One-pot unimolecular reduction of COWayland’s coupling reaction of carbonyl radicals is reversible and gives an equilibrium mixture of [(por)Rh(CO)]• and (por)Rh-CO-CO-Rh(por). Though the latter possesses a C-C bond between the two CO moieties, its extreme steric bulk prevented the activation and transformation of the coupling product. We also made corresponding attempts to add reducing agents (hydrogen, silanes), as well as amines and alcohols to a Rh-CO-CO-Rh complex stabilized by a tether group (-O(CH2)6O- or a m-xylyl diether)17,33,34 between the two porphyrin ligands. Unfortunately, subsequent conversion of the Rh-CO-CO-Rh product was not achieved. Therefore, in the present work, we reduced [(TPP)Rh(CO)]• before subjecting it to a second molecule of CO, leading to successful generation of small-molecule CO reductive coupling products.Addition of triethylsilane (HSi(CH2CH3)3, 5a) to a solution of 1 in C6D6 at room temperature resulted in quantitative formation of 3 (δ (ppm): −40.00 (d, Rh-H, 1JRh-H = 40 Hz)) instantly (Fig. 2A. eq. (1)). Pressurizing a solution of the above system with 8 atm of CO in a J. Young valve NMR tube, then placing the mixture at room temperature for 1 day led to a new doublet at δ 3.17 ppm characteristic of (TPP)Rh-CHO20,23,24,35,36 (Fig. 2A. eq. (2)), together with the gradual consumption of 4 through decomposition to H2 and [(TPP)Rh]2 (Fig. 2A. eq. (2a)-(2d))20,23,24,35,36. Finally, we obtained a mixture of 4, 3 and the poorly soluble [(TPP)Rh]2. Removing the CO gas from the J. Young valve NMR tube caused significant decrease of the concentration of 4 as evidenced by 1H NMR. The above experiments proved that 4 and 3 exist in an equilibrium as mentioned in the Introduction, which increases the difficulty of further transformations of 4. In order to circumvent this difficulty, and in view of the aforementioned lack of reactivity of H2 and silanes without a catalyst, we decided to reduce 4 with 5a under the catalysis of B(C6F5)3. After treatment of the above solution with 2 μL of 5a and 2 mg of B(C6F5)3, we repressurized 1 atm of CO into the J. Young valve NMR tube to afford (TPP)RhCH2OSi(CH2CH3)3 (6a) (Fig. 2A. eq. (3)). The transformation progress was conveniently followed by the appearance of high field 1H NMR resonances of 6a (Supplementary Fig. 1) and a rhodium methylene 13C NMR signal at δ 52.79 ppm (Supplementary Figs. 2 and 3).Fig. 2: One-pot reaction of CO, 5 and 1 catalyzed by B(C6F5)3.A Proposed reaction mechanism and the overall reaction. B Solid state structure of complex 6a·THF. H atoms are omitted for clarity. Gray: carbon, yellow: silicon, red: oxygen, light blue: nitrogen, dark blue-green: rhodium.Interestingly, the yield of 6a was near quantitative, which is much higher than the yield of its precursor, 4. Evidently, the reduction of 4 drove the equilibrium of the reaction of 3 and CO (Fig. 2A. eq. (2)) to the right side, eventually leading to near complete conversion of 3. Combining with the fact that 3 is readily obtainable through the reduction of 1 by silanes, we therefore proposed an one-pot strategy, i.e. performing the reduction of 1, carbonylation of 3, and the hydrosilylation of 4 in one shot, which allowed us to conveniently scale up the reaction. Mixing the benzene solution containing 3, B(C6F5)3 and 5a in a Schlenk flask and then pressurizing 3 atm of CO successfully led to unimolecular reduction of CO in one pot, and stoichiometric production of 6a was achieved after 12 h (Fig. 2A. eq. (4)). Column chromatography readily afforded the pure product, enabling unambiguous structural characterization through single crystal XRD (Fig. 2B).To probe the generality of this transformation, the reactivity of a variety of silanes (5a-5e) was explored (Supplementary Table 1). Bulkier silanes, such as phenyldimethylsilane (HSiMe2Ph, 5d) and methyldiphenylsilane (HSiMePh2, 5e), also underwent this transformation smoothly to give the reduction products in moderate to good yields, manifesting the ease of this transformation. Due to the fact that (TPP)RhCH2OSiR1R2R3 (6a-6e) underwent partial decomposition in the process of column isolation, the isolated yields were about 30 %, lower than those determined by 1H NMR which were nearly quantitative.Release of reduction productsSince the photolysis of (por)Rh-C bonds yields (por)Rh(II) and carbon radicals, a protocol for the release of the C1 reduction product might be easily envisaged by conducting the photolysis of (TPP)RhCH2OSiR1R2R3 (6a-6e) in the presence of a radical trap. Thus, reaction of 6a with excess TEMPO (TEMPO = 2,2,6,6-tetramethylpiperidine-1-oxyl) under visible light irradiation produced TEMPO-Rh(TPP) (Supplementary Fig. 34) and TEMPO-CH2OSi(CH2CH3)3 (9, Supplementary Fig. 35) (Fig. 3A. eq. (5)). Apart from being a route of product release, this reaction also served as an experimental proof that the photolysis of (TPP)RhCH2OSiR1R2R3 proceeded homolytically.Fig. 3: Formation of small-molecule C1 products via the photolysis of 6a.A Visible light promoted reaction of 5a with different reagents. B Formally catalytic reduction of CO mediated by (TPP)RhX (X = I, Br).Because the steric hindrance of TEMPO-Rh(TPP) might prevent the exposure of the rhodium center towards external reagents, which are crucial for recovering the initial state of the catalyst, we have sought to use good hydrogen atom donors instead of TEMPO, such as silanes, in the hope that 3 might be formed instead, hence completing a formal catalytic cycle. To examine the feasibility of this route, excess 5e was added to a solution of 6a in dry C6D6. Surprisingly, after exposure to visible light for 1 h, a 1:1 mixture of (TPP)RhSiMePh2 and 7a was observed by 1H NMR (Fig. 3A. eq. (6)). Compound 7a was obtained by vacuum distillation as a C6D6 solution. A 1H NMR (Supplementary Fig. 36) single peak of CH3OSi(CH2CH3)3 at δ 3.31 ppm together with a singlet of CH3OSi(CH2CH3)3 at δ 50.08 ppm in 13C DEPT NMR (Supplementary Fig. 37) unambiguously confirmed the proposed structure, which was also supported by GC-MS analysis. 6b-6e also underwent this transformation smoothly to give the corresponding products with near quantitative yields (Supplementary Table 1, Supplementary Figs. 38–45).In order to elucidate the sources of H and C atoms in compounds 6a and 7a, isotopic labeling experiments using either (CH3CH2)3SiD or 13CO instead of their natural abundance counterparts were performed, resulting in the formation of (TPP)RhCD2OSi(CH2CH3)3 (6f) and (TPP)Rh13CH2OSi(CH2CH3)3 (6g), respectively. The absence of the doublet at −1.73 ppm in the 1H NMR spectrum of 6f (Supplementary Fig. 30) and the presence of a strong doublet at 52.41 ppm in the 13C NMR spectrum of 6g (Supplementary Fig. 32) confirmed that the -CH2OSi(CH2CH3)3 moiety was attributed to the reduction of CO by silanes. Subsequently, addition of (CH3CH2)3SiD (60 % D) to 6a, followed by irradiation for 2 h at 6 °C, produced DCH2OSi(CH2CH3)3 (7f) (56 % D), which was evidenced by the 1H NMR triplet peak at 3.28 ppm (Supplementary Fig. 46), confirming the origin of the third methyl proton.Unfortunately, (TPP)RhSiMePh2 appeared to be quite inert towards photolysis, unlike its Rh-C homologs: heating at 120 oC or prolonged irradiation under room temperature resulted in no reaction even in the presence of TEMPO as a radical trap. To avoid the generation of (TPP)RhSiMePh2, we turned our attention to another hydrogen source, 3. Paralleling the reactivity of silanes, the photoreaction of 3 and 