Selective Catalytic Frustrated Lewis Pair Hydrogenation of CO₂ in the Presence of Silylhalides

Abstract

The frustrated Lewis pair (FLP) derived from 2,6‐lutidine and B(C₆F₅)₃ is shown to mediate the catalytic hydrogenation of CO₂ using H₂ as the reductant and a silylhalide as an oxophile. The nature of the products can be controlled with the judicious selection of the silylhalide and the solvent. In this fashion, this metal‐free catalysis affords avenues to the selective formation of the disilylacetal (R₃SiOCH₂OSiR₃), methoxysilane (R₃SiOCH₃), methyliodide (CH₃I) and methane (CH₄) under mild conditions. DFT studies illuminate the complexities of the mechanism and account for the observed selectivity.

The frustrated Lewis pair (FLP), 2,6‐lutidine/ B(C₆F₅)₃ mediates the catalytic reduction of CO₂ using H₂ and a silylhalide. Control of the silylhalide and solvent, affords selective avenues to the disilylacetal (R₃SiOCH₂OSiR₃), methoxysilane (R₃SiOCH₃), methyliodide (CH₃I) and methane (CH₄) under mild conditions. The mechanism is studied by DFT and accounts for the observed selectivity.

The dramatic and continuous increase in the atmospheric CO₂ level since the industrial revolution results from the extensive use of fossil fuels and is the major contributor to climate change. This has prompted the scientific community to target a variety of new technologies to reduce emissions or provide alternative energy sources as these offer the most promising avenues to address climate change. Nonetheless, other efforts targeting the capture or use of atmospheric CO₂ have also garnered attention. One potential avenue to the use of atmospheric CO₂ involves reduction via hydrogenation. ^[1] For example, recent reviews have described the conversion of CO₂ to methanol using homogeneous and heterogeneous transition metal‐based catalysts ^[2] while other reports have demonstrated the production of longer chain fuels ^[3] or olefins or higher alcohols. ^[4] In addition to the above metal‐catalyzed processes, there have also been extensive efforts to employ main group reagents to mediate CO₂ reduction processes. A number of studies ^[5] have explored catalytic processes including both base‐mediated and frustrated Lewis pair (FLP) hydrosilylations ^[6] and hydroborations ^[7] of CO₂ while others have probed aminations. ^[8]

Despite the seminal finding in 2009 in which Ashley and O'Hare ^[9] reported the FLP‐mediated reduction of CO₂ to methanol (Scheme 1), albeit in low yield and at 160 °C for 6 days, the direct hydrogenation of CO₂ mediated by a main group species has garnered limited attention. A collaborative effort with the Fontaine group ^[10] described the stoichiometric reactions of the intramolecular FLP, 1‐BMes₂‐2‐NMe₂‐C₆H₄, with H₂ and CO₂ yielding formyl, acetal and methoxy‐borane derivatives (Scheme 1). This study suggested that judicious selection of the combination of the Lewis acid and the base could plausibly lead to catalytic H₂/CO₂ chemistry. More recently, Zhao et al. ^[11] described the hydrogenation of CO₂ in the presence of H₂ and K₂CO₃ using B(C₆F₅)₃ as the catalyst, affording effective turn‐over to K[HCO₂] at comparatively high H₂/CO₂ pressures of 60 bar (Scheme 1). While the achievement of catalytic hydrogenation is impressive, the reduction was limited to the formation of formate product.

Scheme 1.

Direct reactions of CO₂/H₂ mediated by main group reagents.

Pondering an FLP system that would effect reduction beyond formate, we recognized that in earlier studies methanol or methane were obtained exploited hydrosilanes or hydroboranes that provide both a reducing agent and an oxophile.[ ⁶ , ⁷ ] In contrast, use of H₂ as the reducing agent in direct FLP hydrogenations of CO₂ does not provide such an oxygen‐atom scavenger. Thus, we speculated that further hydrogenation of CO₂ could be effected in the presence of a silylhalide. Herein, we report the FLP‐mediated catalytic hydrogenation of CO₂ using H₂ as the reducing agent performed in the presence of a silylhalide which acts as an oxophile. Judicious choices of the silylhalide and reaction solvent are shown to provide fine control over the nature of the products of catalysis.

