl acetal (23). Most importantly, the high kinetic barrier for the acetal cleavage is identified as the crucial enabler to achieve synthetically relevant selectivities for the formaldehyde level, because the barrier of hydride attack at free H 2 C=O is small, leading to the thermodynamically favored over-reduction. Thus, utilization of the competing hydride transfer to either CO 2 or silyl formate enables selectivity control in the two-or four-electron reduction of CO 2 , while differentiating the energy spans for the formaldehyde and the methanol level via the acetal cleavage rather than the hydride transfer seems to open promising strategies for developing effective catalytic systems that target the four-compared to the six-electron reduction of CO 2 .
In general, the present study highlights the potential of mapping out the complex reaction networks of catalytic CO 2 reduction by experimental and computational techniques. Identifying the relative influence of the reaction parameters on the catalytic mechanism may contribute to parameter screening and experiment planning in the design of catalytic systems targeting challenging products. 65 Obviously, general applicable mechanistic principles have to be available for such approaches, and the present study is hoped to provide a valuable contribution in this more general context.
General considerations, experimental methods and synthetic details, copies of NMR spectra, computational data (PDF) des deutschen Volkes" for a fellowship as well as the IMPRS-RECHARGE School.