6a was expected to produce [(TPP)Rh]2, whose reaction with H2 is known to produce 330, thus completing a formal catalytic cycle. Exposure a solution of 6a and 3 in dry C6D6 to visible light for 1 h afforded a dark cloudy orange red solution containing the poorly soluble [(TPP)Rh]2 and (CH3CH2)3SiOCH3 (7a, Fig. 3A. eq. (7)), as revealed by 1H NMR and ESI-MS. The analogous reactions of 6b-6e proceeded similarly with high yields (Supplementary Table 1).Although [(TPP)Rh]2 can react with H2 to complete a formal catalytic cycle (Fig. 2A. eqs. (2a) and (2b)), the reaction is very slow due to the former’s poor solubility. It might thus be desirable to generate a precursor to 3 that is more soluble than [(TPP)Rh]2. Bearing this in mind, we utilized excess BrCCl3 instead of hydrogen sources, and obtained BrCH2OSi(CH2CH3)3 (8) and 2 (Fig. 3A. eq. (8), Supplementary Figs. 47–49) under visible light irradiation for 2 h. Gratifyingly, removing all the components with low boiling point under vacuum, and adding 5a, we obtained 3 again with a doublet peak at −40.00 ppm corresponding to the characteristic of Rh-H in 1H NMR spectrum, suggesting that 2 could be rapidly reduced to 3 by silanes. On the basis of the above observations, a formally catalytic unimolecular reduction of CO was envisioned to proceed through a two-step reaction (Fig. 3B) involving: reduction of CO mediated by the precursor (1 or 2) and 5a in the presence of B(C6F5)3, giving the intermediate 6a; production of 8 and 2 through photolysis of 6a in the presence of BrCCl3, completing the formal catalytic cycle.Reductive coupling of COThe facile homolysis of 6a prompted us to explore the possibility of inserting a second CO molecule into the Rh-C bond, which would constitute a formal reductive coupling of two CO molecules. Pressurizing 8 atm of CO into a dry C6D6 solution containing 6a in a J. Young valve NMR tube, and then irradiating with visible light at 6 oC over 20 h, led to the CO insertion product (TPP)RhCOCH2OSi(CH2CH3)3 (10a) with a yield of 92 % (Fig. 4A. eq. (9)). The existence of a carbonyl group caused the doublet attributable to Rh-CH2OSi(CH2CH3)3 (δ (ppm): −1.72 (d), Supplementary Fig. 1) to shift to lower field (δ (ppm): −1.52 (s), Supplementary Fig. 54) in 1H NMR spectrum and collapse into a singlet due to attenuated JRh-H coupling. A doublet with 1JRh-C = 31.7 Hz at δ 198.00 ppm in 13C NMR spectrum (Supplementary Fig. 55) and a single carbonyl stretch peak at 1735 cm−1 in IR spectrum (Supplementary Fig. 59) further confirmed this assignment. After chromatographic isolation, pure 10a was obtained and crystallized as its THF adduct, enabling the determination of its structure through XRD, as shown in Fig. 4B and Supplementary Table 6.Fig. 4: Reaction of the second molecule of CO with 6a.A Insertion and reversible coordination under light and dark conditions, respectively. B Solid state structure of complex 10a·THF. H atoms are omitted for clarity. Gray: carbon, yellow: silicon, red: oxygen, light blue: nitrogen, dark blue-green: rhodium.To prove the origin of the second carbon atom, an isotope labeling experiment was carried out by replacing CO with 13CO. Accordingly, the production of (TPP)Rh13C(O)CH2OSi(CH2CH3)3 (10f, Fig. 4A. eq. (10)) was confirmed by the splitting of the δ −1.52 ppm methylene signal to a doublet (2JC-H = 2.1 Hz) in 1H NMR spectrum (Supplementary Fig. 78) and the appearance of a strong doublet signal (1JRh-C = 31.2 Hz) at δ 198.00 ppm in 13C NMR spectrum (Supplementary Fig. 79), characteristic of the newly formed carbonyl group. Furthermore, the methylene carbon signal ((TPP)Rh13COCH2OSi(CH2CH3)3) was split into a dd pattern, with coupling constants 1JC-C = 33.2 Hz and 2JRh-C = 4.7 Hz. Thus, introducing light