The activation of H₂ by 2,6‐lutidine/B(C₆F₅)₃ (Scheme 2) ^[12] and subsequent reaction with CO₂ is known to afford the salt [C₅H₃Me₂NH][HCO₂B(C₆F₅)₃]. ^[13] This species was allowed to react with 1 equivalent of Et₃SiI in CDCl₃ resulting in the upfield shift of the formyl proton in the ¹H NMR from 8.31 ppm to 8.17 ppm and the appearance of a ¹¹B{¹H} NMR signal at −0.1 ppm. These data affirm the formation of B(C₆F₅)₃ adduct of silyl formate Et₃SiOC(O)H ^[6c] and are consistent with the cleavage of the B−O bond in the formyl‐borate salt (Scheme 2). Recognizing that the silyl formate‐borane adduct will exist in an equilibrium with free borane, this implies that it should be accessible for further reaction.

Scheme 2.

Control reactions.

We also queried the possibility of reduction of Et₃SiI in the presence of excess base. To this end, Et₃SiI and 2,6‐lutidine were combined under H₂ (4 atm) in the presence of 10 mol % B(C₆F₅)₃ in either CDCl₃ or C₆D₆ and heated at 100 °C for 40 h (Scheme 2). In both cases no reduction of the silylhalide was observed. This suggested that the silylhalide could act as an oxophile in the presence of H₂, for the hydrogenation of CO₂, without the possibility of invoking a hydrosilylation mechanism.

Thus, targeting FLP hydrogenations of CO₂, reactions of 10 equivalents of Lewis base and silylhalide were performed in C₆D₆ or CDCl₃ solution of 10 mol % of B(C₆F₅)₃. In these reactions the substituted pyridines, 2,4,6‐collidine and less basic 2,6‐lutidine were employed and the systems were pressurized with H₂ (4 atm.) and ¹³CO₂ (2 atm.) and heated to 100 °C for up to 60 h. The reactions were monitored by ¹H NMR and ¹³C NMR spectroscopy. Initial reactions using Me₃SiCl and 2,6‐lutidine in C₆D₆ or CDCl₃ (Table 1, entry 1, 2) as the solvent, afforded [C₅H₃Me₂NH][HCO₂B(C₆F₅)₃] ^[13] as the major product as evidenced by the doublet resonance (¹ J _C–H=209 Hz) at 8.37 ppm in the ¹H NMR spectrum and the doublet resonance in the ¹H‐coupled ¹³C NMR at 169.5 ppm. The generally poor reactivity in the presence of Me₃SiCl was attributed to the relatively strong Si−Cl bond and prompted the use of 2,6‐lutidine and Me₃SiBr. This led to an 83 % yield of methoxysilane Me₃SiO¹³CH_3, after 40 h of heating in C₆D₆ (entry 3). In this case, the major product was identified by a ¹H NMR resonance at 3.25 ppm as a doublet (¹ J _C–H=141 Hz), the corresponding ¹³C{¹H} NMR signal is found at 49.9 ppm. ^[6b] Repetition of the experiment in CDCl₃ also led to the selective production of Me₃SiO¹³CH₃ in 73 % yield after 60 h heating (entry 4). The combination of 2,6‐lutidine and Me₃SiI generated ¹³CH₄ in 76 % yield after 60 h (entry 5). As these reactions were done in a sealed J‐Young NMR tube, the methane was identified by ¹³C NMR spectroscopy as a pentet at −4.3 ppm (¹ J _C–H=126 Hz) and further confirmed by an HSQC experiment, revealing a correlation with the ¹H signal at 0.19 ppm. ^[14] Further improvement in the reactivity was seen with use of CDCl₃ as the solvent as ¹³CH₄ was produced in 85 % yield after 20 h at 100 °C (entry 6). Reactions with the more sterically hindered halosilane Et₃SiI afforded the acetal (Et₃SiO)₂ ¹³CH₂ as the dominant product in 72 % yield after heating at 100 °C for 60 h (entry 7). This product exhibited a doublet at 5.06 ppm in the ¹H NMR with a ¹ J _C–H of 162 Hz and a ¹³C{¹H} NMR signal at 84.5 ppm. Interestingly, performance of the reaction in the more polar solvent CDCl₃ (entry 8) afforded ¹³CH₃I in 82 % yield as evidenced by the quartet resonance in the ¹³C NMR at −23.5 ppm with ¹ J _C–H=151 Hz, while the HSQC experiment revealed a correlation with the ¹H signal at 2.16 ppm. ^[15] Use of the more basic 2,4,6‐collidine resulted in a significant reduction in reactivity affording low yields of the acetal and methoxylsilane in C₆D₆ and CDCl₃, respectively (entry 9, 10), likely due to slightly reduced reactivity for CO₂ reduction though better H₂‐activation reactivity is expected.

