Frustrated Lewis Pair Catalyzed Reduction of Carbon Dioxide Using Hydroboranes: New DFT Mechanistic Insights

Catalytic CO2 reduction is attractive for the sustainable production of value‐added fuels and chemicals. Prevented by an unclear mechanistic picture, only a few frustrated Lewis pair (FLP) catalysts are known for the efficient CO2 reduction under mild conditions using hydroboranes as reductant. For the first time, based on extensive DFT calculations, very efficient mechanisms for all steps of the FLP‐catalyzed reduction of CO2 to various products of boryl formate, H2CO, bis(boryl) acetal and methoxyl borane are revealed. Zwitterionic B/P FLP‐H2CO adducts are confirmed as active catalyst via the Lewis‐basic CH2O “oxide” center. Strong O and N Lewis‐bases are very active in promoting hydride transfer from hydroboranes to CO2 and strongly bound to the boryl group of reduced species. This can be modulated by Lewis acids in FLP catalysts for well‐balanced catalytic activity, which is crucial for further design of novel catalytic process.


Frustrated Lewis Pair Catalyzed Reduction of Carbon Dioxide Using Hydroboranes: New DFT Mechanistic Insights
Zheng-Wang Qu,* [a] Hui Zhu, [a] and Stefan Grimme [a] Catalytic CO 2 reduction is attractive for the sustainable production of value-added fuels and chemicals. Prevented by an unclear mechanistic picture, only a few frustrated Lewis pair (FLP) catalysts are known for the efficient CO 2 reduction under mild conditions using hydroboranes as reductant. For the first time, based on extensive DFT calculations, very efficient mechanisms for all steps of the FLP-catalyzed reduction of CO 2 to various products of boryl formate, H 2 CO, bis(boryl) acetal and methoxyl borane are revealed. Zwitterionic B/P FLP-H 2 CO adducts are confirmed as active catalyst via the Lewis-basic CH 2 O "oxide" center. Strong O and N Lewis-bases are very active in promoting hydride transfer from hydroboranes to CO 2 and strongly bound to the boryl group of reduced species. This can be modulated by Lewis acids in FLP catalysts for well-balanced catalytic activity, which is crucial for further design of novel catalytic process.
The worldwide dependence on non-renewable energy sources and chemical feedstock is detrimental to the environment. There is enormous interest in the development of alternative sustainable resources in order to meet our global energy and commodity chemical demands. Carbon dioxide (CO 2 ) is an appealing renewable carbon source for the production of value-added fuels and commodity chemicals due to its low cost, high abundance and relative lack of toxicity. [1] However, since CO 2 is the most oxidized form of carbon that is thermodynamically and kinetically quite stable, catalysis is crucial to the efficient reduction of CO 2 into value-added products using reducing reagents such as dihydrogen (H 2 ), hydroborane and hydrosilane. [2] The direct hydrogenation of CO 2 to methanol is the most atom-efficient route, but currently requires transition-metal-based catalysts (such as pincer iron) under relatively harsh conditions. [3] Alternatively, both metal [4] and non-metal [5] catalysts have been developed for the efficient hydroboration or hydrosilylation of CO 2 under mild conditions. In particular, metal-free frustrated Lewis pair (FLP) systems [5c-e,i,l-o,6] are usually more environment-friendly thus quite attractive for sustainable synthesis of high-value fine chemicals. Despite of stoichiometric bis(boryl) oxide byproduct, the products of CO 2 hydroboration can be useful for the transfer of formyl, methylene and methyl groups into organic compounds. [7] Mechanistic insights into such catalytic CO 2 hydroboration reactions is of great importance for the further design of novel catalytic process.