irradiation helped the system to overcome the high thermal barrier of the cleavage of the strong (por)Rh-C bond, causing the reductive insertion of the second CO molecule to occur under mild conditions with good substrate universality (Supplementary Table 5).To probe the exact mechanism of the insertion process, we treated 6a with 13CO (6a-13CO) in the dark (Fig. 4A. eq. (11)), which resulted in the shift of the methylene peak from −1.72 ppm to −1.84 ppm, consistent with the trans effect of a coordinated CO. Variable-temperature NMR studies showed fluxional behavior. A trend consistent with the trans effect of CO was observed in 1H NMR, where the rhodium-bound methylene peak ((TPP)Rh(13CO)CH2OSi(CH2CH3)3) shifted to lower fields with increasing temperature as the coordinated CO was gradually detached from the rhodium center (Fig. 5). The 13C NMR spectra (Supplementary Fig. 53) displayed a singlet, the chemical shift of which (185.28 ppm) being within the typical range of Rh(III) carbonyl complexes31. Based on the above observations, we conclude that the intermediate (TPP)Rh(CO)(CH2OSi(CH2CH3)3) (6a-CO) was formed, where the CO is coordinated to the trans position of the CH2OSi(CH2CH3)3 ligand.Fig. 5: Variable-temperature 1H NMR monitoring of the (TPP)Rh(13CO)(CH2OSi(CH2CH3)3) methylene signal in toluene-d8.A progressive shift to the lower fields is seen upon increasing the temperature from −50 oC to 70 oC.A variety of mechanisms can be proposed for the CO insertion step. The Rh-C bond of 6a can be photolyzed to yield [(TPP)Rh(II)]• and •CH2OSi(CH2CH3)3 (as suggested by the TEMPO trapping experiment, Fig. 3A, eq. (5)), both of which might react with CO before being trapped by the other open-shell fragment to yield 10a (Fig. 6A, a, b). Alternatively, the reaction might proceed through the photolysis of 6a-CO, forming [(TPP)Rh(CO)]•19 without recourse to the bimolecular reaction of [(TPP)Rh(II)]• and CO (Fig. 6A, c). Computational studies performed by ORCA37 indicate that the reaction of •CH2OSi(CH2CH3)3 with CO in pathway (a) has a barrier of 10.9 kcal/mol, while all other ground state reactions, including the coordination of CO to 6a and [(TPP)Rh(II)]• as well as all radical recombination steps, are barrierless (Fig. 6B). Therefore, pathways (b) and (c) in Fig. 6B are both facile and are probably more important than pathway (a). As shown in Fig. 6B, the equilibrium of 6a and 6a-CO is almost thermoneutral and slightly favors 6a-CO. The result is clearly consistent with the variable-temperature NMR results that both species must have contributed substantially to the equilibrium. Nevertheless, the ground state Rh-C(alkyl) bond dissociation free energy (BDFE) of 6a-CO is less than that of 6a by 4.7 kcal/mol, owing to the strong trans effect of the CO ligand (the (TPP)Rh(CO)-(CH2OSi(CH2CH3)3) bond length, 2.050 Å, is 0.030 Å longer than that of 6a, 2.020 Å; Supplementary Table 8), as well as the spin delocalization effect in [(TPP)Rh(CO)]•19 which stabilizes the dissociation product. Thus, the Rh-C bond cleavage in pathway (c) is thermodynamically less unfavorable than the Rh-C bond cleavage in pathway (b). Finally, we note that our NMR experiments failed to detect any coupling product of two carbon radicals, which can be explained by the persistent radical effect, since [(TPP)Rh(II)]• and [(TPP)Rh(CO)]• are much more stable than the carbon radicals involved in Fig. 6. This may be one of the reasons for the observed high selectivity (> 99 %) of the CO insertion reactions, despite the involvement of highly reactive radicals. An alternative explanation would be that the CO molecule reacts simultaneously with the nascent radicals [(TPP)Rh(II)]• and/or •CH2OSi(CH2CH3)3 before they leave the solvent cage; while