Table 1.

CO₂ hydrogenation in the presence of silylhalides.

Ent	Solv.	Silylhalide[a]	base[a]	t [h]	Major product	Yield[b]
1	C6D6	Me3SiCl	Lut	20	‐	<1 %
2	CDCl3	Me3SiCl	Lut	20	‐	<1 %
3	C6D6	Me3SiBr	Lut	40	MeOSiMe3	83 %
4	CDCl3	Me3SiBr	Lut	60	MeOSiMe3	73 %
5	C6D6	Me3SiI	Lut	60	13CH4	76 %
6	CDCl3	Me3SiI	Lut	20	13CH4	85 %
7	C6D6	Et3SiI	Lut	60	(Et3SiO)2 13CH2	72 %
8	CDCl3	Et3SiI	Lut	40	13CH3I	82 %
9	C6D6	Et3SiI	Col	40	(Et3SiO)2 13CH2	8 %
10	CDCl3	Et3SiI	Col	40	MeOSiEt3	9 %

[a] 0.05 mmol silylhalide and Lewis base were added; Lut=2,6‐lutidine; Col=2,4,6 collidine. [b] Yields are determined by ¹H NMR spectroscopy using 10 μL toluene as internal standard.

The above reactions demonstrate that simple tuning of the reaction conditions for FLP hydrogenation of CO₂ provided variation of the major products. While lutidine was identified as the preferred base in the presence of the Lewis acid catalyst B(C₆F₅)₃, the use of Me₃SiBr produced Me₃SiO¹³CH₃, whereas Me₃SiI afforded primarily ¹³CH₄ as the CO₂ reduction product. The acetal, (Et₃SiO)₂ ¹³CH₂, was formed preferentially when Et₃SiI was employed in C₆D₆ solution. Perhaps most remarkably, however was the impact of the use of Et₃SiI in CDCl₃ which resulted in the formation of ¹³CH₃I as the major product (Scheme 3). ^[16]

Scheme 3.

Summary of major products of CO₂ reduction using the FLP catalyst B(C₆F₅)₃/2,6‐lutidine.

Efforts to probe the reaction affording isotopically enriched methyl iodide prompted us to monitor the reaction of ¹³CO₂ (2 atm) and D₂ (2 atm) in the presence of 2,6‐lutidine, Et₃SiI and 10 mol % B(C₆F₅)₃ in CDCl₃ at 100 °C. At this lower pressure and with the shorter reaction time of 24 h, the reaction was not complete. However, the NMR spectra revealed the formation of isotopologues of the acetal and methoxy species in 33 % yield and 21 % yield, respectively. The three isotopologues of the acetal, (Et₃SiO)₂ ¹³CH₂ and (Et₃SiO)₂ ¹³CHD and (Et₃SiO)₂ ¹³CD₂ were formed in an approximately 1:4:1 ratio. The isotopologue (Et₃SiO)₂ ¹³CHD exhibited a triplet in the ¹³C{¹H} NMR spectrum at 84.0 ppm (¹ J _C–D=25 Hz) as well as a doublet at 5.03 ppm (¹ J _C–H=161 Hz) in the ¹H NMR spectrum; while the (Et₃SiO)₂ ¹³CD₂ was found as a pentet in the ¹³C{¹H} NMR spectrum at 83.6 ppm (¹ J _C–D=25 Hz). The four isotopologues of methoxy, Et₃SiO¹³CH_3, Et₃SiO¹³CH₂D, Et₃SiO¹³CHD₂ and Et₃SiO¹³CD₃ were generated in a 1:5:8:4 ratio, each of them was found in the ¹³C{¹H} NMR spectrum at 50.8 ppm, 50.5 ppm, 50.2 ppm and 49.7 ppm as singlet, triplet, pentet and septet resonance with ¹ J _C–D=22 Hz, respectively. In addition, the NMR data showed the formation of H₂ as a singlet at 4.63 ppm and HD as a triplet at 4.59 ppm (J_H–D=43 Hz) and a triplet at 2.39 ppm (² J _H–D=2 Hz) adjacent the methyl resonance of 2,6‐lutidine, which is corresponding to the mono‐methyl‐deuterated 2,6‐lutidine. These data suggest that competitive to reaction with CO₂, the product of initial activation of D₂, [C₅H₃Me₂ND][DB(C₆F₅)₃], can evolve HD, generating a transient enamine, while tautomerization regenerates lutidine leading to H/D scrambling into the methyl groups of lutidine, the generation of HD and H₂, and the generation of the isotopologues of the CO₂ reduction products (Scheme 4). It is noteworthy that on prolonged reaction for 70 h, the above reaction gave 78 % yield of the expected isotopologues of methyl iodide, CH₃I, CH₂DI, CD₂HI and CD₃I in a 1:5:5:4 ratio. These species are observed in the ¹³C{¹H} NMR spectrum at −23.39 ppm, −23.41 ppm, −23.44 ppm and −23.47 ppm as singlet, triplet, pentet and septet resonances, respectively. The deuterated species exhibited ¹ J _C–D values of 23 Hz.