Catalytic CO 2 reduction is generally thought to be initialized by hydride transfer from the metal or boron center to the CO 2 molecule activated by the catalyst. [4e,k,5a,m,o,6b,8] However, only a few FLP catalysts are known to be active for CO 2 reduction under mild conditions; three typical examples of known hydroboration are shown in Scheme 1. In 2013, Fontaine et al demonstrated the first intramolecular B/P FLP 1 a for CO 2 reduction using catecholborane (HBcat), [5m,n] and subsequently suggested the FLPÀ H 2 CO adduct as the actual catalyst via its Bcat oxygen centers to induce hydride transfer from HBcat. [5j] In 2014, using Lewis-basic phosphine tBu 3 P and reactive 9-borabicyclo[3.3.1]nonane diborane (HBbn) 2 , Stephan et al found that similar H 2 CO-adducts such as tBu 3 PÀ CH 2 OÀ BbnOCHO were also active for CO 2 reduction, leading to several reduced products via an unclear mechanism. [9] More recently in 2016, Cantat et al presented the Si/N FLP 1 b for CO 2 reduction using reactive (HBbn) 2 , and proposed the FLPÀ CO 2 adduct as the key intermediate for initial C=O hydroboration. [5e] However, for the proposed initial steps catalyzed by 1 a and 1 b, the respective DFT-computed free energy barriers of 28.6 and 26.7 kcal/mol [5e,j] only for the first reduction step seem too high to account for the efficient reactions observed even at room temperature.
For a rational design of more active and selective catalysts, a clear and complete mechanistic picture of FLP-catalyzed CO 2 reduction is highly desirable. In this work, based on extensive DFT calculations at the PW6B95-D3 + COSMO-RS//TPSSÀ D3 + COSMO level in THF solution (see below for computational details), very efficient new mechanisms are proposed for the complete catalytic CO 2 reduction steps, catalyzed by the active FLP catalysts mentioned above (Scheme 1).
In line with previous DFT-calculations [5m] and the experimentally observed induction period, [5n] our DFT calculations have found similarly high overall barriers of 43.1 and 36.2 kcal/mol for the direct and the 1 a-catalyzed CO 2 hydroboration using HBcat. Thus, the original FLP 1 a is important for the initial formation of the zwitterionic FLPÀ H 2 CO adduct 1 af that was proposed [5j] as the actual catalyst via Lewis-basic Bcat oxygen atoms. As shown below, the complex 1 af is indeed able to efficiently catalyze the CO 2 reduction via the Lewis-basic CH 2 O "oxide" site instead, which is very active in promoting hydride transfers from hydroborane and modulated by the Lewis-acidic Bcat group for well-balanced catalytic reduction steps over similar barriers.
As seen in Figure 1, HBcat may bind to the CH 2 O "oxide" site of 1 af, leading to transient 2 a that is 2.8 kcal/mol higher in free energy with an evidently elongated (by 0.16 Å) CH 2 OÀ Bcat bond and a highly reductive hydroborate center. Facile hydride transfer from 2 a to CO 2 and its reduced intermediates (boryl formate 4 a and H 2 CO) may occur via similar transition structures TS1a, TS2ac and TS3ac, respectively, leading via intermediate H 2 CO to the methoxyl borane 6 a as the final product. The initial catalytic reduction of CO 2 into 1 af-bound boryl formate HCOOBcat (4 a) is À 7.6 kcal/mol exergonic over a moderate barrier of 24.1 kcal/mol (TS1a); the complex 3 a needs only 1.4 kcal/mol to release the boryl formate 4 a and regenerated 1 af with a recovered CH 2 OÀ Bcat bond. As suggested previously, [5j] the Bcat oxygen atoms of 1 af may also promote hydride transfer from HBcat to CO 2 (via TS1o), which is however kinetically 2.1 kcal/mol less favorable. Similar CH 2 O oxidepromoted hydride transfer from HBcat to 4 a is À 19.3 kcal/mol exergonic over a barrier of 26.7 kcal/mol (TS2ac), forming free H 2 CO (along with O(Bcat) 2 and regenerated 1 af via ionic O···B recombination) that can be more readily reduced by 1 afpromoted hydride transfer (TS3ac) into the methoxyl borane H 3 COBcat (6 a). Alternatively, direct C=O hydroboration of 3 a with HBcat is À 20.4 kcal/mol exergonic over a barrier of 25.4 kcal/mol (TS2a) to form the bis(boryl) acetal H 2 C(OBcat) 2 (5 a) (along with regenerated 1 af) that can be more readily reduced by another HBcat via 1 af-aided OBcat À abstraction from and hydride transfer to 5 a (TS3a), leading again to 6 a as the final product, which is kinetically 1.3 kcal/mol more favorable. In contrast, C=O hydroboration reactions of CO 2 , 4 a and H 2 CO using HBcat without catalyst are prevented by high barriers of 43.1, 35.9 and 32.1 kcal/mol (see ESI Table S1), respectively. The Lewis-basic CH 2 O "oxide" center is thus proven very active in promoting hydride transfer, even partially quenched by the Lewis-acidic Bcat group as base shuttle within the FLP catalyst 1 af to realize well-balanced catalytic steps.