we failed to locate any stable singlet radical pair between these two species, we found that CO can react with the •CH2OSi(CH2CH3)3 radical in the triplet radical pair, with a somewhat smaller barrier (6.8 kcal/mol) than pathway (a). Due to the triplet multiplicity, however, the reaction is not accompanied by concomitant Rh-C bond formation. Therefore, the reaction of CO with the (TPP)Rh…CH2OSi(CH2CH3)3 radical pair is more appropriately described as a special case of pathway (a).Fig. 6: Light-induced CO insertion of (TPP)Rh-CH2OSi(CH2CH3)3.A Schematic depiction of possible mechanisms. B Gibbs free energy profile of the CO insertion of 6a. All ground state reactions shown in this figure for which no transition states are given are barrierless. Green, blue and magenta represent the mechanisms (a), (b) and (c), respectively.To gain further insights into the photolysis steps, we have computed the excited states of 6a, 10a and 6a-CO at their respective ground state equilibrium geometries (Fig. 7). The computed absorptions of 6a and 10a match very well with the respective experimental UV-Vis spectra measured in toluene (Fig. 7A). Natural transition orbital (NTO) analyses reveal that the bright states of 6a and 10a in the visible region (at 423(415) and 534(522) nm for 6a(10a), commonly known as the Soret and Q bands, respectively) are due to excitations from the two highest ligand π orbitals (a1u-like and a2u-like; herein we have named the orbitals by the irreducible representations of the D4h group, as is conventionally done) to the two lowest ligand π* orbitals (eg-like), in accord with the well-known Gouterman four-orbital model for porphyrin complexes. The a2u-like ligand orbital contains progressively larger contributions from the Rh-C(alkyl) σ orbital upon going from 6a to 10a to 6a-CO, reaching a maximum of 28 % in the S4 state of 6a-CO (Fig. 7B); by contrast, while the a1u-like and eg-like orbitals contain small contributions from the Rh dxz and dyz orbitals, in none of the cases do they show noticeable contributions from the σ(Rh-C(alkyl)) orbital. This can be explained by symmetry arguments: among the four Gouterman orbitals, only the a2u-like orbital has no angular node with respect to the approximate four-fold symmetry axis of the porphyrin ligand, which allows it to mix with the Rh-C(alkyl) σ orbital. Therefore, the Rh-C bond orders of 6a, 10a and 6a-CO are reduced by photo-excitation at either the Soret band or the Q band (Fig. 7C), due to the removal of electron density from the a2u-like orbital, and therefore from the Rh-C(alkyl) bonding orbital; this is expected to facilitate Rh-C bond cleavage.Fig. 7: Absorption spectra and excited state compositions of 6a, 10a and 6a-CO.A UV-Vis spectra of 6a (experimental), 10a (experimental) and 6a-CO (computational) (red) with computed (TD-PBE0/x2c-TZVPall) absorptions (black) and dominant NTO transitions. The experimental spectra were recorded in toluene. B Side views of a2u-like NTOs of all excited states, highlighting the contributions of Rh-C bonds. C Orbital energy expectation values of the a2u-like NTOs and relevant NBOs, and the Rh-C Wiberg bond indices of ground and excited states. The energy splitting due to linear combination of the n(CO) and σ(Rh-C(alkyl)) NBOs is estimated from their off-diagonal Fock matrix element.The fact that 6a-CO exhibits the largest σ(Rh-C(alkyl)) component in the a2u-like orbital among the three complexes can be rationalized by the orbital interaction between the lone pair orbital of the trans CO ligand (n(CO)) and the σ(Rh-C(alkyl)) orbital, as revealed by natural bonding orbital (NBO) analysis (Fig. 7C). The interaction results in in-phase and out-of-phase combinations of the two NBOs, the latter being