Scheme 4.

a) Deuteration of CO₂ b) deuteration of lutidine, mediated by B(C₆F₅)₃ under D₂.

Mechanistically, the above reactivity indicates that the present hydrogenation of CO₂ begins with the known FLP activation of H₂ followed by the reaction with CO₂ affording a formyl borate anion. Reaction with the silylhalide affords the silyl‐formate and frees the borane for further activation of H₂. Hydrido‐borate attack of the silyl‐formate and reactions with the silylhalide affords the acetal and subsequently the methyloxy‐silane, although the dominance of these reactions depends on the nature of the silyl‐substituent, the halide and the solvent. In a non‐polar solvent, reaction of the methyloxy‐silane with the hydrido‐borate in the presence of the silylhalide affords methane and the disilylether. In contrast, a polar solvent favors attack by iodide, affording methyl iodide as the dominant product.

This view of the reactivity was further probed by extensive DFT calculations at the dispersion‐corrected PW6B95‐D3/def2‐QZVP + COSMO‐RS// TPSS‐D3/def2‐TZVP + COSMO level of theory in chloroform solution, ^[17] using the typical substrates of 2,6‐lutidine (Lut), H₂, CO₂ and Me₃SiI along with the Lewis‐acid B(C₆F₅)₃ as the catalyst. The final PW6B95‐D3 free energies (in kcal mol⁻¹, at 298 K and 1 M concentration) are discussed.

The activation of H₂ by the separated FLP Lut/B(C₆F₅)₃ (Figure 1 A) is −10.0 kcal mol⁻¹ exergonic over a low free energy barrier of 15.9 kcal mol⁻¹ (via TS1) giving the ion pair [LutH]⁺[HB(C₆F₅)₃]⁻ (A). In CHCl₃ solution, the separated ions are 1.1 kcal mol⁻¹ less stable at room temperature but are easily accessible and even more stable upon heating due to favorable entropic effects. In contrast, both CO₂ and Me₃SiI cannot be activated by the FLP, as the adduct LutCOOB(C₆F₅)₃ and the separated ions of [LutSiMe₃]⁺ and I⁻, are 11.5 and 5.1 kcal mol⁻¹ endergonic, respectively (see Supporting Information). However, CO₂ is easily reduced by A via hydride transfer from [HB(C₆F₅)₃]⁻ to the carbon with H‐bonding of [LutH]⁺ to oxygen and the formation of [LutH]⁺[HCOOB(C₆F₅)₃]⁻ (B) is −5.3 kcal mol⁻¹ exergonic over a free energy barrier of only 18.9 kcal mol⁻¹ (via TS2). Consistent with experiment, the reduction of Me₃SiI with A to form Me₃SiH, [LutH]I and regenerated B(C₆F₅)₃ catalyst is 10.1 kcal mol⁻¹ endergonic and thus thermodynamically prevented (see Supporting Information). On the other hand, the reaction between Me₃SiI and B is −1.6 kcal mol⁻¹ exergonic and proceeds easily over a low barrier of 14.3 kcal mol⁻¹ (via TS3⁻ ). This affords the neutral adduct Me₃SiOCHOB(C₆F₅)₃ (C) that still requires 3.9 kcal mol⁻¹ to eliminate B(C₆F₅)₃ and give Me₃SiOCHO (D). Such trapping of B(C₆F₅)₃ with D effectively increases the free energy barrier to the initial H₂‐activation to 19.8 kcal mol⁻¹ (via TS1), which is thus the rate‐limiting step for the formation of D. For comparison, the Lewis bases Lut, Col, Cl⁻ and Br⁻ also form stable B(C₆F₅) adducts that are −2.0, −4.7, −5.7 and −1.3 kcal mol⁻¹ exergonic in CHCl₃ solution (see Supporting Information), respectively. The higher affinity for Col and Cl⁻ may further inhibit H₂‐activation reactivity.