In contrast, a similar CH 2 O "oxide" site is also found in the tBu 3 PÀ CH 2 OÀ BbnOCHO complex but without additional acidic stabilization, which is also active for catalytic CO 2 reduction using reactive diborane (HBbn) 2 but with a relatively low product selectivity (methoxyl borane H 3 COBbn together with significant amount of HCOOBbn and H 2 C(OBbn) 2 as products). [9] Our DFT calculations (see ESI Figure S3 and Table S2) show that the hydride transfer from the tBu 3 PÀ CH 2 OÀ BbnH complex to CO 2 is À 16.1 kcal/mol exergonic over a rather low barrier of 7.2 kcal/mol, indicating a very active oxide center in promoting hydride transfer. The tBu 3 PCH 2 O center is strongly bound to the boryl group of HCOOBbn, H 2 C(OBbn) 2 , H 3 COBbn and HBbn species by À 21.1, À 7.1, À 3.8 and À 9.6 kcal/mol, respectively, making possible the release of H 2 C(OBbn) 2 and H 3 COBbn (via transient H 2 CO over low barriers within 15 kcal/mol) but not HCOOBbn as product by HBbn up-taking. The C=O hydroboration of the tBu 3 PCH 2 O-bound HCOOBbn is À 4.0 kcal/mol exergonic over a moderate barrier of 17.9 kcal/mol to form the tBu 3 PCH 2 O-bound H 2 C(OBbn) 2 . After fast HBbn up-taking, the further reduction of H 2 C(OBbn) 2 into H 3 COBbn is prevented by a sizeable barrier of 23.2 kcal/mol (see below) thus is the slowest reduction step. More recently, the novel Si/N FLP catalyst 1 b was proven active for the CO 2 reduction using reactive (HBbn) 2 in THF solution.
[5e] The proposed mechanism via H/Cl exchange between HBbn and 1 b [5e] is examined at first.
As shown in Figure 2, the diborane (HBbn) 2 may dissociate with the help of a coordinating THF molecule over a moderate barrier of 23.1 kcal/mol (TS1) into monomeric HBbn species. The CO 2 hydroboration via HBbn is À 4.4 kcal/mol exergonic over a sizable barrier of 30.3 kcal/mol (TS2) to form the boryl formate 3 b (a higher barrier of 36.5 kcal/mol via dimeric (HBbn) 2 , see ESI Table S3) and should be slow at room temperature. On the other hand, the H/Cl exchange between (HBbn) 2 and 1 b is shown to be 15.4 kcal/mol endergonic to form the SiÀ H-bond containing complex 1 bH. Further hydride transfer from 1 bH to CO 2 is À 14.8 kcal/mol exergonic over a low barrier of only 15.7 kcal/mol (TS3) and thus should be efficient even at room temperature. However, this effectively leads to a high barrier of 31.1 kcal/mol for the 1 b-catalyzed CO 2 hydroboration that is kinetically even worse than the case without any catalyst.
Experimentally, 1 b reacts reversibly (in 30 % conversion) with CO 2 to form the neutral adduct 2 b in THF solution.
[5e] As shown in Figure 3, according to our DFT calculations, this reaction is indeed only 0.4 kcal/mol endergonic over a low barrier of 15.5 kcal/mol (TS1b), excluding 2 b as stable off-cycle resting state as suggested previously. [5e] Moreover, the proposed [5e] C=O hydroboration of 2 b via monomeric HBbn is À 4.4 kcal/mol exergonic but prevented by a sizable barrier of 27.1 kcal/mol (TS2b) to form the boryl formate 4 b and regenerated catalyst 1 b. Since the B-to-N binding of 4 b to 1 b is actually À 3.7 kcal/mol exergonic to form the adduct 4 b.1 b, the effective barrier for such catalytic CO 2 hydroboration would be 30.8 kcal/mol thus again kinetically not favorable.