nearly 3 eV higher than the σ(Rh-C(alkyl)) NBO itself. As a result, the out-of-phase NBO combination is energetically very close to the porphyrin a2u-like orbital (here approximately represented by the energy expectation value of the a2u-like NTO), and thus mixes favorably with the latter. Interestingly, the out-of-phase NBO combination has anti-bonding character between Rh and the CO ligand; therefore, removal of electron density from the a2u-like orbital strengthens rather than weakens the Rh-CO bond. This is not only apparent from the Rh-C bond orders (Fig. 7C), but also from the excited state equilibrium geometries and BDFEs. While the Rh-C bond lengths of 6a and 10a do not change noticeably upon going from the S0 state to the S1 state, 6a-CO shows a 0.300 Å increase and a 0.215 Å decrease of the Rh-C(alkyl) and Rh-CO bond lengths, respectively; moreover, the Rh-C(alkyl) BDFE becomes 22.4 kcal/mol lower than the Rh-CO BDFE in the S1 state, even though the former is 32.4 kcal/mol higher than the latter in the S0 state (Supplementary Table 8). Combined with excited state potential energy curve calculations (Supplementary Figs. 96–97), we found that Q band absorption of 6a-CO leads to selective cleavage of the Rh-C(alkyl) bond but not of the Rh-CO bond, contrary to what would be predicted from ground state BDFEs. A similar but smaller trend is found for the T1 state (Supplementary Table 8).Apart from CO coordination, the σ(Rh-C(alkyl)) orbital energy can be also raised by a lengthening of the Rh-C(alkyl) bond, due to the narrowing of the σ-σ* gap. Therefore, the S1 states of all three complexes (formed by irradiating the Q band) gradually acquire σ-π* character when the Rh-C(alkyl) bond stretches beyond the excited state equilibrium. Upon further stretching, the high-lying σ*(Rh-C(alkyl)) drops below the porphyrin π* orbitals, resulting in a further transition into 3(σ-σ*) character and therefore complete rupture of the Rh-C bond (Supplementary Figs. 94–96). For photolysis via Q band absorption, the transition from π-π* character to σ-σ* character is associated with a barrier of 5.3 (6a), 8.4 (10a) and 4.5 (6a-CO) kcal/mol, respectively, suggesting that the Rh-C(alkyl) photolysis of 6a-CO may be slightly more kinetically favorable than 6a (Supplementary Figs. 94–96); this is consistent with the reduced Rh-C(alkyl) bond order in the S1 state due to CO coordination (Fig. 7C). Photolysis via irradiating the Soret band is more complex due to numerous state crossings and may involve the dπ-π* states, as suggested by the excited state potential energy curves. In general, Soret band photolysis is expected to be roughly equally efficient for the three complexes due to the availability of barrierless or nearly barrierless relaxation pathways that lead to Rh-C homolysis (see discussions of Supplementary Figs. 94–96). To summarize, while for Q band absorption we predict a slight advantage of pathway (c) compared to pathway (b), for Soret band absorption both pathways are about equally viable.Release of reductive coupling productsGiven the successful synthesis of the reductively coupled products (TPP)RhCOCH2OSiR1R2R3 (10a-10e), we turned to the release of the reductively coupled fragments. We have previously reported that 4 could react with n-propylamine to produce HCONHnPr and 3 through a four-membered ring transition state (Fig. 8A)19. On account of the similarity in the structures of 4 and 10a-10e, we proposed that the reactions of 10a-10e with amines would probably give the respective amides as well (Fig. 8B). Indeed, adding 0.5 μL of n-propylamine to a 1 mg 10a solution in 300 μL of C6D6, followed by heating at 60 °C for 3 h, resulted in the formation of nPrNHCOCH2OSi(CH2CH3)3 (11a) with a yield of 