Figure 1.

DFT‐computed free energy paths for: A) the lutidine/B(C₆F₅)₃ FLP‐mediated H₂ activation and further reduction of CO₂ into HCOOSiMe₃; B) further reduction into H₂C(OSiMe₃)₂ and even H₃COSiMe₃; C) slower and kinetically competitive formation of CH₃I and CH₄.

Once intermediate D is formed (Figure 1 B), further reduction via silylium transfer from Me₃SiI (via TS4) and subsequent hydride transfer from A (via TS5) to give the acetal H₂C(OSiMe₃)₂ (E) proceeds quickly and is −13.3 kcal mol⁻¹ exergonic. Further silylium transfer from Me₃SiI to E (via TS6) and subsequent hydride transfer from A (via TS7) to give H₃COSiMe₃ (F), O(SiMe₃)₂ and [LutH]I is still possible over a slightly higher barrier of 20.3 kcal mol⁻¹ (via TS6), but is −39.9 kcal mol⁻¹ exergonic. Under moderate heating, both formation of E and F should be kinetically facile. The use of bulkier silanes such as Et₃SiI may enhance the barrier to silylium transfer and thus slow formation of F, making selective acetal formation possible in less polar benzene solution (Table 1, entry 7).

Silylium transfer from Me₃SiI to F to give the cation H₃CO(SiMe₃)₂ ⁺ (G⁺ ) and the I⁻ anion (via TS8, Figure 1 C), is 10.5 kcal mol⁻¹ endergonic over a low barrier of 16.5 kcal mol⁻¹ and thus is kinetically feasible. Further nucleophilic iodide transfers from [LutH]I to G⁺ to give the experimentally observed CH₃I and O(SiMe₃)₂ is −24.3 kcal mol⁻¹ exergonic over a low barrier of 13.9 kcal mol⁻¹ (via TS9⁺ ). The overall formation of CH₃I from F is thus −13.8 kcal mol⁻¹ exergonic over a sizable barrier of 24.4 kcal mol⁻¹, consistent with the moderate heating required experimentally. On the other hand, nucleophilic hydride transfer from A to G⁺ to give CH₄, O(SiMe₃)₂ and regenerate B(C₆F₅)₃ is −49.0 kcal mol⁻¹ exergonic over a low barrier of 14.3 kcal mol⁻¹ (via TS10). Coupled with the facile H₂ activation, the overall formation of CH₄ from F is thus −53.2 kcal mol⁻¹ exergonic over a barrier of 24.8 kcal mol⁻¹. This is thermodynamically more favorable but kinetically comparable with the formation of CH₃I. Indeed, the use of Et₃SiI and Me₃SiI are found to favor iodide and hydride transfer affording CH₃I and CH₄, respectively.

In conclusion, we have achieved metal‐free catalytic hydrogenation of CO₂ using H₂ and a silylhalide as an oxophile in the presence of a FLP derived from lutidine and B(C₆F₅)₃. The judicious selection of the steric demands and nature of the silylhalide and the solvent provides control of these catalytic reductions affording avenues to the selective formation of the methoxysilane, Me₃SiO¹³CH_3, the acetal (Et₃SiO)₂ ¹³CH₂, ¹³CH₄ and ¹³CH₃I. The complexities of the mechanisms involved have been detailed using DFT studies. We are continuing to explore the use of FLPs in reactions of interest.

Supporting Information available: Synthetic and spectral data, computational details and DFT‐computed energies and Cartesian coordinates are deposited.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

T. Wang, M. Xu, A. R. Jupp, Z.-W. Qu, S. Grimme, D. W. Stephan, Angew. Chem. Int. Ed. 2021, 60, 25771.