Again, facile hydride transfer from HBbn to CO 2 promoted by the N-base center of 1 b is found in our DFT calculations. The cooperative binding of HBbn to 1 b is À 0.6 kcal/mol exergonic over a barrier of 20.2 kcal/mol (TS3b), leading to the complex 3 b with the BÀ H bond partially activated by the Si/N centers. Further hydride transfer to CO 2 is À 7.5 kcal/mol exergonic over a low barrier of 18.0 kcal/mol (TS4b) to form the complex 4 b.1 b. When the Lewis-acidic SiMe 2 Cl is replaced by the neutral methyl group, such barrier can be further reduced to only 11.0 kcal/mol, indicating again a highly active N-base site in promoting hydride transfer. Direct C=O hydroboration of 4 b.1 b by HBbn is À 14.0 kcal/mol exergonic over a low barrier of 18.3 kcal/mol (TS5b) to form the bis(boryl) acetal H 2 C(OBbn) 2 5 b and released 1 b, which is kinetically 7.5 kcal/mol more favorable than the alternative pathway via the hydroboration of free 4 b that may lead to H 2 CO instead of 5 b. If present, the addition of H 2 CO to the FLP 1 b is À 14.2 kcal/mol exergonic over a rather low barrier of 11.8 kcal/mol, leading to the separated ions of 1 bf + and Cl À in THF solution. The bis(boryl) acetal 5 b can be directly reduced by HBbn, which is À 26.2 kcal/ mol exergonic over a sizable barrier of 23.2 kcal/mol (via novel cyclic TS6b) to reach the methoxyl borane 6 b. Very similar mechanism is also found when less reactive HBcat is used as reductant, despite that 1 b-bound H 2 C(OBcat) 2 (5 c) is involved in the final (also slowest) reduction using HBcat over a sizable barrier of 28.3 kcal/mol (see ESI Figure S2 and Table S4), suggesting lower catalytic reactivity than 1 af. Our new mechanism of FLP-catalyzed CO 2 reduction is thus kinetically  very efficient, limited by the reduction of the bis(boryl) acetal intermediates.
In summary, for the first time, very efficient new mechanisms are revealed by extensive DFT calculations for the complete steps of FLP-catalyzed CO 2 reduction using hydroboranes. Zwitterionic B/P FLPÀ H 2 CO adducts represent the active catalysts via binding at the Lewis-basic CH 2 O "oxide" center. Strong O and N Lewis bases are very active in promoting hydride transfer to CO 2 and are also strongly bound to the boryl group of reduced species, which can be modulated by Lewisacidic groups such as Bcat as a base shuttle in FLP catalysts for more balanced catalytic activity that are crucial for further design of novel catalytic process.
Computational MethodsAll DFT calculations are performed with the TURBOMOLE 7.3 suite of programs. [10] The structures are fully optimized at the TPSSÀ D3/def2-TZVP + COSMO(THF) level, which combines the TPSS meta-GGA density functional [11] with the BJdamped DFTÀ D3 dispersion correction [12] and the def2-TZVP basis set, [13] using the Conductor-like Screening Model (COSMO) [14] for THF solvent (dielectric constant ɛ = 7.58 and diameter R solv = 3.18 Å). The well-established density-fitting RIÀ J approach [15] is used, which speeds up semi-local DFT functional calculations by a factor of 5-20 at practically no loss of accuracy. Chemically reasonable reaction paths are generated manually and tested in DFT calculations. Useful initial guesses of transition structures are obtained from interpolation between optimized reactant/intermediate/ product structures and constrained optimizations with appropriate reaction coordinates. The optimized structures are characterized by frequency analysis (no imaginary frequency for true minima and only one imaginary frequency for transition states) to provide thermal freeenergy corrections (at 298.15 K and 1 atm) according to the modified ideal gas-rigid rotor-harmonic oscillator model. [16] The connection of the transition state with reactants and products is checked visually by careful examining the vibrational transition mode. When a motion of heavy functional groups such as tBu 3 P is involved in the transition mode, the corresponding imaginary frequencies can be very small (< 50 cm À 1 ) and in such cases the chemical plausibility of the transition mode character was further checked.