41 %, as shown in Fig. 8C. eq. (12). To aid characterization, the isotopically labeled reactant (TPP)Rh13COCH2OSi(CH2CH3)3 was used which produced nPrNH13COCH2OSi(CH2CH3)3 (11b) (Fig. 8C. eq. (13)) with the signature carbon peak at 169.98 ppm in 13C NMR, the signature proton peak at 4.08 ppm (d, 2J13C-H = 4.9 Hz; Supplementary Fig. 76) and an ESI-MS spectrum consistent with the incorporation of one 13C ([M + H]+, 233.17601; Supplementary Fig. 78).Fig. 8: The reaction of (TPP)Rh-COR’ with amines.A Reported mechanism for R’ = H19. B Proposed mechanism for R’ = CH2OSiR1R2R3. C Reaction of (TPP)RhCOCH2OSi(CH2CH3)3 and (TPP)Rh13COCH2OSi(CH2CH3)3, with nPrNH2.In order to circumvent the low yield of the release product, we designed reaction pathways for more efficient release of the C2 fragment. Inspired by the similarity of the Rh-C bonds of 10a-10e with those of 6a-6e, and the computational feasibility of the Rh-C bond photolysis of 10a, we proposed that 10a-10e might also undergo Rh-C bond cleavage under photochemical conditions, which would serve as a facile way of product release. Following the previously established protocols of product release of 6a-6e, TEMPO was first selected as a radical trap. Treating 1 mg of 10a with excess TEMPO (~10 eq.) and exposing to visible light led to the formation of TEMPO-COCH2OSi(CH2CH3)3 (12a) (Fig. 9A. eq. (14)) (Supplementary Fig. 80) and TEMPO-Rh(TPP) (Supplementary Fig. 28) as the major products (over 85 % yield as determined by 1H NMR spectroscopy, Supplementary Fig. 79). The chemical shift of TEMPO-COCH2OSi(CH2CH3)3 (4.22 ppm) appeared downfield from TEMPO-CH2OSi(CH2CH3)3 (3.84 ppm), consistent with the presence of a carbonyl group. Next, the hydrogen sources, namely silanes and 3, were employed as radical traps. Contrary to the high efficacy of the release of the C1 products, however, we failed to trap the carbonyl radical fragments because they are prone to decarbonylation (Fig. 9A. Eqs. (15), (16)). Instead, the C1 product 7a was observed as the exclusive product, manifesting the hydrogen sources as poorer radical traps compared to TEMPO. To our delight, the trapping reaction using BrCCl3 proceeded smoothly to give the release products 2 and BrCOCH2OSi(CH2CH3)3 (13) with near quantitative NMR yield, as shown in Fig. 9A. eq. (17). The structure of 13 was verified by 1H NMR (Supplementary Fig. 81) and ESI-MS (Supplementary Fig. 82). To our understanding, the production of an acyl halide through the reductive coupling of CO, as demonstrated in this example, has not been previously achieved. The present approach thus significantly expands the potential for subsequent transformations of CO-derived C2 products due to the unique reactivity of the acyl halide functionality. Considering that the release of CO reductive coupling products from metal centers is inherently difficult, the straightforward construction of a C-Br bond during product release, whose bond energy does not offer a significant thermodynamic driving force for the reaction, is truly remarkable. Herein, the utilization of light as an energy source was apparently the key to realizing such reactivity.Fig. 9: Formation of small-molecule C2 products via 10.A Visible light promoted reaction of 10 with different reagents. B Formally catalytic reductive coupling of CO mediated by (TPP)RhX (X = I, Br).A three-step formally catalytic bimolecular reduction of CO was proposed, based on the above observations (Fig. 9B). The steps are the reduction of CO mediated by the precursor (TPP)RhX to give the intermediate 6; photolysis of Rh-C bond and insertion of the second molecule of CO to produce 10; production of 13 and 2 through photolysis of 10 in the presence of BrCCl3, completing the formal catalytic cycle.
